1 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4 * config/i386/i386.c (ix86_print_operand): Use conditional
5 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
8 2018-03-02 Richard Biener <rguenther@suse.de>
10 PR tree-optimization/84427
11 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
12 (bitmap_set_subtract_values): Rewrite to handle multiple
15 (prune_clobbered_mems): Likewise.
16 (phi_translate): Take edge instead of pred/phiblock.
17 (phi_translate_1): Likewise.
18 (phi_translate_set): Likewise. Insert all translated
19 exprs for a value into the set, keeping possibly multiple
20 expressions per value.
21 (compute_antic_aux): Adjust for phi_translate changes.
22 When intersecting union the expressions and prune those
23 not in the final value set, keeping possibly multiple
24 expressions per value. Do not use value-insertion
25 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
26 all expressions. Add verification that the value-sets
27 only shrink during iteration.
28 (compute_partial_antic_aux): Adjust for the phi_translate changes.
29 (do_pre_regular_insertion): Likewise.
30 (do_pre_partial_partial_insertion): Likewise.
32 2018-03-02 Richard Biener <rguenther@suse.de>
35 * config/darwin.c (saved_debug_info_level): New static global.
36 (darwin_asm_lto_start): Disable debug info generation for LTO out.
37 (darwin_asm_lto_end): Restore debug info generation settings.
39 2018-03-01 Martin Liska <mliska@suse.cz>
42 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
45 2018-03-01 Richard Biener <rguenther@suse.de>
48 * dwarf2out.c (gen_variable_die): Properly handle late VLA
49 type annotation with LTO when debug was disabled at compile-time.
51 2018-03-01 Matthew Fortune <mfortune@gmail.com>
53 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
55 (mips_final_postscan_insn): Likewise.
57 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
59 PR rtl-optimization/84528
60 * alias.c (init_alias_target): Add commentary.
61 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
62 a unique base value if the frame pointer is not eliminated
65 2018-03-01 Tom de Vries <tom@codesourcery.com>
67 PR rtl-optimization/83327
68 * lra-int.h (hard_regs_spilled_into): Declare.
69 * lra.c (hard_regs_spilled_into): Define.
70 (init_reg_info): Init hard_regs_spilled_into.
71 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
72 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
73 (process_bb_lives): Handle hard_regs_spilled_into.
74 (lra_create_live_ranges_1): Before doing liveness propagation, clear
75 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
77 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
79 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
80 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
81 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
82 * config/rs6000/aix72.h: New file.
84 2018-02-28 Jakub Jelinek <jakub@redhat.com>
86 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
87 instead of warning_at with conditional singular and plural messages
91 * stor-layout.c (update_alignment_for_field): For
92 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
93 && !DECL_PACKED (field), do the alignment update, just use
94 only desired_align instead of MAX (type_align, desired_align)
96 (place_field): Don't do known_align < desired_align handling
97 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
98 is non-NULL, instead do it after rli->prev_field handling and
99 only if not within a bitfield word. For DECL_PACKED (field)
100 use type_align of BITS_PER_UNIT.
102 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
104 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
105 superfluous parentheses and trailing spaces.
107 2018-02-28 Richard Biener <rguenther@suse.de>
109 PR tree-optimization/84584
110 * graphite-scop-detection.c (scop_detection::add_scop): Discard
111 SCoPs with fake exit edge.
113 2018-02-28 Martin Liska <mliska@suse.cz>
116 * timevar.c (timer::print): Fix format to properly print 100%
119 2018-02-28 Richard Biener <rguenther@suse.de>
122 * genmatch.c (capture_info::walk_match): Do not mark
123 captured expressions without operands as expr_p given
124 they act more like predicates and should be subject to
125 "lost tail" side-effect preserving.
127 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
129 PR rtl-optimization/81611
130 * auto-inc-dec.c (attempt_change): Move dead note from
131 mem_insn if it's the next use of regno
132 (find_address): Take address use of reg holding
133 non-incremented value. Add parm to limit search to the named
135 (merge_in_block): Attempt to use a mem insn that is the next
136 use of the original regno.
138 2018-02-27 Martin Sebor <msebor@redhat.com>
141 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
142 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
144 2018-02-27 Martin Sebor <msebor@redhat.com>
147 * diagnostic-core.h (warning_n, error_n, inform_n): Change
148 n argument to unsigned HOST_WIDE_INT.
149 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
150 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
151 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
152 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
154 2018-02-27 Richard Biener <rguenther@suse.de>
156 PR tree-optimization/84512
157 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
158 Do not use the estimate returned from record_stmt_cost for
159 the scalar iteration cost but sum properly using add_stmt_cost.
161 2018-02-27 Richard Biener <rguenther@suse.de>
163 PR tree-optimization/84466
164 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
165 Adjust last change to less strictly validate use operands.
167 2018-02-27 Martin Liska <mliska@suse.cz>
169 PR gcov-profile/84548
170 * gcov.c (process_file): Allow partial overlap and consider it
171 also as group functions.
172 (output_lines): Properly calculate range of lines for a group.
174 2018-02-27 Martin Liska <mliska@suse.cz>
176 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
177 'ggc' suffixes. Change first column width.
178 (timer::print): Fix formatting of the column.
180 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
182 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
183 preserve inline entry blocks for the sake of debug inline
184 entry point markers alone.
185 (remove_unused_locals): Suggest in comments a better place to
186 force the preservation of inline entry blocks that are
187 otherwise unused, but do not preserve them.
189 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
191 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
193 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
196 * config/i386/constraints.md (Bs): Replace
197 ix86_indirect_branch_register with
198 TARGET_INDIRECT_BRANCH_REGISTER.
200 * config/i386/i386.md (indirect_jump): Likewise.
201 (tablejump): Likewise.
202 (*sibcall_memory): Likewise.
203 (*sibcall_value_memory): Likewise.
204 Peepholes of indirect call and jump via memory: Likewise.
205 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
206 (*sibcall_value_GOT_32): Likewise.
207 * config/i386/predicates.md (indirect_branch_operand): Likewise.
208 (GOT_memory_operand): Likewise.
209 (call_insn_operand): Likewise.
210 (sibcall_insn_operand): Likewise.
211 (GOT32_symbol_operand): Likewise.
212 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
214 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
216 PR rtl-optimization/83496
217 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
218 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
219 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
220 redundant insn, if any.
221 (relax_delay_slots): Likewise.
222 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
224 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
226 PR tree-optimization/83965
227 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
228 that grouped statements are part of a reduction chain. Return
229 true if the statement is not marked as a reduction itself but
231 (vect_recog_dot_prod_pattern): Don't check whether the statement
232 is part of a group here.
233 (vect_recog_sad_pattern): Likewise.
234 (vect_recog_widen_sum_pattern): Likewise.
236 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
239 * final.c (rest_of_clean_state): Also look for calls inside sequences.
241 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
244 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
246 (ix86_output_indirect_function_return): New prototype.
247 (ix86_split_simple_return_pop_internal): Likewise.
248 * config/i386/i386.c (indirect_return_via_cx): New.
249 (indirect_return_via_cx_bnd): Likewise.
250 (indirect_thunk_name): Handle return va CX_REG.
251 (output_indirect_thunk_function): Create alias for
252 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
253 (ix86_output_indirect_jmp): Remove the bool argument.
254 (ix86_output_indirect_function_return): New function.
255 (ix86_split_simple_return_pop_internal): Likewise.
256 * config/i386/i386.md (*indirect_jump): Don't pass false
257 to ix86_output_indirect_jmp.
258 (*tablejump_1): Likewise.
259 (simple_return_pop_internal): Change it to define_insn_and_split.
260 Call ix86_split_simple_return_pop_internal to split it for
262 (simple_return_indirect_internal): Call
263 ix86_output_indirect_function_return instead of
264 ix86_output_indirect_jmp.
266 2018-02-26 Jakub Jelinek <jakub@redhat.com>
269 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
270 memset and value initialization afterwards.
272 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
274 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
276 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
279 * common/config/aarch64/aarch64-common.c
280 (aarch_option_optimization_table[]): Switch
281 off fomit-frame-pointer
283 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
284 Chung-Ju Wu <jasonwucj@gmail.com>
286 * config/nds32/nds32-multiple.md (load_multiple): Disallow
288 (store_multiple): Ditto.
290 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
292 * config.gcc: Add --with-cpu support for nds32 target.
293 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
294 * config/nds32/nds32.opt: Add -mcpu= option.
296 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
298 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
299 isel=yes): Warn for these deprecated options.
301 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
303 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
304 ISA_2_5_MASKS_EMBEDDED.
306 2018-02-23 Jakub Jelinek <jakub@redhat.com>
308 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
309 p->max as pointers rather than using iterative_hash_expr.
311 2018-02-23 Carl Love <cel@us.ibm.com>
313 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
314 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
316 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
317 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
318 P8V_BUILTIN_VEC_VUNSIGNED2.
320 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
323 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
324 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
326 * lra-constraints.c (curr_insn_transform): Set up
327 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
329 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
330 * lra-eliminations.c (spill_pseudos): Ditto.
331 (process_insn_for_elimination): Ditto.
332 * lra-lives.c (reg_early_clobber_p): Use the new macros.
333 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
334 LRA_NON_CLOBBERED_ALT.
336 2018-02-22 Martin Sebor <msebor@redhat.com>
338 PR tree-optimization/84480
339 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
340 to maybe_diag_stxncpy_trunc. Call it.
341 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
342 from gimple_fold_builtin_strcpy. Print inlining stack.
343 (handle_builtin_stxncpy): Print inlining stack.
344 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
346 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
349 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
350 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
351 and -fcheck-pointer-bounds are used together.
352 (indirect_thunk_prefix): New enum.
353 (indirect_thunk_need_prefix): New function.
354 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
355 "_nt" instead of "_bnd" for NOTRACK prefix.
356 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
357 (output_indirect_thunk_function): Likewise.
359 (ix86_code_end): Update output_indirect_thunk_function calls.
360 (ix86_output_indirect_branch_via_reg): Replace
361 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
362 (ix86_output_indirect_branch_via_push): Likewise.
363 (ix86_output_function_return): Likewise.
364 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
365 incompatible with -fcf-protection=branch and
366 -fcheck-pointer-bounds.
368 2018-02-22 Steve Ellcey <sellcey@cavium.com>
371 * config/aarch64/aarch64.c (aarch64_print_address_internal):
372 Change gcc_assert call to output_operand_lossage.
374 2018-02-22 Steve Ellcey <sellcey@cavium.com>
376 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
378 2018-02-22 DJ Delorie <dj@redhat.com>
379 Sebastian Perta <sebastian.perta@renesas.com>
380 Oleg Endo <olegendo@gcc.gnu.org>
382 * config/rx/rx.c (rx_rtx_costs): New function.
383 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
385 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
387 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
389 2018-02-22 Martin Liska <mliska@suse.cz>
392 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
393 Add "native" as a possible value.
395 2018-02-22 Martin Liska <mliska@suse.cz>
398 * config/i386/i386.c (ix86_option_override_internal):
399 Add "native" as a possible value for -march and -mtune.
401 2018-02-22 Jakub Jelinek <jakub@redhat.com>
404 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
405 to all type variants.
407 PR tree-optimization/84503
408 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
409 width as info->bitpos + info->bitsize - start.
410 (merged_store_group::merge_overlapping): Simplify width computation.
411 (check_no_overlap): New function.
412 (imm_store_chain_info::try_coalesce_bswap): Compute expected
413 start + width and last_order of the group, fail if check_no_overlap
415 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
416 to group if check_no_overlap fails.
418 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
420 * config/rs6000/altivec.md: Delete contraint arguments to
421 define_expand, define_split, and define_peephole2, and in
422 define_insn_and_split if always unused.
423 * config/rs6000/darwin.md: Ditto.
424 * config/rs6000/dfp.md: Ditto.
425 * config/rs6000/rs6000.md: Ditto.
426 * config/rs6000/sync.md: Ditto.
427 * config/rs6000/vector.md: Ditto.
428 * config/rs6000/vsx.md: Ditto.
430 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
432 * config/rs6000/altivec.md: Write output control strings as braced
433 blocks instead of double-quoted strings.
434 * config/rs6000/darwin.md: Ditto.
435 * config/rs6000/rs6000.md: Ditto.
436 * config/rs6000/vector.md: Ditto.
437 * config/rs6000/vsx.md: Ditto.
439 2018-02-21 Jason Merrill <jason@redhat.com>
441 PR c++/84314 - ICE with templates and fastcall attribute.
442 * attribs.c (build_type_attribute_qual_variant): Remove assert.
444 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
446 * ipa-cp.c (determine_versionability): Fix comment typos.
448 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
451 * ipa-cp.c (determine_versionability): Do not version functions caling
454 2018-02-21 Martin Liska <mliska@suse.cz>
457 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
458 Add "native" as a possible value.
459 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
460 the macro when native cpu detection is available.
462 2018-02-21 Martin Liska <mliska@suse.cz>
465 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
466 Add "native" as a possible value.
467 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
468 when native cpu detection is available.
470 2018-02-21 Jakub Jelinek <jakub@redhat.com>
471 Martin Sebor <msebor@redhat.com>
473 PR tree-optimization/84478
474 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
476 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
477 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
478 support which is conservatively correct, for 2 only stay conservative
479 for maxlen. Formatting and comment capitalization fixes. Add STRICT
480 argument to the 2 argument get_range_strlen, adjust 6 arg
481 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
483 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
484 (gimple_fold_builtin_strlen): Pass true as last argument to
487 2018-02-20 Martin Sebor <msebor@redhat.com>
490 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
491 (builtin_memref::set_base_and_offset): Same. Handle inner references.
492 (builtin_memref::builtin_memref): Factor out parts into
493 set_base_and_offset and call it.
495 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
498 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
499 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
500 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
501 search at the associated MODE_INT.
503 2018-02-20 Jeff Law <law@redhat.com>
506 PR tree-optimization/81592
508 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
509 for range data rather than using global data.
510 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
511 range data rather than using global data.
512 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
513 pass it to children as needed.
514 (struct directive::fmtresult): Similarly.
515 (struct directive::set_width): Similarly.
516 (struct directive::set_precision): Similarly.
517 (format_integer, format_directive, parse_directive): Similarly.
518 (format_none): Accept unnamed vr_values parameter.
519 (format_percent, format_floating, format_character): Similarly.
520 (format_string, format_plain): Similarly.
521 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
522 the EVRP range analyzer for range data rather than using global data.
523 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
524 gimple-ssa-evrp-analyze.h
525 (class sprintf_dom_walker): Add after_dom_children member function.
526 Add evrp_range_analyzer member.
527 (sprintf_dom_walker::before_dom_children): Call into the EVRP
528 range analyzer as needed.
529 (sprintf_dom_walker::after_dom_children): New member function.
530 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
532 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
533 (evrp_range_analyzer::pop_to_marker): Likewise.
535 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
537 PR tree-optimization/84419
538 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
539 with the required type if its current type is compatible but
542 2018-02-20 Jakub Jelinek <jakub@redhat.com>
545 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
548 2018-02-20 Martin Liska <mliska@suse.cz>
551 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
552 possible values if we don't have a hint.
554 2018-02-20 Martin Liska <mliska@suse.cz>
558 * final.c (shorten_branches): Build align_tab array with one
560 * opts.c (finish_options): Add alignment option limit check.
561 (MAX_CODE_ALIGN): Likewise.
562 (MAX_CODE_ALIGN_VALUE): Likewise.
563 * doc/invoke.texi: Document maximum allowed option value for
564 all -falign-* options.
566 2018-02-19 Jakub Jelinek <jakub@redhat.com>
569 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
570 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
571 * var-tracking.c (emit_note_insn_var_location): Remove all references
572 to NOTE_INSN_CALL_ARG_LOCATION.
573 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
574 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
575 Use copy_rtx_if_shared.
576 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
577 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
578 (dwarf2out_var_location): Remove handling of
579 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
581 * final.c (final_scan_insn): Remove all references to
582 NOTE_INSN_CALL_ARG_LOCATION.
583 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
584 before dumping final insns.
585 * except.c (emit_note_eh_region_end): Remove all references to
586 NOTE_INSN_CALL_ARG_LOCATION.
587 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
588 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
589 * config/arc/arc.c (hwloop_optimize): Likewise.
590 * config/arm/arm.c (create_fix_barrier): Likewise.
591 * config/s390/s390.c (s390_chunkify_start): Likewise.
592 * config/sh/sh.c (find_barrier): Likewise.
593 * config/i386/i386.c (rest_of_insert_endbranch,
594 ix86_seh_fixup_eh_fallthru): Likewise.
595 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
596 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
597 * config/frv/frv.c (frv_function_prologue): Likewise.
598 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
600 (note_outside_basic_block_p): Remove all references to
601 NOTE_INSN_CALL_ARG_LOCATION.
602 * gengtype.c (adjust_field_rtx_def): Likewise.
603 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
605 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
606 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
609 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
612 PR tree-optimization/84452
613 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
614 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
617 2018-02-19 Martin Liska <mliska@suse.cz>
620 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
622 2018-02-19 Martin Liska <mliska@suse.cz>
623 Richard Sandiford <richard.sandiford@linaro.org>
625 PR tree-optimization/82491
626 * gimple-fold.c (get_base_constructor): Make earlier bail out
629 2018-02-19 Carl Love <cel@us.ibm.com>
631 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
632 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
634 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
637 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
639 * config/rl78/rl78.md (movdf): New define expand.
641 2018-02-19 Martin Liska <mliska@suse.cz>
644 * doc/invoke.texi: Fix typo.
645 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
647 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
649 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
650 handle rs6000_single_float and rs6000_double_float specially for
653 2018-02-16 Jeff Law <law@redhat.com>
655 * config/rx/rx.c (add_pop_cfi_notes): New function.;
658 2018-02-16 Jakub Jelinek <jakub@redhat.com>
661 * ipa-inline.c (inline_small_functions): Fix a typo.
663 2018-02-16 Nathan Sidwell <nathan@acm.org>
665 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
667 2018-02-16 Carl Love <cel@us.ibm.com>
669 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
670 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
671 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
672 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
673 expansion to P8V_BUILTIN_VEC_FLOAT2.
675 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
677 PR rtl-optimization/70023
678 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
679 src_regno into account.
681 2018-02-16 Carl Love <cel@us.ibm.com>
683 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
684 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
685 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
686 * config/rs6000/rs6000.c: Remove case statements for
687 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
688 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
689 and P9V_BUILTIN_VEC_VINSERT4B.
690 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
691 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
692 * config/rs6000/vsx.md:
693 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
696 2018-02-16 Carl Love <cel@us.ibm.com>
698 * config/rs6000/altivec.h: Add builtin names vec_extract4b
700 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
702 * config/rs6000/rs6000-c.c: Add the definitions for
703 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
704 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
705 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
706 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
707 definition for insert4b and define insn *insert3b_internal.
708 * doc/extend.texi: Add documentation for vec_extract4b.
710 2018-02-16 Nathan Sidwell <nathan@acm.org>
712 * doc/extend.texi (Backwards Compatibility): Mention friend
713 injection. Note for-scope is deprecated.
714 * doc/invoke.texi (-ffriend-injection): Deprecate.
716 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
718 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
719 that moved to I2, also allow destinations that are a paradoxical
720 subreg (instead of a normal reg).
722 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
725 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
728 2018-02-16 Richard Biener <rguenther@suse.de>
730 PR tree-optimization/84037
731 PR tree-optimization/84016
733 * config/i386/i386.c (ix86_builtin_vectorization_cost):
734 Adjust vec_construct for the fact we need additional higher latency
735 128bit inserts for AVX256 and AVX512 vector builds.
736 (ix86_add_stmt_cost): Scale vector construction cost for
739 2018-02-16 Richard Biener <rguenther@suse.de>
741 PR tree-optimization/84417
742 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
743 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
744 (non_rewritable_lvalue_p): Likewise, use poly-ints.
746 2018-02-16 Martin Liska <mliska@suse.cz>
749 * internal-fn.def (ASAN_CHECK): Set proper flags.
750 (ASAN_MARK): Likewise.
752 2018-02-16 Julia Koval <julia.koval@intel.com>
754 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
757 2018-02-16 Jakub Jelinek <jakub@redhat.com>
760 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
761 Use ++iter rather than iter++ for std::list iterators.
762 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
763 defer deleting them until all nodes in the forest are processed. Do
764 free even leaf nodes. Change to_process into auto_vec.
767 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
768 * vec.h (vec_default_construct): Use memset instead of placement new
769 if BROKEN_VALUE_INITIALIZATION is defined.
770 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
771 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
774 PR rtl-optimization/83723
775 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
776 * lra.c (lra_substitute_pseudo): Likewise. If true, use
777 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
779 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
781 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
783 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
785 PR rtl-optimization/81443
786 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
787 from inner REGs to paradoxical SUBREGs.
789 2018-02-16 Richard Biener <rguenther@suse.de>
791 PR tree-optimization/84399
792 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
793 For operands we can analyze at their definition make sure we can
794 analyze them at each use as well.
796 2018-02-16 Richard Biener <rguenther@suse.de>
798 PR tree-optimization/84190
799 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
800 volatile accesses if the decl isn't volatile.
802 2018-02-15 Jason Merrill <jason@redhat.com>
804 PR c++/84314 - ICE with templates and fastcall attribute.
805 * attribs.c (build_type_attribute_qual_variant): Don't clobber
806 TYPE_CANONICAL on an existing type.
808 2018-02-15 Jakub Jelinek <jakub@redhat.com>
810 PR tree-optimization/84383
811 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
812 dstoff nor call operand_equal_p if dstbase is NULL.
814 PR tree-optimization/84334
815 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
816 also a CONSTANT_CLASS_P, punt.
818 2018-02-14 Jim Wilson <jimw@sifive.com>
820 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
821 first SMALL_OPERAND check. New local min_second_step. Move assert
822 to where locals are set. Add TARGET_RVC support.
823 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
825 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
827 * doc/invoke.texi: Correct -Wformat-overflow code sample.
829 2018-02-14 Martin Sebor <msebor@redhat.com>
831 PR tree-optimization/83698
832 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
833 arrays constrain the offset range to their bounds.
834 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
835 (builtin_access::overlap): Avoid setting the size of overlap if it's
837 (maybe_diag_overlap): Also consider arrays when deciding what values
838 of offsets to include in diagnostics.
840 2018-02-14 Martin Sebor <msebor@redhat.com>
843 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
844 that correspond to the kind of a declaration.
846 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
849 * config/pa/pa.md: Load address of PIC label using the linkage table
850 if the label is nonlocal.
852 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
854 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
855 warning message if user requests -maltivec=be.
856 * doc/invoke.texi: Document deprecation of -maltivec=be.
858 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
861 * config/rs6000/rs6000-c.c: Update definitions for
862 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
863 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
865 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
868 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
869 add _get_ssp intrinsics. Remove argument from
870 __builtin_ia32_rdssp[d|q].
871 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
872 * config/i386/i386-builtin.def: Remove argument from
873 __builtin_ia32_rdssp[d|q].
874 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
875 ix86_expand_special_args_builtin for _rdssp[d|q].
876 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
877 Clear register before usage.
878 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
879 Add documentation for new _get_ssp and _inc_ssp intrinsics.
881 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
883 PR tree-optimization/84357
884 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
885 operand 1 of an ARRAY_REF too.
887 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
890 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
891 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
893 (set_of_reg): New struct.
894 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
895 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
896 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
898 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
899 Split into bitclr, bitset, bitinvert patterns if appropriate.
900 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
901 use rx_fuse_in_memory_bitop.
902 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
903 to named insn, correct maximum insn length.
905 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
908 * machmode.def: Define a complex mode for PARTIAL_INT.
909 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
911 * doc/rtl.texi: Document CSPImode.
912 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
914 (msp430_hard_regno_nregs_with_padding): Likewise.
916 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
919 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
921 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
923 PR rtl-optimization/84169
924 * combine.c (try_combine): New variable split_i2i3. Set it to true if
925 we generated a parallel as new i3 and we split that to new i2 and i3
926 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
927 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
928 those to i2, not i1. Partially rewrite this scan code.
930 2018-02-13 Jakub Jelinek <jakub@redhat.com>
933 * stor-layout.c (place_field): For variable length fields, adjust
934 offset_align afterwards not just based on the field's alignment,
935 but also on the size.
938 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
939 of exps and logs in the use_exp2 case.
941 2018-02-13 Jeff Law <law@redhat.com>
943 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
946 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
949 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
953 * common.opt (gas-loc-support, gas-locview-support): New.
954 (ginline-points, ginternal-reset-location-views): New.
955 * doc/invoke.texi: Document them. Use @itemx where intended.
956 (gvariable-location-views): Adjust.
957 * target.def (reset_location_view): New.
958 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
959 (TARGET_RESET_LOCATION_VIEW): New.
960 * doc/tm.texi: Rebuilt.
961 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
962 (dwarf2out_default_as_locview_support): New.
963 (output_asm_line_debug_info): Use option variables.
964 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
965 (output_loc_list): Likewise.
966 (add_high_low_attributes): Check option variables.
967 Don't output entry view attribute in strict mode.
968 (gen_inlined_subroutine_die): Check option variables.
969 (dwarf2out_inline_entry): Likewise.
970 (init_sections_and_labels): Likewise.
971 (dwarf2out_early_finish): Likewise.
972 (maybe_reset_location_view): New, from...
973 (dwarf2out_var_location): ... here. Call it.
974 * debug.h (dwarf2out_default_as_loc_support): Declare.
975 (dwarf2out_default_as_locview_support): Declare.
976 * hooks.c (hook_int_rtx_insn_0): New.
977 * hooks.h (hook_int_rtx_insn_0): Declare.
978 * toplev.c (process_options): Take -gas-loc-support and
979 -gas-locview-support from dwarf2out. Enable
980 -gvariable-location-views by default only with locview
981 assembler support. Enable -ginternal-reset-location-views by
982 default only if the target defines the corresponding hook.
983 Enable -ginline-points by default if location views are
984 enabled; force it disabled if statement frontiers are
986 * tree-inline.c (expand_call_inline): Check option variables.
987 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
989 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
991 PR tree-optimization/84321
992 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
993 handling. Also check whether the anti-range contains any values
994 that satisfy the mask; switch to a VR_RANGE if not.
996 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
999 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
1001 2018-02-13 Martin Jambor <mjambor@suse.cz>
1004 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
1005 of call statements, also set location of a load to a temporary.
1007 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
1009 * config/rl78/rl78.c (add_vector_labels): New function.
1010 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
1011 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
1012 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
1013 which checks that no arguments are passed.
1014 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
1015 * doc/extend.texi: Documentation for the new attribute.
1017 2018-02-13 Andreas Schwab <schwab@suse.de>
1019 * config/riscv/linux.h (CPP_SPEC): Define.
1021 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1024 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
1025 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
1026 OPTION_MASK_ISA_AES as first argument to def_builtin_const
1027 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
1028 instead of OPTION_MASK_ISA_PCLMUL as first argument to
1029 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
1030 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
1031 temporarily for AES and PCLMUL builtins.
1033 PR tree-optimization/84339
1034 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
1035 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
1039 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
1040 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
1041 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
1043 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
1045 * omp-simd-clone.h: New file.
1046 * omp-simd-clone.c: Include omp-simd-clone.h.
1047 (expand_simd_clones): No longer static.
1048 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
1049 cgraph.h and omp-simd-clone.h.
1050 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
1051 (vect_recog_widen_shift_pattern): Formatting fix.
1052 (vect_pattern_recog_1): Don't check optab for calls.
1055 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
1056 operands[2] into a REG before using gen_lowpart on it.
1058 2018-02-12 Jeff Law <law@redhat.com>
1061 * config/sh/sh.c (find_barrier): Consider a sibling call
1064 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
1065 successfully back substituting a reg.
1067 2018-02-12 Richard Biener <rguenther@suse.de>
1069 PR tree-optimization/84037
1070 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
1071 parameter, move visited init to caller.
1072 (vect_slp_analyze_operations): Separate cost from validity
1073 check, initialize visited once for all instances.
1074 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
1076 * tree-vect-stmts.c (vect_model_simple_cost): Make early
1078 (vect_model_promotion_demotion_cost): Likewise.
1079 (vectorizable_bswap): Guard cost modeling with !slp_node
1080 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
1082 (vectorizable_call): Likewise.
1083 (vectorizable_conversion): Likewise.
1084 (vectorizable_assignment): Likewise.
1085 (vectorizable_shift): Likewise.
1086 (vectorizable_operation): Likewise.
1087 (vectorizable_store): Likewise.
1088 (vectorizable_load): Likewise.
1089 (vectorizable_condition): Likewise.
1090 (vectorizable_comparison): Likewise.
1092 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
1095 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
1096 (ASAN_MARK): Fix fnspec to account for return value, change pointer
1097 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
1099 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1102 * params.def (inline-min-speedup): Increase from 8 to 15.
1103 (max-inline-insns-auto): Decrease from 40 to 30.
1104 * ipa-split.c (consider_split): Add some buffer for function to
1105 be considered inlining candidate.
1106 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
1109 2018-02-12 Richard Biener <rguenther@suse.de>
1111 PR tree-optimization/84037
1112 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
1113 matched stmts if we cannot swap the non-matched ones.
1115 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1117 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
1118 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
1119 _mm_maskz_scalef_round_ss): New intrinsics.
1120 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
1121 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
1122 __builtin_ia32_scalefss_round): Remove.
1123 (__builtin_ia32_scalefsd_mask_round,
1124 __builtin_ia32_scalefss_mask_round): New intrinsics.
1125 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
1126 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
1127 ((match_operand:VF_128 2 "<round_nimm_predicate>"
1128 "<round_constraint>")): Changed to ...
1129 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
1130 "<round_scalar_constraint>")): ... this.
1131 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
1132 %0, %1, %2<round_op3>}"): Changed to ...
1133 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
1134 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
1135 %2<round_scalar_mask_op3>}"): ... this.
1136 * config/i386/subst.md (round_scalar_nimm_predicate): New.
1138 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1140 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
1141 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
1142 (_mm_maskz_sqrt_round_ss): New intrinsics.
1143 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
1144 (__builtin_ia32_sqrtsd_mask_round)
1145 (__builtin_ia32_sqrtss_mask_round): New builtins.
1146 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
1147 (__builtin_ia32_sqrtss_round): Remove.
1148 (__builtin_ia32_sqrtsd_mask_round)
1149 (__builtin_ia32_sqrtss_mask_round): New builtins.
1150 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
1151 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
1152 ((match_operand:VF_128 1 "vector_operand"
1153 "xBm,<round_constraint>")): Changed to ...
1154 ((match_operand:VF_128 1 "vector_operand"
1155 "xBm,<round_scalar_constraint>")): ... this.
1156 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
1157 %0, %2, %<iptr>1<round_op3>}): Changed to ...
1158 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
1159 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
1160 %<iptr>1<round_scalar_mask_op3>}): ... this.
1161 ((set_attr "prefix" "<round_prefix>")): Changed to ...
1162 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
1164 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
1167 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
1168 Cast vec_cmpeq result to correct type.
1169 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
1170 Cast vec_cmpgt result to correct type.
1172 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
1174 * final.c (final_scan_insn_1): Renamed from...
1175 (final_scan_insn): ... this. New wrapper, to recover
1176 seen from the outermost call in recursive ones.
1177 * config/sparc/sparc.c (output_return): Drop seen from call.
1178 (output_sibcall): Likewise.
1179 * config/visium/visium.c (output_branch): Likewise.
1181 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
1183 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
1186 2018-02-10 Alan Modra <amodra@gmail.com>
1189 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
1190 Specify LR as an input.
1192 2018-02-10 Jakub Jelinek <jakub@redhat.com>
1195 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
1196 remove_member_access_dummy_vars): New functions.
1197 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
1198 lower_omp_1, execute_lower_omp): Use them.
1200 PR rtl-optimization/84308
1201 * shrink-wrap.c (spread_components): Release todo vector.
1203 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
1205 PR rtl-optimization/57193
1206 * ira-color.c (struct allocno_color_data): Add member
1207 conflict_allocno_hard_prefs.
1208 (update_conflict_allocno_hard_prefs): New.
1209 (bucket_allocno_compare_func): Add a preference based on
1210 conflict_allocno_hard_prefs.
1211 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
1212 (color_allocnos): Remove a dead code. Initiate
1213 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
1215 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1218 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
1219 constraint from =wa to wa. Avoid a subreg on the output operand,
1220 instead use a pseudo and subreg it in a move.
1221 (p9_xxbrd_<mode>): Changed to ...
1222 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
1223 (p9_xxbrd_v2df): New expander.
1224 (p9_xxbrw_<mode>): Changed to ...
1225 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
1226 (p9_xxbrw_v4sf): New expander.
1228 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1230 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
1232 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
1235 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
1236 multiply in 32-bit mode.
1237 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
1238 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
1241 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1243 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
1244 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
1245 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
1246 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
1248 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
1251 * dwarf2out.c (is_trivial_indirect_ref): New function.
1252 (dwarf2out_late_global_decl): Do not generate a location
1253 attribute for variables that have a non-trivial DECL_VALUE_EXPR
1254 and that are not defined in the current unit.
1256 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1258 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
1259 instead of a libcall for UNORDERED.
1261 2018-02-09 Tamar Christina <tamar.christina@arm.com>
1264 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
1265 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
1267 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1270 * config/s390/s390.c (s390_set_current_function): Invoke
1271 s390_indirect_branch_settings also if fndecl didn't change.
1273 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1275 * config/rs6000/rs6000.md (blockage): Set length to zero.
1277 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1279 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
1281 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1284 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
1285 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
1289 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
1290 PARALLEL incoming that failed vt_get_decl_and_offset check.
1293 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
1294 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
1296 (get_variable_section): For decls in named .bss* sections pass true as
1297 second argument to bss_initializer_p.
1299 2018-02-09 Marek Polacek <polacek@redhat.com>
1300 Jakub Jelinek <jakub@redhat.com>
1303 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
1304 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
1305 Sync some changes from cxx_fold_indirect_ref.
1307 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1309 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
1311 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
1312 (BLOCK_INLINE_ENTRY_LABEL): New.
1313 (dwarf2out_var_location): Disregard inline entry markers.
1314 (inline_entry_data): New struct.
1315 (inline_entry_data_hasher): New hashtable type.
1316 (inline_entry_data_hasher::hash): New.
1317 (inline_entry_data_hasher::equal): New.
1318 (inline_entry_data_table): New variable.
1319 (add_high_low_attributes): Add DW_AT_entry_pc and
1320 DW_AT_GNU_entry_view attributes if a pending entry is found
1321 in inline_entry_data_table. Add old entry_pc attribute only
1322 if debug nonbinding markers are disabled.
1323 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
1324 markers are enabled.
1325 (block_within_block_p, dwarf2out_inline_entry): New.
1326 (dwarf2out_finish): Check that no entries remained in
1327 inline_entry_data_table.
1328 * final.c (reemit_insn_block_notes): Handle inline entry notes.
1329 (final_scan_insn, notice_source_line): Likewise.
1330 (rest_of_clean_state): Skip inline entry markers.
1331 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
1333 * gimple.c (gimple_build_debug_inline_entry): New.
1334 * gimple.h (enum gimple_debug_subcode): Add
1335 GIMPLE_DEBUG_INLINE_ENTRY.
1336 (gimple_build_debug_inline_entry): Declare.
1337 (gimple_debug_inline_entry_p): New.
1338 (gimple_debug_nonbind_marker_p): Adjust.
1339 * insn-notes.def (INLINE_ENTRY): New.
1340 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
1341 inline entry marker notes.
1342 (print_insn): Likewise.
1343 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
1344 (INSN_DEBUG_MARKER_KIND): Likewise.
1345 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
1346 * tree-inline.c (expand_call_inline): Build and insert
1347 debug_inline_entry stmt.
1348 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
1349 inline entry blocks early, if nonbind markers are enabled.
1350 (dump_scope_block): Dump fragment info.
1351 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
1352 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
1353 (gimple_build_debug_inline_entry): New.
1354 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
1355 Enable/disable inline entry points too.
1356 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
1357 (DEBUG_INSN): Describe inline entry markers.
1359 * common.opt (gvariable-location-views): New.
1360 (gvariable-location-views=incompat5): New.
1361 * config.in: Rebuilt.
1362 * configure: Rebuilt.
1363 * configure.ac: Test assembler for view support.
1364 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
1365 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
1366 * dwarf2out.c (var_loc_view): New typedef.
1367 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
1368 (dwarf2out_locviews_in_attribute): New.
1369 (dwarf2out_locviews_in_loclist): New.
1370 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
1371 (enum dw_line_info_opcode): Add LI_adv_address.
1372 (struct dw_line_info_table): Add view.
1373 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
1374 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
1375 (zero_view_p): New variable.
1376 (ZERO_VIEW_P): New macro.
1377 (output_asm_line_debug_info): New.
1378 (struct var_loc_node): Add view.
1379 (add_AT_view_list, AT_loc_list): New.
1380 (add_var_loc_to_decl): Add view param. Test it against last.
1381 (new_loc_list): Add view params. Record them.
1382 (AT_loc_list_ptr): Handle loc and view lists.
1383 (view_list_to_loc_list_val_node): New.
1384 (print_dw_val): Handle dw_val_class_view_list.
1385 (size_of_die): Likewise.
1386 (value_format): Likewise.
1387 (loc_list_has_views): New.
1388 (gen_llsym): Set vl_symbol too.
1389 (maybe_gen_llsym, skip_loc_list_entry): New.
1390 (dwarf2out_maybe_output_loclist_view_pair): New.
1391 (output_loc_list): Output view list or entries too.
1392 (output_view_list_offset): New.
1393 (output_die): Handle dw_val_class_view_list.
1394 (output_dwarf_version): New.
1395 (output_compilation_unit_header): Use it.
1396 (output_skeleton_debug_sections): Likewise.
1397 (output_rnglists, output_line_info): Likewise.
1398 (output_pubnames, output_aranges): Update version comments.
1399 (output_one_line_info_table): Output view numbers in asm comments.
1400 (dw_loc_list): Determine current endview, pass it to new_loc_list.
1401 Call maybe_gen_llsym.
1402 (loc_list_from_tree_1): Adjust.
1403 (add_AT_location_description): Create view list attribute if
1404 needed, check it's absent otherwise.
1405 (convert_cfa_to_fb_loc_list): Adjust.
1406 (maybe_emit_file): Call output_asm_line_debug_info for test.
1407 (dwarf2out_var_location): Reset views as needed. Precompute
1408 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
1409 attribute. Set view.
1410 (new_line_info_table): Reset next view.
1411 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
1412 (dwarf2out_source_line): Likewise. Output view resets and labels to
1413 the assembler, or select appropriate line info opcodes.
1414 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
1415 (optimize_string_length): Catch it. Adjust.
1416 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
1417 dw_val_class_view_list, and remove it if no longer needed.
1418 (hash_loc_list): Hash view numbers.
1419 (loc_list_hasher::equal): Compare them.
1420 (optimize_location_lists): Check whether a view list symbol is
1421 needed, and whether the locview attribute is present, and
1422 whether they match. Remove the locview attribute if no longer
1424 (index_location_lists): Call skip_loc_list_entry for test.
1425 (dwarf2out_finish): Call output_asm_line_debug_info for test.
1426 Use output_dwarf_version.
1427 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
1428 (struct dw_val_node): Add val_view_list.
1429 * final.c (SEEN_NEXT_VIEW): New.
1430 (set_next_view_needed): New.
1431 (clear_next_view_needed): New.
1432 (maybe_output_next_view): New.
1433 (final_start_function): Rename to...
1434 (final_start_function_1): ... this. Take pointer to FIRST,
1435 add SEEN parameter. Emit param bindings in the initial view.
1436 (final_start_function): Reintroduce SEEN-less interface.
1437 (final): Rename to...
1438 (final_1): ... this. Take SEEN parameter. Output final pending
1439 next view at the end.
1440 (final): Reintroduce seen-less interface.
1441 (final_scan_insn): Output pending next view before switching
1442 sections or ending a block. Mark the next view as needed when
1443 outputting variable locations. Notify debug backend of section
1444 changes, and of location view changes.
1445 (rest_of_handle_final): Adjust.
1446 * toplev.c (process_options): Autodetect value for debug variable
1447 location views option. Warn on incompat5 without -gdwarf-5.
1448 * doc/invoke.texi (gvariable-location-views): New.
1449 (gvariable-location-views=incompat5): New.
1450 (gno-variable-location-views): New.
1452 2018-02-08 David Malcolm <dmalcolm@redhat.com>
1454 PR tree-optimization/84136
1455 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
1456 that the result of find_edge is non-NULL.
1458 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1461 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
1462 storing integer register in SImode. Fix cost of 256 and 512
1463 byte aligned SSE register store.
1465 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1467 * config/i386/i386.c (ix86_multiplication_cost): Fix
1468 multiplication cost for TARGET_AVX512DQ.
1470 2018-02-08 Marek Polacek <polacek@redhat.com>
1472 PR tree-optimization/84238
1473 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
1476 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1478 PR tree-optimization/84265
1479 * tree-vect-stmts.c (vectorizable_store): Don't treat
1480 VMAT_CONTIGUOUS accesses as grouped.
1481 (vectorizable_load): Likewise.
1483 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1485 PR tree-optimization/81635
1486 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
1487 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
1488 (test_round_for_mask): New functions.
1489 (wide_int_cc_tests): Call test_round_for_mask.
1490 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
1491 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
1492 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
1493 range returned by get_range_info.
1495 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1498 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
1499 * symtab.c: Include builtins.h
1500 (symtab_node::output_to_lto_symbol_table_p): Move here
1501 from lto-streamer-out.c:output_symbol_p.
1502 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
1503 (output_symbol_p): Move all logic to symtab.c
1504 (produce_symtab): Update.
1506 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1508 * config/s390/s390-opts.h (enum indirect_branch): Define.
1509 * config/s390/s390-protos.h (s390_return_addr_from_memory)
1510 (s390_indirect_branch_via_thunk)
1511 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
1512 (enum s390_indirect_branch_type): Define.
1513 * config/s390/s390.c (struct s390_frame_layout, struct
1514 machine_function): Remove.
1515 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
1516 (indirect_branch_table_label_no, indirect_branch_table_name):
1518 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
1519 (enum s390_indirect_branch_option): Define.
1520 (s390_return_addr_from_memory): New function.
1521 (s390_handle_string_attribute): New function.
1522 (s390_attribute_table): Add new attribute handler.
1523 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
1524 (s390_indirect_branch_via_thunk): New function.
1525 (s390_indirect_branch_via_inline_thunk): New function.
1526 (s390_function_ok_for_sibcall): When jumping via thunk disallow
1527 sibling call optimization for non z10 compiles.
1528 (s390_emit_call): Force indirect branch target to be a single
1529 register. Add r1 clobber for non-z10 compiles.
1530 (s390_emit_epilogue): Emit return jump via return_use expander.
1531 (s390_reorg): Handle JUMP_INSNs as execute targets.
1532 (s390_option_override_internal): Perform validity checks for the
1533 new command line options.
1534 (s390_indirect_branch_attrvalue): New function.
1535 (s390_indirect_branch_settings): New function.
1536 (s390_set_current_function): Invoke s390_indirect_branch_settings.
1537 (s390_output_indirect_thunk_function): New function.
1538 (s390_code_end): Implement target hook.
1539 (s390_case_values_threshold): Implement target hook.
1540 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
1542 * config/s390/s390.h (struct s390_frame_layout)
1543 (struct machine_function): Move here from s390.c.
1544 (TARGET_INDIRECT_BRANCH_NOBP_RET)
1545 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
1546 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
1547 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
1548 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
1549 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
1550 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
1551 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
1552 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
1553 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
1554 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
1555 (mnemonic attribute): Add values which aren't recognized
1557 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
1558 pattern for branch conversion. Fix mnemonic attribute.
1559 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
1560 indirect branch via thunk if requested.
1561 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
1562 ("*indirect_jump"): Disable for branch conversion using out of
1564 ("indirect_jump_via_thunk<mode>_z10")
1565 ("indirect_jump_via_thunk<mode>")
1566 ("indirect_jump_via_inlinethunk<mode>_z10")
1567 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
1568 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
1569 ("casesi_jump_via_inlinethunk<mode>_z10")
1570 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
1571 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
1572 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
1573 ("*indirect2_jump"): Disable for branch conversion.
1574 ("casesi_jump"): Turn into expander and expand patterns for branch
1576 ("return_use"): New expander.
1577 ("*return"): Emit return via thunk and rename it to ...
1578 ("*return<mode>"): ... this one.
1579 * config/s390/s390.opt: Add new options and and enum for the
1582 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1584 * lra-constraints.c (match_reload): Unconditionally use
1585 gen_lowpart_SUBREG, rather than selecting between that
1586 and equivalent gen_rtx_SUBREG code.
1588 2018-02-08 Richard Biener <rguenther@suse.de>
1590 PR tree-optimization/84233
1591 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
1592 changed flag instead of boguously re-using phi_inserted.
1594 2018-02-08 Martin Jambor <mjambor@suse.cz>
1596 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
1597 static local variables.
1599 2018-02-08 Richard Biener <rguenther@suse.de>
1601 PR tree-optimization/84278
1602 * tree-vect-stmts.c (vectorizable_store): When looking for
1603 smaller vector types to perform grouped strided loads/stores
1604 make sure the mode is supported by the target.
1605 (vectorizable_load): Likewise.
1607 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1609 * config/aarch64/aarch64.c (aarch64_components_for_bb):
1610 Increase LDP/STP opportunities by adding adjacent callee-saves.
1612 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1614 PR rtl-optimization/84068
1615 PR rtl-optimization/83459
1616 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
1618 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
1620 PR tree-optimization/84224
1621 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
1622 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
1625 2018-02-07 Iain Sandoe <iain@codesourcery.com>
1628 * config/rs6000/altivec.md (*restore_world): Remove LR use.
1629 * config/rs6000/predicates.md (restore_world_operation): Adjust op
1630 count, remove one USE.
1632 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
1634 * doc/install.texi (Configuration): Document the
1635 --with-long-double-format={ibm,ieee} PowerPC configuration
1639 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
1640 Convert from define_expand to be define_insn_and_split. Rework
1641 float/double/_Float128 conversions to QI/HI/SImode to work with
1642 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
1643 conversions to QI/HImode types did a store and then a load to
1644 truncate the value. For conversions to VSX registers, don't split
1645 the insn, instead emit the code directly. Use the code iterator
1646 any_fix to combine signed and unsigned conversions.
1647 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
1648 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
1649 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
1650 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
1651 (fix_<mode>di2_hw): Likewise.
1652 (fixuns_<mode>di2_hw): Likewise.
1653 (fix_<mode>si2_hw): Likewise.
1654 (fixuns_<mode>si2_hw): Likewise.
1655 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
1656 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
1657 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
1658 fix<uns>_trunc<SFDF:mode>si2_p8.
1659 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
1661 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
1662 (fix<uns>_<mode>_mem): Likewise.
1663 (fctiw<u>z_<mode>_mem): Likewise.
1664 (fix<uns>_<mode>_mem): Likewise.
1665 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
1666 the register allocator from doing a direct move to the GPRs to do
1667 a store, and instead use the ISA 3.0 store byte/half-word from
1668 vector register instruction. For IEEE 128-bit floating point,
1669 also optimize stores of 32-bit ints.
1670 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
1672 2018-02-07 Alan Hayward <alan.hayward@arm.com>
1674 * genextract.c (push_pathstr_operand): New function to support
1676 (walk_rtx): Call push_pathstr_operand.
1677 (print_path): Support [a-zA-Z].
1679 2018-02-07 Richard Biener <rguenther@suse.de>
1681 PR tree-optimization/84037
1682 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
1683 (cse_and_gimplify_to_preheader): Declare.
1684 (vect_get_place_in_interleaving_chain): Likewise.
1685 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
1687 (_loop_vec_info::~_loop_vec_info): Delete it.
1688 (cse_and_gimplify_to_preheader): New function.
1689 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
1690 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
1691 (vectorizable_load): Likewise. For grouped stores always base
1692 the IV on the first element.
1693 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
1694 condition before gimplifying.
1696 2018-02-07 Jakub Jelinek <jakub@redhat.com>
1698 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
1699 *DIV_EXPR and *MOD_EXPR.
1701 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
1704 * config/i386/i386.c (ix86_option_override_internal): Mask out
1705 the CF_SET bit when checking -fcf-protection.
1707 2018-02-07 Tom de Vries <tom@codesourcery.com>
1710 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
1713 2018-02-07 Richard Biener <rguenther@suse.de>
1715 PR tree-optimization/84204
1716 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
1719 PR tree-optimization/84205
1720 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
1721 special-case isl_ast_op_zdiv_r.
1723 PR tree-optimization/84223
1724 * graphite-scop-detection.c (gather_bbs::before_dom_children):
1725 Only add conditions from within the region.
1726 (gather_bbs::after_dom_children): Adjust.
1728 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
1731 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
1732 * config/avr/avr.md: Only post-reload split REG-REG moves if
1733 either register is REGERAL_REG_P.
1735 2018-02-07 Jakub Jelinek <jakub@redhat.com>
1737 PR tree-optimization/84235
1738 * tree-ssa-scopedtables.c
1739 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
1740 if the subtraction is performed in floating point type where NaNs are
1741 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
1742 build 1. Formatting fix.
1744 2018-02-06 Jakub Jelinek <jakub@redhat.com>
1747 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
1748 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
1749 and skip it regardless of bb boundaries. Use CALL_P macro,
1750 don't test INSN_P (insn) together with CALL_P or JUMP_P check
1751 unnecessarily, formatting fix.
1753 2018-02-06 Michael Collison <michael.collison@arm.com>
1755 * config/arm/thumb2.md:
1756 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
1757 (*thumb_mov_notscc): Ditto.
1759 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
1762 * config/rs6000/rs6000.md (su code attribute): Use "u" for
1763 unsigned_fix, not "s".
1765 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1767 * configure.ac (gcc_fn_eh_frame_ro): New function.
1768 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
1769 correct .eh_frame permissions.
1770 * configure: Regenerate.
1772 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
1774 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
1777 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1779 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1780 Display warning message for -mno-speculate-indirect-jumps.
1782 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
1784 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
1786 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
1788 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
1790 PR tree-optimization/84225
1791 * tree-eh.c (find_trapping_overflow): Only call
1792 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
1794 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1797 * config/i386/i386.c: Reimplement the check of possible options
1798 -mibt/-mshstk conbination. Change error messages.
1799 * doc/invoke.texi: Fix a typo: remove extra '='.
1801 2018-02-06 Marek Polacek <polacek@redhat.com>
1803 PR tree-optimization/84228
1804 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
1806 2018-02-06 Tamar Christina <tamar.christina@arm.com>
1809 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
1810 emitted arch directives.
1811 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
1812 __ARM_FEATURE_COPROC before changing architectures.
1814 2018-02-06 Richard Biener <rguenther@suse.de>
1816 * config/i386/i386.c (print_reg): Fix typo.
1817 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
1819 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
1821 * configure: Regenerate.
1823 2018-02-05 Martin Sebor <msebor@redhat.com>
1825 PR tree-optimization/83369
1826 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
1829 2018-02-05 Martin Liska <mliska@suse.cz>
1831 * doc/invoke.texi: Cherry-pick upstream r323995.
1833 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
1835 * ira.c (ira_init_register_move_cost): Adjust comment.
1837 2018-02-05 Martin Liska <mliska@suse.cz>
1839 PR gcov-profile/84137
1840 * doc/gcov.texi: Fix typo in documentation.
1842 2018-02-05 Martin Liska <mliska@suse.cz>
1844 PR gcov-profile/83879
1845 * doc/gcov.texi: Document necessity of --dynamic-list-data when
1846 using dlopen functionality.
1848 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
1850 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
1851 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
1852 _mm_maskz_range_ss, _mm_mask_range_round_ss,
1853 _mm_maskz_range_round_ss): New intrinsics.
1854 (__builtin_ia32_rangesd128_round)
1855 (__builtin_ia32_rangess128_round): Remove.
1856 (__builtin_ia32_rangesd128_mask_round,
1857 __builtin_ia32_rangess128_mask_round): New builtins.
1858 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
1859 __builtin_ia32_rangess128_round): Remove.
1860 (__builtin_ia32_rangesd128_mask_round,
1861 __builtin_ia32_rangess128_mask_round): New builtins.
1862 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
1863 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
1864 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
1865 "<round_saeonly_constraint>")): Changed to ...
1866 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
1867 "<round_saeonly_scalar_constraint>")): ... this.
1868 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
1869 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
1870 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
1871 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
1872 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
1874 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
1876 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
1878 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
1879 Remove all values except native, 8540 and 8548.
1881 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
1883 * config/i386/i386.c (ix86_output_function_return): Pass
1884 INVALID_REGNUM, instead of -1, as invalid register number to
1885 indirect_thunk_name and output_indirect_thunk.
1887 2018-02-02 Julia Koval <julia.koval@intel.com>
1889 * config.gcc: Add -march=icelake.
1890 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
1891 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
1892 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
1893 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
1894 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
1895 (processor_target_table): Add icelake.
1896 (ix86_option_override_internal): Handle new PTAs.
1897 (get_builtin_code_for_version): Handle icelake.
1898 (M_INTEL_COREI7_ICELAKE): New.
1899 (fold_builtin_cpu): Handle icelake.
1900 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
1901 * doc/invoke.texi: Add -march=icelake.
1903 2018-02-02 Julia Koval <julia.koval@intel.com>
1905 * config/i386/i386.c (ix86_option_override_internal): Change flags type
1906 to wide_int_bitmask.
1907 * wide-int-bitmask.h: New.
1909 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1912 * config/i386/i386.md: Replace Pmode with word_mode in
1913 builtin_setjmp_setup and builtin_longjmp to support x32.
1915 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
1919 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
1921 (rs6000_supported_cpu_names): New static variable.
1922 (linux_cpu_translation_table): Likewise.
1923 (elf_platform) <cpu>: Define new static variable and use it.
1924 Translate kernel AT_PLATFORM name to canonical name if needed.
1925 Error if platform name is unknown.
1927 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
1930 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
1932 2018-02-01 Jeff Law <law@redhat.com>
1935 * config/i386/i386.c (release_scratch_register_on_entry): Add new
1936 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
1937 the scratch if RELEASE_VIA_POP is false.
1938 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
1939 If we have to save a temporary register, decrement SIZE appropriately.
1940 Pass new arguments to release_scratch_register_on_entry.
1941 (ix86_adjust_stack_and_probe): Likewise.
1942 (ix86_emit_probe_stack_range): Pass new arguments to
1943 release_scratch_register_on_entry.
1945 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
1947 PR rtl-optimization/84157
1948 * combine.c (change_zero_ext): Use REG_P predicate in
1949 front of HARD_REGISTER_P predicate.
1951 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
1953 * config/avr/avr.c (avr_option_override): Move disabling of
1954 -fdelete-null-pointer-checks to...
1955 * common/config/avr/avr-common.c (avr_option_optimization_table):
1958 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1960 PR tree-optimization/81635
1961 * tree-data-ref.c (split_constant_offset_1): For types that
1962 wrap on overflow, try to use range info to prove that wrapping
1965 2018-02-01 Renlin Li <renlin.li@arm.com>
1968 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
1970 (aarch64_register_move_cost): Likewise.
1971 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
1973 (REG_CLASS_NAMES): Likewise.
1974 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
1975 TAILCALL_ADDR_REGS. Remove IP registers.
1976 * config/aarch64/aarch64.md (Ucs): Update register constraint.
1978 2018-02-01 Richard Biener <rguenther@suse.de>
1980 * domwalk.h (dom_walker::dom_walker): Add additional constructor
1981 for specifying RPO order and allow NULL for that.
1982 * domwalk.c (dom_walker::dom_walker): Likewise.
1983 (dom_walker::walk): Handle NULL RPO order.
1984 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
1986 (rewrite_update_dom_walker): Likewise.
1987 (mark_def_dom_walker): Likewise.
1989 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1991 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
1992 (aarch64_maybe_expand_sve_subreg_move): Declare.
1993 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
1994 * config/aarch64/predicates.md (aarch64_any_register_operand): New
1996 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
1997 that are semantically a reverse operation.
1998 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
1999 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
2000 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
2002 (aarch64_can_change_mode_class): For big-endian, forbid changes
2003 between two SVE modes if they have different element sizes.
2005 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2007 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
2008 the TImode handling for big-endian targets.
2010 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2012 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
2013 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
2015 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
2016 Remove BSWAP handing for big-endian targets and use the form of
2017 LD1RQ appropariate for the mode.
2019 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2021 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
2022 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
2025 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2028 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
2029 check for operands that need to go through aarch64_sve_reload_be.
2031 2018-02-01 Jakub Jelinek <jakub@redhat.com>
2033 PR tree-optimization/81661
2034 PR tree-optimization/84117
2035 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
2036 * tree-eh.c: Include gimplify.h.
2037 (find_trapping_overflow, replace_trapping_overflow,
2038 rewrite_to_non_trapping_overflow): New functions.
2039 * tree-vect-loop.c: Include tree-eh.h.
2040 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
2041 * tree-data-ref.c: Include tree-eh.h.
2042 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
2044 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
2046 PR rtl-optimization/84123
2047 * combine.c (change_zero_ext): Check if hard register satisfies
2048 can_change_dest_mode before calling gen_lowpart_SUBREG.
2050 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
2053 * ira.c (ira_init_register_move_cost): Remove assert.
2055 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2057 PR rtl-optimization/84071
2058 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
2059 * doc/tm.texi: Regenerate.
2061 2018-01-31 Richard Biener <rguenther@suse.de>
2063 PR tree-optimization/84132
2064 * tree-data-ref.c (analyze_miv_subscript): Properly
2065 check whether evolution_function_is_affine_multivariate_p
2066 before calling gcd_of_steps_may_divide_p.
2068 2018-01-31 Julia Koval <julia.koval@intel.com>
2071 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
2072 * config/i386/i386.md (rdpid_rex64) New.
2073 (rdpid): Make 32bit only.
2075 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
2078 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
2079 an IDENTIFIER_NODE for FUNCTION_TYPE's.
2081 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2084 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
2086 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
2088 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2090 PR rtl-optimization/84071
2091 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
2092 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
2094 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2096 * config/arc/arc.c (arc_handle_aux_attribute): New function.
2097 (arc_attribute_table): Add 'aux' attribute.
2098 (arc_in_small_data_p): Consider aux like variables.
2099 (arc_is_aux_reg_p): New function.
2100 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
2101 (arc_get_aux_arg): New function.
2102 (prepare_move_operands): Handle aux-register access.
2103 (arc_handle_aux_attribute): New function.
2104 * doc/extend.texi (ARC Variable attributes): Add subsection.
2106 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2108 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
2109 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
2110 (arc_attribute_table): Add 'uncached' attribute.
2111 (arc_print_operand): Print '.di' flag for uncached memory
2113 (arc_in_small_data_p): Do not consider for small data the uncached
2115 (arc_is_uncached_mem_p): New function.
2116 * config/arc/predicates.md (compact_store_memory_operand): Check
2117 for uncached memory accesses.
2118 (nonvol_nonimm_operand): Likewise.
2119 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
2121 2018-01-31 Jakub Jelinek <jakub@redhat.com>
2124 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
2125 falign-loops=): Add Optimization flag.
2127 2018-01-30 Jeff Law <law@redhat.com>
2130 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
2131 INT_REGISTERS_SAVED. Check it prior to calling
2132 get_scratch_register_on_entry.
2133 (ix86_adjust_stack_and_probe): Similarly.
2134 (ix86_emit_probe_stack_range): Similarly.
2135 (ix86_expand_prologue): Corresponding changes.
2137 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2140 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
2141 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
2143 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
2146 * lra-constraints.c (curr_insn_transform): Process AND in the
2149 2018-01-30 Jakub Jelinek <jakub@redhat.com>
2151 PR rtl-optimization/83986
2152 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
2153 dependence against last_pending_memory_flush in addition to
2156 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
2158 PR tree-optimization/81611
2159 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
2162 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2165 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
2168 2018-01-30 Richard Biener <rguenther@suse.de>
2169 Jakub Jelinek <jakub@redhat.com>
2171 PR tree-optimization/84111
2172 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
2173 inner loops added during recursion, as they don't have up-to-date
2176 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2179 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
2180 (can_inline_edge_by_limits_p): ... here.
2181 (can_early_inline_edge_p, check_callers,
2182 update_caller_keys, update_callee_keys, recursive_inlining,
2183 add_new_edges_to_heap, speculation_useful_p,
2184 inline_small_functions,
2185 inline_small_functions, flatten_function,
2186 inline_to_all_callers_1): Update.
2188 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2190 * profile-count.c (profile_count::combine_with_ipa_count): Handle
2193 2018-01-30 Richard Biener <rguenther@suse.de>
2195 PR tree-optimization/83008
2196 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
2197 invariant and constant vector uses in stmts when they need
2200 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2203 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
2204 * configure: Regenerate.
2206 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2208 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
2210 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
2211 Use gen_rtx_REG rather than gen_lowpart.
2213 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2215 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
2216 rather than 0 when creating partial subregs.
2218 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2220 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
2223 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
2226 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
2227 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
2228 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
2229 flags. This restores the settings used before the 2017-07-24.
2230 Turning off pre increment/decrement/modify allows IVOPTS to
2231 optimize DF/SF loops where the index is an int.
2233 2018-01-29 Richard Biener <rguenther@suse.de>
2234 Kelvin Nilsen <kelvin@gcc.gnu.org>
2237 * tree-vect-stmts.c (vectorizable_call): Don't call
2238 targetm.vectorize_builtin_md_vectorized_function if callee is
2241 2018-01-22 Carl Love <cel@us.ibm.com>
2243 * doc/extend.tex: Fix typo in second arg in
2244 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
2246 2018-01-29 Richard Biener <rguenther@suse.de>
2248 PR tree-optimization/84086
2249 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
2250 (flush_ssaname_freelist): When SSA names were released reset
2251 the SCEV hash table.
2253 2018-01-29 Richard Biener <rguenther@suse.de>
2255 PR tree-optimization/84057
2256 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
2257 removed paths when removing edges.
2259 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
2261 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
2262 -mfunction-return=@var{choice}.
2264 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
2267 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
2269 (layout::print_source_line): Likewise.
2270 (test_get_line_width_without_trailing_whitespace): Add test cases.
2272 2018-01-27 Jakub Jelinek <jakub@redhat.com>
2275 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
2278 2018-01-26 Jim Wilson <jimw@sifive.com>
2280 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
2282 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
2285 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2287 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
2288 and CMP + SUB-immediate -> SUBS.
2290 2018-01-26 Martin Sebor <msebor@redhat.com>
2292 PR tree-optimization/83896
2293 * tree-ssa-strlen.c (get_string_len): Rename...
2294 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
2295 Avoid assuming length is constant.
2296 (handle_char_store): Use HOST_WIDE_INT for string length.
2298 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
2301 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
2302 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
2304 2018-01-26 Richard Biener <rguenther@suse.de>
2306 PR rtl-optimization/84003
2307 * dse.c (record_store): Only record redundant stores when
2308 the earlier store aliases at least all accesses the later one does.
2310 2018-01-26 Jakub Jelinek <jakub@redhat.com>
2312 PR rtl-optimization/83985
2313 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
2314 REG_CFA_RESTORE insns.
2315 (delete_unmarked_insns): Don't ignore separate shrink wrapping
2316 REG_CFA_RESTORE insns here.
2319 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
2320 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
2322 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2324 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
2325 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
2326 (arc_init): Likewise.
2327 (arc_override_options): Likewise.
2328 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
2330 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
2332 * config/arc/arc.h (TARGET_DBNZ): Define.
2333 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
2334 properly set the tune attribute.
2335 (dbnz): Use TARGET_DBNZ guard.
2336 * config/arc/arc.opt (mtune): Add core3 option.
2338 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2340 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
2341 recognize new pic like addresses.
2342 (arc_delegitimize_address): Clean up.
2344 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2346 * config/arc/arc-arches.def: Option mrf16 valid for all
2348 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
2349 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
2350 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
2351 * config/arc/arc-tables.opt: Regenerate.
2352 * config/arc/arc.c (arc_conditional_register_usage): Handle
2353 reduced register file case.
2354 (arc_file_start): Set must have build attributes.
2355 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
2357 * config/arc/arc.opt (mrf16): Add new option.
2358 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
2359 * config/arc/genmultilib.awk: Handle new mrf16 option.
2360 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
2361 * config/arc/t-multilib: Regenerate.
2362 * doc/invoke.texi (ARC Options): Document mrf16 option.
2364 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2366 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
2367 * config/arc/arc.c (arc_handle_secure_attribute): New function.
2368 (arc_attribute_table): Add 'secure_call' attribute.
2369 (arc_print_operand): Print secure call operand.
2370 (arc_function_ok_for_sibcall): Don't optimize tail calls when
2372 (arc_is_secure_call_p): New function. * config/arc/arc.md
2373 (call_i): Add support for sjli instruction.
2374 (call_value_i): Likewise.
2375 * config/arc/constraints.md (Csc): New constraint.
2377 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2378 John Eric Martin <John.Martin@emmicro-us.com>
2380 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
2381 * config/arc/arc.c (_arc_jli_section): New struct.
2382 (arc_jli_section): New type.
2383 (rc_jli_sections): New static variable.
2384 (arc_handle_jli_attribute): New function.
2385 (arc_attribute_table): Add jli_always and jli_fixed attribute.
2386 (arc_file_end): New function.
2387 (TARGET_ASM_FILE_END): Define.
2388 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
2389 (arc_add_jli_section): New function.
2390 (jli_call_scan): Likewise.
2391 (arc_reorg): Call jli_call_scan.
2392 (arc_output_addsi): Remove 'S' from printing asm operand.
2393 (arc_is_jli_call_p): New function.
2394 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
2396 (movhi_insn): Likewise.
2397 (movsi_insn): Likewise.
2398 (movsi_set_cc_insn): Likewise.
2399 (loadqi_update): Likewise.
2400 (load_zeroextendqisi_update): Likewise.
2401 (load_signextendqisi_update): Likewise.
2402 (loadhi_update): Likewise.
2403 (load_zeroextendhisi_update): Likewise.
2404 (load_signextendhisi_update): Likewise.
2405 (loadsi_update): Likewise.
2406 (loadsf_update): Likewise.
2407 (movsicc_insn): Likewise.
2408 (bset_insn): Likewise.
2409 (bxor_insn): Likewise.
2410 (bclr_insn): Likewise.
2411 (bmsk_insn): Likewise.
2412 (bicsi3_insn): Likewise.
2413 (cmpsi_cc_c_insn): Likewise.
2414 (movsi_ne): Likewise.
2415 (movsi_cond_exec): Likewise.
2416 (clrsbsi2): Likewise.
2427 (call_i): Remove 'S' asm letter, add jli instruction.
2428 (call_value_i): Likewise.
2429 * config/arc/arc.op (mjli-always): New option.
2430 * config/arc/constraints.md (Cji): New constraint.
2431 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
2433 (subsf3_fpx): Likewise.
2434 (mulsf3_fpx): Likewise.
2435 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
2437 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
2439 * doc/invoke.texi (ARC): Document mjli-always option.
2441 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
2443 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
2444 avoid addition with 0 and use incw and decw where possible.
2446 2018-01-26 Richard Biener <rguenther@suse.de>
2448 PR tree-optimization/81082
2449 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
2450 association if it requires casting to unsigned.
2451 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
2452 from fold_plusminus_mult_expr to catch important cases late when
2453 range info is available.
2455 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2457 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
2458 * configure.ac (hidden_linkonce): New test.
2459 * configure: Regenerate.
2460 * config.in: Regenerate.
2462 2018-01-26 Julia Koval <julia.koval@intel.com>
2464 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
2465 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
2466 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
2467 _mm_mask_bitshuffle_epi64_mask): Fix type.
2468 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
2469 USI_FTYPE_V4DI_V4DI_USI): Remove.
2470 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
2471 __builtin_ia32_vpshufbitqmb256_mask,
2472 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
2473 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
2474 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
2476 2018-01-26 Alan Modra <amodra@gmail.com>
2479 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
2480 UNSPEC_VBPERMQ. Sort other unspecs.
2482 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
2484 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
2486 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
2489 * predict.c (drop_profile): Do not push/pop cfun; update also
2491 (handle_missing_profiles): Fix logic looking for zero profiles.
2493 2018-01-25 Jakub Jelinek <jakub@redhat.com>
2496 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
2497 on functions with #pragma omp declare simd or functions with simd
2499 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
2500 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
2501 Remove trailing \n from warning_at calls.
2503 2018-01-25 Tom de Vries <tom@codesourcery.com>
2506 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2507 for neutered workers.
2509 2018-01-24 Joseph Myers <joseph@codesourcery.com>
2512 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
2513 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
2515 2018-01-24 Jeff Law <law@redhat.com>
2518 * i386.c (get_probe_interval): Move to earlier point.
2519 (ix86_compute_frame_layout): If -fstack-clash-protection and
2520 the frame is larger than the probe interval, then use pushes
2521 to save registers rather than reg->mem moves.
2522 (ix86_expand_prologue): Remove conditional for int_registers_saved
2525 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
2528 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
2529 min/max for never referenced object.
2531 2018-01-24 Jakub Jelinek <jakub@redhat.com>
2534 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
2536 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
2537 attributes from DECL_ATTRIBUTES (decl) without affecting
2538 DECL_ATTRIBUTES (current_function_decl).
2539 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
2540 functions with non-NULL DECL_ABSTRACT_ORIGIN.
2542 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
2544 PR tree-optimization/83979
2545 * fold-const.c (fold_comparison): Use constant_boolean_node
2546 instead of boolean_{true,false}_node.
2548 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
2550 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
2553 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2555 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
2556 Simplify the clause that sets the length attribute.
2557 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
2558 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
2559 clause that sets the length attribute.
2560 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
2562 2018-01-24 Tom de Vries <tom@codesourcery.com>
2565 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
2566 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
2567 Add strict parameter.
2568 (prevent_branch_around_nothing): Insert dummy insn between branch to
2569 label and label with no ptx insn inbetween.
2570 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
2572 2018-01-24 Tom de Vries <tom@codesourcery.com>
2575 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2576 for neutered threads in warp.
2577 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
2579 2018-01-24 Richard Biener <rguenther@suse.de>
2581 PR tree-optimization/83176
2582 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
2585 2018-01-24 Richard Biener <rguenther@suse.de>
2587 PR tree-optimization/82819
2588 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
2589 code generating pluses that are no-ops in the target precision.
2591 2018-01-24 Richard Biener <rguenther@suse.de>
2594 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
2596 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2598 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
2599 to merge probabilities.
2600 * predict.c (probably_never_executed): Also mark as cold functions
2601 with global 0 profile and guessed local profile.
2602 * profile-count.c (profile_probability::combine_with_count): New
2604 * profile-count.h (profile_probability::operator*,
2605 profile_probability::operator*=, profile_probability::operator/,
2606 profile_probability::operator/=): Reduce precision to adjusted
2607 and set value to guessed on contradictory divisions.
2608 (profile_probability::combine_with_freq): Remove.
2609 (profile_probability::combine_wiht_count): Declare.
2610 (profile_count::force_nonzero):: Set to adjusted.
2611 (profile_count::probability_in):: Set quality to adjusted.
2612 * tree-ssa-tail-merge.c (replace_block_by): Use
2615 2018-01-23 Andrew Waterman <andrew@sifive.com>
2616 Jim Wilson <jimw@sifive.com>
2618 * config/riscv/riscv.c (riscv_stack_boundary): New.
2619 (riscv_option_override): Set riscv_stack_boundary. Handle
2620 riscv_preferred_stack_boundary_arg.
2621 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
2622 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
2623 (STACK_BOUNDARY): Set to riscv_stack_boundary.
2624 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
2625 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
2626 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
2628 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
2631 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
2632 of struct ix86_frame.
2633 (ix86_expand_epilogue): Likewise. Add a local variable for
2634 the reg_save_offset field in struct ix86_frame.
2636 2018-01-23 Bin Cheng <bin.cheng@arm.com>
2638 PR tree-optimization/82604
2639 * tree-loop-distribution.c (enum partition_kind): New enum item
2640 PKIND_PARTIAL_MEMSET.
2641 (partition_builtin_p): Support above new enum item.
2642 (generate_code_for_partition): Ditto.
2643 (compute_access_range): Differentiate cases that equality can be
2644 proven at all loops, the innermost loops or no loops.
2645 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
2646 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
2647 (finalize_partitions, distribute_loop): Don't fuse partition of
2648 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
2649 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
2652 2018-01-23 Martin Liska <mliska@suse.cz>
2654 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
2655 order to ignore the predictor.
2656 (PRED_POLYMORPHIC_CALL): Likewise.
2657 (PRED_RECURSIVE_CALL): Likewise.
2659 2018-01-23 Martin Liska <mliska@suse.cz>
2661 * tree-profile.c (tree_profiling): Print function header to
2662 aware reader which function we are working on.
2663 * value-prof.c (gimple_find_values_to_profile): Do not print
2664 not interesting value histograms.
2666 2018-01-23 Martin Liska <mliska@suse.cz>
2668 * profile-count.h (enum profile_quality): Add
2669 profile_uninitialized as the first value. Do not number values
2670 as they are zero based.
2671 (profile_count::verify): Update sanity check.
2672 (profile_probability::verify): Likewise.
2674 2018-01-23 Nathan Sidwell <nathan@acm.org>
2676 * doc/invoke.texi (ffor-scope): Deprecate.
2678 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2680 PR tree-optimization/83510
2681 * domwalk.c (set_all_edges_as_executable): New function.
2682 (dom_walker::dom_walker): Convert bool param
2683 "skip_unreachable_blocks" to enum reachability. Move setup of
2684 edge flags to set_all_edges_as_executable and only do it when
2685 reachability is REACHABLE_BLOCKS.
2686 * domwalk.h (enum dom_walker::reachability): New enum.
2687 (dom_walker::dom_walker): Convert bool param
2688 "skip_unreachable_blocks" to enum reachability.
2689 (set_all_edges_as_executable): New decl.
2690 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
2691 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
2693 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
2694 but converting true to REACHABLE_BLOCKS.
2695 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
2697 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
2698 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
2699 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
2701 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
2702 if check_all_array_refs will be called.
2704 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2706 * tree.c (selftest::test_location_wrappers): Add more test
2709 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2711 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
2712 (selftest::test_bit_in_range): Likewise.
2714 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
2717 * doc/sourcebuild.texi (vect_float): Say that the selector
2718 only describes the situation when -funsafe-math-optimizations is on.
2719 (vect_float_strict): Document.
2721 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
2723 PR tree-optimization/83965
2724 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
2725 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
2726 instead of checking only for a reduction.
2727 (vect_recog_widen_sum_pattern): Likewise.
2729 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2731 * predict.c (probably_never_executed): Only use precise profile info.
2732 (compute_function_frequency): Skip after inlining hack since we now
2733 have quality checking.
2735 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2737 * profile-count.h (profile_probability::very_unlikely,
2738 profile_probability::unlikely, profile_probability::even): Set
2739 precision to guessed.
2741 2018-01-23 Richard Biener <rguenther@suse.de>
2743 PR tree-optimization/83963
2744 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
2745 Properly terminate dominator walk when crossing the exit edge not
2746 when visiting its source block.
2748 2018-01-23 Jakub Jelinek <jakub@redhat.com>
2751 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
2752 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
2754 2018-01-22 Jakub Jelinek <jakub@redhat.com>
2756 PR tree-optimization/83957
2757 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
2758 semicolon after for body surrounded by braces.
2760 PR tree-optimization/83081
2761 * profile-count.h (profile_probability::split): New method.
2762 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
2763 Use profile_probability::split.
2764 (do_compare_rtx_and_jump): Fix adjustment of probabilities
2765 when splitting a single conditional jump into 2.
2767 2018-01-22 David Malcolm <dmalcolm@redhat.com>
2769 PR tree-optimization/69452
2770 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
2773 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2775 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
2776 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
2777 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
2779 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2781 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
2782 * config/rl78/rl78.md (movdi): New define_expand.
2783 * config/rl78/rl78.c (rl78_split_movdi): New function.
2785 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
2788 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
2790 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
2791 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
2792 128-bit to produce an UNSPEC move to get the double word with the
2793 signbit and then a shift directly to do signbit.
2794 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
2795 implementation with a new version that just does either a direct
2796 move or a regular move. Move memory interface to separate insns.
2797 Move insns so they are next to the expander.
2798 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
2799 with signbit move. Split big and little endian case.
2800 (signbit<mode>2_dm_mem_le): Likewise.
2801 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
2802 (signbit<mode>2_dm2): Likewise.
2804 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2806 * config/rl78/rl78.md (anddi3): New define_expand.
2808 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2810 * config/rl78/rl78.md (umindi3): New define_expand.
2812 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2814 * config/rl78/rl78.md (smindi3): New define_expand.
2816 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2818 * config/rl78/rl78.md (smaxdi3): New define_expand.
2820 2018-01-22 Carl Love <cel@us.ibm.com>
2822 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
2823 LVX_V1TI): Add macro expansion.
2824 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
2825 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
2826 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
2827 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
2828 Change check to determine if the instruction is a byte reversing
2829 entry. Fix typo in comment.
2830 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
2831 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
2832 Add def_builtin calls for new builtins.
2833 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
2834 Add define_insn expansion.
2836 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2838 * config/rl78/rl78.md (umaxdi3): New define_expand.
2840 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2842 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
2843 for non-QImode registers.
2845 2018-01-22 Richard Biener <rguenther@suse.de>
2847 PR tree-optimization/83963
2848 * graphite-scop-detection.c (scop_detection::get_sese): Delay
2849 including the loop exit block.
2850 (scop_detection::merge_sese): Likewise.
2851 (scop_detection::add_scop): Do it here instead.
2853 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2855 * doc/sourcebuild.texi (arm_softfloat): Document.
2857 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
2860 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
2861 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
2862 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
2864 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2865 David Edelsohn <dje.gcc@gmail.com>
2868 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
2869 Change "crset eq" to "crset 2".
2870 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
2871 (*call_indirect_aix<mode>_nospec): Likewise.
2872 (*call_value_indirect_aix<mode>_nospec): Likewise.
2873 (*call_indirect_elfv2<mode>_nospec): Likewise.
2874 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
2875 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
2876 change assembly output from . to $.
2877 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
2878 (indirect_jump<mode>_nospec): Change assembly output from . to $.
2879 (*tablejump<mode>_internal1_nospec): Likewise.
2881 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
2884 * config/sh/sh_optimize_sett_clrt.cc:
2885 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
2887 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
2889 PR tree-optimization/83940
2890 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
2891 offset_dt to vect_constant_def rather than vect_unknown_def_type.
2892 (vect_check_load_store_mask): Add a mask_dt_out parameter and
2893 use it to pass back the definition type.
2894 (vect_check_store_rhs): Likewise rhs_dt_out.
2895 (vect_build_gather_load_calls): Add a mask_dt argument and use
2896 it instead of a call to vect_is_simple_use.
2897 (vectorizable_store): Update calls to vect_check_load_store_mask
2898 and vect_check_store_rhs. Use the dt returned by the latter instead
2899 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
2900 instead of calls to vect_is_simple_use. Pass the scalar rather
2901 than the vector operand to vect_is_simple_use when handling
2902 second and subsequent copies of an rhs value.
2903 (vectorizable_load): Update calls to vect_check_load_store_mask
2904 and vect_build_gather_load_calls. Use the cached mask_dt and
2905 gs_info.offset_dt instead of calls to vect_is_simple_use.
2907 2018-01-20 Jakub Jelinek <jakub@redhat.com>
2910 * tree-emutls.c: Include gimplify.h.
2911 (lower_emutls_2): New function.
2912 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
2913 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
2914 it before further processing.
2917 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
2918 UINTVAL (trueop1) instead of INTVAL (op1).
2920 2018-01-19 Jakub Jelinek <jakub@redhat.com>
2924 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
2925 INCOMING_FRAME_SP_OFFSET if not defined.
2926 (scan_trace): Add ENTRY argument. If true and
2927 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
2928 emit a note to adjust the CFA offset.
2929 (create_cfi_notes): Adjust scan_trace callers.
2930 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
2931 INCOMING_FRAME_SP_OFFSET in the CIE.
2932 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
2933 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
2935 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
2936 * doc/tm.texi: Regenerated.
2938 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2940 PR rtl-optimization/83147
2941 * lra-constraints.c (remove_inheritance_pseudos): Use
2942 lra_substitute_pseudo_within_insn.
2944 2018-01-19 Tom de Vries <tom@codesourcery.com>
2945 Cesar Philippidis <cesar@codesourcery.com>
2948 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
2950 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
2953 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
2954 spaces for function labels.
2956 2018-01-19 Martin Liska <mliska@suse.cz>
2958 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
2959 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
2960 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
2961 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
2962 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
2963 (PRED_CONST_RETURN): Change from 69 to 65.
2964 (PRED_NULL_RETURN): Change from 91 to 71.
2965 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
2966 (PRED_LOOP_GUARD): Change from 66 to 73.
2968 2018-01-19 Martin Liska <mliska@suse.cz>
2970 * predict.c (predict_insn_def): Add new assert.
2971 (struct branch_predictor): Change type to signed integer.
2972 (test_prediction_value_range): Amend test to cover
2974 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
2975 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
2976 (PRED_LOOP_ITERATIONS_MAX): Likewise.
2977 (PRED_LOOP_IV_COMPARE): Likewise.
2978 * predict.h (PROB_UNINITIALIZED): Define new constant.
2980 2018-01-19 Martin Liska <mliska@suse.cz>
2982 * predict.c (dump_prediction): Add new format for
2983 analyze_brprob.py script which is enabled with -details
2985 * profile-count.h (precise_p): New function.
2987 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
2989 PR tree-optimization/83922
2990 * tree-vect-loop.c (vect_verify_full_masking): Return false if
2991 there are no statements that need masking.
2992 (vect_active_double_reduction_p): New function.
2993 (vect_analyze_loop_operations): Use it when handling phis that
2994 are not in the loop header.
2996 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
2998 PR tree-optimization/83914
2999 * tree-vect-loop.c (vectorizable_induction): Don't convert
3000 init_expr or apply the peeling adjustment for inductions
3001 that are nested within the vectorized loop.
3003 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3005 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
3008 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3012 * function.h (gimplify_parameters): Add gimple_seq * argument.
3013 * function.c: Include gimple.h and options.h.
3014 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
3015 for the added local temporaries if needed.
3016 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
3017 if there are any parameter cleanups, wrap whole body into a
3018 try/finally with the cleanups.
3020 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
3023 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
3024 Use GET_MODE_CLASS for scalar floating point.
3026 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3030 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3031 Fix call of call_cgraph_insertion_hooks.
3033 2018-01-18 Martin Sebor <msebor@redhat.com>
3035 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
3037 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3040 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
3043 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
3046 * common.opt: (-ffile-prefix-map): New option.
3047 * opts.c (common_handle_option): Defer it.
3048 * opts-global.c (handle_common_deferred_options): Handle it.
3049 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
3050 * file-prefix-map.h: New file.
3051 (remap_debug_filename, add_debug_prefix_map): ...here.
3052 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
3053 * final.c (debug_prefix_map, add_debug_prefix_map
3054 remap_debug_filename): Move to...
3055 * file-prefix-map.c: New file.
3056 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
3057 generalize, get rid of alloca(), use strrchr() instead of strchr().
3058 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
3059 Implement in terms of add_prefix_map().
3060 (remap_macro_filename, remap_debug_filename): Implement in term of
3062 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
3063 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
3064 * dbxout.c: Include file-prefix-map.h.
3065 * varasm.c: Likewise.
3066 * vmsdbgout.c: Likewise.
3067 * xcoffout.c: Likewise.
3068 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
3069 * doc/cppopts.texi (-fmacro-prefix-map): Document.
3070 * doc/invoke.texi (-ffile-prefix-map): Document.
3071 (-fdebug-prefix-map): Update description.
3073 2018-01-18 Martin Liska <mliska@suse.cz>
3075 * config/i386/i386.c (indirect_thunk_name): Document that also
3077 (output_indirect_thunk): Document why both instructions
3078 (pause and lfence) are generated.
3080 2018-01-18 Richard Biener <rguenther@suse.de>
3082 PR tree-optimization/83887
3083 * graphite-scop-detection.c
3084 (scop_detection::get_nearest_dom_with_single_entry): Remove.
3085 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
3086 (scop_detection::merge_sese): Re-implement with a flood-fill
3087 algorithm that properly finds a SESE region if it exists.
3089 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3092 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
3093 pointer_diff optimizations use view_convert instead of convert.
3095 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3097 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3098 Generate different code for -mno-speculate-indirect-jumps.
3099 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3100 (*call_indirect_aix<mode>): Disable for
3101 -mno-speculate-indirect-jumps.
3102 (*call_indirect_aix<mode>_nospec): New define_insn.
3103 (*call_value_indirect_aix<mode>): Disable for
3104 -mno-speculate-indirect-jumps.
3105 (*call_value_indirect_aix<mode>_nospec): New define_insn.
3106 (*sibcall_nonlocal_sysv<mode>): Generate different code for
3107 -mno-speculate-indirect-jumps.
3108 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3110 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3112 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
3113 long double type, set the flags for noting the default long double
3114 type, even if we don't pass or return a long double type.
3116 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
3119 * ipa-inline.c (flatten_function): Do not overwrite final inlining
3122 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
3124 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
3125 support for merge[hl].
3126 (fold_mergehl_helper): New helper function.
3127 (tree-vector-builder.h): New #include for tree_vector_builder usage.
3128 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
3129 (altivec_vmrglw_direct): Add xxmrglw insn.
3131 2018-01-17 Andrew Waterman <andrew@sifive.com>
3133 * config/riscv/riscv.c (riscv_conditional_register_usage): If
3134 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
3136 2018-01-17 David Malcolm <dmalcolm@redhat.com>
3139 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
3140 call the lto_location_cache before reading the
3141 DECL_SOURCE_LOCATION of the types.
3143 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3144 Richard Sandiford <richard.sandiford@linaro.org>
3146 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
3147 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
3148 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
3149 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
3150 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
3152 * config/aarch64/constraints.md (aarch64_movti_operand):
3154 * config/aarch64/predicates.md (Uti): Add new constraint.
3156 2018-01-17 Carl Love <cel@us.ibm.com>
3158 * config/rs6000/vsx.md (define_expand xl_len_r,
3159 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
3160 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
3162 (define_expand, define_insn): Move the shift left from the
3163 define_insn to the define_expand for lxvl and stxvl instructions.
3164 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
3165 and XL_LEN_R definitions to PURE.
3167 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
3169 * config/i386/i386.c (indirect_thunk_name): Declare regno
3170 as unsigned int. Compare regno with INVALID_REGNUM.
3171 (output_indirect_thunk): Ditto.
3172 (output_indirect_thunk_function): Ditto.
3173 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
3174 in the call to output_indirect_thunk_function.
3176 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
3179 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
3180 rather than the size of inner_type to determine the stack slot size
3181 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
3183 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
3186 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
3189 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
3191 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
3192 endian Linux systems to optionally enable multilibs for selecting
3193 the long double type if the user configured an explicit type.
3194 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
3195 have no long double multilibs if not defined.
3196 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
3197 warn if the user used -mabi={ieee,ibm}longdouble and we built
3198 multilibs for long double.
3199 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
3200 appropriate multilib option.
3201 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
3203 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
3204 for building long double multilibs.
3205 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
3207 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
3209 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
3212 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
3214 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
3217 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
3220 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
3223 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3225 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
3226 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
3228 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3230 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
3231 different rtl trees depending on TARGET_64BIT.
3232 (rs6000_gen_lvx): Likewise.
3234 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3236 * config/visium/visium.md (nop): Tweak comment.
3237 (hazard_nop): Likewise.
3239 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3241 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
3242 -mspeculate-indirect-jumps.
3243 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
3244 for -mno-speculate-indirect-jumps.
3245 (*call_indirect_elfv2<mode>_nospec): New define_insn.
3246 (*call_value_indirect_elfv2<mode>): Disable for
3247 -mno-speculate-indirect-jumps.
3248 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
3249 (indirect_jump): Emit different RTL for
3250 -mno-speculate-indirect-jumps.
3251 (*indirect_jump<mode>): Disable for
3252 -mno-speculate-indirect-jumps.
3253 (*indirect_jump<mode>_nospec): New define_insn.
3254 (tablejump): Emit different RTL for
3255 -mno-speculate-indirect-jumps.
3256 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
3257 (tablejumpsi_nospec): New define_expand.
3258 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
3259 (tablejumpdi_nospec): New define_expand.
3260 (*tablejump<mode>_internal1): Disable for
3261 -mno-speculate-indirect-jumps.
3262 (*tablejump<mode>_internal1_nospec): New define_insn.
3263 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
3266 2018-01-16 Artyom Skrobov tyomitch@gmail.com
3268 * caller-save.c (insert_save): Drop unnecessary parameter. All
3271 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3272 Richard Biener <rguenth@suse.de>
3275 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
3276 return early, inline manually is_gimple_sizepos. Make sure if we
3277 call gimplify_expr we don't end up with a gimple constant.
3278 * tree.c (variably_modified_type_p): Don't return true for
3279 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
3280 * gimplify.h (is_gimple_sizepos): Remove.
3282 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3284 PR tree-optimization/83857
3285 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
3286 vectorizable_live_operation for pure SLP statements.
3287 (vectorizable_live_operation): Handle PHIs.
3289 2018-01-16 Richard Biener <rguenther@suse.de>
3291 PR tree-optimization/83867
3292 * tree-vect-stmts.c (vect_transform_stmt): Precompute
3293 nested_in_vect_loop_p since the scalar stmt may get invalidated.
3295 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3298 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
3299 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
3300 If off is not INTEGER_CST, issue a may not be aligned warning
3301 rather than isn't aligned. Use isn%'t rather than isn't.
3302 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
3304 <case MULT_EXPR>: Improve the case when bottom and one of the
3305 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
3306 operand, in that case check if the other operand is multiple of
3307 bottom divided by the INTEGER_CST operand.
3309 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3312 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
3313 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
3314 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
3315 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
3316 * config/pa/pa.c (pa_function_arg_advance): Likewise.
3317 (pa_function_arg, pa_arg_partial_bytes): Likewise.
3318 (pa_function_arg_size): New function.
3320 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3322 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
3323 in a separate statement.
3325 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3327 PR tree-optimization/83847
3328 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
3329 group gathers and scatters.
3331 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3333 PR rtl-optimization/86620
3334 * params.def (max-sched-ready-insns): Bump minimum value to 1.
3336 PR rtl-optimization/83213
3337 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
3338 to last if both are JUMP_INSNs.
3340 PR tree-optimization/83843
3341 * gimple-ssa-store-merging.c
3342 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
3343 store_immediate_info for bswap/nop orig_stores.
3345 2018-01-15 Andrew Waterman <andrew@sifive.com>
3347 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
3349 <UDIV>: Increase cost if !TARGET_DIV.
3351 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
3353 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
3354 (define_attr "cr_logical_3op"): New.
3355 (cceq_ior_compare): Adjust.
3356 (cceq_ior_compare_complement): Adjust.
3357 (*cceq_rev_compare): Adjust.
3358 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
3359 (is_cracked_insn): Adjust.
3360 (insn_must_be_first_in_group): Adjust.
3361 * config/rs6000/40x.md: Adjust.
3362 * config/rs6000/440.md: Adjust.
3363 * config/rs6000/476.md: Adjust.
3364 * config/rs6000/601.md: Adjust.
3365 * config/rs6000/603.md: Adjust.
3366 * config/rs6000/6xx.md: Adjust.
3367 * config/rs6000/7450.md: Adjust.
3368 * config/rs6000/7xx.md: Adjust.
3369 * config/rs6000/8540.md: Adjust.
3370 * config/rs6000/cell.md: Adjust.
3371 * config/rs6000/e300c2c3.md: Adjust.
3372 * config/rs6000/e500mc.md: Adjust.
3373 * config/rs6000/e500mc64.md: Adjust.
3374 * config/rs6000/e5500.md: Adjust.
3375 * config/rs6000/e6500.md: Adjust.
3376 * config/rs6000/mpc.md: Adjust.
3377 * config/rs6000/power4.md: Adjust.
3378 * config/rs6000/power5.md: Adjust.
3379 * config/rs6000/power6.md: Adjust.
3380 * config/rs6000/power7.md: Adjust.
3381 * config/rs6000/power8.md: Adjust.
3382 * config/rs6000/power9.md: Adjust.
3383 * config/rs6000/rs64.md: Adjust.
3384 * config/rs6000/titan.md: Adjust.
3386 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3388 * config/i386/predicates.md (indirect_branch_operand): Rewrite
3389 ix86_indirect_branch_register logic.
3391 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3393 * config/i386/constraints.md (Bs): Update
3394 ix86_indirect_branch_register check. Don't check
3395 ix86_indirect_branch_register with GOT_memory_operand.
3397 * config/i386/predicates.md (GOT_memory_operand): Don't check
3398 ix86_indirect_branch_register here.
3399 (GOT32_symbol_operand): Likewise.
3401 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3403 * config/i386/predicates.md (constant_call_address_operand):
3404 Rewrite ix86_indirect_branch_register logic.
3405 (sibcall_insn_operand): Likewise.
3407 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3409 * config/i386/constraints.md (Bs): Replace
3410 ix86_indirect_branch_thunk_register with
3411 ix86_indirect_branch_register.
3413 * config/i386/i386.md (indirect_jump): Likewise.
3414 (tablejump): Likewise.
3415 (*sibcall_memory): Likewise.
3416 (*sibcall_value_memory): Likewise.
3417 Peepholes of indirect call and jump via memory: Likewise.
3418 * config/i386/i386.opt: Likewise.
3419 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3420 (GOT_memory_operand): Likewise.
3421 (call_insn_operand): Likewise.
3422 (sibcall_insn_operand): Likewise.
3423 (GOT32_symbol_operand): Likewise.
3425 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3428 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
3429 type rather than type addr's type points to.
3430 (expand_omp_atomic_mutex): Likewise.
3431 (expand_omp_atomic): Likewise.
3433 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3436 * config/i386/i386.c (output_indirect_thunk_function): Use
3437 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
3438 for __x86_return_thunk.
3440 2018-01-15 Richard Biener <rguenther@suse.de>
3443 * expmed.c (extract_bit_field_1): Fix typo.
3445 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3448 * config/arm/iterators.md (VF): New mode iterator.
3449 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
3450 Remove integer-related logic from pattern.
3451 (neon_vabd<mode>_3): Likewise.
3453 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3456 * common.opt (fstrict-overflow): No longer an alias.
3457 (fwrapv-pointer): New option.
3458 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
3459 also for pointer types based on flag_wrapv_pointer.
3460 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
3461 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
3462 opts->x_flag_wrapv got set.
3463 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
3464 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
3465 POINTER_TYPE_OVERFLOW_UNDEFINED.
3466 * match.pd: Likewise in address comparison pattern.
3467 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
3469 2018-01-15 Richard Biener <rguenther@suse.de>
3472 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
3473 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
3474 Reset type names to their identifier if their TYPE_DECL doesn't
3475 have linkage (and thus is used for ODR and devirt).
3476 (save_debug_info_for_decl): Remove.
3477 (save_debug_info_for_type): Likewise.
3478 (add_tree_to_fld_list): Adjust.
3479 * tree-pretty-print.c (dump_generic_node): Make dumping of
3480 type names more robust.
3482 2018-01-15 Richard Biener <rguenther@suse.de>
3484 * BASE-VER: Bump to 8.0.1.
3486 2018-01-14 Martin Sebor <msebor@redhat.com>
3489 * builtins.c (check_access): Avoid warning when the no-warning bit
3492 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
3494 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
3495 * ira-color (allocno_hard_regs_compare): Likewise.
3497 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
3500 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
3501 Use .pushsection/.popsection.
3503 2018-01-14 Martin Sebor <msebor@redhat.com>
3506 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
3508 2018-01-14 Jakub Jelinek <jakub@redhat.com>
3510 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
3511 entry from extra_headers.
3512 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
3513 extra_headers, make the list bitwise identical to the i?86-*-* one.
3515 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3517 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
3518 -mcmodel=large with -mindirect-branch=thunk,
3519 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
3520 -mfunction-return=thunk-extern.
3521 * doc/invoke.texi: Document -mcmodel=large is incompatible with
3522 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
3523 -mfunction-return=thunk and -mfunction-return=thunk-extern.
3525 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3527 * config/i386/i386.c (print_reg): Print the name of the full
3528 integer register without '%'.
3529 (ix86_print_operand): Handle 'V'.
3530 * doc/extend.texi: Document 'V' modifier.
3532 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3534 * config/i386/constraints.md (Bs): Disallow memory operand for
3535 -mindirect-branch-register.
3537 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3538 (GOT_memory_operand): Likewise.
3539 (call_insn_operand): Likewise.
3540 (sibcall_insn_operand): Likewise.
3541 (GOT32_symbol_operand): Likewise.
3542 * config/i386/i386.md (indirect_jump): Call convert_memory_address
3543 for -mindirect-branch-register.
3544 (tablejump): Likewise.
3545 (*sibcall_memory): Likewise.
3546 (*sibcall_value_memory): Likewise.
3547 Disallow peepholes of indirect call and jump via memory for
3548 -mindirect-branch-register.
3549 (*call_pop): Replace m with Bw.
3550 (*call_value_pop): Likewise.
3551 (*sibcall_pop_memory): Replace m with Bs.
3552 * config/i386/i386.opt (mindirect-branch-register): New option.
3553 * doc/invoke.texi: Document -mindirect-branch-register option.
3555 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3557 * config/i386/i386-protos.h (ix86_output_function_return): New.
3558 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
3559 set function_return_type.
3560 (indirect_thunk_name): Add ret_p to indicate thunk for function
3562 (output_indirect_thunk_function): Pass false to
3563 indirect_thunk_name.
3564 (ix86_output_indirect_branch_via_reg): Likewise.
3565 (ix86_output_indirect_branch_via_push): Likewise.
3566 (output_indirect_thunk_function): Create alias for function
3567 return thunk if regno < 0.
3568 (ix86_output_function_return): New function.
3569 (ix86_handle_fndecl_attribute): Handle function_return.
3570 (ix86_attribute_table): Add function_return.
3571 * config/i386/i386.h (machine_function): Add
3572 function_return_type.
3573 * config/i386/i386.md (simple_return_internal): Use
3574 ix86_output_function_return.
3575 (simple_return_internal_long): Likewise.
3576 * config/i386/i386.opt (mfunction-return=): New option.
3577 (indirect_branch): Mention -mfunction-return=.
3578 * doc/extend.texi: Document function_return function attribute.
3579 * doc/invoke.texi: Document -mfunction-return= option.
3581 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3583 * config/i386/i386-opts.h (indirect_branch): New.
3584 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
3585 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
3586 with local indirect jump when converting indirect call and jump.
3587 (ix86_set_indirect_branch_type): New.
3588 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
3589 (indirectlabelno): New.
3590 (indirect_thunk_needed): Likewise.
3591 (indirect_thunk_bnd_needed): Likewise.
3592 (indirect_thunks_used): Likewise.
3593 (indirect_thunks_bnd_used): Likewise.
3594 (INDIRECT_LABEL): Likewise.
3595 (indirect_thunk_name): Likewise.
3596 (output_indirect_thunk): Likewise.
3597 (output_indirect_thunk_function): Likewise.
3598 (ix86_output_indirect_branch_via_reg): Likewise.
3599 (ix86_output_indirect_branch_via_push): Likewise.
3600 (ix86_output_indirect_branch): Likewise.
3601 (ix86_output_indirect_jmp): Likewise.
3602 (ix86_code_end): Call output_indirect_thunk_function if needed.
3603 (ix86_output_call_insn): Call ix86_output_indirect_branch if
3605 (ix86_handle_fndecl_attribute): Handle indirect_branch.
3606 (ix86_attribute_table): Add indirect_branch.
3607 * config/i386/i386.h (machine_function): Add indirect_branch_type
3608 and has_local_indirect_jump.
3609 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
3611 (tablejump): Likewise.
3612 (*indirect_jump): Use ix86_output_indirect_jmp.
3613 (*tablejump_1): Likewise.
3614 (simple_return_indirect_internal): Likewise.
3615 * config/i386/i386.opt (mindirect-branch=): New option.
3616 (indirect_branch): New.
3619 (thunk-inline): Likewise.
3620 (thunk-extern): Likewise.
3621 * doc/extend.texi: Document indirect_branch function attribute.
3622 * doc/invoke.texi: Document -mindirect-branch= option.
3624 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
3627 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
3629 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
3631 * ipa-inline.c (want_inline_small_function_p): Return false if
3632 inlining has already failed with CIF_FINAL_ERROR.
3633 (update_caller_keys): Call want_inline_small_function_p before
3635 (update_callee_keys): Likewise.
3637 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3639 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
3641 (rs6000_quadword_masked_address_p): Likewise.
3642 (quad_aligned_load_p): Likewise.
3643 (quad_aligned_store_p): Likewise.
3644 (const_load_sequence_p): Add comment to describe the outer-most loop.
3645 (mimic_memory_attributes_and_flags): New function.
3646 (rs6000_gen_stvx): Likewise.
3647 (replace_swapped_aligned_store): Likewise.
3648 (rs6000_gen_lvx): Likewise.
3649 (replace_swapped_aligned_load): Likewise.
3650 (replace_swapped_load_constant): Capitalize argument name in
3651 comment describing this function.
3652 (rs6000_analyze_swaps): Add a third pass to search for vector loads
3653 and stores that access quad-word aligned addresses and replace
3654 with stvx or lvx instructions when appropriate.
3655 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
3656 New function prototype.
3657 (rs6000_quadword_masked_address_p): Likewise.
3658 (rs6000_gen_lvx): Likewise.
3659 (rs6000_gen_stvx): Likewise.
3660 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
3661 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
3662 when memory address is aligned.
3663 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
3664 this split to select lvx instruction when memory address is aligned.
3665 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
3666 instruction when memory address is aligned.
3667 (*vsx_le_perm_load_v16qi): Likewise.
3668 (four unnamed splitters): Modify to select the stvx instruction
3669 when memory is aligned.
3671 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
3673 * predict.c (determine_unlikely_bbs): Handle correctly BBs
3674 which appears in the queue multiple times.
3676 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3677 Alan Hayward <alan.hayward@arm.com>
3678 David Sherwood <david.sherwood@arm.com>
3680 * tree-vectorizer.h (vec_lower_bound): New structure.
3681 (_loop_vec_info): Add check_nonzero and lower_bounds.
3682 (LOOP_VINFO_CHECK_NONZERO): New macro.
3683 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
3684 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
3685 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
3686 fields. Make seg_len the distance travelled, not including the
3688 (dr_direction_indicator): Declare.
3689 (dr_zero_step_indicator): Likewise.
3690 (dr_known_forward_stride_p): Likewise.
3691 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
3693 (runtime_alias_check_p): Allow runtime alias checks with
3695 (operator ==): Compare access_size and align.
3696 (prune_runtime_alias_test_list): Rework for new distinction between
3697 the access_size and seg_len.
3698 (create_intersect_range_checks_index): Likewise. Cope with polynomial
3700 (get_segment_min_max): New function.
3701 (create_intersect_range_checks): Use it.
3702 (dr_step_indicator): New function.
3703 (dr_direction_indicator): Likewise.
3704 (dr_zero_step_indicator): Likewise.
3705 (dr_known_forward_stride_p): Likewise.
3706 * tree-loop-distribution.c (data_ref_segment_size): Return
3707 DR_STEP * (niters - 1).
3708 (compute_alias_check_pairs): Update call to the dr_with_seg_len
3710 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
3711 (vect_preserves_scalar_order_p): New function, split out from...
3712 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
3713 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
3714 (vect_vfa_access_size): New function.
3715 (vect_vfa_align): Likewise.
3716 (vect_compile_time_alias): Take access_size_a and access_b arguments.
3717 (dump_lower_bound): New function.
3718 (vect_check_lower_bound): Likewise.
3719 (vect_small_gap_p): Likewise.
3720 (vectorizable_with_step_bound_p): Likewise.
3721 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
3722 depencies if the vectorization factor is 1. Convert the checks
3723 for nonzero steps into checks on the bounds of DR_STEP. Try using
3724 a bunds check for variable steps if the minimum required step is
3725 relatively small. Update calls to the dr_with_seg_len
3726 constructor and to vect_compile_time_alias.
3727 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
3729 (vect_loop_versioning): Call it.
3730 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
3732 (vect_estimate_min_profitable_iters): Account for any bounds checks.
3734 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3735 Alan Hayward <alan.hayward@arm.com>
3736 David Sherwood <david.sherwood@arm.com>
3738 * doc/sourcebuild.texi (vect_scatter_store): Document.
3739 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
3741 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
3743 * genopinit.c (main): Add supports_vec_scatter_store and
3744 supports_vec_scatter_store_cached to target_optabs.
3745 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
3746 IFN_MASK_SCATTER_STORE.
3747 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
3749 * internal-fn.h (internal_store_fn_p): Declare.
3750 (internal_fn_stored_value_index): Likewise.
3751 * internal-fn.c (scatter_store_direct): New macro.
3752 (expand_scatter_store_optab_fn): New function.
3753 (direct_scatter_store_optab_supported_p): New macro.
3754 (internal_store_fn_p): New function.
3755 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
3756 IFN_MASK_SCATTER_STORE.
3757 (internal_fn_mask_index): Likewise.
3758 (internal_fn_stored_value_index): New function.
3759 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
3761 * optabs-query.h (supports_vec_scatter_store_p): Declare.
3762 * optabs-query.c (supports_vec_scatter_store_p): New function.
3763 * tree-vectorizer.h (vect_get_store_rhs): Declare.
3764 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
3765 true for scatter stores.
3766 (vect_gather_scatter_fn_p): Handle scatter stores too.
3767 (vect_check_gather_scatter): Consider using scatter stores if
3768 supports_vec_scatter_store_p.
3769 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
3771 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
3772 internal_fn_stored_value_index.
3773 (check_load_store_masking): Handle scatter stores too.
3774 (vect_get_store_rhs): Make public.
3775 (vectorizable_call): Use internal_store_fn_p.
3776 (vectorizable_store): Handle scatter store internal functions.
3777 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
3778 when deciding whether the end of the group has been reached.
3779 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
3780 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
3781 (mask_scatter_store<mode>): New insns.
3783 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3784 Alan Hayward <alan.hayward@arm.com>
3785 David Sherwood <david.sherwood@arm.com>
3787 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
3788 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
3789 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
3791 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
3792 Use vect_truncate_gather_scatter_offset if we can't treat the
3793 operation as a normal gather load or scatter store.
3794 (get_group_load_store_type): Take the gather_scatter_info
3795 as argument. Try using a gather load or scatter store for
3796 single-element groups.
3797 (get_load_store_type): Update calls to get_group_load_store_type
3798 and vect_use_strided_gather_scatters_p.
3800 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3801 Alan Hayward <alan.hayward@arm.com>
3802 David Sherwood <david.sherwood@arm.com>
3804 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
3805 optional tree argument.
3806 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
3808 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
3809 but continue to use the current value as a fallback.
3810 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
3811 to compare the updates.
3812 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
3813 (get_load_store_type): Use it when handling a strided access.
3814 (vect_get_strided_load_store_ops): New function.
3815 (vect_get_data_ptr_increment): Likewise.
3816 (vectorizable_load): Handle strided gather loads. Always pass
3817 a step to vect_create_data_ref_ptr and bump_vector_ptr.
3819 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3820 Alan Hayward <alan.hayward@arm.com>
3821 David Sherwood <david.sherwood@arm.com>
3823 * doc/md.texi (gather_load@var{m}): Document.
3824 (mask_gather_load@var{m}): Likewise.
3825 * genopinit.c (main): Add supports_vec_gather_load and
3826 supports_vec_gather_load_cached to target_optabs.
3827 * optabs-tree.c (init_tree_optimization_optabs): Use
3828 ggc_cleared_alloc to allocate target_optabs.
3829 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
3830 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
3832 * internal-fn.h (internal_load_fn_p): Declare.
3833 (internal_gather_scatter_fn_p): Likewise.
3834 (internal_fn_mask_index): Likewise.
3835 (internal_gather_scatter_fn_supported_p): Likewise.
3836 * internal-fn.c (gather_load_direct): New macro.
3837 (expand_gather_load_optab_fn): New function.
3838 (direct_gather_load_optab_supported_p): New macro.
3839 (direct_internal_fn_optab): New function.
3840 (internal_load_fn_p): Likewise.
3841 (internal_gather_scatter_fn_p): Likewise.
3842 (internal_fn_mask_index): Likewise.
3843 (internal_gather_scatter_fn_supported_p): Likewise.
3844 * optabs-query.c (supports_at_least_one_mode_p): New function.
3845 (supports_vec_gather_load_p): Likewise.
3846 * optabs-query.h (supports_vec_gather_load_p): Declare.
3847 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
3848 and memory_type field.
3849 (NUM_PATTERNS): Bump to 15.
3850 * tree-vect-data-refs.c: Include internal-fn.h.
3851 (vect_gather_scatter_fn_p): New function.
3852 (vect_describe_gather_scatter_call): Likewise.
3853 (vect_check_gather_scatter): Try using internal functions for
3854 gather loads. Recognize existing calls to a gather load function.
3855 (vect_analyze_data_refs): Consider using gather loads if
3856 supports_vec_gather_load_p.
3857 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
3858 (vect_get_gather_scatter_offset_type): Likewise.
3859 (vect_convert_mask_for_vectype): Likewise.
3860 (vect_add_conversion_to_patterm): Likewise.
3861 (vect_try_gather_scatter_pattern): Likewise.
3862 (vect_recog_gather_scatter_pattern): New pattern recognizer.
3863 (vect_vect_recog_func_ptrs): Add it.
3864 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
3865 internal_fn_mask_index and internal_gather_scatter_fn_p.
3866 (check_load_store_masking): Take the gather_scatter_info as an
3867 argument and handle gather loads.
3868 (vect_get_gather_scatter_ops): New function.
3869 (vectorizable_call): Check internal_load_fn_p.
3870 (vectorizable_load): Likewise. Handle gather load internal
3872 (vectorizable_store): Update call to check_load_store_masking.
3873 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
3874 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
3875 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
3876 (aarch64_gather_scale_operand_d): New predicates.
3877 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
3878 (mask_gather_load<mode>): New insns.
3880 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3881 Alan Hayward <alan.hayward@arm.com>
3882 David Sherwood <david.sherwood@arm.com>
3884 * optabs.def (fold_left_plus_optab): New optab.
3885 * doc/md.texi (fold_left_plus_@var{m}): Document.
3886 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
3887 * internal-fn.c (fold_left_direct): Define.
3888 (expand_fold_left_optab_fn): Likewise.
3889 (direct_fold_left_optab_supported_p): Likewise.
3890 * fold-const-call.c (fold_const_fold_left): New function.
3891 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
3892 * tree-parloops.c (valid_reduction_p): New function.
3893 (gather_scalar_reductions): Use it.
3894 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
3895 (vect_finish_replace_stmt): Declare.
3896 * tree-vect-loop.c (fold_left_reduction_fn): New function.
3897 (needs_fold_left_reduction_p): New function, split out from...
3898 (vect_is_simple_reduction): ...here. Accept reductions that
3899 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
3900 (vect_force_simple_reduction): Also store the reduction type in
3901 the assignment's STMT_VINFO_REDUC_TYPE.
3902 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
3903 (merge_with_identity): New function.
3904 (vect_expand_fold_left): Likewise.
3905 (vectorize_fold_left_reduction): Likewise.
3906 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
3907 scalar phi in place for it. Check for target support and reject
3908 cases that would reassociate the operation. Defer the transform
3909 phase to vectorize_fold_left_reduction.
3910 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
3911 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
3912 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
3914 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3916 * tree-if-conv.c (predicate_mem_writes): Remove redundant
3917 call to ifc_temp_var.
3919 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3920 Alan Hayward <alan.hayward@arm.com>
3921 David Sherwood <david.sherwood@arm.com>
3923 * target.def (legitimize_address_displacement): Take the original
3924 offset as a poly_int.
3925 * targhooks.h (default_legitimize_address_displacement): Update
3927 * targhooks.c (default_legitimize_address_displacement): Likewise.
3928 * doc/tm.texi: Regenerate.
3929 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
3930 as an argument, moving assert of ad->disp == ad->disp_term to...
3931 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
3932 Try calling targetm.legitimize_address_displacement before expanding
3933 the address rather than afterwards, and adjust for the new interface.
3934 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
3935 Match the new hook interface. Handle SVE addresses.
3936 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
3939 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3941 * Makefile.in (OBJS): Add early-remat.o.
3942 * target.def (select_early_remat_modes): New hook.
3943 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
3944 * doc/tm.texi: Regenerate.
3945 * targhooks.h (default_select_early_remat_modes): Declare.
3946 * targhooks.c (default_select_early_remat_modes): New function.
3947 * timevar.def (TV_EARLY_REMAT): New timevar.
3948 * passes.def (pass_early_remat): New pass.
3949 * tree-pass.h (make_pass_early_remat): Declare.
3950 * early-remat.c: New file.
3951 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
3953 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
3955 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3956 Alan Hayward <alan.hayward@arm.com>
3957 David Sherwood <david.sherwood@arm.com>
3959 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
3960 vfm1 with a bound_epilog parameter.
3961 (vect_do_peeling): Update calls accordingly, and move the prologue
3962 call earlier in the function. Treat the base bound_epilog as 0 for
3963 fully-masked loops and retain vf - 1 for other loops. Add 1 to
3964 this base when peeling for gaps.
3965 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
3966 with fully-masked loops.
3967 (vect_estimate_min_profitable_iters): Handle the single peeled
3968 iteration in that case.
3970 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3971 Alan Hayward <alan.hayward@arm.com>
3972 David Sherwood <david.sherwood@arm.com>
3974 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
3975 single-element interleaving even if the size is not a power of 2.
3976 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
3977 accesses for single-element interleaving if the group size is
3980 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3981 Alan Hayward <alan.hayward@arm.com>
3982 David Sherwood <david.sherwood@arm.com>
3984 * doc/md.texi (fold_extract_last_@var{m}): Document.
3985 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
3986 * optabs.def (fold_extract_last_optab): New optab.
3987 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
3988 * internal-fn.c (fold_extract_direct): New macro.
3989 (expand_fold_extract_optab_fn): Likewise.
3990 (direct_fold_extract_optab_supported_p): Likewise.
3991 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
3992 * tree-vect-loop.c (vect_model_reduction_cost): Handle
3993 EXTRACT_LAST_REDUCTION.
3994 (get_initial_def_for_reduction): Do not create an initial vector
3995 for EXTRACT_LAST_REDUCTION reductions.
3996 (vectorizable_reduction): Leave the scalar phi in place for
3997 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
3998 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
3999 epilogue code for EXTRACT_LAST_REDUCTION and defer the
4000 transform phase to vectorizable_condition.
4001 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
4003 (vect_finish_stmt_generation): ...here.
4004 (vect_finish_replace_stmt): New function.
4005 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
4006 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
4008 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
4010 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4011 Alan Hayward <alan.hayward@arm.com>
4012 David Sherwood <david.sherwood@arm.com>
4014 * doc/md.texi (extract_last_@var{m}): Document.
4015 * optabs.def (extract_last_optab): New optab.
4016 * internal-fn.def (EXTRACT_LAST): New internal function.
4017 * internal-fn.c (cond_unary_direct): New macro.
4018 (expand_cond_unary_optab_fn): Likewise.
4019 (direct_cond_unary_optab_supported_p): Likewise.
4020 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
4021 loops using EXTRACT_LAST.
4022 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
4023 (extract_last_<mode>): ...this optab.
4024 (vec_extract<mode><Vel>): Update accordingly.
4026 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4027 Alan Hayward <alan.hayward@arm.com>
4028 David Sherwood <david.sherwood@arm.com>
4030 * target.def (empty_mask_is_expensive): New hook.
4031 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
4032 * doc/tm.texi: Regenerate.
4033 * targhooks.h (default_empty_mask_is_expensive): Declare.
4034 * targhooks.c (default_empty_mask_is_expensive): New function.
4035 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
4036 if the target says that empty masks are expensive.
4037 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
4039 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
4041 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4042 Alan Hayward <alan.hayward@arm.com>
4043 David Sherwood <david.sherwood@arm.com>
4045 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
4046 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
4047 (vect_use_loop_mask_for_alignment_p): New function.
4048 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
4049 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
4050 niters_skip argument. Make sure that the first niters_skip elements
4051 of the first iteration are inactive.
4052 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
4053 Update call to vect_set_loop_masks_directly.
4054 (get_misalign_in_elems): New function, split out from...
4055 (vect_gen_prolog_loop_niters): ...here.
4056 (vect_update_init_of_dr): Take a code argument that specifies whether
4057 the adjustment should be added or subtracted.
4058 (vect_update_init_of_drs): Likewise.
4059 (vect_prepare_for_masked_peels): New function.
4060 (vect_do_peeling): Skip prologue peeling if we're using a mask
4061 instead. Update call to vect_update_inits_of_drs.
4062 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4064 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
4065 alignment. Do not include the number of peeled iterations in
4066 the minimum threshold in that case.
4067 (vectorizable_induction): Adjust the start value down by
4068 LOOP_VINFO_MASK_SKIP_NITERS iterations.
4069 (vect_transform_loop): Call vect_prepare_for_masked_peels.
4070 Take the number of skipped iterations into account when calculating
4072 * tree-vect-stmts.c (vect_gen_while_not): New function.
4074 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4075 Alan Hayward <alan.hayward@arm.com>
4076 David Sherwood <david.sherwood@arm.com>
4078 * doc/sourcebuild.texi (vect_fully_masked): Document.
4079 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
4081 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
4083 (vect_analyze_loop_2): ...here. Don't check the vectorization
4084 factor against the number of loop iterations if the loop is
4087 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4088 Alan Hayward <alan.hayward@arm.com>
4089 David Sherwood <david.sherwood@arm.com>
4091 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
4092 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
4093 (dump_groups): Update accordingly.
4094 (iv_use::mem_type): New member variable.
4095 (address_p): New function.
4096 (record_use): Add a mem_type argument and initialize the new
4098 (record_group_use): Add a mem_type argument. Use address_p.
4099 Remove obsolete null checks of base_object. Update call to record_use.
4100 (find_interesting_uses_op): Update call to record_group_use.
4101 (find_interesting_uses_cond): Likewise.
4102 (find_interesting_uses_address): Likewise.
4103 (get_mem_type_for_internal_fn): New function.
4104 (find_address_like_use): Likewise.
4105 (find_interesting_uses_stmt): Try find_address_like_use before
4106 calling find_interesting_uses_op.
4107 (addr_offset_valid_p): Use the iv mem_type field as the type
4108 of the addressed memory.
4109 (add_autoinc_candidates): Likewise.
4110 (get_address_cost): Likewise.
4111 (split_small_address_groups_p): Use address_p.
4112 (split_address_groups): Likewise.
4113 (add_iv_candidate_for_use): Likewise.
4114 (autoinc_possible_for_pair): Likewise.
4115 (rewrite_groups): Likewise.
4116 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
4117 (determine_group_iv_cost): Update after split of USE_ADDRESS.
4118 (get_alias_ptr_type_for_ptr_address): New function.
4119 (rewrite_use_address): Rewrite address uses in calls that were
4120 identified by find_address_like_use.
4122 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4123 Alan Hayward <alan.hayward@arm.com>
4124 David Sherwood <david.sherwood@arm.com>
4126 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
4128 * gimple-expr.h (is_gimple_addressable: Likewise.
4129 * gimple-expr.c (is_gimple_address): Likewise.
4130 * internal-fn.c (expand_call_mem_ref): New function.
4131 (expand_mask_load_optab_fn): Use it.
4132 (expand_mask_store_optab_fn): Likewise.
4134 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4135 Alan Hayward <alan.hayward@arm.com>
4136 David Sherwood <david.sherwood@arm.com>
4138 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
4139 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
4140 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
4141 (cond_umax@var{mode}): Document.
4142 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
4143 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
4144 (cond_umin_optab, cond_umax_optab): New optabs.
4145 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
4146 (COND_IOR, COND_XOR): New internal functions.
4147 * internal-fn.h (get_conditional_internal_fn): Declare.
4148 * internal-fn.c (cond_binary_direct): New macro.
4149 (expand_cond_binary_optab_fn): Likewise.
4150 (direct_cond_binary_optab_supported_p): Likewise.
4151 (get_conditional_internal_fn): New function.
4152 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
4153 Cope with reduction statements that are vectorized as calls rather
4155 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
4156 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
4157 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
4158 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4159 (UNSPEC_COND_EOR): New unspecs.
4160 (optab): Add mappings for them.
4161 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
4162 (sve_int_op, sve_fp_op): New int attributes.
4164 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4165 Alan Hayward <alan.hayward@arm.com>
4166 David Sherwood <david.sherwood@arm.com>
4168 * optabs.def (while_ult_optab): New optab.
4169 * doc/md.texi (while_ult@var{m}@var{n}): Document.
4170 * internal-fn.def (WHILE_ULT): New internal function.
4171 * internal-fn.h (direct_internal_fn_supported_p): New override
4172 that takes two types as argument.
4173 * internal-fn.c (while_direct): New macro.
4174 (expand_while_optab_fn): New function.
4175 (convert_optab_supported_p): Likewise.
4176 (direct_while_optab_supported_p): New macro.
4177 * wide-int.h (wi::udiv_ceil): New function.
4178 * tree-vectorizer.h (rgroup_masks): New structure.
4179 (vec_loop_masks): New typedef.
4180 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
4182 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
4183 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
4184 (vect_max_vf): New function.
4185 (slpeel_make_loop_iterate_ntimes): Delete.
4186 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
4187 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
4188 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
4189 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
4190 internal-fn.h, stor-layout.h and optabs-query.h.
4191 (vect_set_loop_mask): New function.
4192 (add_preheader_seq): Likewise.
4193 (add_header_seq): Likewise.
4194 (interleave_supported_p): Likewise.
4195 (vect_maybe_permute_loop_masks): Likewise.
4196 (vect_set_loop_masks_directly): Likewise.
4197 (vect_set_loop_condition_masked): Likewise.
4198 (vect_set_loop_condition_unmasked): New function, split out from
4199 slpeel_make_loop_iterate_ntimes.
4200 (slpeel_make_loop_iterate_ntimes): Rename to..
4201 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
4202 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
4203 (vect_do_peeling): Update call accordingly.
4204 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
4206 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4207 mask_compare_type, can_fully_mask_p and fully_masked_p.
4208 (release_vec_loop_masks): New function.
4209 (_loop_vec_info): Use it to free the loop masks.
4210 (can_produce_all_loop_masks_p): New function.
4211 (vect_get_max_nscalars_per_iter): Likewise.
4212 (vect_verify_full_masking): Likewise.
4213 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
4214 retries, and free the mask rgroups before retrying. Check loop-wide
4215 reasons for disallowing fully-masked loops. Make the final decision
4216 about whether use a fully-masked loop or not.
4217 (vect_estimate_min_profitable_iters): Do not assume that peeling
4218 for the number of iterations will be needed for fully-masked loops.
4219 (vectorizable_reduction): Disable fully-masked loops.
4220 (vectorizable_live_operation): Likewise.
4221 (vect_halve_mask_nunits): New function.
4222 (vect_double_mask_nunits): Likewise.
4223 (vect_record_loop_mask): Likewise.
4224 (vect_get_loop_mask): Likewise.
4225 (vect_transform_loop): Handle the case in which the final loop
4226 iteration might handle a partial vector. Call vect_set_loop_condition
4227 instead of slpeel_make_loop_iterate_ntimes.
4228 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
4229 (check_load_store_masking): New function.
4230 (prepare_load_store_mask): Likewise.
4231 (vectorizable_store): Handle fully-masked loops.
4232 (vectorizable_load): Likewise.
4233 (supportable_widening_operation): Use vect_halve_mask_nunits for
4235 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
4236 (vect_gen_while): New function.
4237 * config/aarch64/aarch64.md (umax<mode>3): New expander.
4238 (aarch64_uqdec<mode>): New insn.
4240 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4241 Alan Hayward <alan.hayward@arm.com>
4242 David Sherwood <david.sherwood@arm.com>
4244 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
4245 (reduc_xor_scal_optab): New optabs.
4246 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
4247 (reduc_xor_scal_@var{m}): Document.
4248 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
4249 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
4251 * fold-const-call.c (fold_const_call): Handle them.
4252 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
4253 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
4254 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
4255 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
4256 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
4257 (UNSPEC_XORV): New unspecs.
4258 (optab): Add entries for them.
4259 (BITWISEV): New int iterator.
4260 (bit_reduc_op): New int attributes.
4262 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4263 Alan Hayward <alan.hayward@arm.com>
4264 David Sherwood <david.sherwood@arm.com>
4266 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
4267 * internal-fn.def (VEC_SHL_INSERT): New internal function.
4268 * optabs.def (vec_shl_insert_optab): New optab.
4269 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
4270 (duplicate_and_interleave): Likewise.
4271 * tree-vect-loop.c: Include internal-fn.h.
4272 (neutral_op_for_slp_reduction): New function, split out from
4273 get_initial_defs_for_reduction.
4274 (get_initial_def_for_reduction): Handle option 2 for variable-length
4275 vectors by loading the neutral value into a vector and then shifting
4276 the initial value into element 0.
4277 (get_initial_defs_for_reduction): Replace the code argument with
4278 the neutral value calculated by neutral_op_for_slp_reduction.
4279 Use gimple_build_vector for constant-length vectors.
4280 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
4281 but the first group_size elements have a neutral value.
4282 Use duplicate_and_interleave otherwise.
4283 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
4284 Update call to get_initial_defs_for_reduction. Handle SLP
4285 reductions for variable-length vectors by creating one vector
4286 result for each scalar result, with the elements associated
4287 with other scalar results stubbed out with the neutral value.
4288 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
4289 Require IFN_VEC_SHL_INSERT for double reductions on
4290 variable-length vectors, or SLP reductions that have
4291 a neutral value. Require can_duplicate_and_interleave_p
4292 support for variable-length unchained SLP reductions if there
4293 is no neutral value, such as for MIN/MAX reductions. Also require
4294 the number of vector elements to be a multiple of the number of
4295 SLP statements when doing variable-length unchained SLP reductions.
4296 Update call to vect_create_epilog_for_reduction.
4297 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
4298 and remove initial values.
4299 (duplicate_and_interleave): Make public.
4300 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
4301 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
4303 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4304 Alan Hayward <alan.hayward@arm.com>
4305 David Sherwood <david.sherwood@arm.com>
4307 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
4308 (can_duplicate_and_interleave_p): New function.
4309 (vect_get_and_check_slp_defs): Take the vector of statements
4310 rather than just the current one. Remove excess parentheses.
4311 Restriction rejectinon of vect_constant_def and vect_external_def
4312 for variable-length vectors to boolean types, or types for which
4313 can_duplicate_and_interleave_p is false.
4314 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
4315 (duplicate_and_interleave): New function.
4316 (vect_get_constant_vectors): Use gimple_build_vector for
4317 constant-length vectors and suitable variable-length constant
4318 vectors. Use duplicate_and_interleave for other variable-length
4319 vectors. Don't defer the update when inserting new statements.
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 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
4326 min_profitable_iters doesn't go negative.
4328 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4329 Alan Hayward <alan.hayward@arm.com>
4330 David Sherwood <david.sherwood@arm.com>
4332 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
4333 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
4334 * optabs.def (vec_mask_load_lanes_optab): New optab.
4335 (vec_mask_store_lanes_optab): Likewise.
4336 * internal-fn.def (MASK_LOAD_LANES): New internal function.
4337 (MASK_STORE_LANES): Likewise.
4338 * internal-fn.c (mask_load_lanes_direct): New macro.
4339 (mask_store_lanes_direct): Likewise.
4340 (expand_mask_load_optab_fn): Handle masked operations.
4341 (expand_mask_load_lanes_optab_fn): New macro.
4342 (expand_mask_store_optab_fn): Handle masked operations.
4343 (expand_mask_store_lanes_optab_fn): New macro.
4344 (direct_mask_load_lanes_optab_supported_p): Likewise.
4345 (direct_mask_store_lanes_optab_supported_p): Likewise.
4346 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
4348 (vect_load_lanes_supported): Likewise.
4349 * tree-vect-data-refs.c (strip_conversion): New function.
4350 (can_group_stmts_p): Likewise.
4351 (vect_analyze_data_ref_accesses): Use it instead of checking
4352 for a pair of assignments.
4353 (vect_store_lanes_supported): Take a masked_p parameter.
4354 (vect_load_lanes_supported): Likewise.
4355 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
4356 vect_store_lanes_supported and vect_load_lanes_supported.
4357 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4358 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
4359 parameter. Don't allow gaps for masked accesses.
4360 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
4361 and vect_load_lanes_supported.
4362 (get_load_store_type): Take a masked_p parameter and update
4363 call to get_group_load_store_type.
4364 (vectorizable_store): Update call to get_load_store_type.
4365 Handle IFN_MASK_STORE_LANES.
4366 (vectorizable_load): Update call to get_load_store_type.
4367 Handle IFN_MASK_LOAD_LANES.
4369 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4370 Alan Hayward <alan.hayward@arm.com>
4371 David Sherwood <david.sherwood@arm.com>
4373 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
4375 * config/aarch64/aarch64-protos.h
4376 (aarch64_sve_struct_memory_operand_p): Declare.
4377 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
4378 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
4379 (VPRED, vpred): Handle SVE structure modes.
4380 * config/aarch64/constraints.md (Utx): New constraint.
4381 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
4382 (aarch64_sve_struct_nonimmediate_operand): New predicates.
4383 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
4384 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
4385 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
4386 structure modes. Split into pieces after RA.
4387 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
4388 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
4390 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
4391 SVE structure modes.
4392 (aarch64_classify_address): Likewise.
4393 (sizetochar): Move earlier in file.
4394 (aarch64_print_operand): Handle SVE register lists.
4395 (aarch64_array_mode): New function.
4396 (aarch64_sve_struct_memory_operand_p): Likewise.
4397 (TARGET_ARRAY_MODE): Redefine.
4399 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4400 Alan Hayward <alan.hayward@arm.com>
4401 David Sherwood <david.sherwood@arm.com>
4403 * target.def (array_mode): New target hook.
4404 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
4405 * doc/tm.texi: Regenerate.
4406 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
4407 * hooks.c (hook_optmode_mode_uhwi_none): New function.
4408 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
4410 * stor-layout.c (mode_for_array): Likewise. Support polynomial
4413 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4414 Alan Hayward <alan.hayward@arm.com>
4415 David Sherwood <david.sherwood@arm.com>
4417 * fold-const.c (fold_binary_loc): Check the argument types
4418 rather than the result type when testing for a vector operation.
4420 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4422 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
4423 * doc/tm.texi: Regenerate.
4425 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4426 Alan Hayward <alan.hayward@arm.com>
4427 David Sherwood <david.sherwood@arm.com>
4429 * doc/invoke.texi (-msve-vector-bits=): Document new option.
4430 (sve): Document new AArch64 extension.
4431 * doc/md.texi (w): Extend the description of the AArch64
4432 constraint to include SVE vectors.
4433 (Upl, Upa): Document new AArch64 predicate constraints.
4434 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
4436 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
4437 (msve-vector-bits=): New option.
4438 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
4439 SVE when these are disabled.
4440 (sve): New extension.
4441 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
4442 modes. Adjust their number of units based on aarch64_sve_vg.
4443 (MAX_BITSIZE_MODE_ANY_MODE): Define.
4444 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
4445 aarch64_addr_query_type.
4446 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
4447 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
4448 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
4449 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
4450 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
4451 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
4452 (aarch64_simd_imm_zero_p): Delete.
4453 (aarch64_check_zero_based_sve_index_immediate): Declare.
4454 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4455 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4456 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4457 (aarch64_sve_float_mul_immediate_p): Likewise.
4458 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4460 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
4461 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
4462 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
4463 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
4464 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
4465 (aarch64_regmode_natural_size): Likewise.
4466 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
4467 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
4469 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
4470 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
4471 for VG and the SVE predicate registers.
4472 (V_ALIASES): Add a "z"-prefixed alias.
4473 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
4474 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
4475 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
4476 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
4477 (REG_CLASS_NAMES): Add entries for them.
4478 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
4479 and the predicate registers.
4480 (aarch64_sve_vg): Declare.
4481 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
4482 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
4483 (REGMODE_NATURAL_SIZE): Define.
4484 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
4486 * config/aarch64/aarch64.c: Include cfgrtl.h.
4487 (simd_immediate_info): Add a constructor for series vectors,
4488 and an associated step field.
4489 (aarch64_sve_vg): New variable.
4490 (aarch64_dbx_register_number): Handle VG and the predicate registers.
4491 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
4492 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
4493 (VEC_ANY_DATA, VEC_STRUCT): New constants.
4494 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
4495 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
4496 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
4497 (aarch64_get_mask_mode): New functions.
4498 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
4499 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
4500 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
4501 predicate modes and predicate registers. Explicitly restrict
4502 GPRs to modes of 16 bytes or smaller. Only allow FP registers
4503 to store a vector mode if it is recognized by
4504 aarch64_classify_vector_mode.
4505 (aarch64_regmode_natural_size): New function.
4506 (aarch64_hard_regno_caller_save_mode): Return the original mode
4508 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
4509 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
4510 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
4511 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
4513 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
4514 does not overlap dest if the function is frame-related. Handle
4516 (aarch64_split_add_offset): New function.
4517 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
4518 them aarch64_add_offset.
4519 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
4520 and update call to aarch64_sub_sp.
4521 (aarch64_add_cfa_expression): New function.
4522 (aarch64_expand_prologue): Pass extra temporary registers to the
4523 functions above. Handle the case in which we need to emit new
4524 DW_CFA_expressions for registers that were originally saved
4525 relative to the stack pointer, but now have to be expressed
4526 relative to the frame pointer.
4527 (aarch64_output_mi_thunk): Pass extra temporary registers to the
4529 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
4530 IP0 and IP1 values for SVE frames.
4531 (aarch64_expand_vec_series): New function.
4532 (aarch64_expand_sve_widened_duplicate): Likewise.
4533 (aarch64_expand_sve_const_vector): Likewise.
4534 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
4535 Handle SVE constants. Use emit_move_insn to move a force_const_mem
4536 into the register, rather than emitting a SET directly.
4537 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
4538 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
4539 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
4540 (offset_9bit_signed_scaled_p): New functions.
4541 (aarch64_replicate_bitmask_imm): New function.
4542 (aarch64_bitmask_imm): Use it.
4543 (aarch64_cannot_force_const_mem): Reject expressions involving
4544 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
4545 (aarch64_classify_index): Handle SVE indices, by requiring
4546 a plain register index with a scale that matches the element size.
4547 (aarch64_classify_address): Handle SVE addresses. Assert that
4548 the mode of the address is VOIDmode or an integer mode.
4549 Update call to aarch64_classify_symbol.
4550 (aarch64_classify_symbolic_expression): Update call to
4551 aarch64_classify_symbol.
4552 (aarch64_const_vec_all_in_range_p): New function.
4553 (aarch64_print_vector_float_operand): Likewise.
4554 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
4555 "vN" for FP registers with SVE modes. Handle (const ...) vectors
4556 and the FP immediates 1.0 and 0.5.
4557 (aarch64_print_address_internal): Handle SVE addresses.
4558 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
4559 (aarch64_regno_regclass): Handle predicate registers.
4560 (aarch64_secondary_reload): Handle big-endian reloads of SVE
4562 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
4563 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
4564 (aarch64_convert_sve_vector_bits): New function.
4565 (aarch64_override_options): Use it to handle -msve-vector-bits=.
4566 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4568 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
4569 Handle SVE vector and predicate modes. Accept VL-based constants
4570 that need only one temporary register, and VL offsets that require
4571 no temporary registers.
4572 (aarch64_conditional_register_usage): Mark the predicate registers
4573 as fixed if SVE isn't available.
4574 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
4575 Return true for SVE vector and predicate modes.
4576 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
4577 rather than an unsigned int. Handle SVE modes.
4578 (aarch64_preferred_simd_mode): Update call accordingly. Handle
4580 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
4582 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4583 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4584 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4585 (aarch64_sve_float_mul_immediate_p): New functions.
4586 (aarch64_sve_valid_immediate): New function.
4587 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
4588 Explicitly reject structure modes. Check for INDEX constants.
4589 Handle PTRUE and PFALSE constants.
4590 (aarch64_check_zero_based_sve_index_immediate): New function.
4591 (aarch64_simd_imm_zero_p): Delete.
4592 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
4593 vector modes. Accept constants in the range of CNT[BHWD].
4594 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
4595 ask for an Advanced SIMD mode.
4596 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
4597 (aarch64_simd_vector_alignment): Handle SVE predicates.
4598 (aarch64_vectorize_preferred_vector_alignment): New function.
4599 (aarch64_simd_vector_alignment_reachable): Use it instead of
4601 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
4602 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
4604 (MAX_VECT_LEN): Delete.
4605 (expand_vec_perm_d): Add a vec_flags field.
4606 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
4607 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
4608 (aarch64_evpc_ext): Don't apply a big-endian lane correction
4610 (aarch64_evpc_rev): Rename to...
4611 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
4612 (aarch64_evpc_rev_global): New function.
4613 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
4614 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
4616 (aarch64_evpc_sve_tbl): New function.
4617 (aarch64_expand_vec_perm_const_1): Update after rename of
4618 aarch64_evpc_rev. Handle SVE permutes too, trying
4619 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
4620 than aarch64_evpc_tbl.
4621 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
4622 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
4623 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
4624 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
4625 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
4626 (aarch64_expand_sve_vcond): New functions.
4627 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
4628 of aarch64_vector_mode_p.
4629 (aarch64_dwarf_poly_indeterminate_value): New function.
4630 (aarch64_compute_pressure_classes): Likewise.
4631 (aarch64_can_change_mode_class): Likewise.
4632 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
4633 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
4634 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
4635 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
4636 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
4637 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
4638 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
4639 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
4641 (Dn, Dl, Dr): Accept const as well as const_vector.
4642 (Dz): Likewise. Compare against CONST0_RTX.
4643 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
4644 of "vector" where appropriate.
4645 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
4646 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
4647 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
4648 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
4649 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
4650 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
4651 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
4652 (v_int_equiv): Extend to SVE modes.
4653 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
4655 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
4656 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
4657 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
4658 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
4659 (SVE_COND_FP_CMP): New int iterators.
4660 (perm_hilo): Handle the new unpack unspecs.
4661 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
4663 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
4664 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
4665 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
4666 (aarch64_equality_operator, aarch64_constant_vector_operand)
4667 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
4668 (aarch64_sve_nonimmediate_operand): Likewise.
4669 (aarch64_sve_general_operand): Likewise.
4670 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
4671 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
4672 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
4673 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
4674 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
4675 (aarch64_sve_float_arith_immediate): Likewise.
4676 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
4677 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
4678 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
4679 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
4680 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
4681 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
4682 (aarch64_sve_float_arith_operand): Likewise.
4683 (aarch64_sve_float_arith_with_sub_operand): Likewise.
4684 (aarch64_sve_float_mul_operand): Likewise.
4685 (aarch64_sve_vec_perm_operand): Likewise.
4686 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
4687 (aarch64_mov_operand): Accept const_poly_int and const_vector.
4688 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
4689 as well as const_vector.
4690 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
4691 in file. Use CONST0_RTX and CONSTM1_RTX.
4692 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
4693 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
4694 Use aarch64_simd_imm_zero.
4695 * config/aarch64/aarch64-sve.md: New file.
4696 * config/aarch64/aarch64.md: Include it.
4697 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
4698 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
4699 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
4700 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
4701 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
4702 (sve): New attribute.
4703 (enabled): Disable instructions with the sve attribute unless
4705 (movqi, movhi): Pass CONST_POLY_INT operaneds through
4706 aarch64_expand_mov_immediate.
4707 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
4708 CNT[BHSD] immediates.
4709 (movti): Split CONST_POLY_INT moves into two halves.
4710 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
4711 Split additions that need a temporary here if the destination
4712 is the stack pointer.
4713 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
4714 (*add<mode>3_poly_1): New instruction.
4715 (set_clobber_cc): New expander.
4717 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4719 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
4720 parameter and use it instead of GET_MODE_SIZE (innermode). Use
4721 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
4722 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
4723 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
4724 Change innermode from fixed_mode_size to machine_mode.
4725 (simplify_subreg): Update call accordingly. Handle a constant-sized
4726 subreg of a variable-length CONST_VECTOR.
4728 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4729 Alan Hayward <alan.hayward@arm.com>
4730 David Sherwood <david.sherwood@arm.com>
4732 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
4733 (add_offset_to_base): New function, split out from...
4734 (create_mem_ref): ...here. When handling a scale other than 1,
4735 check first whether the address is valid without the offset.
4736 Add it into the base if so, leaving the index and scale as-is.
4738 2018-01-12 Jakub Jelinek <jakub@redhat.com>
4741 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
4742 fold_for_warn before checking if arg2 is INTEGER_CST.
4744 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
4746 * config/rs6000/predicates.md (load_multiple_operation): Delete.
4747 (store_multiple_operation): Delete.
4748 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
4749 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
4750 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
4751 guarded by TARGET_STRING.
4752 (rs6000_output_load_multiple): Delete.
4753 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
4754 OPTION_MASK_STRING / TARGET_STRING handling.
4755 (print_operand) <'N', 'O'>: Add comment that these are unused now.
4756 (const rs6000_opt_masks) <"string">: Change mask to 0.
4757 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
4758 (MASK_STRING): Delete.
4759 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
4761 (load_multiple): Delete.
4768 (store_multiple): Delete.
4775 (movmemsi_8reg): Delete.
4776 (corresponding unnamed define_insn): Delete.
4777 (movmemsi_6reg): Delete.
4778 (corresponding unnamed define_insn): Delete.
4779 (movmemsi_4reg): Delete.
4780 (corresponding unnamed define_insn): Delete.
4781 (movmemsi_2reg): Delete.
4782 (corresponding unnamed define_insn): Delete.
4783 (movmemsi_1reg): Delete.
4784 (corresponding unnamed define_insn): Delete.
4785 * config/rs6000/rs6000.opt (mno-string): New.
4786 (mstring): Replace by deprecation warning stub.
4787 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
4789 2018-01-12 Jakub Jelinek <jakub@redhat.com>
4791 * regrename.c (regrename_do_replace): If replacing the same
4792 reg multiple times, try to reuse last created gen_raw_REG.
4795 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
4796 main to workaround a bug in GDB.
4798 2018-01-12 Tom de Vries <tom@codesourcery.com>
4801 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
4803 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
4805 PR rtl-optimization/80481
4806 * ira-color.c (get_cap_member): New function.
4807 (allocnos_conflict_by_live_ranges_p): Use it.
4808 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
4809 (setup_slot_coalesced_allocno_live_ranges): Ditto.
4811 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
4814 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
4815 (*saddl_se_1): Ditto.
4817 (*ssubl_se_1): Ditto.
4819 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
4821 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
4822 rather than wi::to_widest for DR_INITs.
4823 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
4824 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
4825 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
4827 (vect_analyze_group_access_1): Note that here.
4829 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
4831 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
4832 polynomial type sizes.
4834 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
4836 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
4837 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
4838 (gimple_add_tmp_var): Likewise.
4840 2018-01-12 Martin Liska <mliska@suse.cz>
4842 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
4843 (gimple_alloc_sizes): Likewise.
4844 (dump_gimple_statistics): Use PRIu64 in printf format.
4845 * gimple.h: Change uint64_t to int.
4847 2018-01-12 Martin Liska <mliska@suse.cz>
4849 * tree-core.h: Use uint64_t instead of int.
4850 * tree.c (tree_node_counts): Likewise.
4851 (tree_node_sizes): Likewise.
4852 (dump_tree_statistics): Use PRIu64 in printf format.
4854 2018-01-12 Martin Liska <mliska@suse.cz>
4856 * Makefile.in: As qsort_chk is implemented in vec.c, add
4857 vec.o to linkage of gencfn-macros.
4858 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
4859 passing the info to record_node_allocation_statistics.
4860 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
4862 * ggc-common.c (struct ggc_usage): Add operator== and use
4863 it in operator< and compare function.
4864 * mem-stats.h (struct mem_usage): Likewise.
4865 * vec.c (struct vec_usage): Remove operator< and compare
4866 function. Can be simply inherited.
4868 2018-01-12 Martin Jambor <mjambor@suse.cz>
4871 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
4872 * tree-ssa-math-opts.c: Include domwalk.h.
4873 (convert_mult_to_fma_1): New function.
4874 (fma_transformation_info): New type.
4875 (fma_deferring_state): Likewise.
4876 (cancel_fma_deferring): New function.
4877 (result_of_phi): Likewise.
4878 (last_fma_candidate_feeds_initial_phi): Likewise.
4879 (convert_mult_to_fma): Added deferring logic, split actual
4880 transformation to convert_mult_to_fma_1.
4881 (math_opts_dom_walker): New type.
4882 (math_opts_dom_walker::after_dom_children): New method, body moved
4883 here from pass_optimize_widening_mul::execute, added deferring logic
4885 (pass_optimize_widening_mul::execute): Moved most of code to
4886 math_opts_dom_walker::after_dom_children.
4887 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
4888 * config/i386/i386.c (ix86_option_override_internal): Added
4889 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
4891 2018-01-12 Richard Biener <rguenther@suse.de>
4894 * dwarf2out.c (gen_variable_die): Do not reset old_die for
4895 inline instance vars.
4897 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
4900 * config/rx/rx.c (rx_is_restricted_memory_address):
4903 2018-01-12 Richard Biener <rguenther@suse.de>
4905 PR tree-optimization/80846
4906 * target.def (split_reduction): New target hook.
4907 * targhooks.c (default_split_reduction): New function.
4908 * targhooks.h (default_split_reduction): Declare.
4909 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
4910 target requests first reduce vectors by combining low and high
4912 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
4913 (get_vectype_for_scalar_type_and_size): Export.
4914 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
4915 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
4916 * doc/tm.texi: Regenerate.
4917 * config/i386/i386.c (ix86_split_reduction): Implement
4918 TARGET_VECTORIZE_SPLIT_REDUCTION.
4920 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4923 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
4924 in PIC mode except for TARGET_VXWORKS_RTP.
4925 * config/sparc/sparc.c: Include cfgrtl.h.
4926 (TARGET_INIT_PIC_REG): Define.
4927 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
4928 (sparc_pic_register_p): New predicate.
4929 (sparc_legitimate_address_p): Use it.
4930 (sparc_legitimize_pic_address): Likewise.
4931 (sparc_delegitimize_address): Likewise.
4932 (sparc_mode_dependent_address_p): Likewise.
4933 (gen_load_pcrel_sym): Remove 4th parameter.
4934 (load_got_register): Adjust call to above. Remove obsolete stuff.
4935 (sparc_expand_prologue): Do not call load_got_register here.
4936 (sparc_flat_expand_prologue): Likewise.
4937 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
4938 (sparc_use_pseudo_pic_reg): New function.
4939 (sparc_init_pic_reg): Likewise.
4940 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
4941 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
4943 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
4945 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
4946 Add item for branch_cost.
4948 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4950 PR rtl-optimization/83565
4951 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
4952 not extend the result to a larger mode for rotate operations.
4953 (num_sign_bit_copies1): Likewise.
4955 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4958 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
4960 Use values-Xc.o for -pedantic.
4961 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
4963 2018-01-12 Martin Liska <mliska@suse.cz>
4966 * ipa-devirt.c (final_warning_record::grow_type_warnings):
4968 (possible_polymorphic_call_targets): Use it.
4969 (ipa_devirt): Likewise.
4971 2018-01-12 Martin Liska <mliska@suse.cz>
4973 * profile-count.h (enum profile_quality): Use 0 as invalid
4974 enum value of profile_quality.
4976 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
4978 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
4979 -mext-string options.
4981 2018-01-12 Richard Biener <rguenther@suse.de>
4983 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
4984 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
4985 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
4987 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
4989 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
4991 * configure.ac (--with-long-double-format): Add support for the
4992 configuration option to change the default long double format on
4994 * config.gcc (powerpc*-linux*-*): Likewise.
4995 * configure: Regenerate.
4996 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
4997 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
4998 used without modification.
5000 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5002 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
5003 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
5004 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
5005 MISC_BUILTIN_SPEC_BARRIER.
5006 (rs6000_init_builtins): Likewise.
5007 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
5009 (speculation_barrier): New define_insn.
5010 * doc/extend.texi: Document __builtin_speculation_barrier.
5012 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5015 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
5016 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
5017 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
5019 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
5020 integral modes instead of "ss" and "sd".
5021 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
5022 vectors with 32-bit and 64-bit elements.
5023 (vecdupssescalarmodesuffix): New mode attribute.
5024 (vec_dup<mode>): Use it.
5026 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
5029 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
5030 frame if argument is passed on stack.
5032 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5035 * ree.c (combine_reaching_defs): Optimize also
5036 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
5037 reg2=any_extend(exp); reg1=reg2;, formatting fix.
5039 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5042 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
5044 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5047 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
5048 after they are computed.
5050 2018-01-11 Bin Cheng <bin.cheng@arm.com>
5052 PR tree-optimization/83695
5053 * gimple-loop-linterchange.cc
5054 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
5055 reset cached scev information after interchange.
5056 (pass_linterchange::execute): Remove call to scev_reset_htab.
5058 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5060 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
5061 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
5062 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
5063 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
5064 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
5065 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
5066 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
5067 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
5068 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
5069 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
5070 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
5071 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
5072 (V_lane_reg): Likewise.
5073 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
5075 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
5076 (vfmal_lane_low<mode>_intrinsic,
5077 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
5078 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
5079 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
5080 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
5081 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
5082 vfmsl_lane_high<mode>_intrinsic): New define_insns.
5084 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5086 * config/arm/arm-cpus.in (fp16fml): New feature.
5087 (ALL_SIMD): Add fp16fml.
5088 (armv8.2-a): Add fp16fml as an option.
5089 (armv8.3-a): Likewise.
5090 (armv8.4-a): Add fp16fml as part of fp16.
5091 * config/arm/arm.h (TARGET_FP16FML): Define.
5092 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
5094 * config/arm/arm-modes.def (V2HF): Define.
5095 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
5096 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
5097 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
5098 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
5099 vfmsl_low, vfmsl_high): New set of builtins.
5100 * config/arm/iterators.md (PLUSMINUS): New code iterator.
5101 (vfml_op): New code attribute.
5102 (VFMLHALVES): New int iterator.
5103 (VFML, VFMLSEL): New mode attributes.
5104 (V_reg): Define mapping for V2HF.
5105 (V_hi, V_lo): New mode attributes.
5106 (VF_constraint): Likewise.
5107 (vfml_half, vfml_half_selector): New int attributes.
5108 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
5110 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
5111 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
5113 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
5114 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
5115 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
5116 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
5118 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
5119 Document new effective target and option set.
5121 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5123 * config/arm/arm-cpus.in (armv8_4): New feature.
5124 (ARMv8_4a): New fgroup.
5125 (armv8.4-a): New arch.
5126 * config/arm/arm-tables.opt: Regenerate.
5127 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
5128 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
5129 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
5130 Add matching rules for -march=armv8.4-a and extensions.
5131 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
5133 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
5136 * config/rx/rx.md (BW): New mode attribute.
5137 (sync_lock_test_and_setsi): Add mode suffix to insn output.
5139 2018-01-11 Richard Biener <rguenther@suse.de>
5141 PR tree-optimization/83435
5142 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
5143 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
5144 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
5146 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5147 Alan Hayward <alan.hayward@arm.com>
5148 David Sherwood <david.sherwood@arm.com>
5150 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
5152 (aarch64_classify_address): Initialize it. Track polynomial offsets.
5153 (aarch64_print_address_internal): Use it to check for a zero offset.
5155 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5156 Alan Hayward <alan.hayward@arm.com>
5157 David Sherwood <david.sherwood@arm.com>
5159 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
5160 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
5161 Return a poly_int64 rather than a HOST_WIDE_INT.
5162 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
5163 rather than a HOST_WIDE_INT.
5164 * config/aarch64/aarch64.h (aarch64_frame): Protect with
5165 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
5166 hard_fp_offset, frame_size, initial_adjust, callee_offset and
5167 final_offset from HOST_WIDE_INT to poly_int64.
5168 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
5169 to_constant when getting the number of units in an Advanced SIMD
5171 (aarch64_builtin_vectorized_function): Check for a constant number
5173 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
5175 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
5176 attribute instead of GET_MODE_NUNITS.
5177 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
5178 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
5179 GET_MODE_SIZE for fixed-size registers.
5180 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
5181 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
5182 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
5183 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
5184 (aarch64_print_operand, aarch64_print_address_internal)
5185 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
5186 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
5187 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
5188 Handle polynomial GET_MODE_SIZE.
5189 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
5190 wider than SImode without modification.
5191 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
5192 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
5193 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
5194 passing and returning SVE modes.
5195 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
5196 rather than GEN_INT.
5197 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
5198 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
5199 (aarch64_allocate_and_probe_stack_space): Likewise.
5200 (aarch64_layout_frame): Cope with polynomial offsets.
5201 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
5202 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
5204 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
5205 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
5206 poly_int64 rather than a HOST_WIDE_INT.
5207 (aarch64_get_separate_components, aarch64_process_components)
5208 (aarch64_expand_prologue, aarch64_expand_epilogue)
5209 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
5210 (aarch64_anchor_offset): New function, split out from...
5211 (aarch64_legitimize_address): ...here.
5212 (aarch64_builtin_vectorization_cost): Handle polynomial
5213 TYPE_VECTOR_SUBPARTS.
5214 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
5216 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
5217 number of elements from the PARALLEL rather than the mode.
5218 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
5219 rather than GET_MODE_BITSIZE.
5220 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
5221 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
5222 (aarch64_expand_vec_perm_const_1): Handle polynomial
5223 d->perm.length () and d->perm elements.
5224 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
5225 Apply to_constant to d->perm elements.
5226 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
5227 polynomial CONST_VECTOR_NUNITS.
5228 (aarch64_move_pointer): Take amount as a poly_int64 rather
5230 (aarch64_progress_pointer): Avoid temporary variable.
5231 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
5232 the mode attribute instead of GET_MODE.
5234 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5235 Alan Hayward <alan.hayward@arm.com>
5236 David Sherwood <david.sherwood@arm.com>
5238 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
5239 x exists before using it.
5240 (aarch64_add_constant_internal): Rename to...
5241 (aarch64_add_offset_1): ...this. Replace regnum with separate
5242 src and dest rtxes. Handle the case in which they're different,
5243 including when the offset is zero. Replace scratchreg with an rtx.
5244 Use 2 additions if there is no spare register into which we can
5245 move a 16-bit constant.
5246 (aarch64_add_constant): Delete.
5247 (aarch64_add_offset): Replace reg with separate src and dest
5248 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
5249 Use aarch64_add_offset_1.
5250 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
5251 an rtx rather than an int. Take the delta as a poly_int64
5252 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
5253 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
5254 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
5255 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
5256 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
5258 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
5259 aarch64_add_constant.
5261 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5263 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
5264 Use scalar_float_mode.
5266 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5268 * config/aarch64/aarch64-simd.md
5269 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
5270 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
5271 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
5272 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
5273 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
5274 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
5275 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
5276 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
5277 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
5278 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
5280 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5283 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
5284 targ_options->x_arm_arch_string is non NULL.
5286 2018-01-11 Tamar Christina <tamar.christina@arm.com>
5288 * config/aarch64/aarch64.h
5289 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
5291 2018-01-11 Sudakshina Das <sudi.das@arm.com>
5294 * expmed.c (emit_store_flag_force): Swap if const op0
5295 and change VOIDmode to mode of op0.
5297 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5299 PR rtl-optimization/83761
5300 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
5301 than bytes to mode_for_size.
5303 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5306 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
5307 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
5310 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5313 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
5314 when in layout mode.
5315 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
5316 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
5319 2018-01-10 Michael Collison <michael.collison@arm.com>
5321 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
5322 * config/aarch64/aarch64-option-extension.def: Add
5323 AARCH64_OPT_EXTENSION of 'fp16fml'.
5324 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5325 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
5326 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
5327 * config/aarch64/constraints.md (Ui7): New constraint.
5328 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
5329 (VFMLA_SEL_W): Ditto.
5332 (VFMLA16_LOW): New int iterator.
5333 (VFMLA16_HIGH): Ditto.
5334 (UNSPEC_FMLAL): New unspec.
5335 (UNSPEC_FMLSL): Ditto.
5336 (UNSPEC_FMLAL2): Ditto.
5337 (UNSPEC_FMLSL2): Ditto.
5338 (f16mac): New code attribute.
5339 * config/aarch64/aarch64-simd-builtins.def
5340 (aarch64_fmlal_lowv2sf): Ditto.
5341 (aarch64_fmlsl_lowv2sf): Ditto.
5342 (aarch64_fmlalq_lowv4sf): Ditto.
5343 (aarch64_fmlslq_lowv4sf): Ditto.
5344 (aarch64_fmlal_highv2sf): Ditto.
5345 (aarch64_fmlsl_highv2sf): Ditto.
5346 (aarch64_fmlalq_highv4sf): Ditto.
5347 (aarch64_fmlslq_highv4sf): Ditto.
5348 (aarch64_fmlal_lane_lowv2sf): Ditto.
5349 (aarch64_fmlsl_lane_lowv2sf): Ditto.
5350 (aarch64_fmlal_laneq_lowv2sf): Ditto.
5351 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
5352 (aarch64_fmlalq_lane_lowv4sf): Ditto.
5353 (aarch64_fmlsl_lane_lowv4sf): Ditto.
5354 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
5355 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
5356 (aarch64_fmlal_lane_highv2sf): Ditto.
5357 (aarch64_fmlsl_lane_highv2sf): Ditto.
5358 (aarch64_fmlal_laneq_highv2sf): Ditto.
5359 (aarch64_fmlsl_laneq_highv2sf): Ditto.
5360 (aarch64_fmlalq_lane_highv4sf): Ditto.
5361 (aarch64_fmlsl_lane_highv4sf): Ditto.
5362 (aarch64_fmlalq_laneq_highv4sf): Ditto.
5363 (aarch64_fmlsl_laneq_highv4sf): Ditto.
5364 * config/aarch64/aarch64-simd.md:
5365 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
5366 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5367 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
5368 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5369 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
5370 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
5371 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
5372 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
5373 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
5374 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
5375 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
5376 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
5377 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
5378 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
5379 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
5380 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
5381 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
5382 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
5383 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
5384 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
5385 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
5386 (vfmlsl_low_u32): Ditto.
5387 (vfmlalq_low_u32): Ditto.
5388 (vfmlslq_low_u32): Ditto.
5389 (vfmlal_high_u32): Ditto.
5390 (vfmlsl_high_u32): Ditto.
5391 (vfmlalq_high_u32): Ditto.
5392 (vfmlslq_high_u32): Ditto.
5393 (vfmlal_lane_low_u32): Ditto.
5394 (vfmlsl_lane_low_u32): Ditto.
5395 (vfmlal_laneq_low_u32): Ditto.
5396 (vfmlsl_laneq_low_u32): Ditto.
5397 (vfmlalq_lane_low_u32): Ditto.
5398 (vfmlslq_lane_low_u32): Ditto.
5399 (vfmlalq_laneq_low_u32): Ditto.
5400 (vfmlslq_laneq_low_u32): Ditto.
5401 (vfmlal_lane_high_u32): Ditto.
5402 (vfmlsl_lane_high_u32): Ditto.
5403 (vfmlal_laneq_high_u32): Ditto.
5404 (vfmlsl_laneq_high_u32): Ditto.
5405 (vfmlalq_lane_high_u32): Ditto.
5406 (vfmlslq_lane_high_u32): Ditto.
5407 (vfmlalq_laneq_high_u32): Ditto.
5408 (vfmlslq_laneq_high_u32): Ditto.
5409 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
5410 (AARCH64_FL_FOR_ARCH8_4): New.
5411 (AARCH64_ISA_F16FML): New ISA flag.
5412 (TARGET_F16FML): New feature flag for fp16fml.
5413 (doc/invoke.texi): Document new fp16fml option.
5415 2018-01-10 Michael Collison <michael.collison@arm.com>
5417 * config/aarch64/aarch64-builtins.c:
5418 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
5419 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5420 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
5421 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
5422 (AARCH64_ISA_SHA3): New ISA flag.
5423 (TARGET_SHA3): New feature flag for sha3.
5424 * config/aarch64/iterators.md (sha512_op): New int attribute.
5425 (CRYPTO_SHA512): New int iterator.
5426 (UNSPEC_SHA512H): New unspec.
5427 (UNSPEC_SHA512H2): Ditto.
5428 (UNSPEC_SHA512SU0): Ditto.
5429 (UNSPEC_SHA512SU1): Ditto.
5430 * config/aarch64/aarch64-simd-builtins.def
5431 (aarch64_crypto_sha512hqv2di): New builtin.
5432 (aarch64_crypto_sha512h2qv2di): Ditto.
5433 (aarch64_crypto_sha512su0qv2di): Ditto.
5434 (aarch64_crypto_sha512su1qv2di): Ditto.
5435 (aarch64_eor3qv8hi): Ditto.
5436 (aarch64_rax1qv2di): Ditto.
5437 (aarch64_xarqv2di): Ditto.
5438 (aarch64_bcaxqv8hi): Ditto.
5439 * config/aarch64/aarch64-simd.md:
5440 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
5441 (aarch64_crypto_sha512su0qv2di): Ditto.
5442 (aarch64_crypto_sha512su1qv2di): Ditto.
5443 (aarch64_eor3qv8hi): Ditto.
5444 (aarch64_rax1qv2di): Ditto.
5445 (aarch64_xarqv2di): Ditto.
5446 (aarch64_bcaxqv8hi): Ditto.
5447 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
5448 (vsha512h2q_u64): Ditto.
5449 (vsha512su0q_u64): Ditto.
5450 (vsha512su1q_u64): Ditto.
5451 (veor3q_u16): Ditto.
5452 (vrax1q_u64): Ditto.
5454 (vbcaxq_u16): Ditto.
5455 * config/arm/types.md (crypto_sha512): New type attribute.
5456 (crypto_sha3): Ditto.
5457 (doc/invoke.texi): Document new sha3 option.
5459 2018-01-10 Michael Collison <michael.collison@arm.com>
5461 * config/aarch64/aarch64-builtins.c:
5462 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
5463 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5464 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
5465 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
5466 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
5467 (AARCH64_ISA_SM4): New ISA flag.
5468 (TARGET_SM4): New feature flag for sm4.
5469 * config/aarch64/aarch64-simd-builtins.def
5470 (aarch64_sm3ss1qv4si): Ditto.
5471 (aarch64_sm3tt1aq4si): Ditto.
5472 (aarch64_sm3tt1bq4si): Ditto.
5473 (aarch64_sm3tt2aq4si): Ditto.
5474 (aarch64_sm3tt2bq4si): Ditto.
5475 (aarch64_sm3partw1qv4si): Ditto.
5476 (aarch64_sm3partw2qv4si): Ditto.
5477 (aarch64_sm4eqv4si): Ditto.
5478 (aarch64_sm4ekeyqv4si): Ditto.
5479 * config/aarch64/aarch64-simd.md:
5480 (aarch64_sm3ss1qv4si): Ditto.
5481 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
5482 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
5483 (aarch64_sm4eqv4si): Ditto.
5484 (aarch64_sm4ekeyqv4si): Ditto.
5485 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
5486 (sm3part_op): Ditto.
5487 (CRYPTO_SM3TT): Ditto.
5488 (CRYPTO_SM3PART): Ditto.
5489 (UNSPEC_SM3SS1): New unspec.
5490 (UNSPEC_SM3TT1A): Ditto.
5491 (UNSPEC_SM3TT1B): Ditto.
5492 (UNSPEC_SM3TT2A): Ditto.
5493 (UNSPEC_SM3TT2B): Ditto.
5494 (UNSPEC_SM3PARTW1): Ditto.
5495 (UNSPEC_SM3PARTW2): Ditto.
5496 (UNSPEC_SM4E): Ditto.
5497 (UNSPEC_SM4EKEY): Ditto.
5498 * config/aarch64/constraints.md (Ui2): New constraint.
5499 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
5500 * config/arm/types.md (crypto_sm3): New type attribute.
5501 (crypto_sm4): Ditto.
5502 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
5503 (vsm3tt1aq_u32): Ditto.
5504 (vsm3tt1bq_u32): Ditto.
5505 (vsm3tt2aq_u32): Ditto.
5506 (vsm3tt2bq_u32): Ditto.
5507 (vsm3partw1q_u32): Ditto.
5508 (vsm3partw2q_u32): Ditto.
5509 (vsm4eq_u32): Ditto.
5510 (vsm4ekeyq_u32): Ditto.
5511 (doc/invoke.texi): Document new sm4 option.
5513 2018-01-10 Michael Collison <michael.collison@arm.com>
5515 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
5516 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
5517 (AARCH64_FL_FOR_ARCH8_4): New.
5518 (AARCH64_FL_V8_4): New flag.
5519 (doc/invoke.texi): Document new armv8.4-a option.
5521 2018-01-10 Michael Collison <michael.collison@arm.com>
5523 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5524 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
5525 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
5526 * config/aarch64/aarch64-option-extension.def: Add
5527 AARCH64_OPT_EXTENSION of 'sha2'.
5528 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
5529 (crypto): Disable sha2 and aes if crypto disabled.
5530 (crypto): Enable aes and sha2 if enabled.
5531 (simd): Disable sha2 and aes if simd disabled.
5532 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
5534 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
5535 (TARGET_SHA2): New feature flag for sha2.
5536 (TARGET_AES): New feature flag for aes.
5537 * config/aarch64/aarch64-simd.md:
5538 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
5539 conditional on TARGET_AES.
5540 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
5541 (aarch64_crypto_sha1hsi): Make pattern conditional
5543 (aarch64_crypto_sha1hv4si): Ditto.
5544 (aarch64_be_crypto_sha1hv4si): Ditto.
5545 (aarch64_crypto_sha1su1v4si): Ditto.
5546 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
5547 (aarch64_crypto_sha1su0v4si): Ditto.
5548 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
5549 (aarch64_crypto_sha256su0v4si): Ditto.
5550 (aarch64_crypto_sha256su1v4si): Ditto.
5551 (doc/invoke.texi): Document new aes and sha2 options.
5553 2018-01-10 Martin Sebor <msebor@redhat.com>
5555 PR tree-optimization/83781
5556 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
5559 2018-01-11 Martin Sebor <msebor@gmail.com>
5560 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5562 PR tree-optimization/83501
5563 PR tree-optimization/81703
5565 * tree-ssa-strlen.c (get_string_cst): Rename...
5566 (get_string_len): ...to this. Handle global constants.
5567 (handle_char_store): Adjust.
5569 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
5570 Jim Wilson <jimw@sifive.com>
5572 * config/riscv/riscv-protos.h (riscv_output_return): New.
5573 * config/riscv/riscv.c (struct machine_function): New naked_p field.
5574 (riscv_attribute_table, riscv_output_return),
5575 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
5576 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
5577 (riscv_compute_frame_info): Only compute frame->mask if not a naked
5579 (riscv_expand_prologue): Add early return for naked function.
5580 (riscv_expand_epilogue): Likewise.
5581 (riscv_function_ok_for_sibcall): Return false for naked function.
5582 (riscv_set_current_function): New.
5583 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
5584 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
5585 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
5586 * doc/extend.texi (RISC-V Function Attributes): New.
5588 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
5590 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
5591 check for 128-bit long double before checking TCmode.
5592 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
5593 128-bit long doubles before checking TFmode or TCmode.
5594 (FLOAT128_IBM_P): Likewise.
5596 2018-01-10 Martin Sebor <msebor@redhat.com>
5598 PR tree-optimization/83671
5599 * builtins.c (c_strlen): Unconditionally return zero for the empty
5601 Use -Warray-bounds for warnings.
5602 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
5603 for non-constant array indices with COMPONENT_REF, arrays of
5604 arrays, and pointers to arrays.
5605 (gimple_fold_builtin_strlen): Determine and set length range for
5606 non-constant character arrays.
5608 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
5611 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
5614 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
5616 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
5618 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5621 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
5622 VECTOR_MEM_ALTIVEC_OR_VSX_P.
5623 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
5624 indexed_or_indirect_operand predicate.
5625 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
5626 (*vsx_le_perm_load_v8hi): Likewise.
5627 (*vsx_le_perm_load_v16qi): Likewise.
5628 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
5629 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
5630 (*vsx_le_perm_store_v8hi): Likewise.
5631 (*vsx_le_perm_store_v16qi): Likewise.
5632 (eight unnamed splitters): Likewise.
5634 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5636 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
5637 * config/rs6000/emmintrin.h: Likewise.
5638 * config/rs6000/mmintrin.h: Likewise.
5639 * config/rs6000/xmmintrin.h: Likewise.
5641 2018-01-10 David Malcolm <dmalcolm@redhat.com>
5644 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
5646 * tree.c (tree_nop_conversion): Return true for location wrapper
5648 (maybe_wrap_with_location): New function.
5649 (selftest::check_strip_nops): New function.
5650 (selftest::test_location_wrappers): New function.
5651 (selftest::tree_c_tests): Call it.
5652 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
5653 (maybe_wrap_with_location): New decl.
5654 (EXPR_LOCATION_WRAPPER_P): New macro.
5655 (location_wrapper_p): New inline function.
5656 (tree_strip_any_location_wrapper): New inline function.
5658 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
5661 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
5662 stack_realign_offset for the largest alignment of stack slot
5664 (ix86_find_max_used_stack_alignment): New function.
5665 (ix86_finalize_stack_frame_flags): Use it. Set
5666 max_used_stack_alignment if we don't realign stack.
5667 * config/i386/i386.h (machine_function): Add
5668 max_used_stack_alignment.
5670 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
5672 * config/arm/arm.opt (-mbranch-cost): New option.
5673 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
5676 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
5679 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
5680 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
5682 2018-01-10 Richard Biener <rguenther@suse.de>
5685 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
5686 early out so it also covers the case where we have a non-NULL
5689 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
5691 PR tree-optimization/83753
5692 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
5693 for non-strided grouped accesses if the number of elements is 1.
5695 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5698 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
5699 * i386.h (TARGET_USE_GATHER): Define.
5700 * x86-tune.def (X86_TUNE_USE_GATHER): New.
5702 2018-01-10 Martin Liska <mliska@suse.cz>
5705 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
5706 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
5708 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
5709 CLEANUP_NO_PARTITIONING is not set.
5711 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
5713 * doc/rtl.texi: Remove documentation of (const ...) wrappers
5714 for vectors, as a partial revert of r254296.
5715 * rtl.h (const_vec_p): Delete.
5716 (const_vec_duplicate_p): Don't test for vector CONSTs.
5717 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
5718 * expmed.c (make_tree): Likewise.
5721 * common.md (E, F): Use CONSTANT_P instead of checking for
5723 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
5724 checking for CONST_VECTOR.
5726 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5729 * predict.c (force_edge_cold): Handle in more sane way edges
5732 2018-01-09 Carl Love <cel@us.ibm.com>
5734 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
5736 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
5737 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
5738 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
5739 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
5740 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
5741 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
5742 * config/rs6000/rs6000-protos.h: Add extern defition for
5743 rs6000_generate_float2_double_code.
5744 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
5746 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
5747 (float2_v2df): Add define_expand.
5749 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
5752 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
5753 op_mode in the force_to_mode call.
5755 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5757 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
5758 instead of checking each element individually.
5759 (aarch64_evpc_uzp): Likewise.
5760 (aarch64_evpc_zip): Likewise.
5761 (aarch64_evpc_ext): Likewise.
5762 (aarch64_evpc_rev): Likewise.
5763 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
5764 instead of checking each element individually. Return true without
5766 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
5767 whether all selected elements come from the same input, instead of
5768 checking each element individually. Remove calls to gen_rtx_REG,
5769 start_sequence and end_sequence and instead assert that no rtl is
5772 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5774 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
5775 order of HIGH and CONST checks.
5777 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5779 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
5780 if the destination isn't an SSA_NAME.
5782 2018-01-09 Richard Biener <rguenther@suse.de>
5784 PR tree-optimization/83668
5785 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
5787 (canonicalize_loop_form): ... here, renamed from ...
5788 (canonicalize_loop_closed_ssa_form): ... this and amended to
5789 swap successor edges for loop exit blocks to make us use
5790 the RPO order we need for initial schedule generation.
5792 2018-01-09 Joseph Myers <joseph@codesourcery.com>
5794 PR tree-optimization/64811
5795 * match.pd: When optimizing comparisons with Inf, avoid
5796 introducing or losing exceptions from comparisons with NaN.
5798 2018-01-09 Martin Liska <mliska@suse.cz>
5801 * asan.c (shadow_mem_size): Add gcc_assert.
5803 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
5805 Don't save registers in main().
5808 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
5809 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
5810 * config/avr/avr.c (avr_set_current_function): Don't error if
5811 naked, OS_task or OS_main are specified at the same time.
5812 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
5814 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
5816 * common/config/avr/avr-common.c (avr_option_optimization_table):
5817 Switch on -mmain-is-OS_task for optimizing compilations.
5819 2018-01-09 Richard Biener <rguenther@suse.de>
5821 PR tree-optimization/83572
5822 * graphite.c: Include cfganal.h.
5823 (graphite_transform_loops): Connect infinite loops to exit
5824 and remove fake edges at the end.
5826 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5828 * ipa-inline.c (edge_badness): Revert accidental checkin.
5830 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5833 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
5834 symbols; not inline clones.
5836 2018-01-09 Jakub Jelinek <jakub@redhat.com>
5839 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
5840 hard registers. Formatting fixes.
5842 PR preprocessor/83722
5843 * gcc.c (try_generate_repro): Pass
5844 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
5845 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
5848 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
5849 Kito Cheng <kito.cheng@gmail.com>
5851 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
5852 (riscv_leaf_function_p): Delete.
5853 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
5855 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5857 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
5859 (do_ifelse): New function.
5860 (do_isel): New function.
5861 (do_sub3): New function.
5862 (do_add3): New function.
5863 (do_load_mask_compare): New function.
5864 (do_overlap_load_compare): New function.
5865 (expand_compare_loop): New function.
5866 (expand_block_compare): Call expand_compare_loop() when appropriate.
5867 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
5869 (-mblock-compare-inline-loop-limit): New option.
5871 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5874 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
5875 Reverse order of second and third operands in first alternative.
5876 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
5877 of first and second elements in UNSPEC_VPERMR vector.
5878 (altivec_expand_vec_perm_le): Likewise.
5880 2018-01-08 Jeff Law <law@redhat.com>
5882 PR rtl-optimizatin/81308
5883 * tree-switch-conversion.c (cfg_altered): New file scoped static.
5884 (process_switch): If group_case_labels makes a change, then set
5886 (pass_convert_switch::execute): If a switch is converted, then
5887 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
5889 PR rtl-optimization/81308
5890 * recog.c (split_all_insns): Conditionally cleanup the CFG after
5893 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
5895 PR target/83663 - Revert r255946
5896 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
5897 generation for cases where splatting a value is not useful.
5898 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
5899 across a vec_duplicate and a paradoxical subreg forming a vector
5900 mode to a vec_concat.
5902 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5904 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
5905 -march=armv8.3-a variants.
5906 * config/arm/t-multilib: Likewise.
5907 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
5909 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5911 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
5913 (cceq_ior_compare_complement): Give it a name so I can use it, and
5914 change boolean_or_operator predicate to boolean_operator so it can
5915 be used to generate a crand.
5916 (eqne): New code iterator.
5917 (bd/bd_neg): New code_attrs.
5918 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
5919 a single define_insn.
5920 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
5921 decrement (bdnzt/bdnzf/bdzt/bdzf).
5922 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
5923 with the new names of the branch decrement patterns, and added the
5924 names of the branch decrement conditional patterns.
5926 2018-01-08 Richard Biener <rguenther@suse.de>
5928 PR tree-optimization/83563
5929 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
5932 2018-01-08 Richard Biener <rguenther@suse.de>
5935 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
5937 2018-01-08 Richard Biener <rguenther@suse.de>
5939 PR tree-optimization/83685
5940 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
5941 references to abnormals.
5943 2018-01-08 Richard Biener <rguenther@suse.de>
5946 * dwarf2out.c (output_indirect_strings): Handle empty
5947 skeleton_debug_str_hash.
5948 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
5950 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
5952 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
5953 (emit_store_direct): Likewise.
5954 (arc_trampoline_adjust_address): Likewise.
5955 (arc_asm_trampoline_template): New function.
5956 (arc_initialize_trampoline): Use asm_trampoline_template.
5957 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
5958 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
5959 * config/arc/arc.md (flush_icache): Delete pattern.
5961 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
5963 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
5964 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
5967 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
5970 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
5971 by not USED_FOR_TARGET.
5972 (make_pass_resolve_sw_modes): Likewise.
5974 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
5976 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
5979 2018-01-08 Richard Biener <rguenther@suse.de>
5982 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
5984 2018-01-08 Richard Biener <rguenther@suse.de>
5987 * match.pd ((t * 2) / 2) -> t): Add missing :c.
5989 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
5992 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
5993 basic blocks with a small number of successors.
5994 (convert_control_dep_chain_into_preds): Improve handling of
5996 (dump_predicates): Split apart into...
5997 (dump_pred_chain): ...here...
5998 (dump_pred_info): ...and here.
5999 (can_one_predicate_be_invalidated_p): Add debugging printfs.
6000 (can_chain_union_be_invalidated_p): Improve check for invalidation
6002 (uninit_uses_cannot_happen): Avoid unnecessary if
6003 convert_control_dep_chain_into_preds yielded nothing.
6005 2018-01-06 Martin Sebor <msebor@redhat.com>
6007 PR tree-optimization/83640
6008 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
6009 subtracting negative offset from size.
6010 (builtin_access::overlap): Adjust offset bounds of the access to fall
6011 within the size of the object if possible.
6013 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
6015 PR rtl-optimization/83699
6016 * expmed.c (extract_bit_field_1): Restrict the vector usage of
6017 extract_bit_field_as_subreg to cases in which the extracted
6018 value is also a vector.
6020 * lra-constraints.c (process_alt_operands): Test for the equivalence
6021 substitutions when detecting a possible reload cycle.
6023 2018-01-06 Jakub Jelinek <jakub@redhat.com>
6026 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
6027 by default if flag_selective_schedling{,2}. Formatting fixes.
6029 PR rtl-optimization/83682
6030 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
6031 if it has non-VECTOR_MODE element mode.
6032 (vec_duplicate_p): Likewise.
6035 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
6036 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
6038 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6041 * config/i386/i386-builtin.def
6042 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
6043 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
6044 Require also OPTION_MASK_ISA_AVX512F in addition to
6045 OPTION_MASK_ISA_GFNI.
6046 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
6047 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
6048 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
6049 to OPTION_MASK_ISA_GFNI.
6050 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
6051 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
6052 OPTION_MASK_ISA_AVX512BW.
6053 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
6054 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
6055 addition to OPTION_MASK_ISA_GFNI.
6056 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
6057 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
6058 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
6059 to OPTION_MASK_ISA_GFNI.
6060 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
6061 a requirement for all ISAs rather than any of them with a few
6063 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
6065 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
6066 bitmasks to be enabled with 3 exceptions, instead of requiring any
6067 enabled ISA with lots of exceptions.
6068 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
6069 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
6070 Change avx512bw in isa attribute to avx512f.
6071 * config/i386/sgxintrin.h: Add license boilerplate.
6072 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
6073 to __AVX512F__ and __AVX512VL to __AVX512VL__.
6074 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
6075 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
6077 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
6078 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
6079 temporarily sse2 rather than sse if not enabled already.
6082 * config/i386/sse.md (VI248_VLBW): Rename to ...
6083 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
6084 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
6085 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
6086 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
6087 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
6088 mode iterator instead of VI248_VLBW.
6090 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
6092 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
6093 (record_modified): Skip clobbers; add debug output.
6094 (param_change_prob): Use sreal frequencies.
6096 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6098 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
6099 punt for user-aligned variables.
6101 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6103 * tree-chrec.c (chrec_contains_symbols): Return true for
6106 2018-01-05 Sudakshina Das <sudi.das@arm.com>
6109 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
6110 of (x|y) == x for BICS pattern.
6112 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6114 PR tree-optimization/83605
6115 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
6116 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
6119 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
6121 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
6122 * config/epiphany/rtems.h: New file.
6124 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6125 Uros Bizjak <ubizjak@gmail.com>
6128 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
6129 QIreg_operand instead of register_operand predicate.
6130 * config/i386/i386.c (ix86_rop_should_change_byte_p,
6131 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
6132 comments instead of -fmitigate[-_]rop.
6134 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6137 * cgraphunit.c (symbol_table::compile): Switch to text_section
6138 before calling assembly_start debug hook.
6139 * run-rtl-passes.c (run_rtl_passes): Likewise.
6142 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6144 * tree-vrp.c (extract_range_from_binary_expr_1): Check
6145 range_int_cst_p rather than !symbolic_range_p before calling
6146 extract_range_from_multiplicative_op_1.
6148 2018-01-04 Jeff Law <law@redhat.com>
6150 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
6151 redundant test in assertion.
6153 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6155 * doc/rtl.texi: Document machine_mode wrapper classes.
6157 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6159 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
6162 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6164 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
6165 the VEC_PERM_EXPR fold to fail.
6167 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6170 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
6171 to switched_sections.
6173 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6176 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
6179 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
6182 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
6183 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
6185 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6188 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
6189 is BLKmode and bitpos not zero or mode change is needed.
6191 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6194 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
6197 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
6200 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
6201 instead of MULT rtx. Update all corresponding splitters.
6203 (*ssub<modesuffix>): Ditto.
6205 (*cmp_sadd_di): Update split patterns.
6206 (*cmp_sadd_si): Ditto.
6207 (*cmp_sadd_sidi): Ditto.
6208 (*cmp_ssub_di): Ditto.
6209 (*cmp_ssub_si): Ditto.
6210 (*cmp_ssub_sidi): Ditto.
6211 * config/alpha/predicates.md (const23_operand): New predicate.
6212 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
6213 Look for ASHIFT, not MULT inner operand.
6214 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
6216 2018-01-04 Martin Liska <mliska@suse.cz>
6218 PR gcov-profile/83669
6219 * gcov.c (output_intermediate_file): Add version to intermediate
6221 * doc/gcov.texi: Document new field 'version' in intermediate
6222 file format. Fix location of '-k' option of gcov command.
6224 2018-01-04 Martin Liska <mliska@suse.cz>
6227 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
6229 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6231 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
6233 2018-01-03 Martin Sebor <msebor@redhat.com>
6235 PR tree-optimization/83655
6236 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
6237 checking calls with invalid arguments.
6239 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6241 * tree-vect-stmts.c (vect_get_store_rhs): New function.
6242 (vectorizable_mask_load_store): Delete.
6243 (vectorizable_call): Return false for masked loads and stores.
6244 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
6245 instead of gimple_assign_rhs1.
6246 (vectorizable_load): Handle IFN_MASK_LOAD.
6247 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
6249 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6251 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
6253 (vectorizable_mask_load_store): ...here.
6254 (vectorizable_load): ...and here.
6256 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6258 * tree-vect-stmts.c (vect_build_all_ones_mask)
6259 (vect_build_zero_merge_argument): New functions, split out from...
6260 (vectorizable_load): ...here.
6262 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6264 * tree-vect-stmts.c (vect_check_store_rhs): New function,
6266 (vectorizable_mask_load_store): ...here.
6267 (vectorizable_store): ...and here.
6269 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6271 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
6273 (vectorizable_mask_load_store): ...here.
6275 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6277 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
6278 (vect_model_store_cost): Take a vec_load_store_type instead of a
6280 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
6281 (vect_model_store_cost): Take a vec_load_store_type instead of a
6283 (vectorizable_mask_load_store): Update accordingly.
6284 (vectorizable_store): Likewise.
6285 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
6287 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6289 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
6290 IFN_MASK_LOAD calls here rather than...
6291 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
6293 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6294 Alan Hayward <alan.hayward@arm.com>
6295 David Sherwood <david.sherwood@arm.com>
6297 * expmed.c (extract_bit_field_1): For vector extracts,
6298 fall back to extract_bit_field_as_subreg if vec_extract
6301 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6302 Alan Hayward <alan.hayward@arm.com>
6303 David Sherwood <david.sherwood@arm.com>
6305 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
6306 they are variable or constant sized.
6307 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
6308 slots for constant-sized data.
6310 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6311 Alan Hayward <alan.hayward@arm.com>
6312 David Sherwood <david.sherwood@arm.com>
6314 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
6315 handling COND_EXPRs with boolean comparisons, try to find a better
6316 basis for the mask type than the boolean itself.
6318 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6320 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
6321 is calculated and how it can be overridden.
6322 * genmodes.c (max_bitsize_mode_any_mode): New variable.
6323 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
6325 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
6328 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6329 Alan Hayward <alan.hayward@arm.com>
6330 David Sherwood <david.sherwood@arm.com>
6332 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
6333 Remove the mode argument.
6334 (aarch64_simd_valid_immediate): Remove the mode and inverse
6336 * config/aarch64/iterators.md (bitsize): New iterator.
6337 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
6338 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
6339 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
6340 aarch64_simd_valid_immediate.
6341 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
6342 (aarch64_reg_or_bic_imm): Likewise.
6343 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
6344 with an insn_type enum and msl with a modifier_type enum.
6345 Replace element_width with a scalar_mode. Change the shift
6346 to unsigned int. Add constructors for scalar_float_mode and
6347 scalar_int_mode elements.
6348 (aarch64_vect_float_const_representable_p): Delete.
6349 (aarch64_can_const_movi_rtx_p)
6350 (aarch64_simd_scalar_immediate_valid_for_move)
6351 (aarch64_simd_make_constant): Update call to
6352 aarch64_simd_valid_immediate.
6353 (aarch64_advsimd_valid_immediate_hs): New function.
6354 (aarch64_advsimd_valid_immediate): Likewise.
6355 (aarch64_simd_valid_immediate): Remove mode and inverse
6356 arguments. Rewrite to use the above. Use const_vec_duplicate_p
6357 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
6358 and aarch64_float_const_representable_p on the result.
6359 (aarch64_output_simd_mov_immediate): Remove mode argument.
6360 Update call to aarch64_simd_valid_immediate and use of
6361 simd_immediate_info.
6362 (aarch64_output_scalar_simd_mov_immediate): Update call
6365 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6366 Alan Hayward <alan.hayward@arm.com>
6367 David Sherwood <david.sherwood@arm.com>
6369 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
6370 (mode_nunits): Likewise CONST_MODE_NUNITS.
6371 * machmode.def (ADJUST_NUNITS): Document.
6372 * genmodes.c (mode_data::need_nunits_adj): New field.
6373 (blank_mode): Update accordingly.
6374 (adj_nunits): New variable.
6375 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
6377 (emit_mode_size_inline): Set need_bytesize_adj for all modes
6378 listed in adj_nunits.
6379 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
6380 listed in adj_nunits. Don't emit case statements for such modes.
6381 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
6382 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
6383 nothing if adj_nunits is nonnull.
6384 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
6385 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
6386 (emit_mode_fbit): Update use of print_maybe_const_decl.
6387 (emit_move_size): Likewise. Treat the array as non-const
6389 (emit_mode_adjustments): Handle adj_nunits.
6391 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6393 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
6394 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
6395 (VECTOR_MODES): Use it.
6396 (make_vector_modes): Take the prefix as an argument.
6398 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6399 Alan Hayward <alan.hayward@arm.com>
6400 David Sherwood <david.sherwood@arm.com>
6402 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
6403 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
6404 for MODE_VECTOR_BOOL.
6405 * machmode.def (VECTOR_BOOL_MODE): Document.
6406 * genmodes.c (VECTOR_BOOL_MODE): New macro.
6407 (make_vector_bool_mode): New function.
6408 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
6410 * lto-streamer-in.c (lto_input_mode_table): Likewise.
6411 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
6413 * stor-layout.c (int_mode_for_mode): Likewise.
6414 * tree.c (build_vector_type_for_mode): Likewise.
6415 * varasm.c (output_constant_pool_2): Likewise.
6416 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
6417 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
6418 for MODE_VECTOR_BOOL.
6419 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
6420 of mode class checks.
6421 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
6422 instead of a list of mode class checks.
6423 (expand_vector_scalar_condition): Likewise.
6424 (type_for_widest_vector_mode): Handle BImode as an inner mode.
6426 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6427 Alan Hayward <alan.hayward@arm.com>
6428 David Sherwood <david.sherwood@arm.com>
6430 * machmode.h (mode_size): Change from unsigned short to
6432 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
6433 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6434 or if measurement_type is not polynomial.
6435 (fixed_size_mode::includes_p): Check for constant-sized modes.
6436 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
6437 return a poly_uint16 rather than an unsigned short.
6438 (emit_mode_size): Change the type of mode_size from unsigned short
6439 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
6440 (emit_mode_adjustments): Cope with polynomial vector sizes.
6441 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6443 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6445 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
6446 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
6447 * caller-save.c (setup_save_areas): Likewise.
6448 (replace_reg_with_saved_mem): Likewise.
6449 * calls.c (emit_library_call_value_1): Likewise.
6450 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
6451 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
6452 (gen_lowpart_for_combine): Likewise.
6453 * convert.c (convert_to_integer_1): Likewise.
6454 * cse.c (equiv_constant, cse_insn): Likewise.
6455 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
6456 (cselib_subst_to_values): Likewise.
6457 * dce.c (word_dce_process_block): Likewise.
6458 * df-problems.c (df_word_lr_mark_ref): Likewise.
6459 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
6460 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
6461 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
6462 (rtl_for_decl_location): Likewise.
6463 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
6464 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
6465 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
6466 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
6467 (expand_expr_real_1): Likewise.
6468 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
6469 (pad_below): Likewise.
6470 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6471 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
6472 * ira.c (get_subreg_tracking_sizes): Likewise.
6473 * ira-build.c (ira_create_allocno_objects): Likewise.
6474 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
6475 (ira_sort_regnos_for_alter_reg): Likewise.
6476 * ira-costs.c (record_operand_costs): Likewise.
6477 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
6478 (resolve_simple_move): Likewise.
6479 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
6480 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
6481 (lra_constraints): Likewise.
6482 (CONST_POOL_OK_P): Reject variable-sized modes.
6483 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
6484 (add_pseudo_to_slot, lra_spill): Likewise.
6485 * omp-low.c (omp_clause_aligned_alignment): Likewise.
6486 * optabs-query.c (get_best_extraction_insn): Likewise.
6487 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6488 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
6489 (expand_mult_highpart, valid_multiword_target_p): Likewise.
6490 * recog.c (offsettable_address_addr_space_p): Likewise.
6491 * regcprop.c (maybe_mode_change): Likewise.
6492 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
6493 * regrename.c (build_def_use): Likewise.
6494 * regstat.c (dump_reg_info): Likewise.
6495 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
6496 (find_reloads, find_reloads_subreg_address): Likewise.
6497 * reload1.c (eliminate_regs_1): Likewise.
6498 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
6499 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
6500 (simplify_binary_operation_1, simplify_subreg): Likewise.
6501 * targhooks.c (default_function_arg_padding): Likewise.
6502 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
6503 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
6504 (verify_gimple_assign_ternary): Likewise.
6505 * tree-inline.c (estimate_move_cost): Likewise.
6506 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6507 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
6508 (get_address_cost_ainc): Likewise.
6509 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
6510 (vect_supportable_dr_alignment): Likewise.
6511 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6512 (vectorizable_reduction): Likewise.
6513 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
6514 (vectorizable_operation, vectorizable_load): Likewise.
6515 * tree.c (build_same_sized_truth_vector_type): Likewise.
6516 * valtrack.c (cleanup_auto_inc_dec): Likewise.
6517 * var-tracking.c (emit_note_insn_var_location): Likewise.
6518 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
6519 (ADDR_VEC_ALIGN): Likewise.
6521 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6522 Alan Hayward <alan.hayward@arm.com>
6523 David Sherwood <david.sherwood@arm.com>
6525 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
6527 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6528 or if measurement_type is polynomial.
6529 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
6530 * combine.c (make_extraction): Likewise.
6531 * dse.c (find_shift_sequence): Likewise.
6532 * dwarf2out.c (mem_loc_descriptor): Likewise.
6533 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
6534 (extract_bit_field, extract_low_bits): Likewise.
6535 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
6536 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
6537 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
6538 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
6539 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6540 * reload.c (find_reloads): Likewise.
6541 * reload1.c (alter_reg): Likewise.
6542 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
6543 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
6544 * tree-if-conv.c (predicate_mem_writes): Likewise.
6545 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
6546 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
6547 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
6548 * valtrack.c (dead_debug_insert_temp): Likewise.
6549 * varasm.c (mergeable_constant_section): Likewise.
6550 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
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 * expr.c (expand_assignment): Cope with polynomial mode sizes
6557 when assigning to a CONCAT.
6559 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6560 Alan Hayward <alan.hayward@arm.com>
6561 David Sherwood <david.sherwood@arm.com>
6563 * machmode.h (mode_precision): Change from unsigned short to
6565 (mode_to_precision): Return a poly_uint16 rather than an unsigned
6567 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
6568 or if measurement_type is not polynomial.
6569 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
6570 in which the mode is already known to be a scalar_int_mode.
6571 * genmodes.c (emit_mode_precision): Change the type of mode_precision
6572 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
6574 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6575 for GET_MODE_PRECISION.
6576 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6577 for GET_MODE_PRECISION.
6578 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
6580 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
6581 (expand_field_assignment, make_extraction): Likewise.
6582 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
6583 (get_last_value): Likewise.
6584 * convert.c (convert_to_integer_1): Likewise.
6585 * cse.c (cse_insn): Likewise.
6586 * expr.c (expand_expr_real_1): Likewise.
6587 * lra-constraints.c (simplify_operand_subreg): Likewise.
6588 * optabs-query.c (can_atomic_load_p): Likewise.
6589 * optabs.c (expand_atomic_load): Likewise.
6590 (expand_atomic_store): Likewise.
6591 * ree.c (combine_reaching_defs): Likewise.
6592 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
6593 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
6594 * tree.h (type_has_mode_precision_p): Likewise.
6595 * ubsan.c (instrument_si_overflow): Likewise.
6597 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6598 Alan Hayward <alan.hayward@arm.com>
6599 David Sherwood <david.sherwood@arm.com>
6601 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
6602 polynomial numbers of units.
6603 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
6604 (valid_vector_subparts_p): New function.
6605 (build_vector_type): Remove temporary shim and take the number
6606 of units as a poly_uint64 rather than an int.
6607 (build_opaque_vector_type): Take the number of units as a
6608 poly_uint64 rather than an int.
6609 * tree.c (build_vector_from_ctor): Handle polynomial
6610 TYPE_VECTOR_SUBPARTS.
6611 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
6612 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
6613 (build_vector_from_val): If the number of units is variable,
6614 use build_vec_duplicate_cst for constant operands and
6615 VEC_DUPLICATE_EXPR otherwise.
6616 (make_vector_type): Remove temporary is_constant ().
6617 (build_vector_type, build_opaque_vector_type): Take the number of
6618 units as a poly_uint64 rather than an int.
6619 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
6621 * cfgexpand.c (expand_debug_expr): Likewise.
6622 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
6623 (store_constructor, expand_expr_real_1): Likewise.
6624 (const_scalar_mask_from_tree): Likewise.
6625 * fold-const-call.c (fold_const_reduction): Likewise.
6626 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
6627 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
6628 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
6629 (fold_relational_const): Likewise.
6630 (native_interpret_vector): Likewise. Change the size from an
6631 int to an unsigned int.
6632 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
6633 TYPE_VECTOR_SUBPARTS.
6634 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
6635 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
6636 duplicating a non-constant operand into a variable-length vector.
6637 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
6638 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
6639 * ipa-icf.c (sem_variable::equals): Likewise.
6640 * match.pd: Likewise.
6641 * omp-simd-clone.c (simd_clone_subparts): Likewise.
6642 * print-tree.c (print_node): Likewise.
6643 * stor-layout.c (layout_type): Likewise.
6644 * targhooks.c (default_builtin_vectorization_cost): Likewise.
6645 * tree-cfg.c (verify_gimple_comparison): Likewise.
6646 (verify_gimple_assign_binary): Likewise.
6647 (verify_gimple_assign_ternary): Likewise.
6648 (verify_gimple_assign_single): Likewise.
6649 * tree-pretty-print.c (dump_generic_node): Likewise.
6650 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6651 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
6652 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
6653 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
6654 (vect_shift_permute_load_chain): Likewise.
6655 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
6656 (expand_vector_condition, optimize_vector_constructor): Likewise.
6657 (lower_vec_perm, get_compute_type): Likewise.
6658 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6659 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
6660 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
6661 (vect_recog_mask_conversion_pattern): Likewise.
6662 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
6663 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
6664 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
6665 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
6666 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
6667 (vectorizable_shift, vectorizable_operation, vectorizable_store)
6668 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
6669 (supportable_widening_operation): Likewise.
6670 (supportable_narrowing_operation): Likewise.
6671 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
6673 * varasm.c (output_constant): Likewise.
6675 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6676 Alan Hayward <alan.hayward@arm.com>
6677 David Sherwood <david.sherwood@arm.com>
6679 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
6680 so that both the length == 3 and length != 3 cases set up their
6681 own permute vectors. Add comments explaining why we know the
6682 number of elements is constant.
6683 (vect_permute_load_chain): Likewise.
6685 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6686 Alan Hayward <alan.hayward@arm.com>
6687 David Sherwood <david.sherwood@arm.com>
6689 * machmode.h (mode_nunits): Change from unsigned char to
6691 (ONLY_FIXED_SIZE_MODES): New macro.
6692 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
6693 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
6694 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
6696 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
6697 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
6698 or if measurement_type is not polynomial.
6699 * genmodes.c (ZERO_COEFFS): New macro.
6700 (emit_mode_nunits_inline): Make mode_nunits_inline return a
6702 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
6703 Use ZERO_COEFFS when emitting initializers.
6704 * data-streamer.h (bp_pack_poly_value): New function.
6705 (bp_unpack_poly_value): Likewise.
6706 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6707 for GET_MODE_NUNITS.
6708 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6709 for GET_MODE_NUNITS.
6710 * tree.c (make_vector_type): Remove temporary shim and make
6711 the real function take the number of units as a poly_uint64
6713 (build_vector_type_for_mode): Handle polynomial nunits.
6714 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
6715 * emit-rtl.c (const_vec_series_p_1): Likewise.
6716 (gen_rtx_CONST_VECTOR): Likewise.
6717 * fold-const.c (test_vec_duplicate_folding): Likewise.
6718 * genrecog.c (validate_pattern): Likewise.
6719 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
6720 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6721 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
6722 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
6723 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
6724 * rtlanal.c (subreg_get_info): Likewise.
6725 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
6726 (vect_grouped_load_supported): Likewise.
6727 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
6728 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
6729 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
6730 (simplify_const_unary_operation, simplify_binary_operation_1)
6731 (simplify_const_binary_operation, simplify_ternary_operation)
6732 (test_vector_ops_duplicate, test_vector_ops): Likewise.
6733 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
6734 instead of CONST_VECTOR_NUNITS.
6735 * varasm.c (output_constant_pool_2): Likewise.
6736 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
6737 explicit-encoded elements in the XVEC for variable-length vectors.
6739 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6741 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
6743 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6744 Alan Hayward <alan.hayward@arm.com>
6745 David Sherwood <david.sherwood@arm.com>
6747 * coretypes.h (fixed_size_mode): Declare.
6748 (fixed_size_mode_pod): New typedef.
6749 * builtins.h (target_builtins::x_apply_args_mode)
6750 (target_builtins::x_apply_result_mode): Change type to
6751 fixed_size_mode_pod.
6752 * builtins.c (apply_args_size, apply_result_size, result_vector)
6753 (expand_builtin_apply_args_1, expand_builtin_apply)
6754 (expand_builtin_return): Update accordingly.
6756 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6758 * cse.c (hash_rtx_cb): Hash only the encoded elements.
6759 * cselib.c (cselib_hash_rtx): Likewise.
6760 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
6761 CONST_VECTOR encoding.
6763 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6764 Jeff Law <law@redhat.com>
6767 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
6768 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
6769 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
6770 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
6773 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
6774 explicitly probe *sp in a noreturn function if there were any callee
6775 register saves or frame pointer is needed.
6777 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6780 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
6781 BLKmode for ternary, binary or unary expressions.
6784 * var-tracking.c (delete_vta_debug_insn): New inline function.
6785 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
6786 insns from get_insns () to NULL instead of each bb separately.
6787 Use delete_vta_debug_insn. No longer static.
6788 (vt_debug_insns_local, variable_tracking_main_1): Adjust
6789 delete_vta_debug_insns callers.
6790 * rtl.h (delete_vta_debug_insns): Declare.
6791 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
6792 instead of variable_tracking_main.
6794 2018-01-03 Martin Sebor <msebor@redhat.com>
6796 PR tree-optimization/83603
6797 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
6798 arguments past the endof the argument list in functions declared
6799 without a prototype.
6800 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
6801 Avoid checking when arguments are null.
6803 2018-01-03 Martin Sebor <msebor@redhat.com>
6806 * doc/extend.texi (attribute const): Fix a typo.
6807 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
6808 issuing -Wsuggest-attribute for void functions.
6810 2018-01-03 Martin Sebor <msebor@redhat.com>
6812 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
6813 offset_int::from instead of wide_int::to_shwi.
6814 (maybe_diag_overlap): Remove assertion.
6815 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
6816 * gimple-ssa-sprintf.c (format_directive): Same.
6817 (parse_directive): Same.
6818 (sprintf_dom_walker::compute_format_length): Same.
6819 (try_substitute_return_value): Same.
6821 2018-01-03 Jeff Law <law@redhat.com>
6824 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
6825 non-constant residual for zero at runtime and avoid probing in
6826 that case. Reorganize code for trailing problem to mirror handling
6829 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6831 PR tree-optimization/83501
6832 * tree-ssa-strlen.c (get_string_cst): New.
6833 (handle_char_store): Call get_string_cst.
6835 2018-01-03 Martin Liska <mliska@suse.cz>
6837 PR tree-optimization/83593
6838 * tree-ssa-strlen.c: Include tree-cfg.h.
6839 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
6840 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
6841 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
6843 (strlen_dom_walker::before_dom_children): Call
6844 gimple_purge_dead_eh_edges. Dump tranformation with details
6846 (strlen_dom_walker::before_dom_children): Update call by adding
6847 new argument cleanup_eh.
6848 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
6850 2018-01-03 Martin Liska <mliska@suse.cz>
6853 * cif-code.def (VARIADIC_THUNK): New enum value.
6854 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
6857 2018-01-03 Jan Beulich <jbeulich@suse.com>
6859 * sse.md (mov<mode>_internal): Tighten condition for when to use
6860 vmovdqu<ssescalarsize> for TI and OI modes.
6862 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6864 Update copyright years.
6866 2018-01-03 Martin Liska <mliska@suse.cz>
6869 * ipa-visibility.c (function_and_variable_visibility): Skip
6870 functions with noipa attribure.
6872 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6874 * gcc.c (process_command): Update copyright notice dates.
6875 * gcov-dump.c (print_version): Ditto.
6876 * gcov.c (print_version): Ditto.
6877 * gcov-tool.c (print_version): Ditto.
6878 * gengtype.c (create_file): Ditto.
6879 * doc/cpp.texi: Bump @copying's copyright year.
6880 * doc/cppinternals.texi: Ditto.
6881 * doc/gcc.texi: Ditto.
6882 * doc/gccint.texi: Ditto.
6883 * doc/gcov.texi: Ditto.
6884 * doc/install.texi: Ditto.
6885 * doc/invoke.texi: Ditto.
6887 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6889 * vector-builder.h (vector_builder::m_full_nelts): Change from
6890 unsigned int to poly_uint64.
6891 (vector_builder::full_nelts): Update prototype accordingly.
6892 (vector_builder::new_vector): Likewise.
6893 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
6894 (vector_builder::operator ==): Likewise.
6895 (vector_builder::finalize): Likewise.
6896 * int-vector-builder.h (int_vector_builder::int_vector_builder):
6897 Take the number of elements as a poly_uint64 rather than an
6899 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
6900 from unsigned int to poly_uint64.
6901 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
6902 (vec_perm_indices::new_vector): Likewise.
6903 (vec_perm_indices::length): Likewise.
6904 (vec_perm_indices::nelts_per_input): Likewise.
6905 (vec_perm_indices::input_nelts): Likewise.
6906 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
6907 number of elements per input as a poly_uint64 rather than an
6908 unsigned int. Use the original encoding for variable-length
6909 vectors, rather than clamping each individual element.
6910 For the second and subsequent elements in each pattern,
6911 clamp the step and base before clamping their sum.
6912 (vec_perm_indices::series_p): Handle polynomial element counts.
6913 (vec_perm_indices::all_in_range_p): Likewise.
6914 (vec_perm_indices_to_tree): Likewise.
6915 (vec_perm_indices_to_rtx): Likewise.
6916 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
6917 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
6918 (tree_vector_builder::new_binary_operation): Handle polynomial
6919 element counts. Return false if we need to know the number
6920 of elements at compile time.
6921 * fold-const.c (fold_vec_perm): Punt if the number of elements
6922 isn't known at compile time.
6924 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6926 * vec-perm-indices.h (vec_perm_builder): Change element type
6927 from HOST_WIDE_INT to poly_int64.
6928 (vec_perm_indices::element_type): Update accordingly.
6929 (vec_perm_indices::clamp): Handle polynomial element_types.
6930 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
6931 (vec_perm_indices::all_in_range_p): Likewise.
6932 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
6934 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
6935 polynomial vec_perm_indices element types.
6936 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
6937 * fold-const.c (fold_vec_perm): Likewise.
6938 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
6939 * tree-vect-generic.c (lower_vec_perm): Likewise.
6940 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
6941 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
6942 element type to HOST_WIDE_INT.
6944 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6945 Alan Hayward <alan.hayward@arm.com>
6946 David Sherwood <david.sherwood@arm.com>
6948 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
6949 rather than an int. Use plus_constant.
6950 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
6951 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
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 (emit_call_1, expand_call): Change struct_value_size from
6958 a HOST_WIDE_INT to a poly_int64.
6960 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6961 Alan Hayward <alan.hayward@arm.com>
6962 David Sherwood <david.sherwood@arm.com>
6964 * calls.c (load_register_parameters): Cope with polynomial
6965 mode sizes. Require a constant size for BLKmode parameters
6966 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
6967 forces a parameter to be padded at the lsb end in order to
6968 fill a complete number of words, require the parameter size
6969 to be ordered wrt UNITS_PER_WORD.
6971 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6972 Alan Hayward <alan.hayward@arm.com>
6973 David Sherwood <david.sherwood@arm.com>
6975 * reload1.c (spill_stack_slot_width): Change element type
6976 from unsigned int to poly_uint64_pod.
6977 (alter_reg): Treat mode sizes as polynomial.
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 * reload.c (complex_word_subreg_p): New function.
6984 (reload_inner_reg_of_subreg, push_reload): Use it.
6986 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6987 Alan Hayward <alan.hayward@arm.com>
6988 David Sherwood <david.sherwood@arm.com>
6990 * lra-constraints.c (process_alt_operands): Reject matched
6991 operands whose sizes aren't ordered.
6992 (match_reload): Refer to this check here.
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 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
6999 that the mode size is in the set {1, 2, 4, 8, 16}.
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 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
7006 Use plus_constant instead of gen_rtx_PLUS.
7008 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7009 Alan Hayward <alan.hayward@arm.com>
7010 David Sherwood <david.sherwood@arm.com>
7012 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
7013 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
7014 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
7015 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
7016 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
7017 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
7018 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
7019 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
7020 * config/i386/i386.c (ix86_push_rounding): ...this new function.
7021 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
7023 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
7024 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
7025 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
7026 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
7027 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
7028 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
7029 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
7030 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
7031 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
7032 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
7034 * expr.c (emit_move_resolve_push): Treat the input and result
7035 of PUSH_ROUNDING as a poly_int64.
7036 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
7037 (emit_push_insn): Likewise.
7038 * lra-eliminations.c (mark_not_eliminable): Likewise.
7039 * recog.c (push_operand): Likewise.
7040 * reload1.c (elimination_effects): Likewise.
7041 * rtlanal.c (nonzero_bits1): Likewise.
7042 * calls.c (store_one_arg): Likewise. Require the padding to be
7043 known at compile time.
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 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
7050 Use plus_constant instead of gen_rtx_PLUS.
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 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
7059 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7060 Alan Hayward <alan.hayward@arm.com>
7061 David Sherwood <david.sherwood@arm.com>
7063 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
7064 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
7065 via stack temporaries. Treat the mode size as polynomial too.
7067 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7068 Alan Hayward <alan.hayward@arm.com>
7069 David Sherwood <david.sherwood@arm.com>
7071 * expr.c (expand_expr_real_2): When handling conversions involving
7072 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
7073 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
7074 as a poly_uint64 too.
7076 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7077 Alan Hayward <alan.hayward@arm.com>
7078 David Sherwood <david.sherwood@arm.com>
7080 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
7082 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7083 Alan Hayward <alan.hayward@arm.com>
7084 David Sherwood <david.sherwood@arm.com>
7086 * combine.c (can_change_dest_mode): Handle polynomial
7087 REGMODE_NATURAL_SIZE.
7088 * expmed.c (store_bit_field_1): Likewise.
7089 * expr.c (store_constructor): Likewise.
7090 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
7091 and polynomial REGMODE_NATURAL_SIZE.
7092 (gen_lowpart_common): Likewise.
7093 * reginfo.c (record_subregs_of_mode): Likewise.
7094 * rtlanal.c (read_modify_subreg_p): Likewise.
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 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
7101 numbers of elements.
7103 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7104 Alan Hayward <alan.hayward@arm.com>
7105 David Sherwood <david.sherwood@arm.com>
7107 * match.pd: Cope with polynomial numbers of vector elements.
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 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
7114 in a POINTER_PLUS_EXPR.
7116 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7117 Alan Hayward <alan.hayward@arm.com>
7118 David Sherwood <david.sherwood@arm.com>
7120 * omp-simd-clone.c (simd_clone_subparts): New function.
7121 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
7122 (ipa_simd_modify_function_body): Likewise.
7124 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7125 Alan Hayward <alan.hayward@arm.com>
7126 David Sherwood <david.sherwood@arm.com>
7128 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
7129 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
7130 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
7131 (expand_vector_condition, vector_element): Likewise.
7132 (subparts_gt): New function.
7133 (get_compute_type): Use subparts_gt.
7134 (count_type_subparts): Delete.
7135 (expand_vector_operations_1): Use subparts_gt instead of
7136 count_type_subparts.
7138 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7139 Alan Hayward <alan.hayward@arm.com>
7140 David Sherwood <david.sherwood@arm.com>
7142 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
7143 (vect_compile_time_alias): ...this new function. Do the calculation
7144 on poly_ints rather than trees.
7145 (vect_prune_runtime_alias_test_list): Update call accordingly.
7147 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7148 Alan Hayward <alan.hayward@arm.com>
7149 David Sherwood <david.sherwood@arm.com>
7151 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
7153 (vect_schedule_slp_instance): Likewise.
7155 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7156 Alan Hayward <alan.hayward@arm.com>
7157 David Sherwood <david.sherwood@arm.com>
7159 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
7160 constant and extern definitions for variable-length vectors.
7161 (vect_get_constant_vectors): Note that the number of units
7162 is known to be constant.
7164 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7165 Alan Hayward <alan.hayward@arm.com>
7166 David Sherwood <david.sherwood@arm.com>
7168 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
7169 of units as polynomial. Choose between WIDE and NARROW based
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 (simd_clone_subparts): New function.
7177 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
7179 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7180 Alan Hayward <alan.hayward@arm.com>
7181 David Sherwood <david.sherwood@arm.com>
7183 * tree-vect-stmts.c (vectorizable_call): Treat the number of
7184 vectors as polynomial. Use build_index_vector for
7187 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7188 Alan Hayward <alan.hayward@arm.com>
7189 David Sherwood <david.sherwood@arm.com>
7191 * tree-vect-stmts.c (get_load_store_type): Treat the number of
7192 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
7193 for variable-length vectors.
7194 (vectorizable_mask_load_store): Treat the number of units as
7195 polynomial, asserting that it is constant if the condition has
7196 already been enforced.
7197 (vectorizable_store, vectorizable_load): Likewise.
7199 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7200 Alan Hayward <alan.hayward@arm.com>
7201 David Sherwood <david.sherwood@arm.com>
7203 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
7204 of units as polynomial. Punt if we can't tell at compile time
7205 which vector contains the final result.
7207 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7208 Alan Hayward <alan.hayward@arm.com>
7209 David Sherwood <david.sherwood@arm.com>
7211 * tree-vect-loop.c (vectorizable_induction): Treat the number
7212 of units as polynomial. Punt on SLP inductions. Use an integer
7213 VEC_SERIES_EXPR for variable-length integer reductions. Use a
7214 cast of such a series for variable-length floating-point
7217 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7218 Alan Hayward <alan.hayward@arm.com>
7219 David Sherwood <david.sherwood@arm.com>
7221 * tree.h (build_index_vector): Declare.
7222 * tree.c (build_index_vector): New function.
7223 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
7224 of units as polynomial, forcibly converting it to a constant if
7225 vectorizable_reduction has already enforced the condition.
7226 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
7227 to create a {1,2,3,...} vector.
7228 (vectorizable_reduction): Treat the number of units as polynomial.
7229 Choose vectype_in based on the largest scalar element size rather
7230 than the smallest number of units. Enforce the restrictions
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 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
7238 number of units as polynomial.
7240 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7241 Alan Hayward <alan.hayward@arm.com>
7242 David Sherwood <david.sherwood@arm.com>
7244 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
7245 * target.def (autovectorize_vector_sizes): Return the vector sizes
7246 by pointer, using vector_sizes rather than a bitmask.
7247 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
7248 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
7249 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
7251 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
7252 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
7253 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
7254 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
7255 * omp-general.c (omp_max_vf): Likewise.
7256 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7257 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
7258 * tree-vect-loop.c (vect_analyze_loop): Likewise.
7259 * tree-vect-slp.c (vect_slp_bb): Likewise.
7260 * doc/tm.texi: Regenerate.
7261 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
7263 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
7264 the vector size as a poly_uint64 rather than an unsigned int.
7265 (current_vector_size): Change from an unsigned int to a poly_uint64.
7266 (get_vectype_for_scalar_type): Update accordingly.
7267 * tree.h (build_truth_vector_type): Take the size and number of
7268 units as a poly_uint64 rather than an unsigned int.
7269 (build_vector_type): Add a temporary overload that takes
7270 the number of units as a poly_uint64 rather than an unsigned int.
7271 * tree.c (make_vector_type): Likewise.
7272 (build_truth_vector_type): Take the number of units as a poly_uint64
7273 rather than an unsigned int.
7275 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7276 Alan Hayward <alan.hayward@arm.com>
7277 David Sherwood <david.sherwood@arm.com>
7279 * target.def (get_mask_mode): Take the number of units and length
7280 as poly_uint64s rather than unsigned ints.
7281 * targhooks.h (default_get_mask_mode): Update accordingly.
7282 * targhooks.c (default_get_mask_mode): Likewise.
7283 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
7284 * doc/tm.texi: Regenerate.
7286 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7287 Alan Hayward <alan.hayward@arm.com>
7288 David Sherwood <david.sherwood@arm.com>
7290 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
7291 * omp-general.c (omp_max_vf): Likewise.
7292 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
7293 (expand_omp_simd): Handle polynomial safelen.
7294 * omp-low.c (omplow_simd_context): Add a default constructor.
7295 (omplow_simd_context::max_vf): Change from int to poly_uint64.
7296 (lower_rec_simd_input_clauses): Update accordingly.
7297 (lower_rec_input_clauses): Likewise.
7299 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7300 Alan Hayward <alan.hayward@arm.com>
7301 David Sherwood <david.sherwood@arm.com>
7303 * tree-vectorizer.h (vect_nunits_for_cost): New function.
7304 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
7305 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
7306 (vect_analyze_slp_cost): Likewise.
7307 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
7308 (vect_model_load_cost): Likewise.
7310 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7311 Alan Hayward <alan.hayward@arm.com>
7312 David Sherwood <david.sherwood@arm.com>
7314 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
7315 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
7316 from an unsigned int * to a poly_uint64_pod *.
7317 (calculate_unrolling_factor): New function.
7318 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
7320 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7321 Alan Hayward <alan.hayward@arm.com>
7322 David Sherwood <david.sherwood@arm.com>
7324 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
7325 from an unsigned int to a poly_uint64.
7326 (_loop_vec_info::slp_unrolling_factor): Likewise.
7327 (_loop_vec_info::vectorization_factor): Change from an int
7329 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
7330 (vect_get_num_vectors): New function.
7331 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
7332 (vect_get_num_copies): Use vect_get_num_vectors.
7333 (vect_analyze_data_ref_dependences): Change max_vf from an int *
7334 to an unsigned int *.
7335 (vect_analyze_data_refs): Change min_vf from an int * to a
7337 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7338 than an unsigned HOST_WIDE_INT.
7339 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7340 (vect_analyze_data_ref_dependence): Change max_vf from an int *
7341 to an unsigned int *.
7342 (vect_analyze_data_ref_dependences): Likewise.
7343 (vect_compute_data_ref_alignment): Handle polynomial vf.
7344 (vect_enhance_data_refs_alignment): Likewise.
7345 (vect_prune_runtime_alias_test_list): Likewise.
7346 (vect_shift_permute_load_chain): Likewise.
7347 (vect_supportable_dr_alignment): Likewise.
7348 (dependence_distance_ge_vf): Take the vectorization factor as a
7349 poly_uint64 rather than an unsigned HOST_WIDE_INT.
7350 (vect_analyze_data_refs): Change min_vf from an int * to a
7352 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
7353 vfm1 as a poly_uint64 rather than an int. Make the same change
7354 for the returned bound_scalar.
7355 (vect_gen_vector_loop_niters): Handle polynomial vf.
7356 (vect_do_peeling): Likewise. Update call to
7357 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
7358 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
7360 * tree-vect-loop.c (vect_determine_vectorization_factor)
7361 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
7362 (vect_get_known_peeling_cost): Likewise.
7363 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
7364 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
7365 (vect_transform_loop): Likewise. Use the lowest possible VF when
7366 updating the upper bounds of the loop.
7367 (vect_min_worthwhile_factor): Make static. Return an unsigned int
7369 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
7370 polynomial unroll factors.
7371 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
7372 (vect_make_slp_decision): Likewise.
7373 (vect_supported_load_permutation_p): Likewise, and polynomial
7375 (vect_analyze_slp_cost): Handle polynomial vf.
7376 (vect_slp_analyze_node_operations): Likewise.
7377 (vect_slp_analyze_bb_1): Likewise.
7378 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7379 than an unsigned HOST_WIDE_INT.
7380 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
7381 (vectorizable_load): Handle polynomial vf.
7382 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
7384 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
7386 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7387 Alan Hayward <alan.hayward@arm.com>
7388 David Sherwood <david.sherwood@arm.com>
7390 * match.pd: Handle bit operations involving three constants
7391 and try to fold one pair.
7393 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7395 * tree-vect-loop-manip.c: Include gimple-fold.h.
7396 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
7397 niters_maybe_zero parameters. Handle other cases besides a step of 1.
7398 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
7399 Add a path that uses a step of VF instead of 1, but disable it
7401 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
7402 and niters_no_overflow parameters. Update calls to
7403 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
7404 Create a new SSA name if the latter choses to use a ste other
7405 than zero, and return it via niters_vector_mult_vf_var.
7406 * tree-vect-loop.c (vect_transform_loop): Update calls to
7407 vect_do_peeling, vect_gen_vector_loop_niters and
7408 slpeel_make_loop_iterate_ntimes.
7409 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
7410 (vect_gen_vector_loop_niters): Update declarations after above changes.
7412 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
7414 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
7415 128-bit round to integer instructions.
7416 (ceil<mode>2): Likewise.
7417 (btrunc<mode>2): Likewise.
7418 (round<mode>2): Likewise.
7420 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7422 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
7423 unaligned VSX load/store on P8/P9.
7424 (expand_block_clear): Allow the use of unaligned VSX
7425 load/store on P8/P9.
7427 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7429 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
7431 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
7432 swap associated with both a load and a store.
7434 2018-01-02 Andrew Waterman <andrew@sifive.com>
7436 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
7437 * config/riscv/riscv.md (clear_cache): Use it.
7439 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
7441 * web.c: Remove out-of-date comment.
7443 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7445 * expr.c (fixup_args_size_notes): Check that any existing
7446 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
7447 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
7448 (emit_single_push_insn): ...here.
7450 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7452 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
7453 (const_vector_encoded_nelts): New function.
7454 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
7455 (const_vector_int_elt, const_vector_elt): Declare.
7456 * emit-rtl.c (const_vector_int_elt_1): New function.
7457 (const_vector_elt): Likewise.
7458 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
7459 of CONST_VECTOR_ELT.
7461 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7463 * expr.c: Include rtx-vector-builder.h.
7464 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
7465 directly on the tree encoding.
7466 (const_vector_from_tree): Likewise.
7467 * optabs.c: Include rtx-vector-builder.h.
7468 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
7469 sequence of "u" values.
7470 * vec-perm-indices.c: Include rtx-vector-builder.h.
7471 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
7472 directly on the vec_perm_indices encoding.
7474 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7476 * doc/rtl.texi (const_vector): Describe new encoding scheme.
7477 * Makefile.in (OBJS): Add rtx-vector-builder.o.
7478 * rtx-vector-builder.h: New file.
7479 * rtx-vector-builder.c: Likewise.
7480 * rtl.h (rtx_def::u2): Add a const_vector field.
7481 (CONST_VECTOR_NPATTERNS): New macro.
7482 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
7483 (CONST_VECTOR_DUPLICATE_P): Likewise.
7484 (CONST_VECTOR_STEPPED_P): Likewise.
7485 (CONST_VECTOR_ENCODED_ELT): Likewise.
7486 (const_vec_duplicate_p): Check for a duplicated vector encoding.
7487 (unwrap_const_vec_duplicate): Likewise.
7488 (const_vec_series_p): Check for a non-duplicated vector encoding.
7489 Say that the function only returns true for integer vectors.
7490 * emit-rtl.c: Include rtx-vector-builder.h.
7491 (gen_const_vec_duplicate_1): Delete.
7492 (gen_const_vector): Call gen_const_vec_duplicate instead of
7493 gen_const_vec_duplicate_1.
7494 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
7495 (gen_const_vec_duplicate): Use rtx_vector_builder.
7496 (gen_const_vec_series): Likewise.
7497 (gen_rtx_CONST_VECTOR): Likewise.
7498 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
7499 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7500 Build a new vector rather than modifying a CONST_VECTOR in-place.
7501 (handle_special_swappables): Update call accordingly.
7502 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
7503 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7504 Build a new vector rather than modifying a CONST_VECTOR in-place.
7505 (handle_special_swappables): Update call accordingly.
7507 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7509 * simplify-rtx.c (simplify_const_binary_operation): Use
7510 CONST_VECTOR_ELT instead of XVECEXP.
7512 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7514 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
7515 the selector elements to be different from the data elements
7516 if the selector is a VECTOR_CST.
7517 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
7518 ssizetype for the selector.
7520 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7522 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
7523 before testing each element individually.
7524 * tree-vect-generic.c (lower_vec_perm): Likewise.
7526 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7528 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
7529 * selftest-run-tests.c (selftest::run_tests): Call it.
7530 * vector-builder.h (vector_builder::operator ==): New function.
7531 (vector_builder::operator !=): Likewise.
7532 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
7533 (vec_perm_indices::all_from_input_p): New function.
7534 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7535 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
7536 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
7537 instead of reading the VECTOR_CST directly. Detect whether both
7538 vector inputs are the same before constructing the vec_perm_indices,
7539 and update the number of inputs argument accordingly. Use the
7540 utility functions added above. Only construct sel2 if we need to.
7542 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7544 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
7545 the broadcast of the low byte.
7546 (expand_mult_highpart): Use an explicit encoding for the permutes.
7547 * optabs-query.c (can_mult_highpart_p): Likewise.
7548 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
7549 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7550 (vectorizable_bswap): Likewise.
7551 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
7552 explicit encoding for the power-of-2 permutes.
7553 (vect_permute_store_chain): Likewise.
7554 (vect_grouped_load_supported): Likewise.
7555 (vect_permute_load_chain): Likewise.
7557 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7559 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
7560 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
7561 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
7562 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7563 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
7564 (vect_gen_perm_mask_any): Likewise.
7566 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7568 * int-vector-builder.h: New file.
7569 * vec-perm-indices.h: Include int-vector-builder.h.
7570 (vec_perm_indices): Redefine as an int_vector_builder.
7571 (auto_vec_perm_indices): Delete.
7572 (vec_perm_builder): Redefine as a stand-alone class.
7573 (vec_perm_indices::vec_perm_indices): New function.
7574 (vec_perm_indices::clamp): Likewise.
7575 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
7576 (vec_perm_indices::new_vector): New function.
7577 (vec_perm_indices::new_expanded_vector): Update for new
7578 vec_perm_indices class.
7579 (vec_perm_indices::rotate_inputs): New function.
7580 (vec_perm_indices::all_in_range_p): Operate directly on the
7581 encoded form, without computing elided elements.
7582 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
7583 encoding. Update for new vec_perm_indices class.
7584 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
7585 the given vec_perm_builder.
7586 (expand_vec_perm_var): Update vec_perm_builder constructor.
7587 (expand_mult_highpart): Use vec_perm_builder instead of
7588 auto_vec_perm_indices.
7589 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
7590 vec_perm_indices instead of auto_vec_perm_indices. Use a single
7591 or double series encoding as appropriate.
7592 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
7593 vec_perm_indices instead of auto_vec_perm_indices.
7594 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7595 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7596 (vect_permute_store_chain): Likewise.
7597 (vect_grouped_load_supported): Likewise.
7598 (vect_permute_load_chain): Likewise.
7599 (vect_shift_permute_load_chain): Likewise.
7600 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
7601 (vect_transform_slp_perm_load): Likewise.
7602 (vect_schedule_slp_instance): Likewise.
7603 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7604 (vectorizable_mask_load_store): Likewise.
7605 (vectorizable_bswap): Likewise.
7606 (vectorizable_store): Likewise.
7607 (vectorizable_load): Likewise.
7608 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
7609 vec_perm_indices instead of auto_vec_perm_indices. Use
7610 tree_to_vec_perm_builder to read the vector from a tree.
7611 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
7612 vec_perm_builder instead of a vec_perm_indices.
7613 (have_whole_vector_shift): Use vec_perm_builder and
7614 vec_perm_indices instead of auto_vec_perm_indices. Leave the
7615 truncation to calc_vec_perm_mask_for_shift.
7616 (vect_create_epilog_for_reduction): Likewise.
7617 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
7618 from auto_vec_perm_indices to vec_perm_indices.
7619 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7620 instead of changing individual elements.
7621 (aarch64_vectorize_vec_perm_const): Use new_vector to install
7622 the vector in d.perm.
7623 * config/arm/arm.c (expand_vec_perm_d::perm): Change
7624 from auto_vec_perm_indices to vec_perm_indices.
7625 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7626 instead of changing individual elements.
7627 (arm_vectorize_vec_perm_const): Use new_vector to install
7628 the vector in d.perm.
7629 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
7630 Update vec_perm_builder constructor.
7631 (rs6000_expand_interleave): Likewise.
7632 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
7633 (rs6000_expand_interleave): Likewise.
7635 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7637 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
7638 to qimode could truncate the indices.
7639 * optabs.c (expand_vec_perm_var): Likewise.
7641 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7643 * Makefile.in (OBJS): Add vec-perm-indices.o.
7644 * vec-perm-indices.h: New file.
7645 * vec-perm-indices.c: Likewise.
7646 * target.h (vec_perm_indices): Replace with a forward class
7648 (auto_vec_perm_indices): Move to vec-perm-indices.h.
7649 * optabs.h: Include vec-perm-indices.h.
7650 (expand_vec_perm): Delete.
7651 (selector_fits_mode_p, expand_vec_perm_var): Declare.
7652 (expand_vec_perm_const): Declare.
7653 * target.def (vec_perm_const_ok): Replace with...
7654 (vec_perm_const): ...this new hook.
7655 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
7656 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
7657 * doc/tm.texi: Regenerate.
7658 * optabs.def (vec_perm_const): Delete.
7659 * doc/md.texi (vec_perm_const): Likewise.
7660 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
7661 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
7662 expand_vec_perm for constant permutation vectors. Assert that
7663 the mode of variable permutation vectors is the integer equivalent
7664 of the mode that is being permuted.
7665 * optabs-query.h (selector_fits_mode_p): Declare.
7666 * optabs-query.c: Include vec-perm-indices.h.
7667 (selector_fits_mode_p): New function.
7668 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
7669 is defined, instead of checking whether the vec_perm_const_optab
7670 exists. Use targetm.vectorize.vec_perm_const instead of
7671 targetm.vectorize.vec_perm_const_ok. Check whether the indices
7672 fit in the vector mode before using a variable permute.
7673 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
7674 vec_perm_indices instead of an rtx.
7675 (expand_vec_perm): Replace with...
7676 (expand_vec_perm_const): ...this new function. Take the selector
7677 as a vec_perm_indices rather than an rtx. Also take the mode of
7678 the selector. Update call to shift_amt_for_vec_perm_mask.
7679 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
7680 Use vec_perm_indices::new_expanded_vector to expand the original
7681 selector into bytes. Check whether the indices fit in the vector
7682 mode before using a variable permute.
7683 (expand_vec_perm_var): Make global.
7684 (expand_mult_highpart): Use expand_vec_perm_const.
7685 * fold-const.c: Includes vec-perm-indices.h.
7686 * tree-ssa-forwprop.c: Likewise.
7687 * tree-vect-data-refs.c: Likewise.
7688 * tree-vect-generic.c: Likewise.
7689 * tree-vect-loop.c: Likewise.
7690 * tree-vect-slp.c: Likewise.
7691 * tree-vect-stmts.c: Likewise.
7692 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
7694 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
7695 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
7696 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
7697 (aarch64_vectorize_vec_perm_const): ...this new function.
7698 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7699 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7700 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
7701 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
7702 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7703 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7704 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
7706 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
7707 check for NEON modes.
7708 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
7709 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
7710 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
7711 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
7713 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
7714 the old VEC_PERM_CONST conditions.
7715 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
7716 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
7717 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
7718 (ia64_vectorize_vec_perm_const_ok): Merge into...
7719 (ia64_vectorize_vec_perm_const): ...this new function.
7720 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
7721 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
7722 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
7723 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
7724 * config/mips/mips.c (mips_expand_vec_perm_const)
7725 (mips_vectorize_vec_perm_const_ok): Merge into...
7726 (mips_vectorize_vec_perm_const): ...this new function.
7727 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
7728 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
7729 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
7730 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
7731 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
7732 (rs6000_expand_vec_perm_const): Delete.
7733 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
7735 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7736 (altivec_expand_vec_perm_const_le): Take each operand individually.
7737 Operate on constant selectors rather than rtxes.
7738 (altivec_expand_vec_perm_const): Likewise. Update call to
7739 altivec_expand_vec_perm_const_le.
7740 (rs6000_expand_vec_perm_const): Delete.
7741 (rs6000_vectorize_vec_perm_const_ok): Delete.
7742 (rs6000_vectorize_vec_perm_const): New function.
7743 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
7744 an element count and rtx array.
7745 (rs6000_expand_extract_even): Update call accordingly.
7746 (rs6000_expand_interleave): Likewise.
7747 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
7748 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
7749 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
7750 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
7751 (rs6000_expand_vec_perm_const): Delete.
7752 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7753 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7754 (altivec_expand_vec_perm_const_le): Take each operand individually.
7755 Operate on constant selectors rather than rtxes.
7756 (altivec_expand_vec_perm_const): Likewise. Update call to
7757 altivec_expand_vec_perm_const_le.
7758 (rs6000_expand_vec_perm_const): Delete.
7759 (rs6000_vectorize_vec_perm_const_ok): Delete.
7760 (rs6000_vectorize_vec_perm_const): New function. Remove stray
7761 reference to the SPE evmerge intructions.
7762 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
7763 an element count and rtx array.
7764 (rs6000_expand_extract_even): Update call accordingly.
7765 (rs6000_expand_interleave): Likewise.
7766 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
7767 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
7769 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7771 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7773 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
7774 vector mode and that that mode matches the mode of the data
7776 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
7777 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
7778 directly using expand_vec_perm_1 when forcing selectors into
7780 (expand_vec_perm_var): New function, split out from expand_vec_perm.
7782 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7784 * optabs-query.h (can_vec_perm_p): Delete.
7785 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
7786 * optabs-query.c (can_vec_perm_p): Split into...
7787 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
7788 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
7789 particular selector is valid.
7790 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7791 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7792 (vect_grouped_load_supported): Likewise.
7793 (vect_shift_permute_load_chain): Likewise.
7794 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
7795 (vect_transform_slp_perm_load): Likewise.
7796 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7797 (vectorizable_bswap): Likewise.
7798 (vect_gen_perm_mask_checked): Likewise.
7799 * fold-const.c (fold_ternary_loc): Likewise. Don't take
7800 implementations of variable permutation vectors into account
7801 when deciding which selector to use.
7802 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
7803 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
7804 with a false third argument.
7805 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
7806 to test whether the constant selector is valid and can_vec_perm_var_p
7807 to test whether a variable selector is valid.
7809 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7811 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
7812 * optabs-query.c (can_vec_perm_p): Likewise.
7813 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
7814 instead of vec_perm_indices.
7815 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
7816 (vect_gen_perm_mask_checked): Likewise,
7817 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
7818 (vect_gen_perm_mask_checked): Likewise,
7820 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7822 * optabs-query.h (qimode_for_vec_perm): Declare.
7823 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
7824 (qimode_for_vec_perm): ...this new function.
7825 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
7827 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7829 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
7830 does not have a conditional at the top.
7832 2018-01-02 Richard Biener <rguenther@suse.de>
7834 * ipa-inline.c (big_speedup_p): Fix expression.
7836 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
7839 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
7842 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
7846 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
7847 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
7848 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
7849 cond_taken_branch_cost 3->4.
7851 2018-01-01 Jakub Jelinek <jakub@redhat.com>
7853 PR tree-optimization/83581
7854 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
7855 TODO_cleanup_cfg if any changes have been made.
7858 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
7859 convert_modes if target mode has the right side, but different mode
7863 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
7864 last argument when extracting from CONCAT. If either from_real or
7865 from_imag is NULL, use expansion through memory. If result is not
7866 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
7867 the parts directly to inner mode, if even that fails, use expansion
7871 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
7872 check for bswap in mode rather than HImode and use that in expand_unop
7875 Copyright (C) 2018 Free Software Foundation, Inc.
7877 Copying and distribution of this file, with or without modification,
7878 are permitted in any medium without royalty provided the copyright
7879 notice and this notice are preserved.