1 2018-03-02 Richard Biener <rguenther@suse.de>
3 PR tree-optimization/84427
4 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5 (bitmap_set_subtract_values): Rewrite to handle multiple
8 (prune_clobbered_mems): Likewise.
9 (phi_translate): Take edge instead of pred/phiblock.
10 (phi_translate_1): Likewise.
11 (phi_translate_set): Likewise. Insert all translated
12 exprs for a value into the set, keeping possibly multiple
13 expressions per value.
14 (compute_antic_aux): Adjust for phi_translate changes.
15 When intersecting union the expressions and prune those
16 not in the final value set, keeping possibly multiple
17 expressions per value. Do not use value-insertion
18 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
19 all expressions. Add verification that the value-sets
20 only shrink during iteration.
21 (compute_partial_antic_aux): Adjust for the phi_translate changes.
22 (do_pre_regular_insertion): Likewise.
23 (do_pre_partial_partial_insertion): Likewise.
25 2018-03-02 Richard Biener <rguenther@suse.de>
28 * config/darwin.c (saved_debug_info_level): New static global.
29 (darwin_asm_lto_start): Disable debug info generation for LTO out.
30 (darwin_asm_lto_end): Restore debug info generation settings.
32 2018-03-01 Martin Liska <mliska@suse.cz>
35 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
38 2018-03-01 Richard Biener <rguenther@suse.de>
41 * dwarf2out.c (gen_variable_die): Properly handle late VLA
42 type annotation with LTO when debug was disabled at compile-time.
44 2018-03-01 Matthew Fortune <mfortune@gmail.com>
46 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
48 (mips_final_postscan_insn): Likewise.
50 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
52 PR rtl-optimization/84528
53 * alias.c (init_alias_target): Add commentary.
54 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
55 a unique base value if the frame pointer is not eliminated
58 2018-03-01 Tom de Vries <tom@codesourcery.com>
60 PR rtl-optimization/83327
61 * lra-int.h (hard_regs_spilled_into): Declare.
62 * lra.c (hard_regs_spilled_into): Define.
63 (init_reg_info): Init hard_regs_spilled_into.
64 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
65 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
66 (process_bb_lives): Handle hard_regs_spilled_into.
67 (lra_create_live_ranges_1): Before doing liveness propagation, clear
68 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
70 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
72 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
73 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
74 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
75 * config/rs6000/aix72.h: New file.
77 2018-02-28 Jakub Jelinek <jakub@redhat.com>
79 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
80 instead of warning_at with conditional singular and plural messages
84 * stor-layout.c (update_alignment_for_field): For
85 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
86 && !DECL_PACKED (field), do the alignment update, just use
87 only desired_align instead of MAX (type_align, desired_align)
89 (place_field): Don't do known_align < desired_align handling
90 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
91 is non-NULL, instead do it after rli->prev_field handling and
92 only if not within a bitfield word. For DECL_PACKED (field)
93 use type_align of BITS_PER_UNIT.
95 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
97 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
98 superfluous parentheses and trailing spaces.
100 2018-02-28 Richard Biener <rguenther@suse.de>
102 PR tree-optimization/84584
103 * graphite-scop-detection.c (scop_detection::add_scop): Discard
104 SCoPs with fake exit edge.
106 2018-02-28 Martin Liska <mliska@suse.cz>
109 * timevar.c (timer::print): Fix format to properly print 100%
112 2018-02-28 Richard Biener <rguenther@suse.de>
115 * genmatch.c (capture_info::walk_match): Do not mark
116 captured expressions without operands as expr_p given
117 they act more like predicates and should be subject to
118 "lost tail" side-effect preserving.
120 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
122 PR rtl-optimization/81611
123 * auto-inc-dec.c (attempt_change): Move dead note from
124 mem_insn if it's the next use of regno
125 (find_address): Take address use of reg holding
126 non-incremented value. Add parm to limit search to the named
128 (merge_in_block): Attempt to use a mem insn that is the next
129 use of the original regno.
131 2018-02-27 Martin Sebor <msebor@redhat.com>
134 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
135 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
137 2018-02-27 Martin Sebor <msebor@redhat.com>
140 * diagnostic-core.h (warning_n, error_n, inform_n): Change
141 n argument to unsigned HOST_WIDE_INT.
142 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
143 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
144 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
145 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
147 2018-02-27 Richard Biener <rguenther@suse.de>
149 PR tree-optimization/84512
150 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
151 Do not use the estimate returned from record_stmt_cost for
152 the scalar iteration cost but sum properly using add_stmt_cost.
154 2018-02-27 Richard Biener <rguenther@suse.de>
156 PR tree-optimization/84466
157 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
158 Adjust last change to less strictly validate use operands.
160 2018-02-27 Martin Liska <mliska@suse.cz>
162 PR gcov-profile/84548
163 * gcov.c (process_file): Allow partial overlap and consider it
164 also as group functions.
165 (output_lines): Properly calculate range of lines for a group.
167 2018-02-27 Martin Liska <mliska@suse.cz>
169 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
170 'ggc' suffixes. Change first column width.
171 (timer::print): Fix formatting of the column.
173 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
175 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
176 preserve inline entry blocks for the sake of debug inline
177 entry point markers alone.
178 (remove_unused_locals): Suggest in comments a better place to
179 force the preservation of inline entry blocks that are
180 otherwise unused, but do not preserve them.
182 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
184 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
186 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
189 * config/i386/constraints.md (Bs): Replace
190 ix86_indirect_branch_register with
191 TARGET_INDIRECT_BRANCH_REGISTER.
193 * config/i386/i386.md (indirect_jump): Likewise.
194 (tablejump): Likewise.
195 (*sibcall_memory): Likewise.
196 (*sibcall_value_memory): Likewise.
197 Peepholes of indirect call and jump via memory: Likewise.
198 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
199 (*sibcall_value_GOT_32): Likewise.
200 * config/i386/predicates.md (indirect_branch_operand): Likewise.
201 (GOT_memory_operand): Likewise.
202 (call_insn_operand): Likewise.
203 (sibcall_insn_operand): Likewise.
204 (GOT32_symbol_operand): Likewise.
205 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
207 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
209 PR rtl-optimization/83496
210 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
211 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
212 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
213 redundant insn, if any.
214 (relax_delay_slots): Likewise.
215 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
217 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
219 PR tree-optimization/83965
220 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
221 that grouped statements are part of a reduction chain. Return
222 true if the statement is not marked as a reduction itself but
224 (vect_recog_dot_prod_pattern): Don't check whether the statement
225 is part of a group here.
226 (vect_recog_sad_pattern): Likewise.
227 (vect_recog_widen_sum_pattern): Likewise.
229 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
232 * final.c (rest_of_clean_state): Also look for calls inside sequences.
234 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
237 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
239 (ix86_output_indirect_function_return): New prototype.
240 (ix86_split_simple_return_pop_internal): Likewise.
241 * config/i386/i386.c (indirect_return_via_cx): New.
242 (indirect_return_via_cx_bnd): Likewise.
243 (indirect_thunk_name): Handle return va CX_REG.
244 (output_indirect_thunk_function): Create alias for
245 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
246 (ix86_output_indirect_jmp): Remove the bool argument.
247 (ix86_output_indirect_function_return): New function.
248 (ix86_split_simple_return_pop_internal): Likewise.
249 * config/i386/i386.md (*indirect_jump): Don't pass false
250 to ix86_output_indirect_jmp.
251 (*tablejump_1): Likewise.
252 (simple_return_pop_internal): Change it to define_insn_and_split.
253 Call ix86_split_simple_return_pop_internal to split it for
255 (simple_return_indirect_internal): Call
256 ix86_output_indirect_function_return instead of
257 ix86_output_indirect_jmp.
259 2018-02-26 Jakub Jelinek <jakub@redhat.com>
262 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
263 memset and value initialization afterwards.
265 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
267 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
269 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
272 * common/config/aarch64/aarch64-common.c
273 (aarch_option_optimization_table[]): Switch
274 off fomit-frame-pointer
276 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
277 Chung-Ju Wu <jasonwucj@gmail.com>
279 * config/nds32/nds32-multiple.md (load_multiple): Disallow
281 (store_multiple): Ditto.
283 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
285 * config.gcc: Add --with-cpu support for nds32 target.
286 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
287 * config/nds32/nds32.opt: Add -mcpu= option.
289 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
291 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
292 isel=yes): Warn for these deprecated options.
294 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
296 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
297 ISA_2_5_MASKS_EMBEDDED.
299 2018-02-23 Jakub Jelinek <jakub@redhat.com>
301 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
302 p->max as pointers rather than using iterative_hash_expr.
304 2018-02-23 Carl Love <cel@us.ibm.com>
306 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
307 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
309 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
310 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
311 P8V_BUILTIN_VEC_VUNSIGNED2.
313 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
316 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
317 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
319 * lra-constraints.c (curr_insn_transform): Set up
320 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
322 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
323 * lra-eliminations.c (spill_pseudos): Ditto.
324 (process_insn_for_elimination): Ditto.
325 * lra-lives.c (reg_early_clobber_p): Use the new macros.
326 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
327 LRA_NON_CLOBBERED_ALT.
329 2018-02-22 Martin Sebor <msebor@redhat.com>
331 PR tree-optimization/84480
332 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
333 to maybe_diag_stxncpy_trunc. Call it.
334 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
335 from gimple_fold_builtin_strcpy. Print inlining stack.
336 (handle_builtin_stxncpy): Print inlining stack.
337 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
339 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
342 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
343 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
344 and -fcheck-pointer-bounds are used together.
345 (indirect_thunk_prefix): New enum.
346 (indirect_thunk_need_prefix): New function.
347 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
348 "_nt" instead of "_bnd" for NOTRACK prefix.
349 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
350 (output_indirect_thunk_function): Likewise.
352 (ix86_code_end): Update output_indirect_thunk_function calls.
353 (ix86_output_indirect_branch_via_reg): Replace
354 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
355 (ix86_output_indirect_branch_via_push): Likewise.
356 (ix86_output_function_return): Likewise.
357 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
358 incompatible with -fcf-protection=branch and
359 -fcheck-pointer-bounds.
361 2018-02-22 Steve Ellcey <sellcey@cavium.com>
364 * config/aarch64/aarch64.c (aarch64_print_address_internal):
365 Change gcc_assert call to output_operand_lossage.
367 2018-02-22 Steve Ellcey <sellcey@cavium.com>
369 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
371 2018-02-22 DJ Delorie <dj@redhat.com>
372 Sebastian Perta <sebastian.perta@renesas.com>
373 Oleg Endo <olegendo@gcc.gnu.org>
375 * config/rx/rx.c (rx_rtx_costs): New function.
376 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
378 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
380 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
382 2018-02-22 Martin Liska <mliska@suse.cz>
385 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
386 Add "native" as a possible value.
388 2018-02-22 Martin Liska <mliska@suse.cz>
391 * config/i386/i386.c (ix86_option_override_internal):
392 Add "native" as a possible value for -march and -mtune.
394 2018-02-22 Jakub Jelinek <jakub@redhat.com>
397 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
398 to all type variants.
400 PR tree-optimization/84503
401 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
402 width as info->bitpos + info->bitsize - start.
403 (merged_store_group::merge_overlapping): Simplify width computation.
404 (check_no_overlap): New function.
405 (imm_store_chain_info::try_coalesce_bswap): Compute expected
406 start + width and last_order of the group, fail if check_no_overlap
408 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
409 to group if check_no_overlap fails.
411 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
413 * config/rs6000/altivec.md: Delete contraint arguments to
414 define_expand, define_split, and define_peephole2, and in
415 define_insn_and_split if always unused.
416 * config/rs6000/darwin.md: Ditto.
417 * config/rs6000/dfp.md: Ditto.
418 * config/rs6000/rs6000.md: Ditto.
419 * config/rs6000/sync.md: Ditto.
420 * config/rs6000/vector.md: Ditto.
421 * config/rs6000/vsx.md: Ditto.
423 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
425 * config/rs6000/altivec.md: Write output control strings as braced
426 blocks instead of double-quoted strings.
427 * config/rs6000/darwin.md: Ditto.
428 * config/rs6000/rs6000.md: Ditto.
429 * config/rs6000/vector.md: Ditto.
430 * config/rs6000/vsx.md: Ditto.
432 2018-02-21 Jason Merrill <jason@redhat.com>
434 PR c++/84314 - ICE with templates and fastcall attribute.
435 * attribs.c (build_type_attribute_qual_variant): Remove assert.
437 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
439 * ipa-cp.c (determine_versionability): Fix comment typos.
441 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
444 * ipa-cp.c (determine_versionability): Do not version functions caling
447 2018-02-21 Martin Liska <mliska@suse.cz>
450 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
451 Add "native" as a possible value.
452 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
453 the macro when native cpu detection is available.
455 2018-02-21 Martin Liska <mliska@suse.cz>
458 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
459 Add "native" as a possible value.
460 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
461 when native cpu detection is available.
463 2018-02-21 Jakub Jelinek <jakub@redhat.com>
464 Martin Sebor <msebor@redhat.com>
466 PR tree-optimization/84478
467 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
469 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
470 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
471 support which is conservatively correct, for 2 only stay conservative
472 for maxlen. Formatting and comment capitalization fixes. Add STRICT
473 argument to the 2 argument get_range_strlen, adjust 6 arg
474 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
476 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
477 (gimple_fold_builtin_strlen): Pass true as last argument to
480 2018-02-20 Martin Sebor <msebor@redhat.com>
483 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
484 (builtin_memref::set_base_and_offset): Same. Handle inner references.
485 (builtin_memref::builtin_memref): Factor out parts into
486 set_base_and_offset and call it.
488 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
491 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
492 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
493 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
494 search at the associated MODE_INT.
496 2018-02-20 Jeff Law <law@redhat.com>
499 PR tree-optimization/81592
501 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
502 for range data rather than using global data.
503 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
504 range data rather than using global data.
505 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
506 pass it to children as needed.
507 (struct directive::fmtresult): Similarly.
508 (struct directive::set_width): Similarly.
509 (struct directive::set_precision): Similarly.
510 (format_integer, format_directive, parse_directive): Similarly.
511 (format_none): Accept unnamed vr_values parameter.
512 (format_percent, format_floating, format_character): Similarly.
513 (format_string, format_plain): Similarly.
514 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
515 the EVRP range analyzer for range data rather than using global data.
516 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
517 gimple-ssa-evrp-analyze.h
518 (class sprintf_dom_walker): Add after_dom_children member function.
519 Add evrp_range_analyzer member.
520 (sprintf_dom_walker::before_dom_children): Call into the EVRP
521 range analyzer as needed.
522 (sprintf_dom_walker::after_dom_children): New member function.
523 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
525 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
526 (evrp_range_analyzer::pop_to_marker): Likewise.
528 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
530 PR tree-optimization/84419
531 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
532 with the required type if its current type is compatible but
535 2018-02-20 Jakub Jelinek <jakub@redhat.com>
538 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
541 2018-02-20 Martin Liska <mliska@suse.cz>
544 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
545 possible values if we don't have a hint.
547 2018-02-20 Martin Liska <mliska@suse.cz>
551 * final.c (shorten_branches): Build align_tab array with one
553 * opts.c (finish_options): Add alignment option limit check.
554 (MAX_CODE_ALIGN): Likewise.
555 (MAX_CODE_ALIGN_VALUE): Likewise.
556 * doc/invoke.texi: Document maximum allowed option value for
557 all -falign-* options.
559 2018-02-19 Jakub Jelinek <jakub@redhat.com>
562 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
563 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
564 * var-tracking.c (emit_note_insn_var_location): Remove all references
565 to NOTE_INSN_CALL_ARG_LOCATION.
566 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
567 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
568 Use copy_rtx_if_shared.
569 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
570 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
571 (dwarf2out_var_location): Remove handling of
572 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
574 * final.c (final_scan_insn): Remove all references to
575 NOTE_INSN_CALL_ARG_LOCATION.
576 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
577 before dumping final insns.
578 * except.c (emit_note_eh_region_end): Remove all references to
579 NOTE_INSN_CALL_ARG_LOCATION.
580 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
581 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
582 * config/arc/arc.c (hwloop_optimize): Likewise.
583 * config/arm/arm.c (create_fix_barrier): Likewise.
584 * config/s390/s390.c (s390_chunkify_start): Likewise.
585 * config/sh/sh.c (find_barrier): Likewise.
586 * config/i386/i386.c (rest_of_insert_endbranch,
587 ix86_seh_fixup_eh_fallthru): Likewise.
588 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
589 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
590 * config/frv/frv.c (frv_function_prologue): Likewise.
591 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
593 (note_outside_basic_block_p): Remove all references to
594 NOTE_INSN_CALL_ARG_LOCATION.
595 * gengtype.c (adjust_field_rtx_def): Likewise.
596 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
598 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
599 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
602 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
605 PR tree-optimization/84452
606 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
607 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
610 2018-02-19 Martin Liska <mliska@suse.cz>
613 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
615 2018-02-19 Martin Liska <mliska@suse.cz>
616 Richard Sandiford <richard.sandiford@linaro.org>
618 PR tree-optimization/82491
619 * gimple-fold.c (get_base_constructor): Make earlier bail out
622 2018-02-19 Carl Love <cel@us.ibm.com>
624 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
625 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
627 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
630 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
632 * config/rl78/rl78.md (movdf): New define expand.
634 2018-02-19 Martin Liska <mliska@suse.cz>
637 * doc/invoke.texi: Fix typo.
638 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
640 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
642 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
643 handle rs6000_single_float and rs6000_double_float specially for
646 2018-02-16 Jeff Law <law@redhat.com>
648 * config/rx/rx.c (add_pop_cfi_notes): New function.;
651 2018-02-16 Jakub Jelinek <jakub@redhat.com>
654 * ipa-inline.c (inline_small_functions): Fix a typo.
656 2018-02-16 Nathan Sidwell <nathan@acm.org>
658 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
660 2018-02-16 Carl Love <cel@us.ibm.com>
662 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
663 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
664 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
665 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
666 expansion to P8V_BUILTIN_VEC_FLOAT2.
668 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
670 PR rtl-optimization/70023
671 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
672 src_regno into account.
674 2018-02-16 Carl Love <cel@us.ibm.com>
676 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
677 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
678 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
679 * config/rs6000/rs6000.c: Remove case statements for
680 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
681 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
682 and P9V_BUILTIN_VEC_VINSERT4B.
683 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
684 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
685 * config/rs6000/vsx.md:
686 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
689 2018-02-16 Carl Love <cel@us.ibm.com>
691 * config/rs6000/altivec.h: Add builtin names vec_extract4b
693 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
695 * config/rs6000/rs6000-c.c: Add the definitions for
696 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
697 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
698 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
699 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
700 definition for insert4b and define insn *insert3b_internal.
701 * doc/extend.texi: Add documentation for vec_extract4b.
703 2018-02-16 Nathan Sidwell <nathan@acm.org>
705 * doc/extend.texi (Backwards Compatibility): Mention friend
706 injection. Note for-scope is deprecated.
707 * doc/invoke.texi (-ffriend-injection): Deprecate.
709 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
711 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
712 that moved to I2, also allow destinations that are a paradoxical
713 subreg (instead of a normal reg).
715 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
718 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
721 2018-02-16 Richard Biener <rguenther@suse.de>
723 PR tree-optimization/84037
724 PR tree-optimization/84016
726 * config/i386/i386.c (ix86_builtin_vectorization_cost):
727 Adjust vec_construct for the fact we need additional higher latency
728 128bit inserts for AVX256 and AVX512 vector builds.
729 (ix86_add_stmt_cost): Scale vector construction cost for
732 2018-02-16 Richard Biener <rguenther@suse.de>
734 PR tree-optimization/84417
735 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
736 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
737 (non_rewritable_lvalue_p): Likewise, use poly-ints.
739 2018-02-16 Martin Liska <mliska@suse.cz>
742 * internal-fn.def (ASAN_CHECK): Set proper flags.
743 (ASAN_MARK): Likewise.
745 2018-02-16 Julia Koval <julia.koval@intel.com>
747 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
750 2018-02-16 Jakub Jelinek <jakub@redhat.com>
753 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
754 Use ++iter rather than iter++ for std::list iterators.
755 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
756 defer deleting them until all nodes in the forest are processed. Do
757 free even leaf nodes. Change to_process into auto_vec.
760 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
761 * vec.h (vec_default_construct): Use memset instead of placement new
762 if BROKEN_VALUE_INITIALIZATION is defined.
763 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
764 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
767 PR rtl-optimization/83723
768 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
769 * lra.c (lra_substitute_pseudo): Likewise. If true, use
770 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
772 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
774 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
776 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
778 PR rtl-optimization/81443
779 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
780 from inner REGs to paradoxical SUBREGs.
782 2018-02-16 Richard Biener <rguenther@suse.de>
784 PR tree-optimization/84399
785 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
786 For operands we can analyze at their definition make sure we can
787 analyze them at each use as well.
789 2018-02-16 Richard Biener <rguenther@suse.de>
791 PR tree-optimization/84190
792 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
793 volatile accesses if the decl isn't volatile.
795 2018-02-15 Jason Merrill <jason@redhat.com>
797 PR c++/84314 - ICE with templates and fastcall attribute.
798 * attribs.c (build_type_attribute_qual_variant): Don't clobber
799 TYPE_CANONICAL on an existing type.
801 2018-02-15 Jakub Jelinek <jakub@redhat.com>
803 PR tree-optimization/84383
804 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
805 dstoff nor call operand_equal_p if dstbase is NULL.
807 PR tree-optimization/84334
808 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
809 also a CONSTANT_CLASS_P, punt.
811 2018-02-14 Jim Wilson <jimw@sifive.com>
813 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
814 first SMALL_OPERAND check. New local min_second_step. Move assert
815 to where locals are set. Add TARGET_RVC support.
816 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
818 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
820 * doc/invoke.texi: Correct -Wformat-overflow code sample.
822 2018-02-14 Martin Sebor <msebor@redhat.com>
824 PR tree-optimization/83698
825 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
826 arrays constrain the offset range to their bounds.
827 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
828 (builtin_access::overlap): Avoid setting the size of overlap if it's
830 (maybe_diag_overlap): Also consider arrays when deciding what values
831 of offsets to include in diagnostics.
833 2018-02-14 Martin Sebor <msebor@redhat.com>
836 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
837 that correspond to the kind of a declaration.
839 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
842 * config/pa/pa.md: Load address of PIC label using the linkage table
843 if the label is nonlocal.
845 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
847 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
848 warning message if user requests -maltivec=be.
849 * doc/invoke.texi: Document deprecation of -maltivec=be.
851 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
854 * config/rs6000/rs6000-c.c: Update definitions for
855 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
856 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
858 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
861 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
862 add _get_ssp intrinsics. Remove argument from
863 __builtin_ia32_rdssp[d|q].
864 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
865 * config/i386/i386-builtin.def: Remove argument from
866 __builtin_ia32_rdssp[d|q].
867 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
868 ix86_expand_special_args_builtin for _rdssp[d|q].
869 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
870 Clear register before usage.
871 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
872 Add documentation for new _get_ssp and _inc_ssp intrinsics.
874 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
876 PR tree-optimization/84357
877 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
878 operand 1 of an ARRAY_REF too.
880 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
883 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
884 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
886 (set_of_reg): New struct.
887 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
888 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
889 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
891 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
892 Split into bitclr, bitset, bitinvert patterns if appropriate.
893 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
894 use rx_fuse_in_memory_bitop.
895 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
896 to named insn, correct maximum insn length.
898 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
901 * machmode.def: Define a complex mode for PARTIAL_INT.
902 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
904 * doc/rtl.texi: Document CSPImode.
905 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
907 (msp430_hard_regno_nregs_with_padding): Likewise.
909 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
912 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
914 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
916 PR rtl-optimization/84169
917 * combine.c (try_combine): New variable split_i2i3. Set it to true if
918 we generated a parallel as new i3 and we split that to new i2 and i3
919 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
920 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
921 those to i2, not i1. Partially rewrite this scan code.
923 2018-02-13 Jakub Jelinek <jakub@redhat.com>
926 * stor-layout.c (place_field): For variable length fields, adjust
927 offset_align afterwards not just based on the field's alignment,
928 but also on the size.
931 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
932 of exps and logs in the use_exp2 case.
934 2018-02-13 Jeff Law <law@redhat.com>
936 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
939 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
942 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
946 * common.opt (gas-loc-support, gas-locview-support): New.
947 (ginline-points, ginternal-reset-location-views): New.
948 * doc/invoke.texi: Document them. Use @itemx where intended.
949 (gvariable-location-views): Adjust.
950 * target.def (reset_location_view): New.
951 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
952 (TARGET_RESET_LOCATION_VIEW): New.
953 * doc/tm.texi: Rebuilt.
954 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
955 (dwarf2out_default_as_locview_support): New.
956 (output_asm_line_debug_info): Use option variables.
957 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
958 (output_loc_list): Likewise.
959 (add_high_low_attributes): Check option variables.
960 Don't output entry view attribute in strict mode.
961 (gen_inlined_subroutine_die): Check option variables.
962 (dwarf2out_inline_entry): Likewise.
963 (init_sections_and_labels): Likewise.
964 (dwarf2out_early_finish): Likewise.
965 (maybe_reset_location_view): New, from...
966 (dwarf2out_var_location): ... here. Call it.
967 * debug.h (dwarf2out_default_as_loc_support): Declare.
968 (dwarf2out_default_as_locview_support): Declare.
969 * hooks.c (hook_int_rtx_insn_0): New.
970 * hooks.h (hook_int_rtx_insn_0): Declare.
971 * toplev.c (process_options): Take -gas-loc-support and
972 -gas-locview-support from dwarf2out. Enable
973 -gvariable-location-views by default only with locview
974 assembler support. Enable -ginternal-reset-location-views by
975 default only if the target defines the corresponding hook.
976 Enable -ginline-points by default if location views are
977 enabled; force it disabled if statement frontiers are
979 * tree-inline.c (expand_call_inline): Check option variables.
980 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
982 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
984 PR tree-optimization/84321
985 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
986 handling. Also check whether the anti-range contains any values
987 that satisfy the mask; switch to a VR_RANGE if not.
989 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
992 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
994 2018-02-13 Martin Jambor <mjambor@suse.cz>
997 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
998 of call statements, also set location of a load to a temporary.
1000 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
1002 * config/rl78/rl78.c (add_vector_labels): New function.
1003 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
1004 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
1005 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
1006 which checks that no arguments are passed.
1007 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
1008 * doc/extend.texi: Documentation for the new attribute.
1010 2018-02-13 Andreas Schwab <schwab@suse.de>
1012 * config/riscv/linux.h (CPP_SPEC): Define.
1014 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1017 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
1018 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
1019 OPTION_MASK_ISA_AES as first argument to def_builtin_const
1020 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
1021 instead of OPTION_MASK_ISA_PCLMUL as first argument to
1022 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
1023 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
1024 temporarily for AES and PCLMUL builtins.
1026 PR tree-optimization/84339
1027 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
1028 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
1032 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
1033 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
1034 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
1036 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
1038 * omp-simd-clone.h: New file.
1039 * omp-simd-clone.c: Include omp-simd-clone.h.
1040 (expand_simd_clones): No longer static.
1041 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
1042 cgraph.h and omp-simd-clone.h.
1043 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
1044 (vect_recog_widen_shift_pattern): Formatting fix.
1045 (vect_pattern_recog_1): Don't check optab for calls.
1048 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
1049 operands[2] into a REG before using gen_lowpart on it.
1051 2018-02-12 Jeff Law <law@redhat.com>
1054 * config/sh/sh.c (find_barrier): Consider a sibling call
1057 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
1058 successfully back substituting a reg.
1060 2018-02-12 Richard Biener <rguenther@suse.de>
1062 PR tree-optimization/84037
1063 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
1064 parameter, move visited init to caller.
1065 (vect_slp_analyze_operations): Separate cost from validity
1066 check, initialize visited once for all instances.
1067 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
1069 * tree-vect-stmts.c (vect_model_simple_cost): Make early
1071 (vect_model_promotion_demotion_cost): Likewise.
1072 (vectorizable_bswap): Guard cost modeling with !slp_node
1073 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
1075 (vectorizable_call): Likewise.
1076 (vectorizable_conversion): Likewise.
1077 (vectorizable_assignment): Likewise.
1078 (vectorizable_shift): Likewise.
1079 (vectorizable_operation): Likewise.
1080 (vectorizable_store): Likewise.
1081 (vectorizable_load): Likewise.
1082 (vectorizable_condition): Likewise.
1083 (vectorizable_comparison): Likewise.
1085 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
1088 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
1089 (ASAN_MARK): Fix fnspec to account for return value, change pointer
1090 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
1092 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1095 * params.def (inline-min-speedup): Increase from 8 to 15.
1096 (max-inline-insns-auto): Decrease from 40 to 30.
1097 * ipa-split.c (consider_split): Add some buffer for function to
1098 be considered inlining candidate.
1099 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
1102 2018-02-12 Richard Biener <rguenther@suse.de>
1104 PR tree-optimization/84037
1105 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
1106 matched stmts if we cannot swap the non-matched ones.
1108 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1110 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
1111 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
1112 _mm_maskz_scalef_round_ss): New intrinsics.
1113 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
1114 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
1115 __builtin_ia32_scalefss_round): Remove.
1116 (__builtin_ia32_scalefsd_mask_round,
1117 __builtin_ia32_scalefss_mask_round): New intrinsics.
1118 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
1119 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
1120 ((match_operand:VF_128 2 "<round_nimm_predicate>"
1121 "<round_constraint>")): Changed to ...
1122 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
1123 "<round_scalar_constraint>")): ... this.
1124 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
1125 %0, %1, %2<round_op3>}"): Changed to ...
1126 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
1127 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
1128 %2<round_scalar_mask_op3>}"): ... this.
1129 * config/i386/subst.md (round_scalar_nimm_predicate): New.
1131 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1133 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
1134 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
1135 (_mm_maskz_sqrt_round_ss): New intrinsics.
1136 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
1137 (__builtin_ia32_sqrtsd_mask_round)
1138 (__builtin_ia32_sqrtss_mask_round): New builtins.
1139 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
1140 (__builtin_ia32_sqrtss_round): Remove.
1141 (__builtin_ia32_sqrtsd_mask_round)
1142 (__builtin_ia32_sqrtss_mask_round): New builtins.
1143 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
1144 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
1145 ((match_operand:VF_128 1 "vector_operand"
1146 "xBm,<round_constraint>")): Changed to ...
1147 ((match_operand:VF_128 1 "vector_operand"
1148 "xBm,<round_scalar_constraint>")): ... this.
1149 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
1150 %0, %2, %<iptr>1<round_op3>}): Changed to ...
1151 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
1152 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
1153 %<iptr>1<round_scalar_mask_op3>}): ... this.
1154 ((set_attr "prefix" "<round_prefix>")): Changed to ...
1155 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
1157 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
1160 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
1161 Cast vec_cmpeq result to correct type.
1162 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
1163 Cast vec_cmpgt result to correct type.
1165 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
1167 * final.c (final_scan_insn_1): Renamed from...
1168 (final_scan_insn): ... this. New wrapper, to recover
1169 seen from the outermost call in recursive ones.
1170 * config/sparc/sparc.c (output_return): Drop seen from call.
1171 (output_sibcall): Likewise.
1172 * config/visium/visium.c (output_branch): Likewise.
1174 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
1176 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
1179 2018-02-10 Alan Modra <amodra@gmail.com>
1182 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
1183 Specify LR as an input.
1185 2018-02-10 Jakub Jelinek <jakub@redhat.com>
1188 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
1189 remove_member_access_dummy_vars): New functions.
1190 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
1191 lower_omp_1, execute_lower_omp): Use them.
1193 PR rtl-optimization/84308
1194 * shrink-wrap.c (spread_components): Release todo vector.
1196 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
1198 PR rtl-optimization/57193
1199 * ira-color.c (struct allocno_color_data): Add member
1200 conflict_allocno_hard_prefs.
1201 (update_conflict_allocno_hard_prefs): New.
1202 (bucket_allocno_compare_func): Add a preference based on
1203 conflict_allocno_hard_prefs.
1204 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
1205 (color_allocnos): Remove a dead code. Initiate
1206 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
1208 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1211 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
1212 constraint from =wa to wa. Avoid a subreg on the output operand,
1213 instead use a pseudo and subreg it in a move.
1214 (p9_xxbrd_<mode>): Changed to ...
1215 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
1216 (p9_xxbrd_v2df): New expander.
1217 (p9_xxbrw_<mode>): Changed to ...
1218 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
1219 (p9_xxbrw_v4sf): New expander.
1221 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1223 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
1225 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
1228 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
1229 multiply in 32-bit mode.
1230 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
1231 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
1234 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1236 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
1237 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
1238 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
1239 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
1241 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
1244 * dwarf2out.c (is_trivial_indirect_ref): New function.
1245 (dwarf2out_late_global_decl): Do not generate a location
1246 attribute for variables that have a non-trivial DECL_VALUE_EXPR
1247 and that are not defined in the current unit.
1249 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1251 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
1252 instead of a libcall for UNORDERED.
1254 2018-02-09 Tamar Christina <tamar.christina@arm.com>
1257 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
1258 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
1260 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1263 * config/s390/s390.c (s390_set_current_function): Invoke
1264 s390_indirect_branch_settings also if fndecl didn't change.
1266 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1268 * config/rs6000/rs6000.md (blockage): Set length to zero.
1270 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1272 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
1274 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1277 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
1278 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
1282 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
1283 PARALLEL incoming that failed vt_get_decl_and_offset check.
1286 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
1287 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
1289 (get_variable_section): For decls in named .bss* sections pass true as
1290 second argument to bss_initializer_p.
1292 2018-02-09 Marek Polacek <polacek@redhat.com>
1293 Jakub Jelinek <jakub@redhat.com>
1296 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
1297 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
1298 Sync some changes from cxx_fold_indirect_ref.
1300 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1302 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
1304 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
1305 (BLOCK_INLINE_ENTRY_LABEL): New.
1306 (dwarf2out_var_location): Disregard inline entry markers.
1307 (inline_entry_data): New struct.
1308 (inline_entry_data_hasher): New hashtable type.
1309 (inline_entry_data_hasher::hash): New.
1310 (inline_entry_data_hasher::equal): New.
1311 (inline_entry_data_table): New variable.
1312 (add_high_low_attributes): Add DW_AT_entry_pc and
1313 DW_AT_GNU_entry_view attributes if a pending entry is found
1314 in inline_entry_data_table. Add old entry_pc attribute only
1315 if debug nonbinding markers are disabled.
1316 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
1317 markers are enabled.
1318 (block_within_block_p, dwarf2out_inline_entry): New.
1319 (dwarf2out_finish): Check that no entries remained in
1320 inline_entry_data_table.
1321 * final.c (reemit_insn_block_notes): Handle inline entry notes.
1322 (final_scan_insn, notice_source_line): Likewise.
1323 (rest_of_clean_state): Skip inline entry markers.
1324 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
1326 * gimple.c (gimple_build_debug_inline_entry): New.
1327 * gimple.h (enum gimple_debug_subcode): Add
1328 GIMPLE_DEBUG_INLINE_ENTRY.
1329 (gimple_build_debug_inline_entry): Declare.
1330 (gimple_debug_inline_entry_p): New.
1331 (gimple_debug_nonbind_marker_p): Adjust.
1332 * insn-notes.def (INLINE_ENTRY): New.
1333 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
1334 inline entry marker notes.
1335 (print_insn): Likewise.
1336 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
1337 (INSN_DEBUG_MARKER_KIND): Likewise.
1338 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
1339 * tree-inline.c (expand_call_inline): Build and insert
1340 debug_inline_entry stmt.
1341 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
1342 inline entry blocks early, if nonbind markers are enabled.
1343 (dump_scope_block): Dump fragment info.
1344 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
1345 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
1346 (gimple_build_debug_inline_entry): New.
1347 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
1348 Enable/disable inline entry points too.
1349 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
1350 (DEBUG_INSN): Describe inline entry markers.
1352 * common.opt (gvariable-location-views): New.
1353 (gvariable-location-views=incompat5): New.
1354 * config.in: Rebuilt.
1355 * configure: Rebuilt.
1356 * configure.ac: Test assembler for view support.
1357 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
1358 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
1359 * dwarf2out.c (var_loc_view): New typedef.
1360 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
1361 (dwarf2out_locviews_in_attribute): New.
1362 (dwarf2out_locviews_in_loclist): New.
1363 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
1364 (enum dw_line_info_opcode): Add LI_adv_address.
1365 (struct dw_line_info_table): Add view.
1366 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
1367 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
1368 (zero_view_p): New variable.
1369 (ZERO_VIEW_P): New macro.
1370 (output_asm_line_debug_info): New.
1371 (struct var_loc_node): Add view.
1372 (add_AT_view_list, AT_loc_list): New.
1373 (add_var_loc_to_decl): Add view param. Test it against last.
1374 (new_loc_list): Add view params. Record them.
1375 (AT_loc_list_ptr): Handle loc and view lists.
1376 (view_list_to_loc_list_val_node): New.
1377 (print_dw_val): Handle dw_val_class_view_list.
1378 (size_of_die): Likewise.
1379 (value_format): Likewise.
1380 (loc_list_has_views): New.
1381 (gen_llsym): Set vl_symbol too.
1382 (maybe_gen_llsym, skip_loc_list_entry): New.
1383 (dwarf2out_maybe_output_loclist_view_pair): New.
1384 (output_loc_list): Output view list or entries too.
1385 (output_view_list_offset): New.
1386 (output_die): Handle dw_val_class_view_list.
1387 (output_dwarf_version): New.
1388 (output_compilation_unit_header): Use it.
1389 (output_skeleton_debug_sections): Likewise.
1390 (output_rnglists, output_line_info): Likewise.
1391 (output_pubnames, output_aranges): Update version comments.
1392 (output_one_line_info_table): Output view numbers in asm comments.
1393 (dw_loc_list): Determine current endview, pass it to new_loc_list.
1394 Call maybe_gen_llsym.
1395 (loc_list_from_tree_1): Adjust.
1396 (add_AT_location_description): Create view list attribute if
1397 needed, check it's absent otherwise.
1398 (convert_cfa_to_fb_loc_list): Adjust.
1399 (maybe_emit_file): Call output_asm_line_debug_info for test.
1400 (dwarf2out_var_location): Reset views as needed. Precompute
1401 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
1402 attribute. Set view.
1403 (new_line_info_table): Reset next view.
1404 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
1405 (dwarf2out_source_line): Likewise. Output view resets and labels to
1406 the assembler, or select appropriate line info opcodes.
1407 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
1408 (optimize_string_length): Catch it. Adjust.
1409 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
1410 dw_val_class_view_list, and remove it if no longer needed.
1411 (hash_loc_list): Hash view numbers.
1412 (loc_list_hasher::equal): Compare them.
1413 (optimize_location_lists): Check whether a view list symbol is
1414 needed, and whether the locview attribute is present, and
1415 whether they match. Remove the locview attribute if no longer
1417 (index_location_lists): Call skip_loc_list_entry for test.
1418 (dwarf2out_finish): Call output_asm_line_debug_info for test.
1419 Use output_dwarf_version.
1420 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
1421 (struct dw_val_node): Add val_view_list.
1422 * final.c (SEEN_NEXT_VIEW): New.
1423 (set_next_view_needed): New.
1424 (clear_next_view_needed): New.
1425 (maybe_output_next_view): New.
1426 (final_start_function): Rename to...
1427 (final_start_function_1): ... this. Take pointer to FIRST,
1428 add SEEN parameter. Emit param bindings in the initial view.
1429 (final_start_function): Reintroduce SEEN-less interface.
1430 (final): Rename to...
1431 (final_1): ... this. Take SEEN parameter. Output final pending
1432 next view at the end.
1433 (final): Reintroduce seen-less interface.
1434 (final_scan_insn): Output pending next view before switching
1435 sections or ending a block. Mark the next view as needed when
1436 outputting variable locations. Notify debug backend of section
1437 changes, and of location view changes.
1438 (rest_of_handle_final): Adjust.
1439 * toplev.c (process_options): Autodetect value for debug variable
1440 location views option. Warn on incompat5 without -gdwarf-5.
1441 * doc/invoke.texi (gvariable-location-views): New.
1442 (gvariable-location-views=incompat5): New.
1443 (gno-variable-location-views): New.
1445 2018-02-08 David Malcolm <dmalcolm@redhat.com>
1447 PR tree-optimization/84136
1448 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
1449 that the result of find_edge is non-NULL.
1451 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1454 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
1455 storing integer register in SImode. Fix cost of 256 and 512
1456 byte aligned SSE register store.
1458 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1460 * config/i386/i386.c (ix86_multiplication_cost): Fix
1461 multiplication cost for TARGET_AVX512DQ.
1463 2018-02-08 Marek Polacek <polacek@redhat.com>
1465 PR tree-optimization/84238
1466 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
1469 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1471 PR tree-optimization/84265
1472 * tree-vect-stmts.c (vectorizable_store): Don't treat
1473 VMAT_CONTIGUOUS accesses as grouped.
1474 (vectorizable_load): Likewise.
1476 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1478 PR tree-optimization/81635
1479 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
1480 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
1481 (test_round_for_mask): New functions.
1482 (wide_int_cc_tests): Call test_round_for_mask.
1483 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
1484 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
1485 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
1486 range returned by get_range_info.
1488 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1491 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
1492 * symtab.c: Include builtins.h
1493 (symtab_node::output_to_lto_symbol_table_p): Move here
1494 from lto-streamer-out.c:output_symbol_p.
1495 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
1496 (output_symbol_p): Move all logic to symtab.c
1497 (produce_symtab): Update.
1499 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1501 * config/s390/s390-opts.h (enum indirect_branch): Define.
1502 * config/s390/s390-protos.h (s390_return_addr_from_memory)
1503 (s390_indirect_branch_via_thunk)
1504 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
1505 (enum s390_indirect_branch_type): Define.
1506 * config/s390/s390.c (struct s390_frame_layout, struct
1507 machine_function): Remove.
1508 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
1509 (indirect_branch_table_label_no, indirect_branch_table_name):
1511 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
1512 (enum s390_indirect_branch_option): Define.
1513 (s390_return_addr_from_memory): New function.
1514 (s390_handle_string_attribute): New function.
1515 (s390_attribute_table): Add new attribute handler.
1516 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
1517 (s390_indirect_branch_via_thunk): New function.
1518 (s390_indirect_branch_via_inline_thunk): New function.
1519 (s390_function_ok_for_sibcall): When jumping via thunk disallow
1520 sibling call optimization for non z10 compiles.
1521 (s390_emit_call): Force indirect branch target to be a single
1522 register. Add r1 clobber for non-z10 compiles.
1523 (s390_emit_epilogue): Emit return jump via return_use expander.
1524 (s390_reorg): Handle JUMP_INSNs as execute targets.
1525 (s390_option_override_internal): Perform validity checks for the
1526 new command line options.
1527 (s390_indirect_branch_attrvalue): New function.
1528 (s390_indirect_branch_settings): New function.
1529 (s390_set_current_function): Invoke s390_indirect_branch_settings.
1530 (s390_output_indirect_thunk_function): New function.
1531 (s390_code_end): Implement target hook.
1532 (s390_case_values_threshold): Implement target hook.
1533 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
1535 * config/s390/s390.h (struct s390_frame_layout)
1536 (struct machine_function): Move here from s390.c.
1537 (TARGET_INDIRECT_BRANCH_NOBP_RET)
1538 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
1539 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
1540 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
1541 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
1542 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
1543 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
1544 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
1545 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
1546 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
1547 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
1548 (mnemonic attribute): Add values which aren't recognized
1550 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
1551 pattern for branch conversion. Fix mnemonic attribute.
1552 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
1553 indirect branch via thunk if requested.
1554 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
1555 ("*indirect_jump"): Disable for branch conversion using out of
1557 ("indirect_jump_via_thunk<mode>_z10")
1558 ("indirect_jump_via_thunk<mode>")
1559 ("indirect_jump_via_inlinethunk<mode>_z10")
1560 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
1561 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
1562 ("casesi_jump_via_inlinethunk<mode>_z10")
1563 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
1564 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
1565 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
1566 ("*indirect2_jump"): Disable for branch conversion.
1567 ("casesi_jump"): Turn into expander and expand patterns for branch
1569 ("return_use"): New expander.
1570 ("*return"): Emit return via thunk and rename it to ...
1571 ("*return<mode>"): ... this one.
1572 * config/s390/s390.opt: Add new options and and enum for the
1575 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1577 * lra-constraints.c (match_reload): Unconditionally use
1578 gen_lowpart_SUBREG, rather than selecting between that
1579 and equivalent gen_rtx_SUBREG code.
1581 2018-02-08 Richard Biener <rguenther@suse.de>
1583 PR tree-optimization/84233
1584 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
1585 changed flag instead of boguously re-using phi_inserted.
1587 2018-02-08 Martin Jambor <mjambor@suse.cz>
1589 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
1590 static local variables.
1592 2018-02-08 Richard Biener <rguenther@suse.de>
1594 PR tree-optimization/84278
1595 * tree-vect-stmts.c (vectorizable_store): When looking for
1596 smaller vector types to perform grouped strided loads/stores
1597 make sure the mode is supported by the target.
1598 (vectorizable_load): Likewise.
1600 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1602 * config/aarch64/aarch64.c (aarch64_components_for_bb):
1603 Increase LDP/STP opportunities by adding adjacent callee-saves.
1605 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1607 PR rtl-optimization/84068
1608 PR rtl-optimization/83459
1609 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
1611 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
1613 PR tree-optimization/84224
1614 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
1615 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
1618 2018-02-07 Iain Sandoe <iain@codesourcery.com>
1621 * config/rs6000/altivec.md (*restore_world): Remove LR use.
1622 * config/rs6000/predicates.md (restore_world_operation): Adjust op
1623 count, remove one USE.
1625 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
1627 * doc/install.texi (Configuration): Document the
1628 --with-long-double-format={ibm,ieee} PowerPC configuration
1632 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
1633 Convert from define_expand to be define_insn_and_split. Rework
1634 float/double/_Float128 conversions to QI/HI/SImode to work with
1635 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
1636 conversions to QI/HImode types did a store and then a load to
1637 truncate the value. For conversions to VSX registers, don't split
1638 the insn, instead emit the code directly. Use the code iterator
1639 any_fix to combine signed and unsigned conversions.
1640 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
1641 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
1642 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
1643 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
1644 (fix_<mode>di2_hw): Likewise.
1645 (fixuns_<mode>di2_hw): Likewise.
1646 (fix_<mode>si2_hw): Likewise.
1647 (fixuns_<mode>si2_hw): Likewise.
1648 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
1649 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
1650 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
1651 fix<uns>_trunc<SFDF:mode>si2_p8.
1652 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
1654 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
1655 (fix<uns>_<mode>_mem): Likewise.
1656 (fctiw<u>z_<mode>_mem): Likewise.
1657 (fix<uns>_<mode>_mem): Likewise.
1658 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
1659 the register allocator from doing a direct move to the GPRs to do
1660 a store, and instead use the ISA 3.0 store byte/half-word from
1661 vector register instruction. For IEEE 128-bit floating point,
1662 also optimize stores of 32-bit ints.
1663 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
1665 2018-02-07 Alan Hayward <alan.hayward@arm.com>
1667 * genextract.c (push_pathstr_operand): New function to support
1669 (walk_rtx): Call push_pathstr_operand.
1670 (print_path): Support [a-zA-Z].
1672 2018-02-07 Richard Biener <rguenther@suse.de>
1674 PR tree-optimization/84037
1675 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
1676 (cse_and_gimplify_to_preheader): Declare.
1677 (vect_get_place_in_interleaving_chain): Likewise.
1678 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
1680 (_loop_vec_info::~_loop_vec_info): Delete it.
1681 (cse_and_gimplify_to_preheader): New function.
1682 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
1683 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
1684 (vectorizable_load): Likewise. For grouped stores always base
1685 the IV on the first element.
1686 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
1687 condition before gimplifying.
1689 2018-02-07 Jakub Jelinek <jakub@redhat.com>
1691 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
1692 *DIV_EXPR and *MOD_EXPR.
1694 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
1697 * config/i386/i386.c (ix86_option_override_internal): Mask out
1698 the CF_SET bit when checking -fcf-protection.
1700 2018-02-07 Tom de Vries <tom@codesourcery.com>
1703 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
1706 2018-02-07 Richard Biener <rguenther@suse.de>
1708 PR tree-optimization/84204
1709 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
1712 PR tree-optimization/84205
1713 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
1714 special-case isl_ast_op_zdiv_r.
1716 PR tree-optimization/84223
1717 * graphite-scop-detection.c (gather_bbs::before_dom_children):
1718 Only add conditions from within the region.
1719 (gather_bbs::after_dom_children): Adjust.
1721 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
1724 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
1725 * config/avr/avr.md: Only post-reload split REG-REG moves if
1726 either register is REGERAL_REG_P.
1728 2018-02-07 Jakub Jelinek <jakub@redhat.com>
1730 PR tree-optimization/84235
1731 * tree-ssa-scopedtables.c
1732 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
1733 if the subtraction is performed in floating point type where NaNs are
1734 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
1735 build 1. Formatting fix.
1737 2018-02-06 Jakub Jelinek <jakub@redhat.com>
1740 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
1741 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
1742 and skip it regardless of bb boundaries. Use CALL_P macro,
1743 don't test INSN_P (insn) together with CALL_P or JUMP_P check
1744 unnecessarily, formatting fix.
1746 2018-02-06 Michael Collison <michael.collison@arm.com>
1748 * config/arm/thumb2.md:
1749 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
1750 (*thumb_mov_notscc): Ditto.
1752 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
1755 * config/rs6000/rs6000.md (su code attribute): Use "u" for
1756 unsigned_fix, not "s".
1758 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1760 * configure.ac (gcc_fn_eh_frame_ro): New function.
1761 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
1762 correct .eh_frame permissions.
1763 * configure: Regenerate.
1765 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
1767 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
1770 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1772 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1773 Display warning message for -mno-speculate-indirect-jumps.
1775 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
1777 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
1779 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
1781 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
1783 PR tree-optimization/84225
1784 * tree-eh.c (find_trapping_overflow): Only call
1785 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
1787 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1790 * config/i386/i386.c: Reimplement the check of possible options
1791 -mibt/-mshstk conbination. Change error messages.
1792 * doc/invoke.texi: Fix a typo: remove extra '='.
1794 2018-02-06 Marek Polacek <polacek@redhat.com>
1796 PR tree-optimization/84228
1797 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
1799 2018-02-06 Tamar Christina <tamar.christina@arm.com>
1802 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
1803 emitted arch directives.
1804 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
1805 __ARM_FEATURE_COPROC before changing architectures.
1807 2018-02-06 Richard Biener <rguenther@suse.de>
1809 * config/i386/i386.c (print_reg): Fix typo.
1810 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
1812 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
1814 * configure: Regenerate.
1816 2018-02-05 Martin Sebor <msebor@redhat.com>
1818 PR tree-optimization/83369
1819 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
1822 2018-02-05 Martin Liska <mliska@suse.cz>
1824 * doc/invoke.texi: Cherry-pick upstream r323995.
1826 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
1828 * ira.c (ira_init_register_move_cost): Adjust comment.
1830 2018-02-05 Martin Liska <mliska@suse.cz>
1832 PR gcov-profile/84137
1833 * doc/gcov.texi: Fix typo in documentation.
1835 2018-02-05 Martin Liska <mliska@suse.cz>
1837 PR gcov-profile/83879
1838 * doc/gcov.texi: Document necessity of --dynamic-list-data when
1839 using dlopen functionality.
1841 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
1843 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
1844 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
1845 _mm_maskz_range_ss, _mm_mask_range_round_ss,
1846 _mm_maskz_range_round_ss): New intrinsics.
1847 (__builtin_ia32_rangesd128_round)
1848 (__builtin_ia32_rangess128_round): Remove.
1849 (__builtin_ia32_rangesd128_mask_round,
1850 __builtin_ia32_rangess128_mask_round): New builtins.
1851 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
1852 __builtin_ia32_rangess128_round): Remove.
1853 (__builtin_ia32_rangesd128_mask_round,
1854 __builtin_ia32_rangess128_mask_round): New builtins.
1855 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
1856 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
1857 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
1858 "<round_saeonly_constraint>")): Changed to ...
1859 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
1860 "<round_saeonly_scalar_constraint>")): ... this.
1861 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
1862 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
1863 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
1864 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
1865 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
1867 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
1869 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
1871 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
1872 Remove all values except native, 8540 and 8548.
1874 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
1876 * config/i386/i386.c (ix86_output_function_return): Pass
1877 INVALID_REGNUM, instead of -1, as invalid register number to
1878 indirect_thunk_name and output_indirect_thunk.
1880 2018-02-02 Julia Koval <julia.koval@intel.com>
1882 * config.gcc: Add -march=icelake.
1883 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
1884 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
1885 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
1886 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
1887 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
1888 (processor_target_table): Add icelake.
1889 (ix86_option_override_internal): Handle new PTAs.
1890 (get_builtin_code_for_version): Handle icelake.
1891 (M_INTEL_COREI7_ICELAKE): New.
1892 (fold_builtin_cpu): Handle icelake.
1893 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
1894 * doc/invoke.texi: Add -march=icelake.
1896 2018-02-02 Julia Koval <julia.koval@intel.com>
1898 * config/i386/i386.c (ix86_option_override_internal): Change flags type
1899 to wide_int_bitmask.
1900 * wide-int-bitmask.h: New.
1902 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1905 * config/i386/i386.md: Replace Pmode with word_mode in
1906 builtin_setjmp_setup and builtin_longjmp to support x32.
1908 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
1912 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
1914 (rs6000_supported_cpu_names): New static variable.
1915 (linux_cpu_translation_table): Likewise.
1916 (elf_platform) <cpu>: Define new static variable and use it.
1917 Translate kernel AT_PLATFORM name to canonical name if needed.
1918 Error if platform name is unknown.
1920 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
1923 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
1925 2018-02-01 Jeff Law <law@redhat.com>
1928 * config/i386/i386.c (release_scratch_register_on_entry): Add new
1929 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
1930 the scratch if RELEASE_VIA_POP is false.
1931 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
1932 If we have to save a temporary register, decrement SIZE appropriately.
1933 Pass new arguments to release_scratch_register_on_entry.
1934 (ix86_adjust_stack_and_probe): Likewise.
1935 (ix86_emit_probe_stack_range): Pass new arguments to
1936 release_scratch_register_on_entry.
1938 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
1940 PR rtl-optimization/84157
1941 * combine.c (change_zero_ext): Use REG_P predicate in
1942 front of HARD_REGISTER_P predicate.
1944 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
1946 * config/avr/avr.c (avr_option_override): Move disabling of
1947 -fdelete-null-pointer-checks to...
1948 * common/config/avr/avr-common.c (avr_option_optimization_table):
1951 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1953 PR tree-optimization/81635
1954 * tree-data-ref.c (split_constant_offset_1): For types that
1955 wrap on overflow, try to use range info to prove that wrapping
1958 2018-02-01 Renlin Li <renlin.li@arm.com>
1961 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
1963 (aarch64_register_move_cost): Likewise.
1964 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
1966 (REG_CLASS_NAMES): Likewise.
1967 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
1968 TAILCALL_ADDR_REGS. Remove IP registers.
1969 * config/aarch64/aarch64.md (Ucs): Update register constraint.
1971 2018-02-01 Richard Biener <rguenther@suse.de>
1973 * domwalk.h (dom_walker::dom_walker): Add additional constructor
1974 for specifying RPO order and allow NULL for that.
1975 * domwalk.c (dom_walker::dom_walker): Likewise.
1976 (dom_walker::walk): Handle NULL RPO order.
1977 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
1979 (rewrite_update_dom_walker): Likewise.
1980 (mark_def_dom_walker): Likewise.
1982 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1984 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
1985 (aarch64_maybe_expand_sve_subreg_move): Declare.
1986 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
1987 * config/aarch64/predicates.md (aarch64_any_register_operand): New
1989 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
1990 that are semantically a reverse operation.
1991 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
1992 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
1993 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
1995 (aarch64_can_change_mode_class): For big-endian, forbid changes
1996 between two SVE modes if they have different element sizes.
1998 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2000 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
2001 the TImode handling for big-endian targets.
2003 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2005 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
2006 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
2008 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
2009 Remove BSWAP handing for big-endian targets and use the form of
2010 LD1RQ appropariate for the mode.
2012 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2014 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
2015 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
2018 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2021 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
2022 check for operands that need to go through aarch64_sve_reload_be.
2024 2018-02-01 Jakub Jelinek <jakub@redhat.com>
2026 PR tree-optimization/81661
2027 PR tree-optimization/84117
2028 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
2029 * tree-eh.c: Include gimplify.h.
2030 (find_trapping_overflow, replace_trapping_overflow,
2031 rewrite_to_non_trapping_overflow): New functions.
2032 * tree-vect-loop.c: Include tree-eh.h.
2033 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
2034 * tree-data-ref.c: Include tree-eh.h.
2035 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
2037 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
2039 PR rtl-optimization/84123
2040 * combine.c (change_zero_ext): Check if hard register satisfies
2041 can_change_dest_mode before calling gen_lowpart_SUBREG.
2043 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
2046 * ira.c (ira_init_register_move_cost): Remove assert.
2048 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2050 PR rtl-optimization/84071
2051 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
2052 * doc/tm.texi: Regenerate.
2054 2018-01-31 Richard Biener <rguenther@suse.de>
2056 PR tree-optimization/84132
2057 * tree-data-ref.c (analyze_miv_subscript): Properly
2058 check whether evolution_function_is_affine_multivariate_p
2059 before calling gcd_of_steps_may_divide_p.
2061 2018-01-31 Julia Koval <julia.koval@intel.com>
2064 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
2065 * config/i386/i386.md (rdpid_rex64) New.
2066 (rdpid): Make 32bit only.
2068 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
2071 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
2072 an IDENTIFIER_NODE for FUNCTION_TYPE's.
2074 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2077 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
2079 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
2081 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2083 PR rtl-optimization/84071
2084 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
2085 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
2087 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2089 * config/arc/arc.c (arc_handle_aux_attribute): New function.
2090 (arc_attribute_table): Add 'aux' attribute.
2091 (arc_in_small_data_p): Consider aux like variables.
2092 (arc_is_aux_reg_p): New function.
2093 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
2094 (arc_get_aux_arg): New function.
2095 (prepare_move_operands): Handle aux-register access.
2096 (arc_handle_aux_attribute): New function.
2097 * doc/extend.texi (ARC Variable attributes): Add subsection.
2099 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2101 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
2102 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
2103 (arc_attribute_table): Add 'uncached' attribute.
2104 (arc_print_operand): Print '.di' flag for uncached memory
2106 (arc_in_small_data_p): Do not consider for small data the uncached
2108 (arc_is_uncached_mem_p): New function.
2109 * config/arc/predicates.md (compact_store_memory_operand): Check
2110 for uncached memory accesses.
2111 (nonvol_nonimm_operand): Likewise.
2112 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
2114 2018-01-31 Jakub Jelinek <jakub@redhat.com>
2117 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
2118 falign-loops=): Add Optimization flag.
2120 2018-01-30 Jeff Law <law@redhat.com>
2123 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
2124 INT_REGISTERS_SAVED. Check it prior to calling
2125 get_scratch_register_on_entry.
2126 (ix86_adjust_stack_and_probe): Similarly.
2127 (ix86_emit_probe_stack_range): Similarly.
2128 (ix86_expand_prologue): Corresponding changes.
2130 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2133 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
2134 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
2136 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
2139 * lra-constraints.c (curr_insn_transform): Process AND in the
2142 2018-01-30 Jakub Jelinek <jakub@redhat.com>
2144 PR rtl-optimization/83986
2145 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
2146 dependence against last_pending_memory_flush in addition to
2149 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
2151 PR tree-optimization/81611
2152 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
2155 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2158 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
2161 2018-01-30 Richard Biener <rguenther@suse.de>
2162 Jakub Jelinek <jakub@redhat.com>
2164 PR tree-optimization/84111
2165 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
2166 inner loops added during recursion, as they don't have up-to-date
2169 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2172 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
2173 (can_inline_edge_by_limits_p): ... here.
2174 (can_early_inline_edge_p, check_callers,
2175 update_caller_keys, update_callee_keys, recursive_inlining,
2176 add_new_edges_to_heap, speculation_useful_p,
2177 inline_small_functions,
2178 inline_small_functions, flatten_function,
2179 inline_to_all_callers_1): Update.
2181 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2183 * profile-count.c (profile_count::combine_with_ipa_count): Handle
2186 2018-01-30 Richard Biener <rguenther@suse.de>
2188 PR tree-optimization/83008
2189 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
2190 invariant and constant vector uses in stmts when they need
2193 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2196 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
2197 * configure: Regenerate.
2199 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2201 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
2203 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
2204 Use gen_rtx_REG rather than gen_lowpart.
2206 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2208 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
2209 rather than 0 when creating partial subregs.
2211 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2213 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
2216 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
2219 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
2220 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
2221 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
2222 flags. This restores the settings used before the 2017-07-24.
2223 Turning off pre increment/decrement/modify allows IVOPTS to
2224 optimize DF/SF loops where the index is an int.
2226 2018-01-29 Richard Biener <rguenther@suse.de>
2227 Kelvin Nilsen <kelvin@gcc.gnu.org>
2230 * tree-vect-stmts.c (vectorizable_call): Don't call
2231 targetm.vectorize_builtin_md_vectorized_function if callee is
2234 2018-01-22 Carl Love <cel@us.ibm.com>
2236 * doc/extend.tex: Fix typo in second arg in
2237 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
2239 2018-01-29 Richard Biener <rguenther@suse.de>
2241 PR tree-optimization/84086
2242 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
2243 (flush_ssaname_freelist): When SSA names were released reset
2244 the SCEV hash table.
2246 2018-01-29 Richard Biener <rguenther@suse.de>
2248 PR tree-optimization/84057
2249 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
2250 removed paths when removing edges.
2252 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
2254 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
2255 -mfunction-return=@var{choice}.
2257 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
2260 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
2262 (layout::print_source_line): Likewise.
2263 (test_get_line_width_without_trailing_whitespace): Add test cases.
2265 2018-01-27 Jakub Jelinek <jakub@redhat.com>
2268 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
2271 2018-01-26 Jim Wilson <jimw@sifive.com>
2273 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
2275 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
2278 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2280 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
2281 and CMP + SUB-immediate -> SUBS.
2283 2018-01-26 Martin Sebor <msebor@redhat.com>
2285 PR tree-optimization/83896
2286 * tree-ssa-strlen.c (get_string_len): Rename...
2287 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
2288 Avoid assuming length is constant.
2289 (handle_char_store): Use HOST_WIDE_INT for string length.
2291 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
2294 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
2295 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
2297 2018-01-26 Richard Biener <rguenther@suse.de>
2299 PR rtl-optimization/84003
2300 * dse.c (record_store): Only record redundant stores when
2301 the earlier store aliases at least all accesses the later one does.
2303 2018-01-26 Jakub Jelinek <jakub@redhat.com>
2305 PR rtl-optimization/83985
2306 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
2307 REG_CFA_RESTORE insns.
2308 (delete_unmarked_insns): Don't ignore separate shrink wrapping
2309 REG_CFA_RESTORE insns here.
2312 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
2313 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
2315 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2317 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
2318 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
2319 (arc_init): Likewise.
2320 (arc_override_options): Likewise.
2321 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
2323 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
2325 * config/arc/arc.h (TARGET_DBNZ): Define.
2326 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
2327 properly set the tune attribute.
2328 (dbnz): Use TARGET_DBNZ guard.
2329 * config/arc/arc.opt (mtune): Add core3 option.
2331 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2333 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
2334 recognize new pic like addresses.
2335 (arc_delegitimize_address): Clean up.
2337 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2339 * config/arc/arc-arches.def: Option mrf16 valid for all
2341 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
2342 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
2343 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
2344 * config/arc/arc-tables.opt: Regenerate.
2345 * config/arc/arc.c (arc_conditional_register_usage): Handle
2346 reduced register file case.
2347 (arc_file_start): Set must have build attributes.
2348 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
2350 * config/arc/arc.opt (mrf16): Add new option.
2351 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
2352 * config/arc/genmultilib.awk: Handle new mrf16 option.
2353 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
2354 * config/arc/t-multilib: Regenerate.
2355 * doc/invoke.texi (ARC Options): Document mrf16 option.
2357 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2359 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
2360 * config/arc/arc.c (arc_handle_secure_attribute): New function.
2361 (arc_attribute_table): Add 'secure_call' attribute.
2362 (arc_print_operand): Print secure call operand.
2363 (arc_function_ok_for_sibcall): Don't optimize tail calls when
2365 (arc_is_secure_call_p): New function. * config/arc/arc.md
2366 (call_i): Add support for sjli instruction.
2367 (call_value_i): Likewise.
2368 * config/arc/constraints.md (Csc): New constraint.
2370 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2371 John Eric Martin <John.Martin@emmicro-us.com>
2373 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
2374 * config/arc/arc.c (_arc_jli_section): New struct.
2375 (arc_jli_section): New type.
2376 (rc_jli_sections): New static variable.
2377 (arc_handle_jli_attribute): New function.
2378 (arc_attribute_table): Add jli_always and jli_fixed attribute.
2379 (arc_file_end): New function.
2380 (TARGET_ASM_FILE_END): Define.
2381 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
2382 (arc_add_jli_section): New function.
2383 (jli_call_scan): Likewise.
2384 (arc_reorg): Call jli_call_scan.
2385 (arc_output_addsi): Remove 'S' from printing asm operand.
2386 (arc_is_jli_call_p): New function.
2387 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
2389 (movhi_insn): Likewise.
2390 (movsi_insn): Likewise.
2391 (movsi_set_cc_insn): Likewise.
2392 (loadqi_update): Likewise.
2393 (load_zeroextendqisi_update): Likewise.
2394 (load_signextendqisi_update): Likewise.
2395 (loadhi_update): Likewise.
2396 (load_zeroextendhisi_update): Likewise.
2397 (load_signextendhisi_update): Likewise.
2398 (loadsi_update): Likewise.
2399 (loadsf_update): Likewise.
2400 (movsicc_insn): Likewise.
2401 (bset_insn): Likewise.
2402 (bxor_insn): Likewise.
2403 (bclr_insn): Likewise.
2404 (bmsk_insn): Likewise.
2405 (bicsi3_insn): Likewise.
2406 (cmpsi_cc_c_insn): Likewise.
2407 (movsi_ne): Likewise.
2408 (movsi_cond_exec): Likewise.
2409 (clrsbsi2): Likewise.
2420 (call_i): Remove 'S' asm letter, add jli instruction.
2421 (call_value_i): Likewise.
2422 * config/arc/arc.op (mjli-always): New option.
2423 * config/arc/constraints.md (Cji): New constraint.
2424 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
2426 (subsf3_fpx): Likewise.
2427 (mulsf3_fpx): Likewise.
2428 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
2430 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
2432 * doc/invoke.texi (ARC): Document mjli-always option.
2434 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
2436 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
2437 avoid addition with 0 and use incw and decw where possible.
2439 2018-01-26 Richard Biener <rguenther@suse.de>
2441 PR tree-optimization/81082
2442 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
2443 association if it requires casting to unsigned.
2444 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
2445 from fold_plusminus_mult_expr to catch important cases late when
2446 range info is available.
2448 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2450 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
2451 * configure.ac (hidden_linkonce): New test.
2452 * configure: Regenerate.
2453 * config.in: Regenerate.
2455 2018-01-26 Julia Koval <julia.koval@intel.com>
2457 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
2458 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
2459 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
2460 _mm_mask_bitshuffle_epi64_mask): Fix type.
2461 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
2462 USI_FTYPE_V4DI_V4DI_USI): Remove.
2463 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
2464 __builtin_ia32_vpshufbitqmb256_mask,
2465 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
2466 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
2467 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
2469 2018-01-26 Alan Modra <amodra@gmail.com>
2472 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
2473 UNSPEC_VBPERMQ. Sort other unspecs.
2475 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
2477 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
2479 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
2482 * predict.c (drop_profile): Do not push/pop cfun; update also
2484 (handle_missing_profiles): Fix logic looking for zero profiles.
2486 2018-01-25 Jakub Jelinek <jakub@redhat.com>
2489 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
2490 on functions with #pragma omp declare simd or functions with simd
2492 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
2493 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
2494 Remove trailing \n from warning_at calls.
2496 2018-01-25 Tom de Vries <tom@codesourcery.com>
2499 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2500 for neutered workers.
2502 2018-01-24 Joseph Myers <joseph@codesourcery.com>
2505 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
2506 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
2508 2018-01-24 Jeff Law <law@redhat.com>
2511 * i386.c (get_probe_interval): Move to earlier point.
2512 (ix86_compute_frame_layout): If -fstack-clash-protection and
2513 the frame is larger than the probe interval, then use pushes
2514 to save registers rather than reg->mem moves.
2515 (ix86_expand_prologue): Remove conditional for int_registers_saved
2518 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
2521 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
2522 min/max for never referenced object.
2524 2018-01-24 Jakub Jelinek <jakub@redhat.com>
2527 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
2529 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
2530 attributes from DECL_ATTRIBUTES (decl) without affecting
2531 DECL_ATTRIBUTES (current_function_decl).
2532 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
2533 functions with non-NULL DECL_ABSTRACT_ORIGIN.
2535 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
2537 PR tree-optimization/83979
2538 * fold-const.c (fold_comparison): Use constant_boolean_node
2539 instead of boolean_{true,false}_node.
2541 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
2543 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
2546 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2548 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
2549 Simplify the clause that sets the length attribute.
2550 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
2551 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
2552 clause that sets the length attribute.
2553 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
2555 2018-01-24 Tom de Vries <tom@codesourcery.com>
2558 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
2559 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
2560 Add strict parameter.
2561 (prevent_branch_around_nothing): Insert dummy insn between branch to
2562 label and label with no ptx insn inbetween.
2563 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
2565 2018-01-24 Tom de Vries <tom@codesourcery.com>
2568 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2569 for neutered threads in warp.
2570 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
2572 2018-01-24 Richard Biener <rguenther@suse.de>
2574 PR tree-optimization/83176
2575 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
2578 2018-01-24 Richard Biener <rguenther@suse.de>
2580 PR tree-optimization/82819
2581 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
2582 code generating pluses that are no-ops in the target precision.
2584 2018-01-24 Richard Biener <rguenther@suse.de>
2587 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
2589 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2591 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
2592 to merge probabilities.
2593 * predict.c (probably_never_executed): Also mark as cold functions
2594 with global 0 profile and guessed local profile.
2595 * profile-count.c (profile_probability::combine_with_count): New
2597 * profile-count.h (profile_probability::operator*,
2598 profile_probability::operator*=, profile_probability::operator/,
2599 profile_probability::operator/=): Reduce precision to adjusted
2600 and set value to guessed on contradictory divisions.
2601 (profile_probability::combine_with_freq): Remove.
2602 (profile_probability::combine_wiht_count): Declare.
2603 (profile_count::force_nonzero):: Set to adjusted.
2604 (profile_count::probability_in):: Set quality to adjusted.
2605 * tree-ssa-tail-merge.c (replace_block_by): Use
2608 2018-01-23 Andrew Waterman <andrew@sifive.com>
2609 Jim Wilson <jimw@sifive.com>
2611 * config/riscv/riscv.c (riscv_stack_boundary): New.
2612 (riscv_option_override): Set riscv_stack_boundary. Handle
2613 riscv_preferred_stack_boundary_arg.
2614 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
2615 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
2616 (STACK_BOUNDARY): Set to riscv_stack_boundary.
2617 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
2618 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
2619 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
2621 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
2624 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
2625 of struct ix86_frame.
2626 (ix86_expand_epilogue): Likewise. Add a local variable for
2627 the reg_save_offset field in struct ix86_frame.
2629 2018-01-23 Bin Cheng <bin.cheng@arm.com>
2631 PR tree-optimization/82604
2632 * tree-loop-distribution.c (enum partition_kind): New enum item
2633 PKIND_PARTIAL_MEMSET.
2634 (partition_builtin_p): Support above new enum item.
2635 (generate_code_for_partition): Ditto.
2636 (compute_access_range): Differentiate cases that equality can be
2637 proven at all loops, the innermost loops or no loops.
2638 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
2639 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
2640 (finalize_partitions, distribute_loop): Don't fuse partition of
2641 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
2642 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
2645 2018-01-23 Martin Liska <mliska@suse.cz>
2647 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
2648 order to ignore the predictor.
2649 (PRED_POLYMORPHIC_CALL): Likewise.
2650 (PRED_RECURSIVE_CALL): Likewise.
2652 2018-01-23 Martin Liska <mliska@suse.cz>
2654 * tree-profile.c (tree_profiling): Print function header to
2655 aware reader which function we are working on.
2656 * value-prof.c (gimple_find_values_to_profile): Do not print
2657 not interesting value histograms.
2659 2018-01-23 Martin Liska <mliska@suse.cz>
2661 * profile-count.h (enum profile_quality): Add
2662 profile_uninitialized as the first value. Do not number values
2663 as they are zero based.
2664 (profile_count::verify): Update sanity check.
2665 (profile_probability::verify): Likewise.
2667 2018-01-23 Nathan Sidwell <nathan@acm.org>
2669 * doc/invoke.texi (ffor-scope): Deprecate.
2671 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2673 PR tree-optimization/83510
2674 * domwalk.c (set_all_edges_as_executable): New function.
2675 (dom_walker::dom_walker): Convert bool param
2676 "skip_unreachable_blocks" to enum reachability. Move setup of
2677 edge flags to set_all_edges_as_executable and only do it when
2678 reachability is REACHABLE_BLOCKS.
2679 * domwalk.h (enum dom_walker::reachability): New enum.
2680 (dom_walker::dom_walker): Convert bool param
2681 "skip_unreachable_blocks" to enum reachability.
2682 (set_all_edges_as_executable): New decl.
2683 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
2684 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
2686 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
2687 but converting true to REACHABLE_BLOCKS.
2688 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
2690 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
2691 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
2692 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
2694 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
2695 if check_all_array_refs will be called.
2697 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2699 * tree.c (selftest::test_location_wrappers): Add more test
2702 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2704 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
2705 (selftest::test_bit_in_range): Likewise.
2707 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
2710 * doc/sourcebuild.texi (vect_float): Say that the selector
2711 only describes the situation when -funsafe-math-optimizations is on.
2712 (vect_float_strict): Document.
2714 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
2716 PR tree-optimization/83965
2717 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
2718 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
2719 instead of checking only for a reduction.
2720 (vect_recog_widen_sum_pattern): Likewise.
2722 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2724 * predict.c (probably_never_executed): Only use precise profile info.
2725 (compute_function_frequency): Skip after inlining hack since we now
2726 have quality checking.
2728 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2730 * profile-count.h (profile_probability::very_unlikely,
2731 profile_probability::unlikely, profile_probability::even): Set
2732 precision to guessed.
2734 2018-01-23 Richard Biener <rguenther@suse.de>
2736 PR tree-optimization/83963
2737 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
2738 Properly terminate dominator walk when crossing the exit edge not
2739 when visiting its source block.
2741 2018-01-23 Jakub Jelinek <jakub@redhat.com>
2744 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
2745 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
2747 2018-01-22 Jakub Jelinek <jakub@redhat.com>
2749 PR tree-optimization/83957
2750 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
2751 semicolon after for body surrounded by braces.
2753 PR tree-optimization/83081
2754 * profile-count.h (profile_probability::split): New method.
2755 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
2756 Use profile_probability::split.
2757 (do_compare_rtx_and_jump): Fix adjustment of probabilities
2758 when splitting a single conditional jump into 2.
2760 2018-01-22 David Malcolm <dmalcolm@redhat.com>
2762 PR tree-optimization/69452
2763 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
2766 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2768 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
2769 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
2770 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
2772 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2774 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
2775 * config/rl78/rl78.md (movdi): New define_expand.
2776 * config/rl78/rl78.c (rl78_split_movdi): New function.
2778 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
2781 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
2783 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
2784 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
2785 128-bit to produce an UNSPEC move to get the double word with the
2786 signbit and then a shift directly to do signbit.
2787 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
2788 implementation with a new version that just does either a direct
2789 move or a regular move. Move memory interface to separate insns.
2790 Move insns so they are next to the expander.
2791 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
2792 with signbit move. Split big and little endian case.
2793 (signbit<mode>2_dm_mem_le): Likewise.
2794 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
2795 (signbit<mode>2_dm2): Likewise.
2797 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2799 * config/rl78/rl78.md (anddi3): New define_expand.
2801 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2803 * config/rl78/rl78.md (umindi3): New define_expand.
2805 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2807 * config/rl78/rl78.md (smindi3): New define_expand.
2809 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2811 * config/rl78/rl78.md (smaxdi3): New define_expand.
2813 2018-01-22 Carl Love <cel@us.ibm.com>
2815 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
2816 LVX_V1TI): Add macro expansion.
2817 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
2818 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
2819 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
2820 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
2821 Change check to determine if the instruction is a byte reversing
2822 entry. Fix typo in comment.
2823 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
2824 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
2825 Add def_builtin calls for new builtins.
2826 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
2827 Add define_insn expansion.
2829 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2831 * config/rl78/rl78.md (umaxdi3): New define_expand.
2833 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2835 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
2836 for non-QImode registers.
2838 2018-01-22 Richard Biener <rguenther@suse.de>
2840 PR tree-optimization/83963
2841 * graphite-scop-detection.c (scop_detection::get_sese): Delay
2842 including the loop exit block.
2843 (scop_detection::merge_sese): Likewise.
2844 (scop_detection::add_scop): Do it here instead.
2846 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2848 * doc/sourcebuild.texi (arm_softfloat): Document.
2850 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
2853 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
2854 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
2855 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
2857 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2858 David Edelsohn <dje.gcc@gmail.com>
2861 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
2862 Change "crset eq" to "crset 2".
2863 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
2864 (*call_indirect_aix<mode>_nospec): Likewise.
2865 (*call_value_indirect_aix<mode>_nospec): Likewise.
2866 (*call_indirect_elfv2<mode>_nospec): Likewise.
2867 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
2868 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
2869 change assembly output from . to $.
2870 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
2871 (indirect_jump<mode>_nospec): Change assembly output from . to $.
2872 (*tablejump<mode>_internal1_nospec): Likewise.
2874 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
2877 * config/sh/sh_optimize_sett_clrt.cc:
2878 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
2880 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
2882 PR tree-optimization/83940
2883 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
2884 offset_dt to vect_constant_def rather than vect_unknown_def_type.
2885 (vect_check_load_store_mask): Add a mask_dt_out parameter and
2886 use it to pass back the definition type.
2887 (vect_check_store_rhs): Likewise rhs_dt_out.
2888 (vect_build_gather_load_calls): Add a mask_dt argument and use
2889 it instead of a call to vect_is_simple_use.
2890 (vectorizable_store): Update calls to vect_check_load_store_mask
2891 and vect_check_store_rhs. Use the dt returned by the latter instead
2892 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
2893 instead of calls to vect_is_simple_use. Pass the scalar rather
2894 than the vector operand to vect_is_simple_use when handling
2895 second and subsequent copies of an rhs value.
2896 (vectorizable_load): Update calls to vect_check_load_store_mask
2897 and vect_build_gather_load_calls. Use the cached mask_dt and
2898 gs_info.offset_dt instead of calls to vect_is_simple_use.
2900 2018-01-20 Jakub Jelinek <jakub@redhat.com>
2903 * tree-emutls.c: Include gimplify.h.
2904 (lower_emutls_2): New function.
2905 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
2906 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
2907 it before further processing.
2910 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
2911 UINTVAL (trueop1) instead of INTVAL (op1).
2913 2018-01-19 Jakub Jelinek <jakub@redhat.com>
2917 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
2918 INCOMING_FRAME_SP_OFFSET if not defined.
2919 (scan_trace): Add ENTRY argument. If true and
2920 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
2921 emit a note to adjust the CFA offset.
2922 (create_cfi_notes): Adjust scan_trace callers.
2923 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
2924 INCOMING_FRAME_SP_OFFSET in the CIE.
2925 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
2926 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
2928 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
2929 * doc/tm.texi: Regenerated.
2931 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2933 PR rtl-optimization/83147
2934 * lra-constraints.c (remove_inheritance_pseudos): Use
2935 lra_substitute_pseudo_within_insn.
2937 2018-01-19 Tom de Vries <tom@codesourcery.com>
2938 Cesar Philippidis <cesar@codesourcery.com>
2941 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
2943 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
2946 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
2947 spaces for function labels.
2949 2018-01-19 Martin Liska <mliska@suse.cz>
2951 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
2952 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
2953 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
2954 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
2955 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
2956 (PRED_CONST_RETURN): Change from 69 to 65.
2957 (PRED_NULL_RETURN): Change from 91 to 71.
2958 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
2959 (PRED_LOOP_GUARD): Change from 66 to 73.
2961 2018-01-19 Martin Liska <mliska@suse.cz>
2963 * predict.c (predict_insn_def): Add new assert.
2964 (struct branch_predictor): Change type to signed integer.
2965 (test_prediction_value_range): Amend test to cover
2967 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
2968 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
2969 (PRED_LOOP_ITERATIONS_MAX): Likewise.
2970 (PRED_LOOP_IV_COMPARE): Likewise.
2971 * predict.h (PROB_UNINITIALIZED): Define new constant.
2973 2018-01-19 Martin Liska <mliska@suse.cz>
2975 * predict.c (dump_prediction): Add new format for
2976 analyze_brprob.py script which is enabled with -details
2978 * profile-count.h (precise_p): New function.
2980 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
2982 PR tree-optimization/83922
2983 * tree-vect-loop.c (vect_verify_full_masking): Return false if
2984 there are no statements that need masking.
2985 (vect_active_double_reduction_p): New function.
2986 (vect_analyze_loop_operations): Use it when handling phis that
2987 are not in the loop header.
2989 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
2991 PR tree-optimization/83914
2992 * tree-vect-loop.c (vectorizable_induction): Don't convert
2993 init_expr or apply the peeling adjustment for inductions
2994 that are nested within the vectorized loop.
2996 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2998 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
3001 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3005 * function.h (gimplify_parameters): Add gimple_seq * argument.
3006 * function.c: Include gimple.h and options.h.
3007 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
3008 for the added local temporaries if needed.
3009 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
3010 if there are any parameter cleanups, wrap whole body into a
3011 try/finally with the cleanups.
3013 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
3016 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
3017 Use GET_MODE_CLASS for scalar floating point.
3019 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3023 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3024 Fix call of call_cgraph_insertion_hooks.
3026 2018-01-18 Martin Sebor <msebor@redhat.com>
3028 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
3030 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3033 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
3036 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
3039 * common.opt: (-ffile-prefix-map): New option.
3040 * opts.c (common_handle_option): Defer it.
3041 * opts-global.c (handle_common_deferred_options): Handle it.
3042 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
3043 * file-prefix-map.h: New file.
3044 (remap_debug_filename, add_debug_prefix_map): ...here.
3045 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
3046 * final.c (debug_prefix_map, add_debug_prefix_map
3047 remap_debug_filename): Move to...
3048 * file-prefix-map.c: New file.
3049 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
3050 generalize, get rid of alloca(), use strrchr() instead of strchr().
3051 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
3052 Implement in terms of add_prefix_map().
3053 (remap_macro_filename, remap_debug_filename): Implement in term of
3055 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
3056 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
3057 * dbxout.c: Include file-prefix-map.h.
3058 * varasm.c: Likewise.
3059 * vmsdbgout.c: Likewise.
3060 * xcoffout.c: Likewise.
3061 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
3062 * doc/cppopts.texi (-fmacro-prefix-map): Document.
3063 * doc/invoke.texi (-ffile-prefix-map): Document.
3064 (-fdebug-prefix-map): Update description.
3066 2018-01-18 Martin Liska <mliska@suse.cz>
3068 * config/i386/i386.c (indirect_thunk_name): Document that also
3070 (output_indirect_thunk): Document why both instructions
3071 (pause and lfence) are generated.
3073 2018-01-18 Richard Biener <rguenther@suse.de>
3075 PR tree-optimization/83887
3076 * graphite-scop-detection.c
3077 (scop_detection::get_nearest_dom_with_single_entry): Remove.
3078 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
3079 (scop_detection::merge_sese): Re-implement with a flood-fill
3080 algorithm that properly finds a SESE region if it exists.
3082 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3085 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
3086 pointer_diff optimizations use view_convert instead of convert.
3088 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3090 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3091 Generate different code for -mno-speculate-indirect-jumps.
3092 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3093 (*call_indirect_aix<mode>): Disable for
3094 -mno-speculate-indirect-jumps.
3095 (*call_indirect_aix<mode>_nospec): New define_insn.
3096 (*call_value_indirect_aix<mode>): Disable for
3097 -mno-speculate-indirect-jumps.
3098 (*call_value_indirect_aix<mode>_nospec): New define_insn.
3099 (*sibcall_nonlocal_sysv<mode>): Generate different code for
3100 -mno-speculate-indirect-jumps.
3101 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3103 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3105 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
3106 long double type, set the flags for noting the default long double
3107 type, even if we don't pass or return a long double type.
3109 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
3112 * ipa-inline.c (flatten_function): Do not overwrite final inlining
3115 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
3117 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
3118 support for merge[hl].
3119 (fold_mergehl_helper): New helper function.
3120 (tree-vector-builder.h): New #include for tree_vector_builder usage.
3121 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
3122 (altivec_vmrglw_direct): Add xxmrglw insn.
3124 2018-01-17 Andrew Waterman <andrew@sifive.com>
3126 * config/riscv/riscv.c (riscv_conditional_register_usage): If
3127 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
3129 2018-01-17 David Malcolm <dmalcolm@redhat.com>
3132 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
3133 call the lto_location_cache before reading the
3134 DECL_SOURCE_LOCATION of the types.
3136 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3137 Richard Sandiford <richard.sandiford@linaro.org>
3139 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
3140 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
3141 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
3142 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
3143 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
3145 * config/aarch64/constraints.md (aarch64_movti_operand):
3147 * config/aarch64/predicates.md (Uti): Add new constraint.
3149 2018-01-17 Carl Love <cel@us.ibm.com>
3151 * config/rs6000/vsx.md (define_expand xl_len_r,
3152 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
3153 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
3155 (define_expand, define_insn): Move the shift left from the
3156 define_insn to the define_expand for lxvl and stxvl instructions.
3157 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
3158 and XL_LEN_R definitions to PURE.
3160 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
3162 * config/i386/i386.c (indirect_thunk_name): Declare regno
3163 as unsigned int. Compare regno with INVALID_REGNUM.
3164 (output_indirect_thunk): Ditto.
3165 (output_indirect_thunk_function): Ditto.
3166 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
3167 in the call to output_indirect_thunk_function.
3169 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
3172 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
3173 rather than the size of inner_type to determine the stack slot size
3174 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
3176 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
3179 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
3182 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
3184 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
3185 endian Linux systems to optionally enable multilibs for selecting
3186 the long double type if the user configured an explicit type.
3187 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
3188 have no long double multilibs if not defined.
3189 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
3190 warn if the user used -mabi={ieee,ibm}longdouble and we built
3191 multilibs for long double.
3192 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
3193 appropriate multilib option.
3194 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
3196 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
3197 for building long double multilibs.
3198 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
3200 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
3202 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
3205 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
3207 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
3210 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
3213 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
3216 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3218 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
3219 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
3221 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3223 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
3224 different rtl trees depending on TARGET_64BIT.
3225 (rs6000_gen_lvx): Likewise.
3227 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3229 * config/visium/visium.md (nop): Tweak comment.
3230 (hazard_nop): Likewise.
3232 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3234 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
3235 -mspeculate-indirect-jumps.
3236 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
3237 for -mno-speculate-indirect-jumps.
3238 (*call_indirect_elfv2<mode>_nospec): New define_insn.
3239 (*call_value_indirect_elfv2<mode>): Disable for
3240 -mno-speculate-indirect-jumps.
3241 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
3242 (indirect_jump): Emit different RTL for
3243 -mno-speculate-indirect-jumps.
3244 (*indirect_jump<mode>): Disable for
3245 -mno-speculate-indirect-jumps.
3246 (*indirect_jump<mode>_nospec): New define_insn.
3247 (tablejump): Emit different RTL for
3248 -mno-speculate-indirect-jumps.
3249 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
3250 (tablejumpsi_nospec): New define_expand.
3251 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
3252 (tablejumpdi_nospec): New define_expand.
3253 (*tablejump<mode>_internal1): Disable for
3254 -mno-speculate-indirect-jumps.
3255 (*tablejump<mode>_internal1_nospec): New define_insn.
3256 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
3259 2018-01-16 Artyom Skrobov tyomitch@gmail.com
3261 * caller-save.c (insert_save): Drop unnecessary parameter. All
3264 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3265 Richard Biener <rguenth@suse.de>
3268 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
3269 return early, inline manually is_gimple_sizepos. Make sure if we
3270 call gimplify_expr we don't end up with a gimple constant.
3271 * tree.c (variably_modified_type_p): Don't return true for
3272 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
3273 * gimplify.h (is_gimple_sizepos): Remove.
3275 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3277 PR tree-optimization/83857
3278 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
3279 vectorizable_live_operation for pure SLP statements.
3280 (vectorizable_live_operation): Handle PHIs.
3282 2018-01-16 Richard Biener <rguenther@suse.de>
3284 PR tree-optimization/83867
3285 * tree-vect-stmts.c (vect_transform_stmt): Precompute
3286 nested_in_vect_loop_p since the scalar stmt may get invalidated.
3288 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3291 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
3292 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
3293 If off is not INTEGER_CST, issue a may not be aligned warning
3294 rather than isn't aligned. Use isn%'t rather than isn't.
3295 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
3297 <case MULT_EXPR>: Improve the case when bottom and one of the
3298 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
3299 operand, in that case check if the other operand is multiple of
3300 bottom divided by the INTEGER_CST operand.
3302 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3305 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
3306 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
3307 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
3308 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
3309 * config/pa/pa.c (pa_function_arg_advance): Likewise.
3310 (pa_function_arg, pa_arg_partial_bytes): Likewise.
3311 (pa_function_arg_size): New function.
3313 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3315 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
3316 in a separate statement.
3318 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3320 PR tree-optimization/83847
3321 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
3322 group gathers and scatters.
3324 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3326 PR rtl-optimization/86620
3327 * params.def (max-sched-ready-insns): Bump minimum value to 1.
3329 PR rtl-optimization/83213
3330 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
3331 to last if both are JUMP_INSNs.
3333 PR tree-optimization/83843
3334 * gimple-ssa-store-merging.c
3335 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
3336 store_immediate_info for bswap/nop orig_stores.
3338 2018-01-15 Andrew Waterman <andrew@sifive.com>
3340 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
3342 <UDIV>: Increase cost if !TARGET_DIV.
3344 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
3346 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
3347 (define_attr "cr_logical_3op"): New.
3348 (cceq_ior_compare): Adjust.
3349 (cceq_ior_compare_complement): Adjust.
3350 (*cceq_rev_compare): Adjust.
3351 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
3352 (is_cracked_insn): Adjust.
3353 (insn_must_be_first_in_group): Adjust.
3354 * config/rs6000/40x.md: Adjust.
3355 * config/rs6000/440.md: Adjust.
3356 * config/rs6000/476.md: Adjust.
3357 * config/rs6000/601.md: Adjust.
3358 * config/rs6000/603.md: Adjust.
3359 * config/rs6000/6xx.md: Adjust.
3360 * config/rs6000/7450.md: Adjust.
3361 * config/rs6000/7xx.md: Adjust.
3362 * config/rs6000/8540.md: Adjust.
3363 * config/rs6000/cell.md: Adjust.
3364 * config/rs6000/e300c2c3.md: Adjust.
3365 * config/rs6000/e500mc.md: Adjust.
3366 * config/rs6000/e500mc64.md: Adjust.
3367 * config/rs6000/e5500.md: Adjust.
3368 * config/rs6000/e6500.md: Adjust.
3369 * config/rs6000/mpc.md: Adjust.
3370 * config/rs6000/power4.md: Adjust.
3371 * config/rs6000/power5.md: Adjust.
3372 * config/rs6000/power6.md: Adjust.
3373 * config/rs6000/power7.md: Adjust.
3374 * config/rs6000/power8.md: Adjust.
3375 * config/rs6000/power9.md: Adjust.
3376 * config/rs6000/rs64.md: Adjust.
3377 * config/rs6000/titan.md: Adjust.
3379 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3381 * config/i386/predicates.md (indirect_branch_operand): Rewrite
3382 ix86_indirect_branch_register logic.
3384 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3386 * config/i386/constraints.md (Bs): Update
3387 ix86_indirect_branch_register check. Don't check
3388 ix86_indirect_branch_register with GOT_memory_operand.
3390 * config/i386/predicates.md (GOT_memory_operand): Don't check
3391 ix86_indirect_branch_register here.
3392 (GOT32_symbol_operand): Likewise.
3394 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3396 * config/i386/predicates.md (constant_call_address_operand):
3397 Rewrite ix86_indirect_branch_register logic.
3398 (sibcall_insn_operand): Likewise.
3400 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3402 * config/i386/constraints.md (Bs): Replace
3403 ix86_indirect_branch_thunk_register with
3404 ix86_indirect_branch_register.
3406 * config/i386/i386.md (indirect_jump): Likewise.
3407 (tablejump): Likewise.
3408 (*sibcall_memory): Likewise.
3409 (*sibcall_value_memory): Likewise.
3410 Peepholes of indirect call and jump via memory: Likewise.
3411 * config/i386/i386.opt: Likewise.
3412 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3413 (GOT_memory_operand): Likewise.
3414 (call_insn_operand): Likewise.
3415 (sibcall_insn_operand): Likewise.
3416 (GOT32_symbol_operand): Likewise.
3418 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3421 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
3422 type rather than type addr's type points to.
3423 (expand_omp_atomic_mutex): Likewise.
3424 (expand_omp_atomic): Likewise.
3426 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3429 * config/i386/i386.c (output_indirect_thunk_function): Use
3430 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
3431 for __x86_return_thunk.
3433 2018-01-15 Richard Biener <rguenther@suse.de>
3436 * expmed.c (extract_bit_field_1): Fix typo.
3438 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3441 * config/arm/iterators.md (VF): New mode iterator.
3442 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
3443 Remove integer-related logic from pattern.
3444 (neon_vabd<mode>_3): Likewise.
3446 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3449 * common.opt (fstrict-overflow): No longer an alias.
3450 (fwrapv-pointer): New option.
3451 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
3452 also for pointer types based on flag_wrapv_pointer.
3453 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
3454 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
3455 opts->x_flag_wrapv got set.
3456 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
3457 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
3458 POINTER_TYPE_OVERFLOW_UNDEFINED.
3459 * match.pd: Likewise in address comparison pattern.
3460 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
3462 2018-01-15 Richard Biener <rguenther@suse.de>
3465 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
3466 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
3467 Reset type names to their identifier if their TYPE_DECL doesn't
3468 have linkage (and thus is used for ODR and devirt).
3469 (save_debug_info_for_decl): Remove.
3470 (save_debug_info_for_type): Likewise.
3471 (add_tree_to_fld_list): Adjust.
3472 * tree-pretty-print.c (dump_generic_node): Make dumping of
3473 type names more robust.
3475 2018-01-15 Richard Biener <rguenther@suse.de>
3477 * BASE-VER: Bump to 8.0.1.
3479 2018-01-14 Martin Sebor <msebor@redhat.com>
3482 * builtins.c (check_access): Avoid warning when the no-warning bit
3485 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
3487 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
3488 * ira-color (allocno_hard_regs_compare): Likewise.
3490 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
3493 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
3494 Use .pushsection/.popsection.
3496 2018-01-14 Martin Sebor <msebor@redhat.com>
3499 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
3501 2018-01-14 Jakub Jelinek <jakub@redhat.com>
3503 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
3504 entry from extra_headers.
3505 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
3506 extra_headers, make the list bitwise identical to the i?86-*-* one.
3508 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3510 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
3511 -mcmodel=large with -mindirect-branch=thunk,
3512 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
3513 -mfunction-return=thunk-extern.
3514 * doc/invoke.texi: Document -mcmodel=large is incompatible with
3515 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
3516 -mfunction-return=thunk and -mfunction-return=thunk-extern.
3518 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3520 * config/i386/i386.c (print_reg): Print the name of the full
3521 integer register without '%'.
3522 (ix86_print_operand): Handle 'V'.
3523 * doc/extend.texi: Document 'V' modifier.
3525 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3527 * config/i386/constraints.md (Bs): Disallow memory operand for
3528 -mindirect-branch-register.
3530 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3531 (GOT_memory_operand): Likewise.
3532 (call_insn_operand): Likewise.
3533 (sibcall_insn_operand): Likewise.
3534 (GOT32_symbol_operand): Likewise.
3535 * config/i386/i386.md (indirect_jump): Call convert_memory_address
3536 for -mindirect-branch-register.
3537 (tablejump): Likewise.
3538 (*sibcall_memory): Likewise.
3539 (*sibcall_value_memory): Likewise.
3540 Disallow peepholes of indirect call and jump via memory for
3541 -mindirect-branch-register.
3542 (*call_pop): Replace m with Bw.
3543 (*call_value_pop): Likewise.
3544 (*sibcall_pop_memory): Replace m with Bs.
3545 * config/i386/i386.opt (mindirect-branch-register): New option.
3546 * doc/invoke.texi: Document -mindirect-branch-register option.
3548 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3550 * config/i386/i386-protos.h (ix86_output_function_return): New.
3551 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
3552 set function_return_type.
3553 (indirect_thunk_name): Add ret_p to indicate thunk for function
3555 (output_indirect_thunk_function): Pass false to
3556 indirect_thunk_name.
3557 (ix86_output_indirect_branch_via_reg): Likewise.
3558 (ix86_output_indirect_branch_via_push): Likewise.
3559 (output_indirect_thunk_function): Create alias for function
3560 return thunk if regno < 0.
3561 (ix86_output_function_return): New function.
3562 (ix86_handle_fndecl_attribute): Handle function_return.
3563 (ix86_attribute_table): Add function_return.
3564 * config/i386/i386.h (machine_function): Add
3565 function_return_type.
3566 * config/i386/i386.md (simple_return_internal): Use
3567 ix86_output_function_return.
3568 (simple_return_internal_long): Likewise.
3569 * config/i386/i386.opt (mfunction-return=): New option.
3570 (indirect_branch): Mention -mfunction-return=.
3571 * doc/extend.texi: Document function_return function attribute.
3572 * doc/invoke.texi: Document -mfunction-return= option.
3574 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3576 * config/i386/i386-opts.h (indirect_branch): New.
3577 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
3578 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
3579 with local indirect jump when converting indirect call and jump.
3580 (ix86_set_indirect_branch_type): New.
3581 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
3582 (indirectlabelno): New.
3583 (indirect_thunk_needed): Likewise.
3584 (indirect_thunk_bnd_needed): Likewise.
3585 (indirect_thunks_used): Likewise.
3586 (indirect_thunks_bnd_used): Likewise.
3587 (INDIRECT_LABEL): Likewise.
3588 (indirect_thunk_name): Likewise.
3589 (output_indirect_thunk): Likewise.
3590 (output_indirect_thunk_function): Likewise.
3591 (ix86_output_indirect_branch_via_reg): Likewise.
3592 (ix86_output_indirect_branch_via_push): Likewise.
3593 (ix86_output_indirect_branch): Likewise.
3594 (ix86_output_indirect_jmp): Likewise.
3595 (ix86_code_end): Call output_indirect_thunk_function if needed.
3596 (ix86_output_call_insn): Call ix86_output_indirect_branch if
3598 (ix86_handle_fndecl_attribute): Handle indirect_branch.
3599 (ix86_attribute_table): Add indirect_branch.
3600 * config/i386/i386.h (machine_function): Add indirect_branch_type
3601 and has_local_indirect_jump.
3602 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
3604 (tablejump): Likewise.
3605 (*indirect_jump): Use ix86_output_indirect_jmp.
3606 (*tablejump_1): Likewise.
3607 (simple_return_indirect_internal): Likewise.
3608 * config/i386/i386.opt (mindirect-branch=): New option.
3609 (indirect_branch): New.
3612 (thunk-inline): Likewise.
3613 (thunk-extern): Likewise.
3614 * doc/extend.texi: Document indirect_branch function attribute.
3615 * doc/invoke.texi: Document -mindirect-branch= option.
3617 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
3620 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
3622 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
3624 * ipa-inline.c (want_inline_small_function_p): Return false if
3625 inlining has already failed with CIF_FINAL_ERROR.
3626 (update_caller_keys): Call want_inline_small_function_p before
3628 (update_callee_keys): Likewise.
3630 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3632 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
3634 (rs6000_quadword_masked_address_p): Likewise.
3635 (quad_aligned_load_p): Likewise.
3636 (quad_aligned_store_p): Likewise.
3637 (const_load_sequence_p): Add comment to describe the outer-most loop.
3638 (mimic_memory_attributes_and_flags): New function.
3639 (rs6000_gen_stvx): Likewise.
3640 (replace_swapped_aligned_store): Likewise.
3641 (rs6000_gen_lvx): Likewise.
3642 (replace_swapped_aligned_load): Likewise.
3643 (replace_swapped_load_constant): Capitalize argument name in
3644 comment describing this function.
3645 (rs6000_analyze_swaps): Add a third pass to search for vector loads
3646 and stores that access quad-word aligned addresses and replace
3647 with stvx or lvx instructions when appropriate.
3648 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
3649 New function prototype.
3650 (rs6000_quadword_masked_address_p): Likewise.
3651 (rs6000_gen_lvx): Likewise.
3652 (rs6000_gen_stvx): Likewise.
3653 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
3654 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
3655 when memory address is aligned.
3656 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
3657 this split to select lvx instruction when memory address is aligned.
3658 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
3659 instruction when memory address is aligned.
3660 (*vsx_le_perm_load_v16qi): Likewise.
3661 (four unnamed splitters): Modify to select the stvx instruction
3662 when memory is aligned.
3664 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
3666 * predict.c (determine_unlikely_bbs): Handle correctly BBs
3667 which appears in the queue multiple times.
3669 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3670 Alan Hayward <alan.hayward@arm.com>
3671 David Sherwood <david.sherwood@arm.com>
3673 * tree-vectorizer.h (vec_lower_bound): New structure.
3674 (_loop_vec_info): Add check_nonzero and lower_bounds.
3675 (LOOP_VINFO_CHECK_NONZERO): New macro.
3676 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
3677 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
3678 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
3679 fields. Make seg_len the distance travelled, not including the
3681 (dr_direction_indicator): Declare.
3682 (dr_zero_step_indicator): Likewise.
3683 (dr_known_forward_stride_p): Likewise.
3684 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
3686 (runtime_alias_check_p): Allow runtime alias checks with
3688 (operator ==): Compare access_size and align.
3689 (prune_runtime_alias_test_list): Rework for new distinction between
3690 the access_size and seg_len.
3691 (create_intersect_range_checks_index): Likewise. Cope with polynomial
3693 (get_segment_min_max): New function.
3694 (create_intersect_range_checks): Use it.
3695 (dr_step_indicator): New function.
3696 (dr_direction_indicator): Likewise.
3697 (dr_zero_step_indicator): Likewise.
3698 (dr_known_forward_stride_p): Likewise.
3699 * tree-loop-distribution.c (data_ref_segment_size): Return
3700 DR_STEP * (niters - 1).
3701 (compute_alias_check_pairs): Update call to the dr_with_seg_len
3703 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
3704 (vect_preserves_scalar_order_p): New function, split out from...
3705 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
3706 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
3707 (vect_vfa_access_size): New function.
3708 (vect_vfa_align): Likewise.
3709 (vect_compile_time_alias): Take access_size_a and access_b arguments.
3710 (dump_lower_bound): New function.
3711 (vect_check_lower_bound): Likewise.
3712 (vect_small_gap_p): Likewise.
3713 (vectorizable_with_step_bound_p): Likewise.
3714 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
3715 depencies if the vectorization factor is 1. Convert the checks
3716 for nonzero steps into checks on the bounds of DR_STEP. Try using
3717 a bunds check for variable steps if the minimum required step is
3718 relatively small. Update calls to the dr_with_seg_len
3719 constructor and to vect_compile_time_alias.
3720 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
3722 (vect_loop_versioning): Call it.
3723 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
3725 (vect_estimate_min_profitable_iters): Account for any bounds checks.
3727 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3728 Alan Hayward <alan.hayward@arm.com>
3729 David Sherwood <david.sherwood@arm.com>
3731 * doc/sourcebuild.texi (vect_scatter_store): Document.
3732 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
3734 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
3736 * genopinit.c (main): Add supports_vec_scatter_store and
3737 supports_vec_scatter_store_cached to target_optabs.
3738 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
3739 IFN_MASK_SCATTER_STORE.
3740 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
3742 * internal-fn.h (internal_store_fn_p): Declare.
3743 (internal_fn_stored_value_index): Likewise.
3744 * internal-fn.c (scatter_store_direct): New macro.
3745 (expand_scatter_store_optab_fn): New function.
3746 (direct_scatter_store_optab_supported_p): New macro.
3747 (internal_store_fn_p): New function.
3748 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
3749 IFN_MASK_SCATTER_STORE.
3750 (internal_fn_mask_index): Likewise.
3751 (internal_fn_stored_value_index): New function.
3752 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
3754 * optabs-query.h (supports_vec_scatter_store_p): Declare.
3755 * optabs-query.c (supports_vec_scatter_store_p): New function.
3756 * tree-vectorizer.h (vect_get_store_rhs): Declare.
3757 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
3758 true for scatter stores.
3759 (vect_gather_scatter_fn_p): Handle scatter stores too.
3760 (vect_check_gather_scatter): Consider using scatter stores if
3761 supports_vec_scatter_store_p.
3762 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
3764 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
3765 internal_fn_stored_value_index.
3766 (check_load_store_masking): Handle scatter stores too.
3767 (vect_get_store_rhs): Make public.
3768 (vectorizable_call): Use internal_store_fn_p.
3769 (vectorizable_store): Handle scatter store internal functions.
3770 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
3771 when deciding whether the end of the group has been reached.
3772 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
3773 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
3774 (mask_scatter_store<mode>): New insns.
3776 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3777 Alan Hayward <alan.hayward@arm.com>
3778 David Sherwood <david.sherwood@arm.com>
3780 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
3781 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
3782 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
3784 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
3785 Use vect_truncate_gather_scatter_offset if we can't treat the
3786 operation as a normal gather load or scatter store.
3787 (get_group_load_store_type): Take the gather_scatter_info
3788 as argument. Try using a gather load or scatter store for
3789 single-element groups.
3790 (get_load_store_type): Update calls to get_group_load_store_type
3791 and vect_use_strided_gather_scatters_p.
3793 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3794 Alan Hayward <alan.hayward@arm.com>
3795 David Sherwood <david.sherwood@arm.com>
3797 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
3798 optional tree argument.
3799 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
3801 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
3802 but continue to use the current value as a fallback.
3803 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
3804 to compare the updates.
3805 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
3806 (get_load_store_type): Use it when handling a strided access.
3807 (vect_get_strided_load_store_ops): New function.
3808 (vect_get_data_ptr_increment): Likewise.
3809 (vectorizable_load): Handle strided gather loads. Always pass
3810 a step to vect_create_data_ref_ptr and bump_vector_ptr.
3812 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3813 Alan Hayward <alan.hayward@arm.com>
3814 David Sherwood <david.sherwood@arm.com>
3816 * doc/md.texi (gather_load@var{m}): Document.
3817 (mask_gather_load@var{m}): Likewise.
3818 * genopinit.c (main): Add supports_vec_gather_load and
3819 supports_vec_gather_load_cached to target_optabs.
3820 * optabs-tree.c (init_tree_optimization_optabs): Use
3821 ggc_cleared_alloc to allocate target_optabs.
3822 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
3823 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
3825 * internal-fn.h (internal_load_fn_p): Declare.
3826 (internal_gather_scatter_fn_p): Likewise.
3827 (internal_fn_mask_index): Likewise.
3828 (internal_gather_scatter_fn_supported_p): Likewise.
3829 * internal-fn.c (gather_load_direct): New macro.
3830 (expand_gather_load_optab_fn): New function.
3831 (direct_gather_load_optab_supported_p): New macro.
3832 (direct_internal_fn_optab): New function.
3833 (internal_load_fn_p): Likewise.
3834 (internal_gather_scatter_fn_p): Likewise.
3835 (internal_fn_mask_index): Likewise.
3836 (internal_gather_scatter_fn_supported_p): Likewise.
3837 * optabs-query.c (supports_at_least_one_mode_p): New function.
3838 (supports_vec_gather_load_p): Likewise.
3839 * optabs-query.h (supports_vec_gather_load_p): Declare.
3840 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
3841 and memory_type field.
3842 (NUM_PATTERNS): Bump to 15.
3843 * tree-vect-data-refs.c: Include internal-fn.h.
3844 (vect_gather_scatter_fn_p): New function.
3845 (vect_describe_gather_scatter_call): Likewise.
3846 (vect_check_gather_scatter): Try using internal functions for
3847 gather loads. Recognize existing calls to a gather load function.
3848 (vect_analyze_data_refs): Consider using gather loads if
3849 supports_vec_gather_load_p.
3850 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
3851 (vect_get_gather_scatter_offset_type): Likewise.
3852 (vect_convert_mask_for_vectype): Likewise.
3853 (vect_add_conversion_to_patterm): Likewise.
3854 (vect_try_gather_scatter_pattern): Likewise.
3855 (vect_recog_gather_scatter_pattern): New pattern recognizer.
3856 (vect_vect_recog_func_ptrs): Add it.
3857 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
3858 internal_fn_mask_index and internal_gather_scatter_fn_p.
3859 (check_load_store_masking): Take the gather_scatter_info as an
3860 argument and handle gather loads.
3861 (vect_get_gather_scatter_ops): New function.
3862 (vectorizable_call): Check internal_load_fn_p.
3863 (vectorizable_load): Likewise. Handle gather load internal
3865 (vectorizable_store): Update call to check_load_store_masking.
3866 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
3867 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
3868 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
3869 (aarch64_gather_scale_operand_d): New predicates.
3870 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
3871 (mask_gather_load<mode>): New insns.
3873 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3874 Alan Hayward <alan.hayward@arm.com>
3875 David Sherwood <david.sherwood@arm.com>
3877 * optabs.def (fold_left_plus_optab): New optab.
3878 * doc/md.texi (fold_left_plus_@var{m}): Document.
3879 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
3880 * internal-fn.c (fold_left_direct): Define.
3881 (expand_fold_left_optab_fn): Likewise.
3882 (direct_fold_left_optab_supported_p): Likewise.
3883 * fold-const-call.c (fold_const_fold_left): New function.
3884 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
3885 * tree-parloops.c (valid_reduction_p): New function.
3886 (gather_scalar_reductions): Use it.
3887 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
3888 (vect_finish_replace_stmt): Declare.
3889 * tree-vect-loop.c (fold_left_reduction_fn): New function.
3890 (needs_fold_left_reduction_p): New function, split out from...
3891 (vect_is_simple_reduction): ...here. Accept reductions that
3892 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
3893 (vect_force_simple_reduction): Also store the reduction type in
3894 the assignment's STMT_VINFO_REDUC_TYPE.
3895 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
3896 (merge_with_identity): New function.
3897 (vect_expand_fold_left): Likewise.
3898 (vectorize_fold_left_reduction): Likewise.
3899 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
3900 scalar phi in place for it. Check for target support and reject
3901 cases that would reassociate the operation. Defer the transform
3902 phase to vectorize_fold_left_reduction.
3903 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
3904 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
3905 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
3907 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3909 * tree-if-conv.c (predicate_mem_writes): Remove redundant
3910 call to ifc_temp_var.
3912 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3913 Alan Hayward <alan.hayward@arm.com>
3914 David Sherwood <david.sherwood@arm.com>
3916 * target.def (legitimize_address_displacement): Take the original
3917 offset as a poly_int.
3918 * targhooks.h (default_legitimize_address_displacement): Update
3920 * targhooks.c (default_legitimize_address_displacement): Likewise.
3921 * doc/tm.texi: Regenerate.
3922 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
3923 as an argument, moving assert of ad->disp == ad->disp_term to...
3924 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
3925 Try calling targetm.legitimize_address_displacement before expanding
3926 the address rather than afterwards, and adjust for the new interface.
3927 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
3928 Match the new hook interface. Handle SVE addresses.
3929 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
3932 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3934 * Makefile.in (OBJS): Add early-remat.o.
3935 * target.def (select_early_remat_modes): New hook.
3936 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
3937 * doc/tm.texi: Regenerate.
3938 * targhooks.h (default_select_early_remat_modes): Declare.
3939 * targhooks.c (default_select_early_remat_modes): New function.
3940 * timevar.def (TV_EARLY_REMAT): New timevar.
3941 * passes.def (pass_early_remat): New pass.
3942 * tree-pass.h (make_pass_early_remat): Declare.
3943 * early-remat.c: New file.
3944 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
3946 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
3948 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3949 Alan Hayward <alan.hayward@arm.com>
3950 David Sherwood <david.sherwood@arm.com>
3952 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
3953 vfm1 with a bound_epilog parameter.
3954 (vect_do_peeling): Update calls accordingly, and move the prologue
3955 call earlier in the function. Treat the base bound_epilog as 0 for
3956 fully-masked loops and retain vf - 1 for other loops. Add 1 to
3957 this base when peeling for gaps.
3958 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
3959 with fully-masked loops.
3960 (vect_estimate_min_profitable_iters): Handle the single peeled
3961 iteration in that case.
3963 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3964 Alan Hayward <alan.hayward@arm.com>
3965 David Sherwood <david.sherwood@arm.com>
3967 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
3968 single-element interleaving even if the size is not a power of 2.
3969 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
3970 accesses for single-element interleaving if the group size is
3973 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3974 Alan Hayward <alan.hayward@arm.com>
3975 David Sherwood <david.sherwood@arm.com>
3977 * doc/md.texi (fold_extract_last_@var{m}): Document.
3978 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
3979 * optabs.def (fold_extract_last_optab): New optab.
3980 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
3981 * internal-fn.c (fold_extract_direct): New macro.
3982 (expand_fold_extract_optab_fn): Likewise.
3983 (direct_fold_extract_optab_supported_p): Likewise.
3984 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
3985 * tree-vect-loop.c (vect_model_reduction_cost): Handle
3986 EXTRACT_LAST_REDUCTION.
3987 (get_initial_def_for_reduction): Do not create an initial vector
3988 for EXTRACT_LAST_REDUCTION reductions.
3989 (vectorizable_reduction): Leave the scalar phi in place for
3990 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
3991 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
3992 epilogue code for EXTRACT_LAST_REDUCTION and defer the
3993 transform phase to vectorizable_condition.
3994 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
3996 (vect_finish_stmt_generation): ...here.
3997 (vect_finish_replace_stmt): New function.
3998 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
3999 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
4001 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
4003 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4004 Alan Hayward <alan.hayward@arm.com>
4005 David Sherwood <david.sherwood@arm.com>
4007 * doc/md.texi (extract_last_@var{m}): Document.
4008 * optabs.def (extract_last_optab): New optab.
4009 * internal-fn.def (EXTRACT_LAST): New internal function.
4010 * internal-fn.c (cond_unary_direct): New macro.
4011 (expand_cond_unary_optab_fn): Likewise.
4012 (direct_cond_unary_optab_supported_p): Likewise.
4013 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
4014 loops using EXTRACT_LAST.
4015 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
4016 (extract_last_<mode>): ...this optab.
4017 (vec_extract<mode><Vel>): Update accordingly.
4019 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4020 Alan Hayward <alan.hayward@arm.com>
4021 David Sherwood <david.sherwood@arm.com>
4023 * target.def (empty_mask_is_expensive): New hook.
4024 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
4025 * doc/tm.texi: Regenerate.
4026 * targhooks.h (default_empty_mask_is_expensive): Declare.
4027 * targhooks.c (default_empty_mask_is_expensive): New function.
4028 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
4029 if the target says that empty masks are expensive.
4030 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
4032 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
4034 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4035 Alan Hayward <alan.hayward@arm.com>
4036 David Sherwood <david.sherwood@arm.com>
4038 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
4039 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
4040 (vect_use_loop_mask_for_alignment_p): New function.
4041 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
4042 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
4043 niters_skip argument. Make sure that the first niters_skip elements
4044 of the first iteration are inactive.
4045 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
4046 Update call to vect_set_loop_masks_directly.
4047 (get_misalign_in_elems): New function, split out from...
4048 (vect_gen_prolog_loop_niters): ...here.
4049 (vect_update_init_of_dr): Take a code argument that specifies whether
4050 the adjustment should be added or subtracted.
4051 (vect_update_init_of_drs): Likewise.
4052 (vect_prepare_for_masked_peels): New function.
4053 (vect_do_peeling): Skip prologue peeling if we're using a mask
4054 instead. Update call to vect_update_inits_of_drs.
4055 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4057 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
4058 alignment. Do not include the number of peeled iterations in
4059 the minimum threshold in that case.
4060 (vectorizable_induction): Adjust the start value down by
4061 LOOP_VINFO_MASK_SKIP_NITERS iterations.
4062 (vect_transform_loop): Call vect_prepare_for_masked_peels.
4063 Take the number of skipped iterations into account when calculating
4065 * tree-vect-stmts.c (vect_gen_while_not): New function.
4067 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4068 Alan Hayward <alan.hayward@arm.com>
4069 David Sherwood <david.sherwood@arm.com>
4071 * doc/sourcebuild.texi (vect_fully_masked): Document.
4072 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
4074 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
4076 (vect_analyze_loop_2): ...here. Don't check the vectorization
4077 factor against the number of loop iterations if the loop is
4080 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4081 Alan Hayward <alan.hayward@arm.com>
4082 David Sherwood <david.sherwood@arm.com>
4084 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
4085 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
4086 (dump_groups): Update accordingly.
4087 (iv_use::mem_type): New member variable.
4088 (address_p): New function.
4089 (record_use): Add a mem_type argument and initialize the new
4091 (record_group_use): Add a mem_type argument. Use address_p.
4092 Remove obsolete null checks of base_object. Update call to record_use.
4093 (find_interesting_uses_op): Update call to record_group_use.
4094 (find_interesting_uses_cond): Likewise.
4095 (find_interesting_uses_address): Likewise.
4096 (get_mem_type_for_internal_fn): New function.
4097 (find_address_like_use): Likewise.
4098 (find_interesting_uses_stmt): Try find_address_like_use before
4099 calling find_interesting_uses_op.
4100 (addr_offset_valid_p): Use the iv mem_type field as the type
4101 of the addressed memory.
4102 (add_autoinc_candidates): Likewise.
4103 (get_address_cost): Likewise.
4104 (split_small_address_groups_p): Use address_p.
4105 (split_address_groups): Likewise.
4106 (add_iv_candidate_for_use): Likewise.
4107 (autoinc_possible_for_pair): Likewise.
4108 (rewrite_groups): Likewise.
4109 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
4110 (determine_group_iv_cost): Update after split of USE_ADDRESS.
4111 (get_alias_ptr_type_for_ptr_address): New function.
4112 (rewrite_use_address): Rewrite address uses in calls that were
4113 identified by find_address_like_use.
4115 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4116 Alan Hayward <alan.hayward@arm.com>
4117 David Sherwood <david.sherwood@arm.com>
4119 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
4121 * gimple-expr.h (is_gimple_addressable: Likewise.
4122 * gimple-expr.c (is_gimple_address): Likewise.
4123 * internal-fn.c (expand_call_mem_ref): New function.
4124 (expand_mask_load_optab_fn): Use it.
4125 (expand_mask_store_optab_fn): Likewise.
4127 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4128 Alan Hayward <alan.hayward@arm.com>
4129 David Sherwood <david.sherwood@arm.com>
4131 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
4132 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
4133 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
4134 (cond_umax@var{mode}): Document.
4135 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
4136 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
4137 (cond_umin_optab, cond_umax_optab): New optabs.
4138 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
4139 (COND_IOR, COND_XOR): New internal functions.
4140 * internal-fn.h (get_conditional_internal_fn): Declare.
4141 * internal-fn.c (cond_binary_direct): New macro.
4142 (expand_cond_binary_optab_fn): Likewise.
4143 (direct_cond_binary_optab_supported_p): Likewise.
4144 (get_conditional_internal_fn): New function.
4145 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
4146 Cope with reduction statements that are vectorized as calls rather
4148 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
4149 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
4150 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
4151 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4152 (UNSPEC_COND_EOR): New unspecs.
4153 (optab): Add mappings for them.
4154 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
4155 (sve_int_op, sve_fp_op): New int attributes.
4157 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4158 Alan Hayward <alan.hayward@arm.com>
4159 David Sherwood <david.sherwood@arm.com>
4161 * optabs.def (while_ult_optab): New optab.
4162 * doc/md.texi (while_ult@var{m}@var{n}): Document.
4163 * internal-fn.def (WHILE_ULT): New internal function.
4164 * internal-fn.h (direct_internal_fn_supported_p): New override
4165 that takes two types as argument.
4166 * internal-fn.c (while_direct): New macro.
4167 (expand_while_optab_fn): New function.
4168 (convert_optab_supported_p): Likewise.
4169 (direct_while_optab_supported_p): New macro.
4170 * wide-int.h (wi::udiv_ceil): New function.
4171 * tree-vectorizer.h (rgroup_masks): New structure.
4172 (vec_loop_masks): New typedef.
4173 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
4175 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
4176 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
4177 (vect_max_vf): New function.
4178 (slpeel_make_loop_iterate_ntimes): Delete.
4179 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
4180 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
4181 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
4182 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
4183 internal-fn.h, stor-layout.h and optabs-query.h.
4184 (vect_set_loop_mask): New function.
4185 (add_preheader_seq): Likewise.
4186 (add_header_seq): Likewise.
4187 (interleave_supported_p): Likewise.
4188 (vect_maybe_permute_loop_masks): Likewise.
4189 (vect_set_loop_masks_directly): Likewise.
4190 (vect_set_loop_condition_masked): Likewise.
4191 (vect_set_loop_condition_unmasked): New function, split out from
4192 slpeel_make_loop_iterate_ntimes.
4193 (slpeel_make_loop_iterate_ntimes): Rename to..
4194 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
4195 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
4196 (vect_do_peeling): Update call accordingly.
4197 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
4199 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4200 mask_compare_type, can_fully_mask_p and fully_masked_p.
4201 (release_vec_loop_masks): New function.
4202 (_loop_vec_info): Use it to free the loop masks.
4203 (can_produce_all_loop_masks_p): New function.
4204 (vect_get_max_nscalars_per_iter): Likewise.
4205 (vect_verify_full_masking): Likewise.
4206 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
4207 retries, and free the mask rgroups before retrying. Check loop-wide
4208 reasons for disallowing fully-masked loops. Make the final decision
4209 about whether use a fully-masked loop or not.
4210 (vect_estimate_min_profitable_iters): Do not assume that peeling
4211 for the number of iterations will be needed for fully-masked loops.
4212 (vectorizable_reduction): Disable fully-masked loops.
4213 (vectorizable_live_operation): Likewise.
4214 (vect_halve_mask_nunits): New function.
4215 (vect_double_mask_nunits): Likewise.
4216 (vect_record_loop_mask): Likewise.
4217 (vect_get_loop_mask): Likewise.
4218 (vect_transform_loop): Handle the case in which the final loop
4219 iteration might handle a partial vector. Call vect_set_loop_condition
4220 instead of slpeel_make_loop_iterate_ntimes.
4221 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
4222 (check_load_store_masking): New function.
4223 (prepare_load_store_mask): Likewise.
4224 (vectorizable_store): Handle fully-masked loops.
4225 (vectorizable_load): Likewise.
4226 (supportable_widening_operation): Use vect_halve_mask_nunits for
4228 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
4229 (vect_gen_while): New function.
4230 * config/aarch64/aarch64.md (umax<mode>3): New expander.
4231 (aarch64_uqdec<mode>): New insn.
4233 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4234 Alan Hayward <alan.hayward@arm.com>
4235 David Sherwood <david.sherwood@arm.com>
4237 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
4238 (reduc_xor_scal_optab): New optabs.
4239 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
4240 (reduc_xor_scal_@var{m}): Document.
4241 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
4242 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
4244 * fold-const-call.c (fold_const_call): Handle them.
4245 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
4246 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
4247 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
4248 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
4249 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
4250 (UNSPEC_XORV): New unspecs.
4251 (optab): Add entries for them.
4252 (BITWISEV): New int iterator.
4253 (bit_reduc_op): New int attributes.
4255 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4256 Alan Hayward <alan.hayward@arm.com>
4257 David Sherwood <david.sherwood@arm.com>
4259 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
4260 * internal-fn.def (VEC_SHL_INSERT): New internal function.
4261 * optabs.def (vec_shl_insert_optab): New optab.
4262 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
4263 (duplicate_and_interleave): Likewise.
4264 * tree-vect-loop.c: Include internal-fn.h.
4265 (neutral_op_for_slp_reduction): New function, split out from
4266 get_initial_defs_for_reduction.
4267 (get_initial_def_for_reduction): Handle option 2 for variable-length
4268 vectors by loading the neutral value into a vector and then shifting
4269 the initial value into element 0.
4270 (get_initial_defs_for_reduction): Replace the code argument with
4271 the neutral value calculated by neutral_op_for_slp_reduction.
4272 Use gimple_build_vector for constant-length vectors.
4273 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
4274 but the first group_size elements have a neutral value.
4275 Use duplicate_and_interleave otherwise.
4276 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
4277 Update call to get_initial_defs_for_reduction. Handle SLP
4278 reductions for variable-length vectors by creating one vector
4279 result for each scalar result, with the elements associated
4280 with other scalar results stubbed out with the neutral value.
4281 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
4282 Require IFN_VEC_SHL_INSERT for double reductions on
4283 variable-length vectors, or SLP reductions that have
4284 a neutral value. Require can_duplicate_and_interleave_p
4285 support for variable-length unchained SLP reductions if there
4286 is no neutral value, such as for MIN/MAX reductions. Also require
4287 the number of vector elements to be a multiple of the number of
4288 SLP statements when doing variable-length unchained SLP reductions.
4289 Update call to vect_create_epilog_for_reduction.
4290 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
4291 and remove initial values.
4292 (duplicate_and_interleave): Make public.
4293 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
4294 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
4296 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4297 Alan Hayward <alan.hayward@arm.com>
4298 David Sherwood <david.sherwood@arm.com>
4300 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
4301 (can_duplicate_and_interleave_p): New function.
4302 (vect_get_and_check_slp_defs): Take the vector of statements
4303 rather than just the current one. Remove excess parentheses.
4304 Restriction rejectinon of vect_constant_def and vect_external_def
4305 for variable-length vectors to boolean types, or types for which
4306 can_duplicate_and_interleave_p is false.
4307 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
4308 (duplicate_and_interleave): New function.
4309 (vect_get_constant_vectors): Use gimple_build_vector for
4310 constant-length vectors and suitable variable-length constant
4311 vectors. Use duplicate_and_interleave for other variable-length
4312 vectors. Don't defer the update when inserting new statements.
4314 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4315 Alan Hayward <alan.hayward@arm.com>
4316 David Sherwood <david.sherwood@arm.com>
4318 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
4319 min_profitable_iters doesn't go negative.
4321 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4322 Alan Hayward <alan.hayward@arm.com>
4323 David Sherwood <david.sherwood@arm.com>
4325 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
4326 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
4327 * optabs.def (vec_mask_load_lanes_optab): New optab.
4328 (vec_mask_store_lanes_optab): Likewise.
4329 * internal-fn.def (MASK_LOAD_LANES): New internal function.
4330 (MASK_STORE_LANES): Likewise.
4331 * internal-fn.c (mask_load_lanes_direct): New macro.
4332 (mask_store_lanes_direct): Likewise.
4333 (expand_mask_load_optab_fn): Handle masked operations.
4334 (expand_mask_load_lanes_optab_fn): New macro.
4335 (expand_mask_store_optab_fn): Handle masked operations.
4336 (expand_mask_store_lanes_optab_fn): New macro.
4337 (direct_mask_load_lanes_optab_supported_p): Likewise.
4338 (direct_mask_store_lanes_optab_supported_p): Likewise.
4339 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
4341 (vect_load_lanes_supported): Likewise.
4342 * tree-vect-data-refs.c (strip_conversion): New function.
4343 (can_group_stmts_p): Likewise.
4344 (vect_analyze_data_ref_accesses): Use it instead of checking
4345 for a pair of assignments.
4346 (vect_store_lanes_supported): Take a masked_p parameter.
4347 (vect_load_lanes_supported): Likewise.
4348 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
4349 vect_store_lanes_supported and vect_load_lanes_supported.
4350 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4351 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
4352 parameter. Don't allow gaps for masked accesses.
4353 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
4354 and vect_load_lanes_supported.
4355 (get_load_store_type): Take a masked_p parameter and update
4356 call to get_group_load_store_type.
4357 (vectorizable_store): Update call to get_load_store_type.
4358 Handle IFN_MASK_STORE_LANES.
4359 (vectorizable_load): Update call to get_load_store_type.
4360 Handle IFN_MASK_LOAD_LANES.
4362 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4363 Alan Hayward <alan.hayward@arm.com>
4364 David Sherwood <david.sherwood@arm.com>
4366 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
4368 * config/aarch64/aarch64-protos.h
4369 (aarch64_sve_struct_memory_operand_p): Declare.
4370 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
4371 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
4372 (VPRED, vpred): Handle SVE structure modes.
4373 * config/aarch64/constraints.md (Utx): New constraint.
4374 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
4375 (aarch64_sve_struct_nonimmediate_operand): New predicates.
4376 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
4377 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
4378 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
4379 structure modes. Split into pieces after RA.
4380 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
4381 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
4383 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
4384 SVE structure modes.
4385 (aarch64_classify_address): Likewise.
4386 (sizetochar): Move earlier in file.
4387 (aarch64_print_operand): Handle SVE register lists.
4388 (aarch64_array_mode): New function.
4389 (aarch64_sve_struct_memory_operand_p): Likewise.
4390 (TARGET_ARRAY_MODE): Redefine.
4392 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4393 Alan Hayward <alan.hayward@arm.com>
4394 David Sherwood <david.sherwood@arm.com>
4396 * target.def (array_mode): New target hook.
4397 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
4398 * doc/tm.texi: Regenerate.
4399 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
4400 * hooks.c (hook_optmode_mode_uhwi_none): New function.
4401 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
4403 * stor-layout.c (mode_for_array): Likewise. Support polynomial
4406 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4407 Alan Hayward <alan.hayward@arm.com>
4408 David Sherwood <david.sherwood@arm.com>
4410 * fold-const.c (fold_binary_loc): Check the argument types
4411 rather than the result type when testing for a vector operation.
4413 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4415 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
4416 * doc/tm.texi: Regenerate.
4418 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4419 Alan Hayward <alan.hayward@arm.com>
4420 David Sherwood <david.sherwood@arm.com>
4422 * doc/invoke.texi (-msve-vector-bits=): Document new option.
4423 (sve): Document new AArch64 extension.
4424 * doc/md.texi (w): Extend the description of the AArch64
4425 constraint to include SVE vectors.
4426 (Upl, Upa): Document new AArch64 predicate constraints.
4427 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
4429 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
4430 (msve-vector-bits=): New option.
4431 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
4432 SVE when these are disabled.
4433 (sve): New extension.
4434 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
4435 modes. Adjust their number of units based on aarch64_sve_vg.
4436 (MAX_BITSIZE_MODE_ANY_MODE): Define.
4437 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
4438 aarch64_addr_query_type.
4439 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
4440 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
4441 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
4442 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
4443 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
4444 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
4445 (aarch64_simd_imm_zero_p): Delete.
4446 (aarch64_check_zero_based_sve_index_immediate): Declare.
4447 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4448 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4449 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4450 (aarch64_sve_float_mul_immediate_p): Likewise.
4451 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4453 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
4454 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
4455 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
4456 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
4457 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
4458 (aarch64_regmode_natural_size): Likewise.
4459 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
4460 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
4462 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
4463 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
4464 for VG and the SVE predicate registers.
4465 (V_ALIASES): Add a "z"-prefixed alias.
4466 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
4467 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
4468 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
4469 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
4470 (REG_CLASS_NAMES): Add entries for them.
4471 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
4472 and the predicate registers.
4473 (aarch64_sve_vg): Declare.
4474 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
4475 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
4476 (REGMODE_NATURAL_SIZE): Define.
4477 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
4479 * config/aarch64/aarch64.c: Include cfgrtl.h.
4480 (simd_immediate_info): Add a constructor for series vectors,
4481 and an associated step field.
4482 (aarch64_sve_vg): New variable.
4483 (aarch64_dbx_register_number): Handle VG and the predicate registers.
4484 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
4485 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
4486 (VEC_ANY_DATA, VEC_STRUCT): New constants.
4487 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
4488 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
4489 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
4490 (aarch64_get_mask_mode): New functions.
4491 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
4492 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
4493 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
4494 predicate modes and predicate registers. Explicitly restrict
4495 GPRs to modes of 16 bytes or smaller. Only allow FP registers
4496 to store a vector mode if it is recognized by
4497 aarch64_classify_vector_mode.
4498 (aarch64_regmode_natural_size): New function.
4499 (aarch64_hard_regno_caller_save_mode): Return the original mode
4501 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
4502 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
4503 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
4504 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
4506 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
4507 does not overlap dest if the function is frame-related. Handle
4509 (aarch64_split_add_offset): New function.
4510 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
4511 them aarch64_add_offset.
4512 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
4513 and update call to aarch64_sub_sp.
4514 (aarch64_add_cfa_expression): New function.
4515 (aarch64_expand_prologue): Pass extra temporary registers to the
4516 functions above. Handle the case in which we need to emit new
4517 DW_CFA_expressions for registers that were originally saved
4518 relative to the stack pointer, but now have to be expressed
4519 relative to the frame pointer.
4520 (aarch64_output_mi_thunk): Pass extra temporary registers to the
4522 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
4523 IP0 and IP1 values for SVE frames.
4524 (aarch64_expand_vec_series): New function.
4525 (aarch64_expand_sve_widened_duplicate): Likewise.
4526 (aarch64_expand_sve_const_vector): Likewise.
4527 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
4528 Handle SVE constants. Use emit_move_insn to move a force_const_mem
4529 into the register, rather than emitting a SET directly.
4530 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
4531 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
4532 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
4533 (offset_9bit_signed_scaled_p): New functions.
4534 (aarch64_replicate_bitmask_imm): New function.
4535 (aarch64_bitmask_imm): Use it.
4536 (aarch64_cannot_force_const_mem): Reject expressions involving
4537 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
4538 (aarch64_classify_index): Handle SVE indices, by requiring
4539 a plain register index with a scale that matches the element size.
4540 (aarch64_classify_address): Handle SVE addresses. Assert that
4541 the mode of the address is VOIDmode or an integer mode.
4542 Update call to aarch64_classify_symbol.
4543 (aarch64_classify_symbolic_expression): Update call to
4544 aarch64_classify_symbol.
4545 (aarch64_const_vec_all_in_range_p): New function.
4546 (aarch64_print_vector_float_operand): Likewise.
4547 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
4548 "vN" for FP registers with SVE modes. Handle (const ...) vectors
4549 and the FP immediates 1.0 and 0.5.
4550 (aarch64_print_address_internal): Handle SVE addresses.
4551 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
4552 (aarch64_regno_regclass): Handle predicate registers.
4553 (aarch64_secondary_reload): Handle big-endian reloads of SVE
4555 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
4556 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
4557 (aarch64_convert_sve_vector_bits): New function.
4558 (aarch64_override_options): Use it to handle -msve-vector-bits=.
4559 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4561 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
4562 Handle SVE vector and predicate modes. Accept VL-based constants
4563 that need only one temporary register, and VL offsets that require
4564 no temporary registers.
4565 (aarch64_conditional_register_usage): Mark the predicate registers
4566 as fixed if SVE isn't available.
4567 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
4568 Return true for SVE vector and predicate modes.
4569 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
4570 rather than an unsigned int. Handle SVE modes.
4571 (aarch64_preferred_simd_mode): Update call accordingly. Handle
4573 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
4575 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4576 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4577 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4578 (aarch64_sve_float_mul_immediate_p): New functions.
4579 (aarch64_sve_valid_immediate): New function.
4580 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
4581 Explicitly reject structure modes. Check for INDEX constants.
4582 Handle PTRUE and PFALSE constants.
4583 (aarch64_check_zero_based_sve_index_immediate): New function.
4584 (aarch64_simd_imm_zero_p): Delete.
4585 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
4586 vector modes. Accept constants in the range of CNT[BHWD].
4587 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
4588 ask for an Advanced SIMD mode.
4589 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
4590 (aarch64_simd_vector_alignment): Handle SVE predicates.
4591 (aarch64_vectorize_preferred_vector_alignment): New function.
4592 (aarch64_simd_vector_alignment_reachable): Use it instead of
4594 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
4595 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
4597 (MAX_VECT_LEN): Delete.
4598 (expand_vec_perm_d): Add a vec_flags field.
4599 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
4600 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
4601 (aarch64_evpc_ext): Don't apply a big-endian lane correction
4603 (aarch64_evpc_rev): Rename to...
4604 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
4605 (aarch64_evpc_rev_global): New function.
4606 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
4607 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
4609 (aarch64_evpc_sve_tbl): New function.
4610 (aarch64_expand_vec_perm_const_1): Update after rename of
4611 aarch64_evpc_rev. Handle SVE permutes too, trying
4612 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
4613 than aarch64_evpc_tbl.
4614 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
4615 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
4616 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
4617 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
4618 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
4619 (aarch64_expand_sve_vcond): New functions.
4620 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
4621 of aarch64_vector_mode_p.
4622 (aarch64_dwarf_poly_indeterminate_value): New function.
4623 (aarch64_compute_pressure_classes): Likewise.
4624 (aarch64_can_change_mode_class): Likewise.
4625 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
4626 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
4627 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
4628 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
4629 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
4630 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
4631 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
4632 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
4634 (Dn, Dl, Dr): Accept const as well as const_vector.
4635 (Dz): Likewise. Compare against CONST0_RTX.
4636 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
4637 of "vector" where appropriate.
4638 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
4639 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
4640 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
4641 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
4642 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
4643 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
4644 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
4645 (v_int_equiv): Extend to SVE modes.
4646 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
4648 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
4649 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
4650 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
4651 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
4652 (SVE_COND_FP_CMP): New int iterators.
4653 (perm_hilo): Handle the new unpack unspecs.
4654 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
4656 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
4657 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
4658 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
4659 (aarch64_equality_operator, aarch64_constant_vector_operand)
4660 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
4661 (aarch64_sve_nonimmediate_operand): Likewise.
4662 (aarch64_sve_general_operand): Likewise.
4663 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
4664 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
4665 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
4666 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
4667 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
4668 (aarch64_sve_float_arith_immediate): Likewise.
4669 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
4670 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
4671 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
4672 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
4673 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
4674 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
4675 (aarch64_sve_float_arith_operand): Likewise.
4676 (aarch64_sve_float_arith_with_sub_operand): Likewise.
4677 (aarch64_sve_float_mul_operand): Likewise.
4678 (aarch64_sve_vec_perm_operand): Likewise.
4679 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
4680 (aarch64_mov_operand): Accept const_poly_int and const_vector.
4681 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
4682 as well as const_vector.
4683 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
4684 in file. Use CONST0_RTX and CONSTM1_RTX.
4685 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
4686 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
4687 Use aarch64_simd_imm_zero.
4688 * config/aarch64/aarch64-sve.md: New file.
4689 * config/aarch64/aarch64.md: Include it.
4690 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
4691 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
4692 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
4693 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
4694 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
4695 (sve): New attribute.
4696 (enabled): Disable instructions with the sve attribute unless
4698 (movqi, movhi): Pass CONST_POLY_INT operaneds through
4699 aarch64_expand_mov_immediate.
4700 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
4701 CNT[BHSD] immediates.
4702 (movti): Split CONST_POLY_INT moves into two halves.
4703 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
4704 Split additions that need a temporary here if the destination
4705 is the stack pointer.
4706 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
4707 (*add<mode>3_poly_1): New instruction.
4708 (set_clobber_cc): New expander.
4710 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4712 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
4713 parameter and use it instead of GET_MODE_SIZE (innermode). Use
4714 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
4715 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
4716 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
4717 Change innermode from fixed_mode_size to machine_mode.
4718 (simplify_subreg): Update call accordingly. Handle a constant-sized
4719 subreg of a variable-length CONST_VECTOR.
4721 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4722 Alan Hayward <alan.hayward@arm.com>
4723 David Sherwood <david.sherwood@arm.com>
4725 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
4726 (add_offset_to_base): New function, split out from...
4727 (create_mem_ref): ...here. When handling a scale other than 1,
4728 check first whether the address is valid without the offset.
4729 Add it into the base if so, leaving the index and scale as-is.
4731 2018-01-12 Jakub Jelinek <jakub@redhat.com>
4734 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
4735 fold_for_warn before checking if arg2 is INTEGER_CST.
4737 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
4739 * config/rs6000/predicates.md (load_multiple_operation): Delete.
4740 (store_multiple_operation): Delete.
4741 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
4742 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
4743 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
4744 guarded by TARGET_STRING.
4745 (rs6000_output_load_multiple): Delete.
4746 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
4747 OPTION_MASK_STRING / TARGET_STRING handling.
4748 (print_operand) <'N', 'O'>: Add comment that these are unused now.
4749 (const rs6000_opt_masks) <"string">: Change mask to 0.
4750 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
4751 (MASK_STRING): Delete.
4752 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
4754 (load_multiple): Delete.
4761 (store_multiple): Delete.
4768 (movmemsi_8reg): Delete.
4769 (corresponding unnamed define_insn): Delete.
4770 (movmemsi_6reg): Delete.
4771 (corresponding unnamed define_insn): Delete.
4772 (movmemsi_4reg): Delete.
4773 (corresponding unnamed define_insn): Delete.
4774 (movmemsi_2reg): Delete.
4775 (corresponding unnamed define_insn): Delete.
4776 (movmemsi_1reg): Delete.
4777 (corresponding unnamed define_insn): Delete.
4778 * config/rs6000/rs6000.opt (mno-string): New.
4779 (mstring): Replace by deprecation warning stub.
4780 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
4782 2018-01-12 Jakub Jelinek <jakub@redhat.com>
4784 * regrename.c (regrename_do_replace): If replacing the same
4785 reg multiple times, try to reuse last created gen_raw_REG.
4788 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
4789 main to workaround a bug in GDB.
4791 2018-01-12 Tom de Vries <tom@codesourcery.com>
4794 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
4796 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
4798 PR rtl-optimization/80481
4799 * ira-color.c (get_cap_member): New function.
4800 (allocnos_conflict_by_live_ranges_p): Use it.
4801 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
4802 (setup_slot_coalesced_allocno_live_ranges): Ditto.
4804 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
4807 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
4808 (*saddl_se_1): Ditto.
4810 (*ssubl_se_1): Ditto.
4812 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
4814 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
4815 rather than wi::to_widest for DR_INITs.
4816 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
4817 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
4818 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
4820 (vect_analyze_group_access_1): Note that here.
4822 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
4824 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
4825 polynomial type sizes.
4827 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
4829 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
4830 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
4831 (gimple_add_tmp_var): Likewise.
4833 2018-01-12 Martin Liska <mliska@suse.cz>
4835 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
4836 (gimple_alloc_sizes): Likewise.
4837 (dump_gimple_statistics): Use PRIu64 in printf format.
4838 * gimple.h: Change uint64_t to int.
4840 2018-01-12 Martin Liska <mliska@suse.cz>
4842 * tree-core.h: Use uint64_t instead of int.
4843 * tree.c (tree_node_counts): Likewise.
4844 (tree_node_sizes): Likewise.
4845 (dump_tree_statistics): Use PRIu64 in printf format.
4847 2018-01-12 Martin Liska <mliska@suse.cz>
4849 * Makefile.in: As qsort_chk is implemented in vec.c, add
4850 vec.o to linkage of gencfn-macros.
4851 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
4852 passing the info to record_node_allocation_statistics.
4853 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
4855 * ggc-common.c (struct ggc_usage): Add operator== and use
4856 it in operator< and compare function.
4857 * mem-stats.h (struct mem_usage): Likewise.
4858 * vec.c (struct vec_usage): Remove operator< and compare
4859 function. Can be simply inherited.
4861 2018-01-12 Martin Jambor <mjambor@suse.cz>
4864 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
4865 * tree-ssa-math-opts.c: Include domwalk.h.
4866 (convert_mult_to_fma_1): New function.
4867 (fma_transformation_info): New type.
4868 (fma_deferring_state): Likewise.
4869 (cancel_fma_deferring): New function.
4870 (result_of_phi): Likewise.
4871 (last_fma_candidate_feeds_initial_phi): Likewise.
4872 (convert_mult_to_fma): Added deferring logic, split actual
4873 transformation to convert_mult_to_fma_1.
4874 (math_opts_dom_walker): New type.
4875 (math_opts_dom_walker::after_dom_children): New method, body moved
4876 here from pass_optimize_widening_mul::execute, added deferring logic
4878 (pass_optimize_widening_mul::execute): Moved most of code to
4879 math_opts_dom_walker::after_dom_children.
4880 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
4881 * config/i386/i386.c (ix86_option_override_internal): Added
4882 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
4884 2018-01-12 Richard Biener <rguenther@suse.de>
4887 * dwarf2out.c (gen_variable_die): Do not reset old_die for
4888 inline instance vars.
4890 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
4893 * config/rx/rx.c (rx_is_restricted_memory_address):
4896 2018-01-12 Richard Biener <rguenther@suse.de>
4898 PR tree-optimization/80846
4899 * target.def (split_reduction): New target hook.
4900 * targhooks.c (default_split_reduction): New function.
4901 * targhooks.h (default_split_reduction): Declare.
4902 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
4903 target requests first reduce vectors by combining low and high
4905 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
4906 (get_vectype_for_scalar_type_and_size): Export.
4907 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
4908 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
4909 * doc/tm.texi: Regenerate.
4910 * config/i386/i386.c (ix86_split_reduction): Implement
4911 TARGET_VECTORIZE_SPLIT_REDUCTION.
4913 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4916 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
4917 in PIC mode except for TARGET_VXWORKS_RTP.
4918 * config/sparc/sparc.c: Include cfgrtl.h.
4919 (TARGET_INIT_PIC_REG): Define.
4920 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
4921 (sparc_pic_register_p): New predicate.
4922 (sparc_legitimate_address_p): Use it.
4923 (sparc_legitimize_pic_address): Likewise.
4924 (sparc_delegitimize_address): Likewise.
4925 (sparc_mode_dependent_address_p): Likewise.
4926 (gen_load_pcrel_sym): Remove 4th parameter.
4927 (load_got_register): Adjust call to above. Remove obsolete stuff.
4928 (sparc_expand_prologue): Do not call load_got_register here.
4929 (sparc_flat_expand_prologue): Likewise.
4930 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
4931 (sparc_use_pseudo_pic_reg): New function.
4932 (sparc_init_pic_reg): Likewise.
4933 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
4934 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
4936 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
4938 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
4939 Add item for branch_cost.
4941 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4943 PR rtl-optimization/83565
4944 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
4945 not extend the result to a larger mode for rotate operations.
4946 (num_sign_bit_copies1): Likewise.
4948 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4951 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
4953 Use values-Xc.o for -pedantic.
4954 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
4956 2018-01-12 Martin Liska <mliska@suse.cz>
4959 * ipa-devirt.c (final_warning_record::grow_type_warnings):
4961 (possible_polymorphic_call_targets): Use it.
4962 (ipa_devirt): Likewise.
4964 2018-01-12 Martin Liska <mliska@suse.cz>
4966 * profile-count.h (enum profile_quality): Use 0 as invalid
4967 enum value of profile_quality.
4969 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
4971 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
4972 -mext-string options.
4974 2018-01-12 Richard Biener <rguenther@suse.de>
4976 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
4977 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
4978 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
4980 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
4982 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
4984 * configure.ac (--with-long-double-format): Add support for the
4985 configuration option to change the default long double format on
4987 * config.gcc (powerpc*-linux*-*): Likewise.
4988 * configure: Regenerate.
4989 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
4990 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
4991 used without modification.
4993 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4995 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
4996 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
4997 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
4998 MISC_BUILTIN_SPEC_BARRIER.
4999 (rs6000_init_builtins): Likewise.
5000 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
5002 (speculation_barrier): New define_insn.
5003 * doc/extend.texi: Document __builtin_speculation_barrier.
5005 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5008 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
5009 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
5010 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
5012 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
5013 integral modes instead of "ss" and "sd".
5014 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
5015 vectors with 32-bit and 64-bit elements.
5016 (vecdupssescalarmodesuffix): New mode attribute.
5017 (vec_dup<mode>): Use it.
5019 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
5022 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
5023 frame if argument is passed on stack.
5025 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5028 * ree.c (combine_reaching_defs): Optimize also
5029 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
5030 reg2=any_extend(exp); reg1=reg2;, formatting fix.
5032 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5035 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
5037 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5040 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
5041 after they are computed.
5043 2018-01-11 Bin Cheng <bin.cheng@arm.com>
5045 PR tree-optimization/83695
5046 * gimple-loop-linterchange.cc
5047 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
5048 reset cached scev information after interchange.
5049 (pass_linterchange::execute): Remove call to scev_reset_htab.
5051 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5053 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
5054 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
5055 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
5056 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
5057 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
5058 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
5059 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
5060 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
5061 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
5062 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
5063 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
5064 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
5065 (V_lane_reg): Likewise.
5066 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
5068 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
5069 (vfmal_lane_low<mode>_intrinsic,
5070 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
5071 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
5072 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
5073 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
5074 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
5075 vfmsl_lane_high<mode>_intrinsic): New define_insns.
5077 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5079 * config/arm/arm-cpus.in (fp16fml): New feature.
5080 (ALL_SIMD): Add fp16fml.
5081 (armv8.2-a): Add fp16fml as an option.
5082 (armv8.3-a): Likewise.
5083 (armv8.4-a): Add fp16fml as part of fp16.
5084 * config/arm/arm.h (TARGET_FP16FML): Define.
5085 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
5087 * config/arm/arm-modes.def (V2HF): Define.
5088 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
5089 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
5090 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
5091 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
5092 vfmsl_low, vfmsl_high): New set of builtins.
5093 * config/arm/iterators.md (PLUSMINUS): New code iterator.
5094 (vfml_op): New code attribute.
5095 (VFMLHALVES): New int iterator.
5096 (VFML, VFMLSEL): New mode attributes.
5097 (V_reg): Define mapping for V2HF.
5098 (V_hi, V_lo): New mode attributes.
5099 (VF_constraint): Likewise.
5100 (vfml_half, vfml_half_selector): New int attributes.
5101 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
5103 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
5104 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
5106 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
5107 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
5108 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
5109 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
5111 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
5112 Document new effective target and option set.
5114 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5116 * config/arm/arm-cpus.in (armv8_4): New feature.
5117 (ARMv8_4a): New fgroup.
5118 (armv8.4-a): New arch.
5119 * config/arm/arm-tables.opt: Regenerate.
5120 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
5121 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
5122 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
5123 Add matching rules for -march=armv8.4-a and extensions.
5124 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
5126 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
5129 * config/rx/rx.md (BW): New mode attribute.
5130 (sync_lock_test_and_setsi): Add mode suffix to insn output.
5132 2018-01-11 Richard Biener <rguenther@suse.de>
5134 PR tree-optimization/83435
5135 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
5136 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
5137 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
5139 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5140 Alan Hayward <alan.hayward@arm.com>
5141 David Sherwood <david.sherwood@arm.com>
5143 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
5145 (aarch64_classify_address): Initialize it. Track polynomial offsets.
5146 (aarch64_print_address_internal): Use it to check for a zero offset.
5148 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5149 Alan Hayward <alan.hayward@arm.com>
5150 David Sherwood <david.sherwood@arm.com>
5152 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
5153 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
5154 Return a poly_int64 rather than a HOST_WIDE_INT.
5155 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
5156 rather than a HOST_WIDE_INT.
5157 * config/aarch64/aarch64.h (aarch64_frame): Protect with
5158 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
5159 hard_fp_offset, frame_size, initial_adjust, callee_offset and
5160 final_offset from HOST_WIDE_INT to poly_int64.
5161 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
5162 to_constant when getting the number of units in an Advanced SIMD
5164 (aarch64_builtin_vectorized_function): Check for a constant number
5166 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
5168 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
5169 attribute instead of GET_MODE_NUNITS.
5170 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
5171 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
5172 GET_MODE_SIZE for fixed-size registers.
5173 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
5174 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
5175 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
5176 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
5177 (aarch64_print_operand, aarch64_print_address_internal)
5178 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
5179 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
5180 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
5181 Handle polynomial GET_MODE_SIZE.
5182 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
5183 wider than SImode without modification.
5184 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
5185 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
5186 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
5187 passing and returning SVE modes.
5188 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
5189 rather than GEN_INT.
5190 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
5191 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
5192 (aarch64_allocate_and_probe_stack_space): Likewise.
5193 (aarch64_layout_frame): Cope with polynomial offsets.
5194 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
5195 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
5197 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
5198 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
5199 poly_int64 rather than a HOST_WIDE_INT.
5200 (aarch64_get_separate_components, aarch64_process_components)
5201 (aarch64_expand_prologue, aarch64_expand_epilogue)
5202 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
5203 (aarch64_anchor_offset): New function, split out from...
5204 (aarch64_legitimize_address): ...here.
5205 (aarch64_builtin_vectorization_cost): Handle polynomial
5206 TYPE_VECTOR_SUBPARTS.
5207 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
5209 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
5210 number of elements from the PARALLEL rather than the mode.
5211 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
5212 rather than GET_MODE_BITSIZE.
5213 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
5214 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
5215 (aarch64_expand_vec_perm_const_1): Handle polynomial
5216 d->perm.length () and d->perm elements.
5217 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
5218 Apply to_constant to d->perm elements.
5219 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
5220 polynomial CONST_VECTOR_NUNITS.
5221 (aarch64_move_pointer): Take amount as a poly_int64 rather
5223 (aarch64_progress_pointer): Avoid temporary variable.
5224 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
5225 the mode attribute instead of GET_MODE.
5227 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5228 Alan Hayward <alan.hayward@arm.com>
5229 David Sherwood <david.sherwood@arm.com>
5231 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
5232 x exists before using it.
5233 (aarch64_add_constant_internal): Rename to...
5234 (aarch64_add_offset_1): ...this. Replace regnum with separate
5235 src and dest rtxes. Handle the case in which they're different,
5236 including when the offset is zero. Replace scratchreg with an rtx.
5237 Use 2 additions if there is no spare register into which we can
5238 move a 16-bit constant.
5239 (aarch64_add_constant): Delete.
5240 (aarch64_add_offset): Replace reg with separate src and dest
5241 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
5242 Use aarch64_add_offset_1.
5243 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
5244 an rtx rather than an int. Take the delta as a poly_int64
5245 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
5246 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
5247 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
5248 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
5249 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
5251 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
5252 aarch64_add_constant.
5254 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5256 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
5257 Use scalar_float_mode.
5259 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5261 * config/aarch64/aarch64-simd.md
5262 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
5263 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
5264 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
5265 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
5266 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
5267 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
5268 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
5269 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
5270 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
5271 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
5273 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5276 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
5277 targ_options->x_arm_arch_string is non NULL.
5279 2018-01-11 Tamar Christina <tamar.christina@arm.com>
5281 * config/aarch64/aarch64.h
5282 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
5284 2018-01-11 Sudakshina Das <sudi.das@arm.com>
5287 * expmed.c (emit_store_flag_force): Swap if const op0
5288 and change VOIDmode to mode of op0.
5290 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5292 PR rtl-optimization/83761
5293 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
5294 than bytes to mode_for_size.
5296 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5299 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
5300 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
5303 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5306 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
5307 when in layout mode.
5308 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
5309 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
5312 2018-01-10 Michael Collison <michael.collison@arm.com>
5314 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
5315 * config/aarch64/aarch64-option-extension.def: Add
5316 AARCH64_OPT_EXTENSION of 'fp16fml'.
5317 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5318 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
5319 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
5320 * config/aarch64/constraints.md (Ui7): New constraint.
5321 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
5322 (VFMLA_SEL_W): Ditto.
5325 (VFMLA16_LOW): New int iterator.
5326 (VFMLA16_HIGH): Ditto.
5327 (UNSPEC_FMLAL): New unspec.
5328 (UNSPEC_FMLSL): Ditto.
5329 (UNSPEC_FMLAL2): Ditto.
5330 (UNSPEC_FMLSL2): Ditto.
5331 (f16mac): New code attribute.
5332 * config/aarch64/aarch64-simd-builtins.def
5333 (aarch64_fmlal_lowv2sf): Ditto.
5334 (aarch64_fmlsl_lowv2sf): Ditto.
5335 (aarch64_fmlalq_lowv4sf): Ditto.
5336 (aarch64_fmlslq_lowv4sf): Ditto.
5337 (aarch64_fmlal_highv2sf): Ditto.
5338 (aarch64_fmlsl_highv2sf): Ditto.
5339 (aarch64_fmlalq_highv4sf): Ditto.
5340 (aarch64_fmlslq_highv4sf): Ditto.
5341 (aarch64_fmlal_lane_lowv2sf): Ditto.
5342 (aarch64_fmlsl_lane_lowv2sf): Ditto.
5343 (aarch64_fmlal_laneq_lowv2sf): Ditto.
5344 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
5345 (aarch64_fmlalq_lane_lowv4sf): Ditto.
5346 (aarch64_fmlsl_lane_lowv4sf): Ditto.
5347 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
5348 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
5349 (aarch64_fmlal_lane_highv2sf): Ditto.
5350 (aarch64_fmlsl_lane_highv2sf): Ditto.
5351 (aarch64_fmlal_laneq_highv2sf): Ditto.
5352 (aarch64_fmlsl_laneq_highv2sf): Ditto.
5353 (aarch64_fmlalq_lane_highv4sf): Ditto.
5354 (aarch64_fmlsl_lane_highv4sf): Ditto.
5355 (aarch64_fmlalq_laneq_highv4sf): Ditto.
5356 (aarch64_fmlsl_laneq_highv4sf): Ditto.
5357 * config/aarch64/aarch64-simd.md:
5358 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
5359 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5360 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
5361 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5362 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
5363 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
5364 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
5365 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
5366 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
5367 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
5368 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
5369 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
5370 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
5371 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
5372 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
5373 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
5374 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
5375 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
5376 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
5377 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
5378 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
5379 (vfmlsl_low_u32): Ditto.
5380 (vfmlalq_low_u32): Ditto.
5381 (vfmlslq_low_u32): Ditto.
5382 (vfmlal_high_u32): Ditto.
5383 (vfmlsl_high_u32): Ditto.
5384 (vfmlalq_high_u32): Ditto.
5385 (vfmlslq_high_u32): Ditto.
5386 (vfmlal_lane_low_u32): Ditto.
5387 (vfmlsl_lane_low_u32): Ditto.
5388 (vfmlal_laneq_low_u32): Ditto.
5389 (vfmlsl_laneq_low_u32): Ditto.
5390 (vfmlalq_lane_low_u32): Ditto.
5391 (vfmlslq_lane_low_u32): Ditto.
5392 (vfmlalq_laneq_low_u32): Ditto.
5393 (vfmlslq_laneq_low_u32): Ditto.
5394 (vfmlal_lane_high_u32): Ditto.
5395 (vfmlsl_lane_high_u32): Ditto.
5396 (vfmlal_laneq_high_u32): Ditto.
5397 (vfmlsl_laneq_high_u32): Ditto.
5398 (vfmlalq_lane_high_u32): Ditto.
5399 (vfmlslq_lane_high_u32): Ditto.
5400 (vfmlalq_laneq_high_u32): Ditto.
5401 (vfmlslq_laneq_high_u32): Ditto.
5402 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
5403 (AARCH64_FL_FOR_ARCH8_4): New.
5404 (AARCH64_ISA_F16FML): New ISA flag.
5405 (TARGET_F16FML): New feature flag for fp16fml.
5406 (doc/invoke.texi): Document new fp16fml option.
5408 2018-01-10 Michael Collison <michael.collison@arm.com>
5410 * config/aarch64/aarch64-builtins.c:
5411 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
5412 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5413 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
5414 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
5415 (AARCH64_ISA_SHA3): New ISA flag.
5416 (TARGET_SHA3): New feature flag for sha3.
5417 * config/aarch64/iterators.md (sha512_op): New int attribute.
5418 (CRYPTO_SHA512): New int iterator.
5419 (UNSPEC_SHA512H): New unspec.
5420 (UNSPEC_SHA512H2): Ditto.
5421 (UNSPEC_SHA512SU0): Ditto.
5422 (UNSPEC_SHA512SU1): Ditto.
5423 * config/aarch64/aarch64-simd-builtins.def
5424 (aarch64_crypto_sha512hqv2di): New builtin.
5425 (aarch64_crypto_sha512h2qv2di): Ditto.
5426 (aarch64_crypto_sha512su0qv2di): Ditto.
5427 (aarch64_crypto_sha512su1qv2di): Ditto.
5428 (aarch64_eor3qv8hi): Ditto.
5429 (aarch64_rax1qv2di): Ditto.
5430 (aarch64_xarqv2di): Ditto.
5431 (aarch64_bcaxqv8hi): Ditto.
5432 * config/aarch64/aarch64-simd.md:
5433 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
5434 (aarch64_crypto_sha512su0qv2di): Ditto.
5435 (aarch64_crypto_sha512su1qv2di): Ditto.
5436 (aarch64_eor3qv8hi): Ditto.
5437 (aarch64_rax1qv2di): Ditto.
5438 (aarch64_xarqv2di): Ditto.
5439 (aarch64_bcaxqv8hi): Ditto.
5440 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
5441 (vsha512h2q_u64): Ditto.
5442 (vsha512su0q_u64): Ditto.
5443 (vsha512su1q_u64): Ditto.
5444 (veor3q_u16): Ditto.
5445 (vrax1q_u64): Ditto.
5447 (vbcaxq_u16): Ditto.
5448 * config/arm/types.md (crypto_sha512): New type attribute.
5449 (crypto_sha3): Ditto.
5450 (doc/invoke.texi): Document new sha3 option.
5452 2018-01-10 Michael Collison <michael.collison@arm.com>
5454 * config/aarch64/aarch64-builtins.c:
5455 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
5456 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5457 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
5458 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
5459 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
5460 (AARCH64_ISA_SM4): New ISA flag.
5461 (TARGET_SM4): New feature flag for sm4.
5462 * config/aarch64/aarch64-simd-builtins.def
5463 (aarch64_sm3ss1qv4si): Ditto.
5464 (aarch64_sm3tt1aq4si): Ditto.
5465 (aarch64_sm3tt1bq4si): Ditto.
5466 (aarch64_sm3tt2aq4si): Ditto.
5467 (aarch64_sm3tt2bq4si): Ditto.
5468 (aarch64_sm3partw1qv4si): Ditto.
5469 (aarch64_sm3partw2qv4si): Ditto.
5470 (aarch64_sm4eqv4si): Ditto.
5471 (aarch64_sm4ekeyqv4si): Ditto.
5472 * config/aarch64/aarch64-simd.md:
5473 (aarch64_sm3ss1qv4si): Ditto.
5474 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
5475 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
5476 (aarch64_sm4eqv4si): Ditto.
5477 (aarch64_sm4ekeyqv4si): Ditto.
5478 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
5479 (sm3part_op): Ditto.
5480 (CRYPTO_SM3TT): Ditto.
5481 (CRYPTO_SM3PART): Ditto.
5482 (UNSPEC_SM3SS1): New unspec.
5483 (UNSPEC_SM3TT1A): Ditto.
5484 (UNSPEC_SM3TT1B): Ditto.
5485 (UNSPEC_SM3TT2A): Ditto.
5486 (UNSPEC_SM3TT2B): Ditto.
5487 (UNSPEC_SM3PARTW1): Ditto.
5488 (UNSPEC_SM3PARTW2): Ditto.
5489 (UNSPEC_SM4E): Ditto.
5490 (UNSPEC_SM4EKEY): Ditto.
5491 * config/aarch64/constraints.md (Ui2): New constraint.
5492 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
5493 * config/arm/types.md (crypto_sm3): New type attribute.
5494 (crypto_sm4): Ditto.
5495 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
5496 (vsm3tt1aq_u32): Ditto.
5497 (vsm3tt1bq_u32): Ditto.
5498 (vsm3tt2aq_u32): Ditto.
5499 (vsm3tt2bq_u32): Ditto.
5500 (vsm3partw1q_u32): Ditto.
5501 (vsm3partw2q_u32): Ditto.
5502 (vsm4eq_u32): Ditto.
5503 (vsm4ekeyq_u32): Ditto.
5504 (doc/invoke.texi): Document new sm4 option.
5506 2018-01-10 Michael Collison <michael.collison@arm.com>
5508 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
5509 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
5510 (AARCH64_FL_FOR_ARCH8_4): New.
5511 (AARCH64_FL_V8_4): New flag.
5512 (doc/invoke.texi): Document new armv8.4-a option.
5514 2018-01-10 Michael Collison <michael.collison@arm.com>
5516 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5517 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
5518 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
5519 * config/aarch64/aarch64-option-extension.def: Add
5520 AARCH64_OPT_EXTENSION of 'sha2'.
5521 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
5522 (crypto): Disable sha2 and aes if crypto disabled.
5523 (crypto): Enable aes and sha2 if enabled.
5524 (simd): Disable sha2 and aes if simd disabled.
5525 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
5527 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
5528 (TARGET_SHA2): New feature flag for sha2.
5529 (TARGET_AES): New feature flag for aes.
5530 * config/aarch64/aarch64-simd.md:
5531 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
5532 conditional on TARGET_AES.
5533 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
5534 (aarch64_crypto_sha1hsi): Make pattern conditional
5536 (aarch64_crypto_sha1hv4si): Ditto.
5537 (aarch64_be_crypto_sha1hv4si): Ditto.
5538 (aarch64_crypto_sha1su1v4si): Ditto.
5539 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
5540 (aarch64_crypto_sha1su0v4si): Ditto.
5541 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
5542 (aarch64_crypto_sha256su0v4si): Ditto.
5543 (aarch64_crypto_sha256su1v4si): Ditto.
5544 (doc/invoke.texi): Document new aes and sha2 options.
5546 2018-01-10 Martin Sebor <msebor@redhat.com>
5548 PR tree-optimization/83781
5549 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
5552 2018-01-11 Martin Sebor <msebor@gmail.com>
5553 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5555 PR tree-optimization/83501
5556 PR tree-optimization/81703
5558 * tree-ssa-strlen.c (get_string_cst): Rename...
5559 (get_string_len): ...to this. Handle global constants.
5560 (handle_char_store): Adjust.
5562 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
5563 Jim Wilson <jimw@sifive.com>
5565 * config/riscv/riscv-protos.h (riscv_output_return): New.
5566 * config/riscv/riscv.c (struct machine_function): New naked_p field.
5567 (riscv_attribute_table, riscv_output_return),
5568 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
5569 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
5570 (riscv_compute_frame_info): Only compute frame->mask if not a naked
5572 (riscv_expand_prologue): Add early return for naked function.
5573 (riscv_expand_epilogue): Likewise.
5574 (riscv_function_ok_for_sibcall): Return false for naked function.
5575 (riscv_set_current_function): New.
5576 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
5577 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
5578 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
5579 * doc/extend.texi (RISC-V Function Attributes): New.
5581 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
5583 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
5584 check for 128-bit long double before checking TCmode.
5585 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
5586 128-bit long doubles before checking TFmode or TCmode.
5587 (FLOAT128_IBM_P): Likewise.
5589 2018-01-10 Martin Sebor <msebor@redhat.com>
5591 PR tree-optimization/83671
5592 * builtins.c (c_strlen): Unconditionally return zero for the empty
5594 Use -Warray-bounds for warnings.
5595 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
5596 for non-constant array indices with COMPONENT_REF, arrays of
5597 arrays, and pointers to arrays.
5598 (gimple_fold_builtin_strlen): Determine and set length range for
5599 non-constant character arrays.
5601 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
5604 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
5607 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
5609 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
5611 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5614 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
5615 VECTOR_MEM_ALTIVEC_OR_VSX_P.
5616 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
5617 indexed_or_indirect_operand predicate.
5618 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
5619 (*vsx_le_perm_load_v8hi): Likewise.
5620 (*vsx_le_perm_load_v16qi): Likewise.
5621 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
5622 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
5623 (*vsx_le_perm_store_v8hi): Likewise.
5624 (*vsx_le_perm_store_v16qi): Likewise.
5625 (eight unnamed splitters): Likewise.
5627 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5629 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
5630 * config/rs6000/emmintrin.h: Likewise.
5631 * config/rs6000/mmintrin.h: Likewise.
5632 * config/rs6000/xmmintrin.h: Likewise.
5634 2018-01-10 David Malcolm <dmalcolm@redhat.com>
5637 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
5639 * tree.c (tree_nop_conversion): Return true for location wrapper
5641 (maybe_wrap_with_location): New function.
5642 (selftest::check_strip_nops): New function.
5643 (selftest::test_location_wrappers): New function.
5644 (selftest::tree_c_tests): Call it.
5645 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
5646 (maybe_wrap_with_location): New decl.
5647 (EXPR_LOCATION_WRAPPER_P): New macro.
5648 (location_wrapper_p): New inline function.
5649 (tree_strip_any_location_wrapper): New inline function.
5651 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
5654 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
5655 stack_realign_offset for the largest alignment of stack slot
5657 (ix86_find_max_used_stack_alignment): New function.
5658 (ix86_finalize_stack_frame_flags): Use it. Set
5659 max_used_stack_alignment if we don't realign stack.
5660 * config/i386/i386.h (machine_function): Add
5661 max_used_stack_alignment.
5663 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
5665 * config/arm/arm.opt (-mbranch-cost): New option.
5666 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
5669 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
5672 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
5673 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
5675 2018-01-10 Richard Biener <rguenther@suse.de>
5678 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
5679 early out so it also covers the case where we have a non-NULL
5682 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
5684 PR tree-optimization/83753
5685 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
5686 for non-strided grouped accesses if the number of elements is 1.
5688 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5691 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
5692 * i386.h (TARGET_USE_GATHER): Define.
5693 * x86-tune.def (X86_TUNE_USE_GATHER): New.
5695 2018-01-10 Martin Liska <mliska@suse.cz>
5698 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
5699 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
5701 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
5702 CLEANUP_NO_PARTITIONING is not set.
5704 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
5706 * doc/rtl.texi: Remove documentation of (const ...) wrappers
5707 for vectors, as a partial revert of r254296.
5708 * rtl.h (const_vec_p): Delete.
5709 (const_vec_duplicate_p): Don't test for vector CONSTs.
5710 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
5711 * expmed.c (make_tree): Likewise.
5714 * common.md (E, F): Use CONSTANT_P instead of checking for
5716 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
5717 checking for CONST_VECTOR.
5719 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5722 * predict.c (force_edge_cold): Handle in more sane way edges
5725 2018-01-09 Carl Love <cel@us.ibm.com>
5727 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
5729 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
5730 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
5731 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
5732 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
5733 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
5734 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
5735 * config/rs6000/rs6000-protos.h: Add extern defition for
5736 rs6000_generate_float2_double_code.
5737 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
5739 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
5740 (float2_v2df): Add define_expand.
5742 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
5745 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
5746 op_mode in the force_to_mode call.
5748 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5750 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
5751 instead of checking each element individually.
5752 (aarch64_evpc_uzp): Likewise.
5753 (aarch64_evpc_zip): Likewise.
5754 (aarch64_evpc_ext): Likewise.
5755 (aarch64_evpc_rev): Likewise.
5756 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
5757 instead of checking each element individually. Return true without
5759 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
5760 whether all selected elements come from the same input, instead of
5761 checking each element individually. Remove calls to gen_rtx_REG,
5762 start_sequence and end_sequence and instead assert that no rtl is
5765 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5767 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
5768 order of HIGH and CONST checks.
5770 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5772 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
5773 if the destination isn't an SSA_NAME.
5775 2018-01-09 Richard Biener <rguenther@suse.de>
5777 PR tree-optimization/83668
5778 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
5780 (canonicalize_loop_form): ... here, renamed from ...
5781 (canonicalize_loop_closed_ssa_form): ... this and amended to
5782 swap successor edges for loop exit blocks to make us use
5783 the RPO order we need for initial schedule generation.
5785 2018-01-09 Joseph Myers <joseph@codesourcery.com>
5787 PR tree-optimization/64811
5788 * match.pd: When optimizing comparisons with Inf, avoid
5789 introducing or losing exceptions from comparisons with NaN.
5791 2018-01-09 Martin Liska <mliska@suse.cz>
5794 * asan.c (shadow_mem_size): Add gcc_assert.
5796 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
5798 Don't save registers in main().
5801 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
5802 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
5803 * config/avr/avr.c (avr_set_current_function): Don't error if
5804 naked, OS_task or OS_main are specified at the same time.
5805 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
5807 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
5809 * common/config/avr/avr-common.c (avr_option_optimization_table):
5810 Switch on -mmain-is-OS_task for optimizing compilations.
5812 2018-01-09 Richard Biener <rguenther@suse.de>
5814 PR tree-optimization/83572
5815 * graphite.c: Include cfganal.h.
5816 (graphite_transform_loops): Connect infinite loops to exit
5817 and remove fake edges at the end.
5819 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5821 * ipa-inline.c (edge_badness): Revert accidental checkin.
5823 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5826 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
5827 symbols; not inline clones.
5829 2018-01-09 Jakub Jelinek <jakub@redhat.com>
5832 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
5833 hard registers. Formatting fixes.
5835 PR preprocessor/83722
5836 * gcc.c (try_generate_repro): Pass
5837 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
5838 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
5841 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
5842 Kito Cheng <kito.cheng@gmail.com>
5844 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
5845 (riscv_leaf_function_p): Delete.
5846 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
5848 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5850 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
5852 (do_ifelse): New function.
5853 (do_isel): New function.
5854 (do_sub3): New function.
5855 (do_add3): New function.
5856 (do_load_mask_compare): New function.
5857 (do_overlap_load_compare): New function.
5858 (expand_compare_loop): New function.
5859 (expand_block_compare): Call expand_compare_loop() when appropriate.
5860 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
5862 (-mblock-compare-inline-loop-limit): New option.
5864 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5867 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
5868 Reverse order of second and third operands in first alternative.
5869 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
5870 of first and second elements in UNSPEC_VPERMR vector.
5871 (altivec_expand_vec_perm_le): Likewise.
5873 2018-01-08 Jeff Law <law@redhat.com>
5875 PR rtl-optimizatin/81308
5876 * tree-switch-conversion.c (cfg_altered): New file scoped static.
5877 (process_switch): If group_case_labels makes a change, then set
5879 (pass_convert_switch::execute): If a switch is converted, then
5880 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
5882 PR rtl-optimization/81308
5883 * recog.c (split_all_insns): Conditionally cleanup the CFG after
5886 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
5888 PR target/83663 - Revert r255946
5889 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
5890 generation for cases where splatting a value is not useful.
5891 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
5892 across a vec_duplicate and a paradoxical subreg forming a vector
5893 mode to a vec_concat.
5895 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5897 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
5898 -march=armv8.3-a variants.
5899 * config/arm/t-multilib: Likewise.
5900 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
5902 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5904 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
5906 (cceq_ior_compare_complement): Give it a name so I can use it, and
5907 change boolean_or_operator predicate to boolean_operator so it can
5908 be used to generate a crand.
5909 (eqne): New code iterator.
5910 (bd/bd_neg): New code_attrs.
5911 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
5912 a single define_insn.
5913 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
5914 decrement (bdnzt/bdnzf/bdzt/bdzf).
5915 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
5916 with the new names of the branch decrement patterns, and added the
5917 names of the branch decrement conditional patterns.
5919 2018-01-08 Richard Biener <rguenther@suse.de>
5921 PR tree-optimization/83563
5922 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
5925 2018-01-08 Richard Biener <rguenther@suse.de>
5928 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
5930 2018-01-08 Richard Biener <rguenther@suse.de>
5932 PR tree-optimization/83685
5933 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
5934 references to abnormals.
5936 2018-01-08 Richard Biener <rguenther@suse.de>
5939 * dwarf2out.c (output_indirect_strings): Handle empty
5940 skeleton_debug_str_hash.
5941 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
5943 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
5945 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
5946 (emit_store_direct): Likewise.
5947 (arc_trampoline_adjust_address): Likewise.
5948 (arc_asm_trampoline_template): New function.
5949 (arc_initialize_trampoline): Use asm_trampoline_template.
5950 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
5951 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
5952 * config/arc/arc.md (flush_icache): Delete pattern.
5954 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
5956 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
5957 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
5960 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
5963 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
5964 by not USED_FOR_TARGET.
5965 (make_pass_resolve_sw_modes): Likewise.
5967 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
5969 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
5972 2018-01-08 Richard Biener <rguenther@suse.de>
5975 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
5977 2018-01-08 Richard Biener <rguenther@suse.de>
5980 * match.pd ((t * 2) / 2) -> t): Add missing :c.
5982 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
5985 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
5986 basic blocks with a small number of successors.
5987 (convert_control_dep_chain_into_preds): Improve handling of
5989 (dump_predicates): Split apart into...
5990 (dump_pred_chain): ...here...
5991 (dump_pred_info): ...and here.
5992 (can_one_predicate_be_invalidated_p): Add debugging printfs.
5993 (can_chain_union_be_invalidated_p): Improve check for invalidation
5995 (uninit_uses_cannot_happen): Avoid unnecessary if
5996 convert_control_dep_chain_into_preds yielded nothing.
5998 2018-01-06 Martin Sebor <msebor@redhat.com>
6000 PR tree-optimization/83640
6001 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
6002 subtracting negative offset from size.
6003 (builtin_access::overlap): Adjust offset bounds of the access to fall
6004 within the size of the object if possible.
6006 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
6008 PR rtl-optimization/83699
6009 * expmed.c (extract_bit_field_1): Restrict the vector usage of
6010 extract_bit_field_as_subreg to cases in which the extracted
6011 value is also a vector.
6013 * lra-constraints.c (process_alt_operands): Test for the equivalence
6014 substitutions when detecting a possible reload cycle.
6016 2018-01-06 Jakub Jelinek <jakub@redhat.com>
6019 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
6020 by default if flag_selective_schedling{,2}. Formatting fixes.
6022 PR rtl-optimization/83682
6023 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
6024 if it has non-VECTOR_MODE element mode.
6025 (vec_duplicate_p): Likewise.
6028 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
6029 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
6031 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6034 * config/i386/i386-builtin.def
6035 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
6036 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
6037 Require also OPTION_MASK_ISA_AVX512F in addition to
6038 OPTION_MASK_ISA_GFNI.
6039 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
6040 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
6041 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
6042 to OPTION_MASK_ISA_GFNI.
6043 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
6044 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
6045 OPTION_MASK_ISA_AVX512BW.
6046 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
6047 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
6048 addition to OPTION_MASK_ISA_GFNI.
6049 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
6050 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
6051 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
6052 to OPTION_MASK_ISA_GFNI.
6053 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
6054 a requirement for all ISAs rather than any of them with a few
6056 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
6058 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
6059 bitmasks to be enabled with 3 exceptions, instead of requiring any
6060 enabled ISA with lots of exceptions.
6061 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
6062 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
6063 Change avx512bw in isa attribute to avx512f.
6064 * config/i386/sgxintrin.h: Add license boilerplate.
6065 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
6066 to __AVX512F__ and __AVX512VL to __AVX512VL__.
6067 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
6068 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
6070 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
6071 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
6072 temporarily sse2 rather than sse if not enabled already.
6075 * config/i386/sse.md (VI248_VLBW): Rename to ...
6076 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
6077 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
6078 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
6079 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
6080 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
6081 mode iterator instead of VI248_VLBW.
6083 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
6085 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
6086 (record_modified): Skip clobbers; add debug output.
6087 (param_change_prob): Use sreal frequencies.
6089 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6091 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
6092 punt for user-aligned variables.
6094 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6096 * tree-chrec.c (chrec_contains_symbols): Return true for
6099 2018-01-05 Sudakshina Das <sudi.das@arm.com>
6102 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
6103 of (x|y) == x for BICS pattern.
6105 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6107 PR tree-optimization/83605
6108 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
6109 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
6112 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
6114 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
6115 * config/epiphany/rtems.h: New file.
6117 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6118 Uros Bizjak <ubizjak@gmail.com>
6121 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
6122 QIreg_operand instead of register_operand predicate.
6123 * config/i386/i386.c (ix86_rop_should_change_byte_p,
6124 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
6125 comments instead of -fmitigate[-_]rop.
6127 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6130 * cgraphunit.c (symbol_table::compile): Switch to text_section
6131 before calling assembly_start debug hook.
6132 * run-rtl-passes.c (run_rtl_passes): Likewise.
6135 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6137 * tree-vrp.c (extract_range_from_binary_expr_1): Check
6138 range_int_cst_p rather than !symbolic_range_p before calling
6139 extract_range_from_multiplicative_op_1.
6141 2018-01-04 Jeff Law <law@redhat.com>
6143 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
6144 redundant test in assertion.
6146 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6148 * doc/rtl.texi: Document machine_mode wrapper classes.
6150 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6152 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
6155 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6157 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
6158 the VEC_PERM_EXPR fold to fail.
6160 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6163 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
6164 to switched_sections.
6166 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6169 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
6172 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
6175 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
6176 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
6178 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6181 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
6182 is BLKmode and bitpos not zero or mode change is needed.
6184 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6187 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
6190 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
6193 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
6194 instead of MULT rtx. Update all corresponding splitters.
6196 (*ssub<modesuffix>): Ditto.
6198 (*cmp_sadd_di): Update split patterns.
6199 (*cmp_sadd_si): Ditto.
6200 (*cmp_sadd_sidi): Ditto.
6201 (*cmp_ssub_di): Ditto.
6202 (*cmp_ssub_si): Ditto.
6203 (*cmp_ssub_sidi): Ditto.
6204 * config/alpha/predicates.md (const23_operand): New predicate.
6205 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
6206 Look for ASHIFT, not MULT inner operand.
6207 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
6209 2018-01-04 Martin Liska <mliska@suse.cz>
6211 PR gcov-profile/83669
6212 * gcov.c (output_intermediate_file): Add version to intermediate
6214 * doc/gcov.texi: Document new field 'version' in intermediate
6215 file format. Fix location of '-k' option of gcov command.
6217 2018-01-04 Martin Liska <mliska@suse.cz>
6220 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
6222 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6224 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
6226 2018-01-03 Martin Sebor <msebor@redhat.com>
6228 PR tree-optimization/83655
6229 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
6230 checking calls with invalid arguments.
6232 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6234 * tree-vect-stmts.c (vect_get_store_rhs): New function.
6235 (vectorizable_mask_load_store): Delete.
6236 (vectorizable_call): Return false for masked loads and stores.
6237 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
6238 instead of gimple_assign_rhs1.
6239 (vectorizable_load): Handle IFN_MASK_LOAD.
6240 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
6242 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6244 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
6246 (vectorizable_mask_load_store): ...here.
6247 (vectorizable_load): ...and here.
6249 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6251 * tree-vect-stmts.c (vect_build_all_ones_mask)
6252 (vect_build_zero_merge_argument): New functions, split out from...
6253 (vectorizable_load): ...here.
6255 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6257 * tree-vect-stmts.c (vect_check_store_rhs): New function,
6259 (vectorizable_mask_load_store): ...here.
6260 (vectorizable_store): ...and here.
6262 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6264 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
6266 (vectorizable_mask_load_store): ...here.
6268 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6270 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
6271 (vect_model_store_cost): Take a vec_load_store_type instead of a
6273 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
6274 (vect_model_store_cost): Take a vec_load_store_type instead of a
6276 (vectorizable_mask_load_store): Update accordingly.
6277 (vectorizable_store): Likewise.
6278 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
6280 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6282 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
6283 IFN_MASK_LOAD calls here rather than...
6284 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
6286 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6287 Alan Hayward <alan.hayward@arm.com>
6288 David Sherwood <david.sherwood@arm.com>
6290 * expmed.c (extract_bit_field_1): For vector extracts,
6291 fall back to extract_bit_field_as_subreg if vec_extract
6294 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6295 Alan Hayward <alan.hayward@arm.com>
6296 David Sherwood <david.sherwood@arm.com>
6298 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
6299 they are variable or constant sized.
6300 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
6301 slots for constant-sized data.
6303 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6304 Alan Hayward <alan.hayward@arm.com>
6305 David Sherwood <david.sherwood@arm.com>
6307 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
6308 handling COND_EXPRs with boolean comparisons, try to find a better
6309 basis for the mask type than the boolean itself.
6311 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6313 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
6314 is calculated and how it can be overridden.
6315 * genmodes.c (max_bitsize_mode_any_mode): New variable.
6316 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
6318 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
6321 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6322 Alan Hayward <alan.hayward@arm.com>
6323 David Sherwood <david.sherwood@arm.com>
6325 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
6326 Remove the mode argument.
6327 (aarch64_simd_valid_immediate): Remove the mode and inverse
6329 * config/aarch64/iterators.md (bitsize): New iterator.
6330 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
6331 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
6332 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
6333 aarch64_simd_valid_immediate.
6334 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
6335 (aarch64_reg_or_bic_imm): Likewise.
6336 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
6337 with an insn_type enum and msl with a modifier_type enum.
6338 Replace element_width with a scalar_mode. Change the shift
6339 to unsigned int. Add constructors for scalar_float_mode and
6340 scalar_int_mode elements.
6341 (aarch64_vect_float_const_representable_p): Delete.
6342 (aarch64_can_const_movi_rtx_p)
6343 (aarch64_simd_scalar_immediate_valid_for_move)
6344 (aarch64_simd_make_constant): Update call to
6345 aarch64_simd_valid_immediate.
6346 (aarch64_advsimd_valid_immediate_hs): New function.
6347 (aarch64_advsimd_valid_immediate): Likewise.
6348 (aarch64_simd_valid_immediate): Remove mode and inverse
6349 arguments. Rewrite to use the above. Use const_vec_duplicate_p
6350 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
6351 and aarch64_float_const_representable_p on the result.
6352 (aarch64_output_simd_mov_immediate): Remove mode argument.
6353 Update call to aarch64_simd_valid_immediate and use of
6354 simd_immediate_info.
6355 (aarch64_output_scalar_simd_mov_immediate): Update call
6358 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6359 Alan Hayward <alan.hayward@arm.com>
6360 David Sherwood <david.sherwood@arm.com>
6362 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
6363 (mode_nunits): Likewise CONST_MODE_NUNITS.
6364 * machmode.def (ADJUST_NUNITS): Document.
6365 * genmodes.c (mode_data::need_nunits_adj): New field.
6366 (blank_mode): Update accordingly.
6367 (adj_nunits): New variable.
6368 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
6370 (emit_mode_size_inline): Set need_bytesize_adj for all modes
6371 listed in adj_nunits.
6372 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
6373 listed in adj_nunits. Don't emit case statements for such modes.
6374 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
6375 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
6376 nothing if adj_nunits is nonnull.
6377 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
6378 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
6379 (emit_mode_fbit): Update use of print_maybe_const_decl.
6380 (emit_move_size): Likewise. Treat the array as non-const
6382 (emit_mode_adjustments): Handle adj_nunits.
6384 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6386 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
6387 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
6388 (VECTOR_MODES): Use it.
6389 (make_vector_modes): Take the prefix as an argument.
6391 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6392 Alan Hayward <alan.hayward@arm.com>
6393 David Sherwood <david.sherwood@arm.com>
6395 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
6396 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
6397 for MODE_VECTOR_BOOL.
6398 * machmode.def (VECTOR_BOOL_MODE): Document.
6399 * genmodes.c (VECTOR_BOOL_MODE): New macro.
6400 (make_vector_bool_mode): New function.
6401 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
6403 * lto-streamer-in.c (lto_input_mode_table): Likewise.
6404 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
6406 * stor-layout.c (int_mode_for_mode): Likewise.
6407 * tree.c (build_vector_type_for_mode): Likewise.
6408 * varasm.c (output_constant_pool_2): Likewise.
6409 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
6410 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
6411 for MODE_VECTOR_BOOL.
6412 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
6413 of mode class checks.
6414 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
6415 instead of a list of mode class checks.
6416 (expand_vector_scalar_condition): Likewise.
6417 (type_for_widest_vector_mode): Handle BImode as an inner mode.
6419 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6420 Alan Hayward <alan.hayward@arm.com>
6421 David Sherwood <david.sherwood@arm.com>
6423 * machmode.h (mode_size): Change from unsigned short to
6425 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
6426 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6427 or if measurement_type is not polynomial.
6428 (fixed_size_mode::includes_p): Check for constant-sized modes.
6429 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
6430 return a poly_uint16 rather than an unsigned short.
6431 (emit_mode_size): Change the type of mode_size from unsigned short
6432 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
6433 (emit_mode_adjustments): Cope with polynomial vector sizes.
6434 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6436 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6438 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
6439 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
6440 * caller-save.c (setup_save_areas): Likewise.
6441 (replace_reg_with_saved_mem): Likewise.
6442 * calls.c (emit_library_call_value_1): Likewise.
6443 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
6444 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
6445 (gen_lowpart_for_combine): Likewise.
6446 * convert.c (convert_to_integer_1): Likewise.
6447 * cse.c (equiv_constant, cse_insn): Likewise.
6448 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
6449 (cselib_subst_to_values): Likewise.
6450 * dce.c (word_dce_process_block): Likewise.
6451 * df-problems.c (df_word_lr_mark_ref): Likewise.
6452 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
6453 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
6454 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
6455 (rtl_for_decl_location): Likewise.
6456 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
6457 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
6458 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
6459 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
6460 (expand_expr_real_1): Likewise.
6461 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
6462 (pad_below): Likewise.
6463 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6464 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
6465 * ira.c (get_subreg_tracking_sizes): Likewise.
6466 * ira-build.c (ira_create_allocno_objects): Likewise.
6467 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
6468 (ira_sort_regnos_for_alter_reg): Likewise.
6469 * ira-costs.c (record_operand_costs): Likewise.
6470 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
6471 (resolve_simple_move): Likewise.
6472 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
6473 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
6474 (lra_constraints): Likewise.
6475 (CONST_POOL_OK_P): Reject variable-sized modes.
6476 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
6477 (add_pseudo_to_slot, lra_spill): Likewise.
6478 * omp-low.c (omp_clause_aligned_alignment): Likewise.
6479 * optabs-query.c (get_best_extraction_insn): Likewise.
6480 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6481 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
6482 (expand_mult_highpart, valid_multiword_target_p): Likewise.
6483 * recog.c (offsettable_address_addr_space_p): Likewise.
6484 * regcprop.c (maybe_mode_change): Likewise.
6485 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
6486 * regrename.c (build_def_use): Likewise.
6487 * regstat.c (dump_reg_info): Likewise.
6488 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
6489 (find_reloads, find_reloads_subreg_address): Likewise.
6490 * reload1.c (eliminate_regs_1): Likewise.
6491 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
6492 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
6493 (simplify_binary_operation_1, simplify_subreg): Likewise.
6494 * targhooks.c (default_function_arg_padding): Likewise.
6495 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
6496 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
6497 (verify_gimple_assign_ternary): Likewise.
6498 * tree-inline.c (estimate_move_cost): Likewise.
6499 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6500 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
6501 (get_address_cost_ainc): Likewise.
6502 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
6503 (vect_supportable_dr_alignment): Likewise.
6504 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6505 (vectorizable_reduction): Likewise.
6506 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
6507 (vectorizable_operation, vectorizable_load): Likewise.
6508 * tree.c (build_same_sized_truth_vector_type): Likewise.
6509 * valtrack.c (cleanup_auto_inc_dec): Likewise.
6510 * var-tracking.c (emit_note_insn_var_location): Likewise.
6511 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
6512 (ADDR_VEC_ALIGN): Likewise.
6514 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6515 Alan Hayward <alan.hayward@arm.com>
6516 David Sherwood <david.sherwood@arm.com>
6518 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
6520 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6521 or if measurement_type is polynomial.
6522 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
6523 * combine.c (make_extraction): Likewise.
6524 * dse.c (find_shift_sequence): Likewise.
6525 * dwarf2out.c (mem_loc_descriptor): Likewise.
6526 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
6527 (extract_bit_field, extract_low_bits): Likewise.
6528 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
6529 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
6530 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
6531 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
6532 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6533 * reload.c (find_reloads): Likewise.
6534 * reload1.c (alter_reg): Likewise.
6535 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
6536 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
6537 * tree-if-conv.c (predicate_mem_writes): Likewise.
6538 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
6539 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
6540 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
6541 * valtrack.c (dead_debug_insert_temp): Likewise.
6542 * varasm.c (mergeable_constant_section): Likewise.
6543 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
6545 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6546 Alan Hayward <alan.hayward@arm.com>
6547 David Sherwood <david.sherwood@arm.com>
6549 * expr.c (expand_assignment): Cope with polynomial mode sizes
6550 when assigning to a CONCAT.
6552 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6553 Alan Hayward <alan.hayward@arm.com>
6554 David Sherwood <david.sherwood@arm.com>
6556 * machmode.h (mode_precision): Change from unsigned short to
6558 (mode_to_precision): Return a poly_uint16 rather than an unsigned
6560 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
6561 or if measurement_type is not polynomial.
6562 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
6563 in which the mode is already known to be a scalar_int_mode.
6564 * genmodes.c (emit_mode_precision): Change the type of mode_precision
6565 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
6567 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6568 for GET_MODE_PRECISION.
6569 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6570 for GET_MODE_PRECISION.
6571 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
6573 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
6574 (expand_field_assignment, make_extraction): Likewise.
6575 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
6576 (get_last_value): Likewise.
6577 * convert.c (convert_to_integer_1): Likewise.
6578 * cse.c (cse_insn): Likewise.
6579 * expr.c (expand_expr_real_1): Likewise.
6580 * lra-constraints.c (simplify_operand_subreg): Likewise.
6581 * optabs-query.c (can_atomic_load_p): Likewise.
6582 * optabs.c (expand_atomic_load): Likewise.
6583 (expand_atomic_store): Likewise.
6584 * ree.c (combine_reaching_defs): Likewise.
6585 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
6586 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
6587 * tree.h (type_has_mode_precision_p): Likewise.
6588 * ubsan.c (instrument_si_overflow): Likewise.
6590 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6591 Alan Hayward <alan.hayward@arm.com>
6592 David Sherwood <david.sherwood@arm.com>
6594 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
6595 polynomial numbers of units.
6596 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
6597 (valid_vector_subparts_p): New function.
6598 (build_vector_type): Remove temporary shim and take the number
6599 of units as a poly_uint64 rather than an int.
6600 (build_opaque_vector_type): Take the number of units as a
6601 poly_uint64 rather than an int.
6602 * tree.c (build_vector_from_ctor): Handle polynomial
6603 TYPE_VECTOR_SUBPARTS.
6604 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
6605 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
6606 (build_vector_from_val): If the number of units is variable,
6607 use build_vec_duplicate_cst for constant operands and
6608 VEC_DUPLICATE_EXPR otherwise.
6609 (make_vector_type): Remove temporary is_constant ().
6610 (build_vector_type, build_opaque_vector_type): Take the number of
6611 units as a poly_uint64 rather than an int.
6612 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
6614 * cfgexpand.c (expand_debug_expr): Likewise.
6615 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
6616 (store_constructor, expand_expr_real_1): Likewise.
6617 (const_scalar_mask_from_tree): Likewise.
6618 * fold-const-call.c (fold_const_reduction): Likewise.
6619 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
6620 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
6621 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
6622 (fold_relational_const): Likewise.
6623 (native_interpret_vector): Likewise. Change the size from an
6624 int to an unsigned int.
6625 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
6626 TYPE_VECTOR_SUBPARTS.
6627 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
6628 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
6629 duplicating a non-constant operand into a variable-length vector.
6630 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
6631 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
6632 * ipa-icf.c (sem_variable::equals): Likewise.
6633 * match.pd: Likewise.
6634 * omp-simd-clone.c (simd_clone_subparts): Likewise.
6635 * print-tree.c (print_node): Likewise.
6636 * stor-layout.c (layout_type): Likewise.
6637 * targhooks.c (default_builtin_vectorization_cost): Likewise.
6638 * tree-cfg.c (verify_gimple_comparison): Likewise.
6639 (verify_gimple_assign_binary): Likewise.
6640 (verify_gimple_assign_ternary): Likewise.
6641 (verify_gimple_assign_single): Likewise.
6642 * tree-pretty-print.c (dump_generic_node): Likewise.
6643 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6644 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
6645 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
6646 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
6647 (vect_shift_permute_load_chain): Likewise.
6648 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
6649 (expand_vector_condition, optimize_vector_constructor): Likewise.
6650 (lower_vec_perm, get_compute_type): Likewise.
6651 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6652 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
6653 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
6654 (vect_recog_mask_conversion_pattern): Likewise.
6655 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
6656 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
6657 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
6658 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
6659 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
6660 (vectorizable_shift, vectorizable_operation, vectorizable_store)
6661 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
6662 (supportable_widening_operation): Likewise.
6663 (supportable_narrowing_operation): Likewise.
6664 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
6666 * varasm.c (output_constant): Likewise.
6668 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6669 Alan Hayward <alan.hayward@arm.com>
6670 David Sherwood <david.sherwood@arm.com>
6672 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
6673 so that both the length == 3 and length != 3 cases set up their
6674 own permute vectors. Add comments explaining why we know the
6675 number of elements is constant.
6676 (vect_permute_load_chain): Likewise.
6678 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6679 Alan Hayward <alan.hayward@arm.com>
6680 David Sherwood <david.sherwood@arm.com>
6682 * machmode.h (mode_nunits): Change from unsigned char to
6684 (ONLY_FIXED_SIZE_MODES): New macro.
6685 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
6686 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
6687 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
6689 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
6690 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
6691 or if measurement_type is not polynomial.
6692 * genmodes.c (ZERO_COEFFS): New macro.
6693 (emit_mode_nunits_inline): Make mode_nunits_inline return a
6695 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
6696 Use ZERO_COEFFS when emitting initializers.
6697 * data-streamer.h (bp_pack_poly_value): New function.
6698 (bp_unpack_poly_value): Likewise.
6699 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6700 for GET_MODE_NUNITS.
6701 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6702 for GET_MODE_NUNITS.
6703 * tree.c (make_vector_type): Remove temporary shim and make
6704 the real function take the number of units as a poly_uint64
6706 (build_vector_type_for_mode): Handle polynomial nunits.
6707 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
6708 * emit-rtl.c (const_vec_series_p_1): Likewise.
6709 (gen_rtx_CONST_VECTOR): Likewise.
6710 * fold-const.c (test_vec_duplicate_folding): Likewise.
6711 * genrecog.c (validate_pattern): Likewise.
6712 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
6713 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6714 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
6715 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
6716 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
6717 * rtlanal.c (subreg_get_info): Likewise.
6718 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
6719 (vect_grouped_load_supported): Likewise.
6720 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
6721 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
6722 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
6723 (simplify_const_unary_operation, simplify_binary_operation_1)
6724 (simplify_const_binary_operation, simplify_ternary_operation)
6725 (test_vector_ops_duplicate, test_vector_ops): Likewise.
6726 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
6727 instead of CONST_VECTOR_NUNITS.
6728 * varasm.c (output_constant_pool_2): Likewise.
6729 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
6730 explicit-encoded elements in the XVEC for variable-length vectors.
6732 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6734 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
6736 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6737 Alan Hayward <alan.hayward@arm.com>
6738 David Sherwood <david.sherwood@arm.com>
6740 * coretypes.h (fixed_size_mode): Declare.
6741 (fixed_size_mode_pod): New typedef.
6742 * builtins.h (target_builtins::x_apply_args_mode)
6743 (target_builtins::x_apply_result_mode): Change type to
6744 fixed_size_mode_pod.
6745 * builtins.c (apply_args_size, apply_result_size, result_vector)
6746 (expand_builtin_apply_args_1, expand_builtin_apply)
6747 (expand_builtin_return): Update accordingly.
6749 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6751 * cse.c (hash_rtx_cb): Hash only the encoded elements.
6752 * cselib.c (cselib_hash_rtx): Likewise.
6753 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
6754 CONST_VECTOR encoding.
6756 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6757 Jeff Law <law@redhat.com>
6760 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
6761 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
6762 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
6763 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
6766 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
6767 explicitly probe *sp in a noreturn function if there were any callee
6768 register saves or frame pointer is needed.
6770 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6773 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
6774 BLKmode for ternary, binary or unary expressions.
6777 * var-tracking.c (delete_vta_debug_insn): New inline function.
6778 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
6779 insns from get_insns () to NULL instead of each bb separately.
6780 Use delete_vta_debug_insn. No longer static.
6781 (vt_debug_insns_local, variable_tracking_main_1): Adjust
6782 delete_vta_debug_insns callers.
6783 * rtl.h (delete_vta_debug_insns): Declare.
6784 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
6785 instead of variable_tracking_main.
6787 2018-01-03 Martin Sebor <msebor@redhat.com>
6789 PR tree-optimization/83603
6790 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
6791 arguments past the endof the argument list in functions declared
6792 without a prototype.
6793 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
6794 Avoid checking when arguments are null.
6796 2018-01-03 Martin Sebor <msebor@redhat.com>
6799 * doc/extend.texi (attribute const): Fix a typo.
6800 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
6801 issuing -Wsuggest-attribute for void functions.
6803 2018-01-03 Martin Sebor <msebor@redhat.com>
6805 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
6806 offset_int::from instead of wide_int::to_shwi.
6807 (maybe_diag_overlap): Remove assertion.
6808 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
6809 * gimple-ssa-sprintf.c (format_directive): Same.
6810 (parse_directive): Same.
6811 (sprintf_dom_walker::compute_format_length): Same.
6812 (try_substitute_return_value): Same.
6814 2018-01-03 Jeff Law <law@redhat.com>
6817 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
6818 non-constant residual for zero at runtime and avoid probing in
6819 that case. Reorganize code for trailing problem to mirror handling
6822 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6824 PR tree-optimization/83501
6825 * tree-ssa-strlen.c (get_string_cst): New.
6826 (handle_char_store): Call get_string_cst.
6828 2018-01-03 Martin Liska <mliska@suse.cz>
6830 PR tree-optimization/83593
6831 * tree-ssa-strlen.c: Include tree-cfg.h.
6832 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
6833 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
6834 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
6836 (strlen_dom_walker::before_dom_children): Call
6837 gimple_purge_dead_eh_edges. Dump tranformation with details
6839 (strlen_dom_walker::before_dom_children): Update call by adding
6840 new argument cleanup_eh.
6841 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
6843 2018-01-03 Martin Liska <mliska@suse.cz>
6846 * cif-code.def (VARIADIC_THUNK): New enum value.
6847 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
6850 2018-01-03 Jan Beulich <jbeulich@suse.com>
6852 * sse.md (mov<mode>_internal): Tighten condition for when to use
6853 vmovdqu<ssescalarsize> for TI and OI modes.
6855 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6857 Update copyright years.
6859 2018-01-03 Martin Liska <mliska@suse.cz>
6862 * ipa-visibility.c (function_and_variable_visibility): Skip
6863 functions with noipa attribure.
6865 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6867 * gcc.c (process_command): Update copyright notice dates.
6868 * gcov-dump.c (print_version): Ditto.
6869 * gcov.c (print_version): Ditto.
6870 * gcov-tool.c (print_version): Ditto.
6871 * gengtype.c (create_file): Ditto.
6872 * doc/cpp.texi: Bump @copying's copyright year.
6873 * doc/cppinternals.texi: Ditto.
6874 * doc/gcc.texi: Ditto.
6875 * doc/gccint.texi: Ditto.
6876 * doc/gcov.texi: Ditto.
6877 * doc/install.texi: Ditto.
6878 * doc/invoke.texi: Ditto.
6880 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6882 * vector-builder.h (vector_builder::m_full_nelts): Change from
6883 unsigned int to poly_uint64.
6884 (vector_builder::full_nelts): Update prototype accordingly.
6885 (vector_builder::new_vector): Likewise.
6886 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
6887 (vector_builder::operator ==): Likewise.
6888 (vector_builder::finalize): Likewise.
6889 * int-vector-builder.h (int_vector_builder::int_vector_builder):
6890 Take the number of elements as a poly_uint64 rather than an
6892 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
6893 from unsigned int to poly_uint64.
6894 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
6895 (vec_perm_indices::new_vector): Likewise.
6896 (vec_perm_indices::length): Likewise.
6897 (vec_perm_indices::nelts_per_input): Likewise.
6898 (vec_perm_indices::input_nelts): Likewise.
6899 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
6900 number of elements per input as a poly_uint64 rather than an
6901 unsigned int. Use the original encoding for variable-length
6902 vectors, rather than clamping each individual element.
6903 For the second and subsequent elements in each pattern,
6904 clamp the step and base before clamping their sum.
6905 (vec_perm_indices::series_p): Handle polynomial element counts.
6906 (vec_perm_indices::all_in_range_p): Likewise.
6907 (vec_perm_indices_to_tree): Likewise.
6908 (vec_perm_indices_to_rtx): Likewise.
6909 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
6910 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
6911 (tree_vector_builder::new_binary_operation): Handle polynomial
6912 element counts. Return false if we need to know the number
6913 of elements at compile time.
6914 * fold-const.c (fold_vec_perm): Punt if the number of elements
6915 isn't known at compile time.
6917 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6919 * vec-perm-indices.h (vec_perm_builder): Change element type
6920 from HOST_WIDE_INT to poly_int64.
6921 (vec_perm_indices::element_type): Update accordingly.
6922 (vec_perm_indices::clamp): Handle polynomial element_types.
6923 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
6924 (vec_perm_indices::all_in_range_p): Likewise.
6925 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
6927 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
6928 polynomial vec_perm_indices element types.
6929 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
6930 * fold-const.c (fold_vec_perm): Likewise.
6931 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
6932 * tree-vect-generic.c (lower_vec_perm): Likewise.
6933 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
6934 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
6935 element type to HOST_WIDE_INT.
6937 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6938 Alan Hayward <alan.hayward@arm.com>
6939 David Sherwood <david.sherwood@arm.com>
6941 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
6942 rather than an int. Use plus_constant.
6943 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
6944 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
6946 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6947 Alan Hayward <alan.hayward@arm.com>
6948 David Sherwood <david.sherwood@arm.com>
6950 * calls.c (emit_call_1, expand_call): Change struct_value_size from
6951 a HOST_WIDE_INT to a poly_int64.
6953 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6954 Alan Hayward <alan.hayward@arm.com>
6955 David Sherwood <david.sherwood@arm.com>
6957 * calls.c (load_register_parameters): Cope with polynomial
6958 mode sizes. Require a constant size for BLKmode parameters
6959 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
6960 forces a parameter to be padded at the lsb end in order to
6961 fill a complete number of words, require the parameter size
6962 to be ordered wrt UNITS_PER_WORD.
6964 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6965 Alan Hayward <alan.hayward@arm.com>
6966 David Sherwood <david.sherwood@arm.com>
6968 * reload1.c (spill_stack_slot_width): Change element type
6969 from unsigned int to poly_uint64_pod.
6970 (alter_reg): Treat mode sizes as polynomial.
6972 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6973 Alan Hayward <alan.hayward@arm.com>
6974 David Sherwood <david.sherwood@arm.com>
6976 * reload.c (complex_word_subreg_p): New function.
6977 (reload_inner_reg_of_subreg, push_reload): Use it.
6979 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6980 Alan Hayward <alan.hayward@arm.com>
6981 David Sherwood <david.sherwood@arm.com>
6983 * lra-constraints.c (process_alt_operands): Reject matched
6984 operands whose sizes aren't ordered.
6985 (match_reload): Refer to this check here.
6987 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6988 Alan Hayward <alan.hayward@arm.com>
6989 David Sherwood <david.sherwood@arm.com>
6991 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
6992 that the mode size is in the set {1, 2, 4, 8, 16}.
6994 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6995 Alan Hayward <alan.hayward@arm.com>
6996 David Sherwood <david.sherwood@arm.com>
6998 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
6999 Use plus_constant instead of gen_rtx_PLUS.
7001 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7002 Alan Hayward <alan.hayward@arm.com>
7003 David Sherwood <david.sherwood@arm.com>
7005 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
7006 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
7007 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
7008 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
7009 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
7010 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
7011 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
7012 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
7013 * config/i386/i386.c (ix86_push_rounding): ...this new function.
7014 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
7016 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
7017 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
7018 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
7019 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
7020 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
7021 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
7022 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
7023 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
7024 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
7025 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
7027 * expr.c (emit_move_resolve_push): Treat the input and result
7028 of PUSH_ROUNDING as a poly_int64.
7029 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
7030 (emit_push_insn): Likewise.
7031 * lra-eliminations.c (mark_not_eliminable): Likewise.
7032 * recog.c (push_operand): Likewise.
7033 * reload1.c (elimination_effects): Likewise.
7034 * rtlanal.c (nonzero_bits1): Likewise.
7035 * calls.c (store_one_arg): Likewise. Require the padding to be
7036 known at compile time.
7038 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7039 Alan Hayward <alan.hayward@arm.com>
7040 David Sherwood <david.sherwood@arm.com>
7042 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
7043 Use plus_constant instead of gen_rtx_PLUS.
7045 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7046 Alan Hayward <alan.hayward@arm.com>
7047 David Sherwood <david.sherwood@arm.com>
7049 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
7052 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7053 Alan Hayward <alan.hayward@arm.com>
7054 David Sherwood <david.sherwood@arm.com>
7056 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
7057 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
7058 via stack temporaries. Treat the mode size as polynomial too.
7060 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7061 Alan Hayward <alan.hayward@arm.com>
7062 David Sherwood <david.sherwood@arm.com>
7064 * expr.c (expand_expr_real_2): When handling conversions involving
7065 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
7066 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
7067 as a poly_uint64 too.
7069 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7070 Alan Hayward <alan.hayward@arm.com>
7071 David Sherwood <david.sherwood@arm.com>
7073 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
7075 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7076 Alan Hayward <alan.hayward@arm.com>
7077 David Sherwood <david.sherwood@arm.com>
7079 * combine.c (can_change_dest_mode): Handle polynomial
7080 REGMODE_NATURAL_SIZE.
7081 * expmed.c (store_bit_field_1): Likewise.
7082 * expr.c (store_constructor): Likewise.
7083 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
7084 and polynomial REGMODE_NATURAL_SIZE.
7085 (gen_lowpart_common): Likewise.
7086 * reginfo.c (record_subregs_of_mode): Likewise.
7087 * rtlanal.c (read_modify_subreg_p): Likewise.
7089 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7090 Alan Hayward <alan.hayward@arm.com>
7091 David Sherwood <david.sherwood@arm.com>
7093 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
7094 numbers of elements.
7096 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7097 Alan Hayward <alan.hayward@arm.com>
7098 David Sherwood <david.sherwood@arm.com>
7100 * match.pd: Cope with polynomial numbers of vector elements.
7102 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7103 Alan Hayward <alan.hayward@arm.com>
7104 David Sherwood <david.sherwood@arm.com>
7106 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
7107 in a POINTER_PLUS_EXPR.
7109 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7110 Alan Hayward <alan.hayward@arm.com>
7111 David Sherwood <david.sherwood@arm.com>
7113 * omp-simd-clone.c (simd_clone_subparts): New function.
7114 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
7115 (ipa_simd_modify_function_body): Likewise.
7117 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7118 Alan Hayward <alan.hayward@arm.com>
7119 David Sherwood <david.sherwood@arm.com>
7121 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
7122 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
7123 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
7124 (expand_vector_condition, vector_element): Likewise.
7125 (subparts_gt): New function.
7126 (get_compute_type): Use subparts_gt.
7127 (count_type_subparts): Delete.
7128 (expand_vector_operations_1): Use subparts_gt instead of
7129 count_type_subparts.
7131 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7132 Alan Hayward <alan.hayward@arm.com>
7133 David Sherwood <david.sherwood@arm.com>
7135 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
7136 (vect_compile_time_alias): ...this new function. Do the calculation
7137 on poly_ints rather than trees.
7138 (vect_prune_runtime_alias_test_list): Update call accordingly.
7140 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7141 Alan Hayward <alan.hayward@arm.com>
7142 David Sherwood <david.sherwood@arm.com>
7144 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
7146 (vect_schedule_slp_instance): Likewise.
7148 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7149 Alan Hayward <alan.hayward@arm.com>
7150 David Sherwood <david.sherwood@arm.com>
7152 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
7153 constant and extern definitions for variable-length vectors.
7154 (vect_get_constant_vectors): Note that the number of units
7155 is known to be constant.
7157 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7158 Alan Hayward <alan.hayward@arm.com>
7159 David Sherwood <david.sherwood@arm.com>
7161 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
7162 of units as polynomial. Choose between WIDE and NARROW based
7165 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7166 Alan Hayward <alan.hayward@arm.com>
7167 David Sherwood <david.sherwood@arm.com>
7169 * tree-vect-stmts.c (simd_clone_subparts): New function.
7170 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
7172 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7173 Alan Hayward <alan.hayward@arm.com>
7174 David Sherwood <david.sherwood@arm.com>
7176 * tree-vect-stmts.c (vectorizable_call): Treat the number of
7177 vectors as polynomial. Use build_index_vector for
7180 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7181 Alan Hayward <alan.hayward@arm.com>
7182 David Sherwood <david.sherwood@arm.com>
7184 * tree-vect-stmts.c (get_load_store_type): Treat the number of
7185 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
7186 for variable-length vectors.
7187 (vectorizable_mask_load_store): Treat the number of units as
7188 polynomial, asserting that it is constant if the condition has
7189 already been enforced.
7190 (vectorizable_store, vectorizable_load): Likewise.
7192 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7193 Alan Hayward <alan.hayward@arm.com>
7194 David Sherwood <david.sherwood@arm.com>
7196 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
7197 of units as polynomial. Punt if we can't tell at compile time
7198 which vector contains the final result.
7200 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7201 Alan Hayward <alan.hayward@arm.com>
7202 David Sherwood <david.sherwood@arm.com>
7204 * tree-vect-loop.c (vectorizable_induction): Treat the number
7205 of units as polynomial. Punt on SLP inductions. Use an integer
7206 VEC_SERIES_EXPR for variable-length integer reductions. Use a
7207 cast of such a series for variable-length floating-point
7210 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7211 Alan Hayward <alan.hayward@arm.com>
7212 David Sherwood <david.sherwood@arm.com>
7214 * tree.h (build_index_vector): Declare.
7215 * tree.c (build_index_vector): New function.
7216 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
7217 of units as polynomial, forcibly converting it to a constant if
7218 vectorizable_reduction has already enforced the condition.
7219 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
7220 to create a {1,2,3,...} vector.
7221 (vectorizable_reduction): Treat the number of units as polynomial.
7222 Choose vectype_in based on the largest scalar element size rather
7223 than the smallest number of units. Enforce the restrictions
7226 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7227 Alan Hayward <alan.hayward@arm.com>
7228 David Sherwood <david.sherwood@arm.com>
7230 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
7231 number of units as polynomial.
7233 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7234 Alan Hayward <alan.hayward@arm.com>
7235 David Sherwood <david.sherwood@arm.com>
7237 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
7238 * target.def (autovectorize_vector_sizes): Return the vector sizes
7239 by pointer, using vector_sizes rather than a bitmask.
7240 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
7241 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
7242 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
7244 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
7245 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
7246 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
7247 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
7248 * omp-general.c (omp_max_vf): Likewise.
7249 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7250 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
7251 * tree-vect-loop.c (vect_analyze_loop): Likewise.
7252 * tree-vect-slp.c (vect_slp_bb): Likewise.
7253 * doc/tm.texi: Regenerate.
7254 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
7256 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
7257 the vector size as a poly_uint64 rather than an unsigned int.
7258 (current_vector_size): Change from an unsigned int to a poly_uint64.
7259 (get_vectype_for_scalar_type): Update accordingly.
7260 * tree.h (build_truth_vector_type): Take the size and number of
7261 units as a poly_uint64 rather than an unsigned int.
7262 (build_vector_type): Add a temporary overload that takes
7263 the number of units as a poly_uint64 rather than an unsigned int.
7264 * tree.c (make_vector_type): Likewise.
7265 (build_truth_vector_type): Take the number of units as a poly_uint64
7266 rather than an unsigned int.
7268 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7269 Alan Hayward <alan.hayward@arm.com>
7270 David Sherwood <david.sherwood@arm.com>
7272 * target.def (get_mask_mode): Take the number of units and length
7273 as poly_uint64s rather than unsigned ints.
7274 * targhooks.h (default_get_mask_mode): Update accordingly.
7275 * targhooks.c (default_get_mask_mode): Likewise.
7276 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
7277 * doc/tm.texi: Regenerate.
7279 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7280 Alan Hayward <alan.hayward@arm.com>
7281 David Sherwood <david.sherwood@arm.com>
7283 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
7284 * omp-general.c (omp_max_vf): Likewise.
7285 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
7286 (expand_omp_simd): Handle polynomial safelen.
7287 * omp-low.c (omplow_simd_context): Add a default constructor.
7288 (omplow_simd_context::max_vf): Change from int to poly_uint64.
7289 (lower_rec_simd_input_clauses): Update accordingly.
7290 (lower_rec_input_clauses): Likewise.
7292 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7293 Alan Hayward <alan.hayward@arm.com>
7294 David Sherwood <david.sherwood@arm.com>
7296 * tree-vectorizer.h (vect_nunits_for_cost): New function.
7297 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
7298 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
7299 (vect_analyze_slp_cost): Likewise.
7300 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
7301 (vect_model_load_cost): Likewise.
7303 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7304 Alan Hayward <alan.hayward@arm.com>
7305 David Sherwood <david.sherwood@arm.com>
7307 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
7308 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
7309 from an unsigned int * to a poly_uint64_pod *.
7310 (calculate_unrolling_factor): New function.
7311 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
7313 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7314 Alan Hayward <alan.hayward@arm.com>
7315 David Sherwood <david.sherwood@arm.com>
7317 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
7318 from an unsigned int to a poly_uint64.
7319 (_loop_vec_info::slp_unrolling_factor): Likewise.
7320 (_loop_vec_info::vectorization_factor): Change from an int
7322 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
7323 (vect_get_num_vectors): New function.
7324 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
7325 (vect_get_num_copies): Use vect_get_num_vectors.
7326 (vect_analyze_data_ref_dependences): Change max_vf from an int *
7327 to an unsigned int *.
7328 (vect_analyze_data_refs): Change min_vf from an int * to a
7330 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7331 than an unsigned HOST_WIDE_INT.
7332 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7333 (vect_analyze_data_ref_dependence): Change max_vf from an int *
7334 to an unsigned int *.
7335 (vect_analyze_data_ref_dependences): Likewise.
7336 (vect_compute_data_ref_alignment): Handle polynomial vf.
7337 (vect_enhance_data_refs_alignment): Likewise.
7338 (vect_prune_runtime_alias_test_list): Likewise.
7339 (vect_shift_permute_load_chain): Likewise.
7340 (vect_supportable_dr_alignment): Likewise.
7341 (dependence_distance_ge_vf): Take the vectorization factor as a
7342 poly_uint64 rather than an unsigned HOST_WIDE_INT.
7343 (vect_analyze_data_refs): Change min_vf from an int * to a
7345 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
7346 vfm1 as a poly_uint64 rather than an int. Make the same change
7347 for the returned bound_scalar.
7348 (vect_gen_vector_loop_niters): Handle polynomial vf.
7349 (vect_do_peeling): Likewise. Update call to
7350 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
7351 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
7353 * tree-vect-loop.c (vect_determine_vectorization_factor)
7354 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
7355 (vect_get_known_peeling_cost): Likewise.
7356 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
7357 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
7358 (vect_transform_loop): Likewise. Use the lowest possible VF when
7359 updating the upper bounds of the loop.
7360 (vect_min_worthwhile_factor): Make static. Return an unsigned int
7362 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
7363 polynomial unroll factors.
7364 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
7365 (vect_make_slp_decision): Likewise.
7366 (vect_supported_load_permutation_p): Likewise, and polynomial
7368 (vect_analyze_slp_cost): Handle polynomial vf.
7369 (vect_slp_analyze_node_operations): Likewise.
7370 (vect_slp_analyze_bb_1): Likewise.
7371 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7372 than an unsigned HOST_WIDE_INT.
7373 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
7374 (vectorizable_load): Handle polynomial vf.
7375 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
7377 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
7379 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7380 Alan Hayward <alan.hayward@arm.com>
7381 David Sherwood <david.sherwood@arm.com>
7383 * match.pd: Handle bit operations involving three constants
7384 and try to fold one pair.
7386 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7388 * tree-vect-loop-manip.c: Include gimple-fold.h.
7389 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
7390 niters_maybe_zero parameters. Handle other cases besides a step of 1.
7391 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
7392 Add a path that uses a step of VF instead of 1, but disable it
7394 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
7395 and niters_no_overflow parameters. Update calls to
7396 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
7397 Create a new SSA name if the latter choses to use a ste other
7398 than zero, and return it via niters_vector_mult_vf_var.
7399 * tree-vect-loop.c (vect_transform_loop): Update calls to
7400 vect_do_peeling, vect_gen_vector_loop_niters and
7401 slpeel_make_loop_iterate_ntimes.
7402 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
7403 (vect_gen_vector_loop_niters): Update declarations after above changes.
7405 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
7407 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
7408 128-bit round to integer instructions.
7409 (ceil<mode>2): Likewise.
7410 (btrunc<mode>2): Likewise.
7411 (round<mode>2): Likewise.
7413 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7415 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
7416 unaligned VSX load/store on P8/P9.
7417 (expand_block_clear): Allow the use of unaligned VSX
7418 load/store on P8/P9.
7420 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7422 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
7424 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
7425 swap associated with both a load and a store.
7427 2018-01-02 Andrew Waterman <andrew@sifive.com>
7429 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
7430 * config/riscv/riscv.md (clear_cache): Use it.
7432 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
7434 * web.c: Remove out-of-date comment.
7436 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7438 * expr.c (fixup_args_size_notes): Check that any existing
7439 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
7440 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
7441 (emit_single_push_insn): ...here.
7443 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7445 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
7446 (const_vector_encoded_nelts): New function.
7447 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
7448 (const_vector_int_elt, const_vector_elt): Declare.
7449 * emit-rtl.c (const_vector_int_elt_1): New function.
7450 (const_vector_elt): Likewise.
7451 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
7452 of CONST_VECTOR_ELT.
7454 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7456 * expr.c: Include rtx-vector-builder.h.
7457 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
7458 directly on the tree encoding.
7459 (const_vector_from_tree): Likewise.
7460 * optabs.c: Include rtx-vector-builder.h.
7461 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
7462 sequence of "u" values.
7463 * vec-perm-indices.c: Include rtx-vector-builder.h.
7464 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
7465 directly on the vec_perm_indices encoding.
7467 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7469 * doc/rtl.texi (const_vector): Describe new encoding scheme.
7470 * Makefile.in (OBJS): Add rtx-vector-builder.o.
7471 * rtx-vector-builder.h: New file.
7472 * rtx-vector-builder.c: Likewise.
7473 * rtl.h (rtx_def::u2): Add a const_vector field.
7474 (CONST_VECTOR_NPATTERNS): New macro.
7475 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
7476 (CONST_VECTOR_DUPLICATE_P): Likewise.
7477 (CONST_VECTOR_STEPPED_P): Likewise.
7478 (CONST_VECTOR_ENCODED_ELT): Likewise.
7479 (const_vec_duplicate_p): Check for a duplicated vector encoding.
7480 (unwrap_const_vec_duplicate): Likewise.
7481 (const_vec_series_p): Check for a non-duplicated vector encoding.
7482 Say that the function only returns true for integer vectors.
7483 * emit-rtl.c: Include rtx-vector-builder.h.
7484 (gen_const_vec_duplicate_1): Delete.
7485 (gen_const_vector): Call gen_const_vec_duplicate instead of
7486 gen_const_vec_duplicate_1.
7487 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
7488 (gen_const_vec_duplicate): Use rtx_vector_builder.
7489 (gen_const_vec_series): Likewise.
7490 (gen_rtx_CONST_VECTOR): Likewise.
7491 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
7492 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7493 Build a new vector rather than modifying a CONST_VECTOR in-place.
7494 (handle_special_swappables): Update call accordingly.
7495 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
7496 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7497 Build a new vector rather than modifying a CONST_VECTOR in-place.
7498 (handle_special_swappables): Update call accordingly.
7500 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7502 * simplify-rtx.c (simplify_const_binary_operation): Use
7503 CONST_VECTOR_ELT instead of XVECEXP.
7505 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7507 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
7508 the selector elements to be different from the data elements
7509 if the selector is a VECTOR_CST.
7510 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
7511 ssizetype for the selector.
7513 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7515 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
7516 before testing each element individually.
7517 * tree-vect-generic.c (lower_vec_perm): Likewise.
7519 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7521 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
7522 * selftest-run-tests.c (selftest::run_tests): Call it.
7523 * vector-builder.h (vector_builder::operator ==): New function.
7524 (vector_builder::operator !=): Likewise.
7525 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
7526 (vec_perm_indices::all_from_input_p): New function.
7527 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7528 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
7529 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
7530 instead of reading the VECTOR_CST directly. Detect whether both
7531 vector inputs are the same before constructing the vec_perm_indices,
7532 and update the number of inputs argument accordingly. Use the
7533 utility functions added above. Only construct sel2 if we need to.
7535 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7537 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
7538 the broadcast of the low byte.
7539 (expand_mult_highpart): Use an explicit encoding for the permutes.
7540 * optabs-query.c (can_mult_highpart_p): Likewise.
7541 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
7542 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7543 (vectorizable_bswap): Likewise.
7544 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
7545 explicit encoding for the power-of-2 permutes.
7546 (vect_permute_store_chain): Likewise.
7547 (vect_grouped_load_supported): Likewise.
7548 (vect_permute_load_chain): Likewise.
7550 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7552 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
7553 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
7554 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
7555 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7556 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
7557 (vect_gen_perm_mask_any): Likewise.
7559 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7561 * int-vector-builder.h: New file.
7562 * vec-perm-indices.h: Include int-vector-builder.h.
7563 (vec_perm_indices): Redefine as an int_vector_builder.
7564 (auto_vec_perm_indices): Delete.
7565 (vec_perm_builder): Redefine as a stand-alone class.
7566 (vec_perm_indices::vec_perm_indices): New function.
7567 (vec_perm_indices::clamp): Likewise.
7568 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
7569 (vec_perm_indices::new_vector): New function.
7570 (vec_perm_indices::new_expanded_vector): Update for new
7571 vec_perm_indices class.
7572 (vec_perm_indices::rotate_inputs): New function.
7573 (vec_perm_indices::all_in_range_p): Operate directly on the
7574 encoded form, without computing elided elements.
7575 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
7576 encoding. Update for new vec_perm_indices class.
7577 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
7578 the given vec_perm_builder.
7579 (expand_vec_perm_var): Update vec_perm_builder constructor.
7580 (expand_mult_highpart): Use vec_perm_builder instead of
7581 auto_vec_perm_indices.
7582 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
7583 vec_perm_indices instead of auto_vec_perm_indices. Use a single
7584 or double series encoding as appropriate.
7585 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
7586 vec_perm_indices instead of auto_vec_perm_indices.
7587 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7588 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7589 (vect_permute_store_chain): Likewise.
7590 (vect_grouped_load_supported): Likewise.
7591 (vect_permute_load_chain): Likewise.
7592 (vect_shift_permute_load_chain): Likewise.
7593 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
7594 (vect_transform_slp_perm_load): Likewise.
7595 (vect_schedule_slp_instance): Likewise.
7596 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7597 (vectorizable_mask_load_store): Likewise.
7598 (vectorizable_bswap): Likewise.
7599 (vectorizable_store): Likewise.
7600 (vectorizable_load): Likewise.
7601 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
7602 vec_perm_indices instead of auto_vec_perm_indices. Use
7603 tree_to_vec_perm_builder to read the vector from a tree.
7604 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
7605 vec_perm_builder instead of a vec_perm_indices.
7606 (have_whole_vector_shift): Use vec_perm_builder and
7607 vec_perm_indices instead of auto_vec_perm_indices. Leave the
7608 truncation to calc_vec_perm_mask_for_shift.
7609 (vect_create_epilog_for_reduction): Likewise.
7610 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
7611 from auto_vec_perm_indices to vec_perm_indices.
7612 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7613 instead of changing individual elements.
7614 (aarch64_vectorize_vec_perm_const): Use new_vector to install
7615 the vector in d.perm.
7616 * config/arm/arm.c (expand_vec_perm_d::perm): Change
7617 from auto_vec_perm_indices to vec_perm_indices.
7618 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7619 instead of changing individual elements.
7620 (arm_vectorize_vec_perm_const): Use new_vector to install
7621 the vector in d.perm.
7622 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
7623 Update vec_perm_builder constructor.
7624 (rs6000_expand_interleave): Likewise.
7625 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
7626 (rs6000_expand_interleave): Likewise.
7628 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7630 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
7631 to qimode could truncate the indices.
7632 * optabs.c (expand_vec_perm_var): Likewise.
7634 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7636 * Makefile.in (OBJS): Add vec-perm-indices.o.
7637 * vec-perm-indices.h: New file.
7638 * vec-perm-indices.c: Likewise.
7639 * target.h (vec_perm_indices): Replace with a forward class
7641 (auto_vec_perm_indices): Move to vec-perm-indices.h.
7642 * optabs.h: Include vec-perm-indices.h.
7643 (expand_vec_perm): Delete.
7644 (selector_fits_mode_p, expand_vec_perm_var): Declare.
7645 (expand_vec_perm_const): Declare.
7646 * target.def (vec_perm_const_ok): Replace with...
7647 (vec_perm_const): ...this new hook.
7648 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
7649 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
7650 * doc/tm.texi: Regenerate.
7651 * optabs.def (vec_perm_const): Delete.
7652 * doc/md.texi (vec_perm_const): Likewise.
7653 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
7654 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
7655 expand_vec_perm for constant permutation vectors. Assert that
7656 the mode of variable permutation vectors is the integer equivalent
7657 of the mode that is being permuted.
7658 * optabs-query.h (selector_fits_mode_p): Declare.
7659 * optabs-query.c: Include vec-perm-indices.h.
7660 (selector_fits_mode_p): New function.
7661 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
7662 is defined, instead of checking whether the vec_perm_const_optab
7663 exists. Use targetm.vectorize.vec_perm_const instead of
7664 targetm.vectorize.vec_perm_const_ok. Check whether the indices
7665 fit in the vector mode before using a variable permute.
7666 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
7667 vec_perm_indices instead of an rtx.
7668 (expand_vec_perm): Replace with...
7669 (expand_vec_perm_const): ...this new function. Take the selector
7670 as a vec_perm_indices rather than an rtx. Also take the mode of
7671 the selector. Update call to shift_amt_for_vec_perm_mask.
7672 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
7673 Use vec_perm_indices::new_expanded_vector to expand the original
7674 selector into bytes. Check whether the indices fit in the vector
7675 mode before using a variable permute.
7676 (expand_vec_perm_var): Make global.
7677 (expand_mult_highpart): Use expand_vec_perm_const.
7678 * fold-const.c: Includes vec-perm-indices.h.
7679 * tree-ssa-forwprop.c: Likewise.
7680 * tree-vect-data-refs.c: Likewise.
7681 * tree-vect-generic.c: Likewise.
7682 * tree-vect-loop.c: Likewise.
7683 * tree-vect-slp.c: Likewise.
7684 * tree-vect-stmts.c: Likewise.
7685 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
7687 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
7688 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
7689 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
7690 (aarch64_vectorize_vec_perm_const): ...this new function.
7691 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7692 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7693 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
7694 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
7695 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7696 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7697 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
7699 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
7700 check for NEON modes.
7701 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
7702 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
7703 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
7704 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
7706 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
7707 the old VEC_PERM_CONST conditions.
7708 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
7709 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
7710 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
7711 (ia64_vectorize_vec_perm_const_ok): Merge into...
7712 (ia64_vectorize_vec_perm_const): ...this new function.
7713 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
7714 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
7715 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
7716 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
7717 * config/mips/mips.c (mips_expand_vec_perm_const)
7718 (mips_vectorize_vec_perm_const_ok): Merge into...
7719 (mips_vectorize_vec_perm_const): ...this new function.
7720 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
7721 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
7722 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
7723 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
7724 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
7725 (rs6000_expand_vec_perm_const): Delete.
7726 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
7728 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7729 (altivec_expand_vec_perm_const_le): Take each operand individually.
7730 Operate on constant selectors rather than rtxes.
7731 (altivec_expand_vec_perm_const): Likewise. Update call to
7732 altivec_expand_vec_perm_const_le.
7733 (rs6000_expand_vec_perm_const): Delete.
7734 (rs6000_vectorize_vec_perm_const_ok): Delete.
7735 (rs6000_vectorize_vec_perm_const): New function.
7736 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
7737 an element count and rtx array.
7738 (rs6000_expand_extract_even): Update call accordingly.
7739 (rs6000_expand_interleave): Likewise.
7740 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
7741 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
7742 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
7743 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
7744 (rs6000_expand_vec_perm_const): Delete.
7745 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7746 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7747 (altivec_expand_vec_perm_const_le): Take each operand individually.
7748 Operate on constant selectors rather than rtxes.
7749 (altivec_expand_vec_perm_const): Likewise. Update call to
7750 altivec_expand_vec_perm_const_le.
7751 (rs6000_expand_vec_perm_const): Delete.
7752 (rs6000_vectorize_vec_perm_const_ok): Delete.
7753 (rs6000_vectorize_vec_perm_const): New function. Remove stray
7754 reference to the SPE evmerge intructions.
7755 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
7756 an element count and rtx array.
7757 (rs6000_expand_extract_even): Update call accordingly.
7758 (rs6000_expand_interleave): Likewise.
7759 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
7760 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
7762 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7764 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7766 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
7767 vector mode and that that mode matches the mode of the data
7769 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
7770 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
7771 directly using expand_vec_perm_1 when forcing selectors into
7773 (expand_vec_perm_var): New function, split out from expand_vec_perm.
7775 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7777 * optabs-query.h (can_vec_perm_p): Delete.
7778 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
7779 * optabs-query.c (can_vec_perm_p): Split into...
7780 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
7781 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
7782 particular selector is valid.
7783 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7784 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7785 (vect_grouped_load_supported): Likewise.
7786 (vect_shift_permute_load_chain): Likewise.
7787 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
7788 (vect_transform_slp_perm_load): Likewise.
7789 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7790 (vectorizable_bswap): Likewise.
7791 (vect_gen_perm_mask_checked): Likewise.
7792 * fold-const.c (fold_ternary_loc): Likewise. Don't take
7793 implementations of variable permutation vectors into account
7794 when deciding which selector to use.
7795 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
7796 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
7797 with a false third argument.
7798 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
7799 to test whether the constant selector is valid and can_vec_perm_var_p
7800 to test whether a variable selector is valid.
7802 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7804 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
7805 * optabs-query.c (can_vec_perm_p): Likewise.
7806 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
7807 instead of vec_perm_indices.
7808 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
7809 (vect_gen_perm_mask_checked): Likewise,
7810 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
7811 (vect_gen_perm_mask_checked): Likewise,
7813 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7815 * optabs-query.h (qimode_for_vec_perm): Declare.
7816 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
7817 (qimode_for_vec_perm): ...this new function.
7818 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
7820 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7822 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
7823 does not have a conditional at the top.
7825 2018-01-02 Richard Biener <rguenther@suse.de>
7827 * ipa-inline.c (big_speedup_p): Fix expression.
7829 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
7832 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
7835 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
7839 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
7840 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
7841 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
7842 cond_taken_branch_cost 3->4.
7844 2018-01-01 Jakub Jelinek <jakub@redhat.com>
7846 PR tree-optimization/83581
7847 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
7848 TODO_cleanup_cfg if any changes have been made.
7851 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
7852 convert_modes if target mode has the right side, but different mode
7856 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
7857 last argument when extracting from CONCAT. If either from_real or
7858 from_imag is NULL, use expansion through memory. If result is not
7859 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
7860 the parts directly to inner mode, if even that fails, use expansion
7864 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
7865 check for bswap in mode rather than HImode and use that in expand_unop
7868 Copyright (C) 2018 Free Software Foundation, Inc.
7870 Copying and distribution of this file, with or without modification,
7871 are permitted in any medium without royalty provided the copyright
7872 notice and this notice are preserved.