1 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3 PR tree-optimization/84634
4 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5 masks and masked_loop_p with a single loop_masks, making sure it's
6 null for bb vectorization.
8 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
10 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
11 (vect_analyze_data_ref_access): Use loop->safe_len rather than
12 loop->force_vectorize to check whether there is no alias.
14 2018-03-02 Jakub Jelinek <jakub@redhat.com>
17 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
19 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
21 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
22 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
23 instead of a loop around prev_real_insn.
24 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
28 * config/i386/i386.c (ix86_print_operand): Use conditional
29 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
32 2018-03-02 Richard Biener <rguenther@suse.de>
34 PR tree-optimization/84427
35 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
36 (bitmap_set_subtract_values): Rewrite to handle multiple
39 (prune_clobbered_mems): Likewise.
40 (phi_translate): Take edge instead of pred/phiblock.
41 (phi_translate_1): Likewise.
42 (phi_translate_set): Likewise. Insert all translated
43 exprs for a value into the set, keeping possibly multiple
44 expressions per value.
45 (compute_antic_aux): Adjust for phi_translate changes.
46 When intersecting union the expressions and prune those
47 not in the final value set, keeping possibly multiple
48 expressions per value. Do not use value-insertion
49 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
50 all expressions. Add verification that the value-sets
51 only shrink during iteration.
52 (compute_partial_antic_aux): Adjust for the phi_translate changes.
53 (do_pre_regular_insertion): Likewise.
54 (do_pre_partial_partial_insertion): Likewise.
56 2018-03-02 Richard Biener <rguenther@suse.de>
59 * config/darwin.c (saved_debug_info_level): New static global.
60 (darwin_asm_lto_start): Disable debug info generation for LTO out.
61 (darwin_asm_lto_end): Restore debug info generation settings.
63 2018-03-01 Martin Liska <mliska@suse.cz>
66 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
69 2018-03-01 Richard Biener <rguenther@suse.de>
72 * dwarf2out.c (gen_variable_die): Properly handle late VLA
73 type annotation with LTO when debug was disabled at compile-time.
75 2018-03-01 Matthew Fortune <mfortune@gmail.com>
77 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
79 (mips_final_postscan_insn): Likewise.
81 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
83 PR rtl-optimization/84528
84 * alias.c (init_alias_target): Add commentary.
85 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
86 a unique base value if the frame pointer is not eliminated
89 2018-03-01 Tom de Vries <tom@codesourcery.com>
91 PR rtl-optimization/83327
92 * lra-int.h (hard_regs_spilled_into): Declare.
93 * lra.c (hard_regs_spilled_into): Define.
94 (init_reg_info): Init hard_regs_spilled_into.
95 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
96 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
97 (process_bb_lives): Handle hard_regs_spilled_into.
98 (lra_create_live_ranges_1): Before doing liveness propagation, clear
99 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
101 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
103 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
104 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
105 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
106 * config/rs6000/aix72.h: New file.
108 2018-02-28 Jakub Jelinek <jakub@redhat.com>
110 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
111 instead of warning_at with conditional singular and plural messages
115 * stor-layout.c (update_alignment_for_field): For
116 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
117 && !DECL_PACKED (field), do the alignment update, just use
118 only desired_align instead of MAX (type_align, desired_align)
120 (place_field): Don't do known_align < desired_align handling
121 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
122 is non-NULL, instead do it after rli->prev_field handling and
123 only if not within a bitfield word. For DECL_PACKED (field)
124 use type_align of BITS_PER_UNIT.
126 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
128 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
129 superfluous parentheses and trailing spaces.
131 2018-02-28 Richard Biener <rguenther@suse.de>
133 PR tree-optimization/84584
134 * graphite-scop-detection.c (scop_detection::add_scop): Discard
135 SCoPs with fake exit edge.
137 2018-02-28 Martin Liska <mliska@suse.cz>
140 * timevar.c (timer::print): Fix format to properly print 100%
143 2018-02-28 Richard Biener <rguenther@suse.de>
146 * genmatch.c (capture_info::walk_match): Do not mark
147 captured expressions without operands as expr_p given
148 they act more like predicates and should be subject to
149 "lost tail" side-effect preserving.
151 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
153 PR rtl-optimization/81611
154 * auto-inc-dec.c (attempt_change): Move dead note from
155 mem_insn if it's the next use of regno
156 (find_address): Take address use of reg holding
157 non-incremented value. Add parm to limit search to the named
159 (merge_in_block): Attempt to use a mem insn that is the next
160 use of the original regno.
162 2018-02-27 Martin Sebor <msebor@redhat.com>
165 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
166 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
168 2018-02-27 Martin Sebor <msebor@redhat.com>
171 * diagnostic-core.h (warning_n, error_n, inform_n): Change
172 n argument to unsigned HOST_WIDE_INT.
173 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
174 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
175 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
176 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
178 2018-02-27 Richard Biener <rguenther@suse.de>
180 PR tree-optimization/84512
181 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
182 Do not use the estimate returned from record_stmt_cost for
183 the scalar iteration cost but sum properly using add_stmt_cost.
185 2018-02-27 Richard Biener <rguenther@suse.de>
187 PR tree-optimization/84466
188 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
189 Adjust last change to less strictly validate use operands.
191 2018-02-27 Martin Liska <mliska@suse.cz>
193 PR gcov-profile/84548
194 * gcov.c (process_file): Allow partial overlap and consider it
195 also as group functions.
196 (output_lines): Properly calculate range of lines for a group.
198 2018-02-27 Martin Liska <mliska@suse.cz>
200 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
201 'ggc' suffixes. Change first column width.
202 (timer::print): Fix formatting of the column.
204 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
206 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
207 preserve inline entry blocks for the sake of debug inline
208 entry point markers alone.
209 (remove_unused_locals): Suggest in comments a better place to
210 force the preservation of inline entry blocks that are
211 otherwise unused, but do not preserve them.
213 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
215 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
217 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
220 * config/i386/constraints.md (Bs): Replace
221 ix86_indirect_branch_register with
222 TARGET_INDIRECT_BRANCH_REGISTER.
224 * config/i386/i386.md (indirect_jump): Likewise.
225 (tablejump): Likewise.
226 (*sibcall_memory): Likewise.
227 (*sibcall_value_memory): Likewise.
228 Peepholes of indirect call and jump via memory: Likewise.
229 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
230 (*sibcall_value_GOT_32): Likewise.
231 * config/i386/predicates.md (indirect_branch_operand): Likewise.
232 (GOT_memory_operand): Likewise.
233 (call_insn_operand): Likewise.
234 (sibcall_insn_operand): Likewise.
235 (GOT32_symbol_operand): Likewise.
236 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
238 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
240 PR rtl-optimization/83496
241 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
242 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
243 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
244 redundant insn, if any.
245 (relax_delay_slots): Likewise.
246 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
248 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
250 PR tree-optimization/83965
251 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
252 that grouped statements are part of a reduction chain. Return
253 true if the statement is not marked as a reduction itself but
255 (vect_recog_dot_prod_pattern): Don't check whether the statement
256 is part of a group here.
257 (vect_recog_sad_pattern): Likewise.
258 (vect_recog_widen_sum_pattern): Likewise.
260 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
263 * final.c (rest_of_clean_state): Also look for calls inside sequences.
265 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
268 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
270 (ix86_output_indirect_function_return): New prototype.
271 (ix86_split_simple_return_pop_internal): Likewise.
272 * config/i386/i386.c (indirect_return_via_cx): New.
273 (indirect_return_via_cx_bnd): Likewise.
274 (indirect_thunk_name): Handle return va CX_REG.
275 (output_indirect_thunk_function): Create alias for
276 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
277 (ix86_output_indirect_jmp): Remove the bool argument.
278 (ix86_output_indirect_function_return): New function.
279 (ix86_split_simple_return_pop_internal): Likewise.
280 * config/i386/i386.md (*indirect_jump): Don't pass false
281 to ix86_output_indirect_jmp.
282 (*tablejump_1): Likewise.
283 (simple_return_pop_internal): Change it to define_insn_and_split.
284 Call ix86_split_simple_return_pop_internal to split it for
286 (simple_return_indirect_internal): Call
287 ix86_output_indirect_function_return instead of
288 ix86_output_indirect_jmp.
290 2018-02-26 Jakub Jelinek <jakub@redhat.com>
293 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
294 memset and value initialization afterwards.
296 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
298 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
300 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
303 * common/config/aarch64/aarch64-common.c
304 (aarch_option_optimization_table[]): Switch
305 off fomit-frame-pointer
307 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
308 Chung-Ju Wu <jasonwucj@gmail.com>
310 * config/nds32/nds32-multiple.md (load_multiple): Disallow
312 (store_multiple): Ditto.
314 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
316 * config.gcc: Add --with-cpu support for nds32 target.
317 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
318 * config/nds32/nds32.opt: Add -mcpu= option.
320 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
322 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
323 isel=yes): Warn for these deprecated options.
325 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
327 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
328 ISA_2_5_MASKS_EMBEDDED.
330 2018-02-23 Jakub Jelinek <jakub@redhat.com>
332 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
333 p->max as pointers rather than using iterative_hash_expr.
335 2018-02-23 Carl Love <cel@us.ibm.com>
337 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
338 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
340 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
341 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
342 P8V_BUILTIN_VEC_VUNSIGNED2.
344 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
347 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
348 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
350 * lra-constraints.c (curr_insn_transform): Set up
351 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
353 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
354 * lra-eliminations.c (spill_pseudos): Ditto.
355 (process_insn_for_elimination): Ditto.
356 * lra-lives.c (reg_early_clobber_p): Use the new macros.
357 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
358 LRA_NON_CLOBBERED_ALT.
360 2018-02-22 Martin Sebor <msebor@redhat.com>
362 PR tree-optimization/84480
363 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
364 to maybe_diag_stxncpy_trunc. Call it.
365 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
366 from gimple_fold_builtin_strcpy. Print inlining stack.
367 (handle_builtin_stxncpy): Print inlining stack.
368 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
370 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
373 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
374 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
375 and -fcheck-pointer-bounds are used together.
376 (indirect_thunk_prefix): New enum.
377 (indirect_thunk_need_prefix): New function.
378 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
379 "_nt" instead of "_bnd" for NOTRACK prefix.
380 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
381 (output_indirect_thunk_function): Likewise.
383 (ix86_code_end): Update output_indirect_thunk_function calls.
384 (ix86_output_indirect_branch_via_reg): Replace
385 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
386 (ix86_output_indirect_branch_via_push): Likewise.
387 (ix86_output_function_return): Likewise.
388 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
389 incompatible with -fcf-protection=branch and
390 -fcheck-pointer-bounds.
392 2018-02-22 Steve Ellcey <sellcey@cavium.com>
395 * config/aarch64/aarch64.c (aarch64_print_address_internal):
396 Change gcc_assert call to output_operand_lossage.
398 2018-02-22 Steve Ellcey <sellcey@cavium.com>
400 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
402 2018-02-22 DJ Delorie <dj@redhat.com>
403 Sebastian Perta <sebastian.perta@renesas.com>
404 Oleg Endo <olegendo@gcc.gnu.org>
406 * config/rx/rx.c (rx_rtx_costs): New function.
407 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
409 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
411 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
413 2018-02-22 Martin Liska <mliska@suse.cz>
416 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
417 Add "native" as a possible value.
419 2018-02-22 Martin Liska <mliska@suse.cz>
422 * config/i386/i386.c (ix86_option_override_internal):
423 Add "native" as a possible value for -march and -mtune.
425 2018-02-22 Jakub Jelinek <jakub@redhat.com>
428 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
429 to all type variants.
431 PR tree-optimization/84503
432 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
433 width as info->bitpos + info->bitsize - start.
434 (merged_store_group::merge_overlapping): Simplify width computation.
435 (check_no_overlap): New function.
436 (imm_store_chain_info::try_coalesce_bswap): Compute expected
437 start + width and last_order of the group, fail if check_no_overlap
439 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
440 to group if check_no_overlap fails.
442 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
444 * config/rs6000/altivec.md: Delete contraint arguments to
445 define_expand, define_split, and define_peephole2, and in
446 define_insn_and_split if always unused.
447 * config/rs6000/darwin.md: Ditto.
448 * config/rs6000/dfp.md: Ditto.
449 * config/rs6000/rs6000.md: Ditto.
450 * config/rs6000/sync.md: Ditto.
451 * config/rs6000/vector.md: Ditto.
452 * config/rs6000/vsx.md: Ditto.
454 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
456 * config/rs6000/altivec.md: Write output control strings as braced
457 blocks instead of double-quoted strings.
458 * config/rs6000/darwin.md: Ditto.
459 * config/rs6000/rs6000.md: Ditto.
460 * config/rs6000/vector.md: Ditto.
461 * config/rs6000/vsx.md: Ditto.
463 2018-02-21 Jason Merrill <jason@redhat.com>
465 PR c++/84314 - ICE with templates and fastcall attribute.
466 * attribs.c (build_type_attribute_qual_variant): Remove assert.
468 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
470 * ipa-cp.c (determine_versionability): Fix comment typos.
472 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
475 * ipa-cp.c (determine_versionability): Do not version functions caling
478 2018-02-21 Martin Liska <mliska@suse.cz>
481 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
482 Add "native" as a possible value.
483 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
484 the macro when native cpu detection is available.
486 2018-02-21 Martin Liska <mliska@suse.cz>
489 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
490 Add "native" as a possible value.
491 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
492 when native cpu detection is available.
494 2018-02-21 Jakub Jelinek <jakub@redhat.com>
495 Martin Sebor <msebor@redhat.com>
497 PR tree-optimization/84478
498 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
500 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
501 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
502 support which is conservatively correct, for 2 only stay conservative
503 for maxlen. Formatting and comment capitalization fixes. Add STRICT
504 argument to the 2 argument get_range_strlen, adjust 6 arg
505 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
507 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
508 (gimple_fold_builtin_strlen): Pass true as last argument to
511 2018-02-20 Martin Sebor <msebor@redhat.com>
514 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
515 (builtin_memref::set_base_and_offset): Same. Handle inner references.
516 (builtin_memref::builtin_memref): Factor out parts into
517 set_base_and_offset and call it.
519 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
522 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
523 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
524 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
525 search at the associated MODE_INT.
527 2018-02-20 Jeff Law <law@redhat.com>
530 PR tree-optimization/81592
532 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
533 for range data rather than using global data.
534 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
535 range data rather than using global data.
536 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
537 pass it to children as needed.
538 (struct directive::fmtresult): Similarly.
539 (struct directive::set_width): Similarly.
540 (struct directive::set_precision): Similarly.
541 (format_integer, format_directive, parse_directive): Similarly.
542 (format_none): Accept unnamed vr_values parameter.
543 (format_percent, format_floating, format_character): Similarly.
544 (format_string, format_plain): Similarly.
545 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
546 the EVRP range analyzer for range data rather than using global data.
547 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
548 gimple-ssa-evrp-analyze.h
549 (class sprintf_dom_walker): Add after_dom_children member function.
550 Add evrp_range_analyzer member.
551 (sprintf_dom_walker::before_dom_children): Call into the EVRP
552 range analyzer as needed.
553 (sprintf_dom_walker::after_dom_children): New member function.
554 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
556 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
557 (evrp_range_analyzer::pop_to_marker): Likewise.
559 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
561 PR tree-optimization/84419
562 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
563 with the required type if its current type is compatible but
566 2018-02-20 Jakub Jelinek <jakub@redhat.com>
569 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
572 2018-02-20 Martin Liska <mliska@suse.cz>
575 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
576 possible values if we don't have a hint.
578 2018-02-20 Martin Liska <mliska@suse.cz>
582 * final.c (shorten_branches): Build align_tab array with one
584 * opts.c (finish_options): Add alignment option limit check.
585 (MAX_CODE_ALIGN): Likewise.
586 (MAX_CODE_ALIGN_VALUE): Likewise.
587 * doc/invoke.texi: Document maximum allowed option value for
588 all -falign-* options.
590 2018-02-19 Jakub Jelinek <jakub@redhat.com>
593 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
594 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
595 * var-tracking.c (emit_note_insn_var_location): Remove all references
596 to NOTE_INSN_CALL_ARG_LOCATION.
597 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
598 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
599 Use copy_rtx_if_shared.
600 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
601 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
602 (dwarf2out_var_location): Remove handling of
603 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
605 * final.c (final_scan_insn): Remove all references to
606 NOTE_INSN_CALL_ARG_LOCATION.
607 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
608 before dumping final insns.
609 * except.c (emit_note_eh_region_end): Remove all references to
610 NOTE_INSN_CALL_ARG_LOCATION.
611 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
612 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
613 * config/arc/arc.c (hwloop_optimize): Likewise.
614 * config/arm/arm.c (create_fix_barrier): Likewise.
615 * config/s390/s390.c (s390_chunkify_start): Likewise.
616 * config/sh/sh.c (find_barrier): Likewise.
617 * config/i386/i386.c (rest_of_insert_endbranch,
618 ix86_seh_fixup_eh_fallthru): Likewise.
619 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
620 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
621 * config/frv/frv.c (frv_function_prologue): Likewise.
622 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
624 (note_outside_basic_block_p): Remove all references to
625 NOTE_INSN_CALL_ARG_LOCATION.
626 * gengtype.c (adjust_field_rtx_def): Likewise.
627 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
629 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
630 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
633 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
636 PR tree-optimization/84452
637 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
638 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
641 2018-02-19 Martin Liska <mliska@suse.cz>
644 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
646 2018-02-19 Martin Liska <mliska@suse.cz>
647 Richard Sandiford <richard.sandiford@linaro.org>
649 PR tree-optimization/82491
650 * gimple-fold.c (get_base_constructor): Make earlier bail out
653 2018-02-19 Carl Love <cel@us.ibm.com>
655 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
656 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
658 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
661 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
663 * config/rl78/rl78.md (movdf): New define expand.
665 2018-02-19 Martin Liska <mliska@suse.cz>
668 * doc/invoke.texi: Fix typo.
669 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
671 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
673 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
674 handle rs6000_single_float and rs6000_double_float specially for
677 2018-02-16 Jeff Law <law@redhat.com>
679 * config/rx/rx.c (add_pop_cfi_notes): New function.;
682 2018-02-16 Jakub Jelinek <jakub@redhat.com>
685 * ipa-inline.c (inline_small_functions): Fix a typo.
687 2018-02-16 Nathan Sidwell <nathan@acm.org>
689 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
691 2018-02-16 Carl Love <cel@us.ibm.com>
693 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
694 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
695 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
696 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
697 expansion to P8V_BUILTIN_VEC_FLOAT2.
699 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
701 PR rtl-optimization/70023
702 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
703 src_regno into account.
705 2018-02-16 Carl Love <cel@us.ibm.com>
707 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
708 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
709 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
710 * config/rs6000/rs6000.c: Remove case statements for
711 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
712 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
713 and P9V_BUILTIN_VEC_VINSERT4B.
714 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
715 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
716 * config/rs6000/vsx.md:
717 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
720 2018-02-16 Carl Love <cel@us.ibm.com>
722 * config/rs6000/altivec.h: Add builtin names vec_extract4b
724 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
726 * config/rs6000/rs6000-c.c: Add the definitions for
727 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
728 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
729 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
730 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
731 definition for insert4b and define insn *insert3b_internal.
732 * doc/extend.texi: Add documentation for vec_extract4b.
734 2018-02-16 Nathan Sidwell <nathan@acm.org>
736 * doc/extend.texi (Backwards Compatibility): Mention friend
737 injection. Note for-scope is deprecated.
738 * doc/invoke.texi (-ffriend-injection): Deprecate.
740 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
742 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
743 that moved to I2, also allow destinations that are a paradoxical
744 subreg (instead of a normal reg).
746 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
749 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
752 2018-02-16 Richard Biener <rguenther@suse.de>
754 PR tree-optimization/84037
755 PR tree-optimization/84016
757 * config/i386/i386.c (ix86_builtin_vectorization_cost):
758 Adjust vec_construct for the fact we need additional higher latency
759 128bit inserts for AVX256 and AVX512 vector builds.
760 (ix86_add_stmt_cost): Scale vector construction cost for
763 2018-02-16 Richard Biener <rguenther@suse.de>
765 PR tree-optimization/84417
766 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
767 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
768 (non_rewritable_lvalue_p): Likewise, use poly-ints.
770 2018-02-16 Martin Liska <mliska@suse.cz>
773 * internal-fn.def (ASAN_CHECK): Set proper flags.
774 (ASAN_MARK): Likewise.
776 2018-02-16 Julia Koval <julia.koval@intel.com>
778 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
781 2018-02-16 Jakub Jelinek <jakub@redhat.com>
784 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
785 Use ++iter rather than iter++ for std::list iterators.
786 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
787 defer deleting them until all nodes in the forest are processed. Do
788 free even leaf nodes. Change to_process into auto_vec.
791 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
792 * vec.h (vec_default_construct): Use memset instead of placement new
793 if BROKEN_VALUE_INITIALIZATION is defined.
794 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
795 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
798 PR rtl-optimization/83723
799 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
800 * lra.c (lra_substitute_pseudo): Likewise. If true, use
801 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
803 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
805 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
807 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
809 PR rtl-optimization/81443
810 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
811 from inner REGs to paradoxical SUBREGs.
813 2018-02-16 Richard Biener <rguenther@suse.de>
815 PR tree-optimization/84399
816 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
817 For operands we can analyze at their definition make sure we can
818 analyze them at each use as well.
820 2018-02-16 Richard Biener <rguenther@suse.de>
822 PR tree-optimization/84190
823 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
824 volatile accesses if the decl isn't volatile.
826 2018-02-15 Jason Merrill <jason@redhat.com>
828 PR c++/84314 - ICE with templates and fastcall attribute.
829 * attribs.c (build_type_attribute_qual_variant): Don't clobber
830 TYPE_CANONICAL on an existing type.
832 2018-02-15 Jakub Jelinek <jakub@redhat.com>
834 PR tree-optimization/84383
835 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
836 dstoff nor call operand_equal_p if dstbase is NULL.
838 PR tree-optimization/84334
839 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
840 also a CONSTANT_CLASS_P, punt.
842 2018-02-14 Jim Wilson <jimw@sifive.com>
844 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
845 first SMALL_OPERAND check. New local min_second_step. Move assert
846 to where locals are set. Add TARGET_RVC support.
847 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
849 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
851 * doc/invoke.texi: Correct -Wformat-overflow code sample.
853 2018-02-14 Martin Sebor <msebor@redhat.com>
855 PR tree-optimization/83698
856 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
857 arrays constrain the offset range to their bounds.
858 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
859 (builtin_access::overlap): Avoid setting the size of overlap if it's
861 (maybe_diag_overlap): Also consider arrays when deciding what values
862 of offsets to include in diagnostics.
864 2018-02-14 Martin Sebor <msebor@redhat.com>
867 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
868 that correspond to the kind of a declaration.
870 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
873 * config/pa/pa.md: Load address of PIC label using the linkage table
874 if the label is nonlocal.
876 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
878 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
879 warning message if user requests -maltivec=be.
880 * doc/invoke.texi: Document deprecation of -maltivec=be.
882 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
885 * config/rs6000/rs6000-c.c: Update definitions for
886 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
887 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
889 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
892 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
893 add _get_ssp intrinsics. Remove argument from
894 __builtin_ia32_rdssp[d|q].
895 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
896 * config/i386/i386-builtin.def: Remove argument from
897 __builtin_ia32_rdssp[d|q].
898 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
899 ix86_expand_special_args_builtin for _rdssp[d|q].
900 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
901 Clear register before usage.
902 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
903 Add documentation for new _get_ssp and _inc_ssp intrinsics.
905 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
907 PR tree-optimization/84357
908 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
909 operand 1 of an ARRAY_REF too.
911 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
914 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
915 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
917 (set_of_reg): New struct.
918 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
919 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
920 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
922 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
923 Split into bitclr, bitset, bitinvert patterns if appropriate.
924 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
925 use rx_fuse_in_memory_bitop.
926 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
927 to named insn, correct maximum insn length.
929 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
932 * machmode.def: Define a complex mode for PARTIAL_INT.
933 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
935 * doc/rtl.texi: Document CSPImode.
936 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
938 (msp430_hard_regno_nregs_with_padding): Likewise.
940 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
943 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
945 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
947 PR rtl-optimization/84169
948 * combine.c (try_combine): New variable split_i2i3. Set it to true if
949 we generated a parallel as new i3 and we split that to new i2 and i3
950 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
951 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
952 those to i2, not i1. Partially rewrite this scan code.
954 2018-02-13 Jakub Jelinek <jakub@redhat.com>
957 * stor-layout.c (place_field): For variable length fields, adjust
958 offset_align afterwards not just based on the field's alignment,
959 but also on the size.
962 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
963 of exps and logs in the use_exp2 case.
965 2018-02-13 Jeff Law <law@redhat.com>
967 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
970 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
973 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
977 * common.opt (gas-loc-support, gas-locview-support): New.
978 (ginline-points, ginternal-reset-location-views): New.
979 * doc/invoke.texi: Document them. Use @itemx where intended.
980 (gvariable-location-views): Adjust.
981 * target.def (reset_location_view): New.
982 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
983 (TARGET_RESET_LOCATION_VIEW): New.
984 * doc/tm.texi: Rebuilt.
985 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
986 (dwarf2out_default_as_locview_support): New.
987 (output_asm_line_debug_info): Use option variables.
988 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
989 (output_loc_list): Likewise.
990 (add_high_low_attributes): Check option variables.
991 Don't output entry view attribute in strict mode.
992 (gen_inlined_subroutine_die): Check option variables.
993 (dwarf2out_inline_entry): Likewise.
994 (init_sections_and_labels): Likewise.
995 (dwarf2out_early_finish): Likewise.
996 (maybe_reset_location_view): New, from...
997 (dwarf2out_var_location): ... here. Call it.
998 * debug.h (dwarf2out_default_as_loc_support): Declare.
999 (dwarf2out_default_as_locview_support): Declare.
1000 * hooks.c (hook_int_rtx_insn_0): New.
1001 * hooks.h (hook_int_rtx_insn_0): Declare.
1002 * toplev.c (process_options): Take -gas-loc-support and
1003 -gas-locview-support from dwarf2out. Enable
1004 -gvariable-location-views by default only with locview
1005 assembler support. Enable -ginternal-reset-location-views by
1006 default only if the target defines the corresponding hook.
1007 Enable -ginline-points by default if location views are
1008 enabled; force it disabled if statement frontiers are
1010 * tree-inline.c (expand_call_inline): Check option variables.
1011 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1013 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
1015 PR tree-optimization/84321
1016 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
1017 handling. Also check whether the anti-range contains any values
1018 that satisfy the mask; switch to a VR_RANGE if not.
1020 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
1023 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
1025 2018-02-13 Martin Jambor <mjambor@suse.cz>
1028 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
1029 of call statements, also set location of a load to a temporary.
1031 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
1033 * config/rl78/rl78.c (add_vector_labels): New function.
1034 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
1035 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
1036 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
1037 which checks that no arguments are passed.
1038 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
1039 * doc/extend.texi: Documentation for the new attribute.
1041 2018-02-13 Andreas Schwab <schwab@suse.de>
1043 * config/riscv/linux.h (CPP_SPEC): Define.
1045 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1048 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
1049 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
1050 OPTION_MASK_ISA_AES as first argument to def_builtin_const
1051 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
1052 instead of OPTION_MASK_ISA_PCLMUL as first argument to
1053 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
1054 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
1055 temporarily for AES and PCLMUL builtins.
1057 PR tree-optimization/84339
1058 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
1059 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
1063 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
1064 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
1065 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
1067 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
1069 * omp-simd-clone.h: New file.
1070 * omp-simd-clone.c: Include omp-simd-clone.h.
1071 (expand_simd_clones): No longer static.
1072 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
1073 cgraph.h and omp-simd-clone.h.
1074 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
1075 (vect_recog_widen_shift_pattern): Formatting fix.
1076 (vect_pattern_recog_1): Don't check optab for calls.
1079 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
1080 operands[2] into a REG before using gen_lowpart on it.
1082 2018-02-12 Jeff Law <law@redhat.com>
1085 * config/sh/sh.c (find_barrier): Consider a sibling call
1088 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
1089 successfully back substituting a reg.
1091 2018-02-12 Richard Biener <rguenther@suse.de>
1093 PR tree-optimization/84037
1094 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
1095 parameter, move visited init to caller.
1096 (vect_slp_analyze_operations): Separate cost from validity
1097 check, initialize visited once for all instances.
1098 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
1100 * tree-vect-stmts.c (vect_model_simple_cost): Make early
1102 (vect_model_promotion_demotion_cost): Likewise.
1103 (vectorizable_bswap): Guard cost modeling with !slp_node
1104 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
1106 (vectorizable_call): Likewise.
1107 (vectorizable_conversion): Likewise.
1108 (vectorizable_assignment): Likewise.
1109 (vectorizable_shift): Likewise.
1110 (vectorizable_operation): Likewise.
1111 (vectorizable_store): Likewise.
1112 (vectorizable_load): Likewise.
1113 (vectorizable_condition): Likewise.
1114 (vectorizable_comparison): Likewise.
1116 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
1119 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
1120 (ASAN_MARK): Fix fnspec to account for return value, change pointer
1121 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
1123 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1126 * params.def (inline-min-speedup): Increase from 8 to 15.
1127 (max-inline-insns-auto): Decrease from 40 to 30.
1128 * ipa-split.c (consider_split): Add some buffer for function to
1129 be considered inlining candidate.
1130 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
1133 2018-02-12 Richard Biener <rguenther@suse.de>
1135 PR tree-optimization/84037
1136 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
1137 matched stmts if we cannot swap the non-matched ones.
1139 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1141 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
1142 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
1143 _mm_maskz_scalef_round_ss): New intrinsics.
1144 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
1145 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
1146 __builtin_ia32_scalefss_round): Remove.
1147 (__builtin_ia32_scalefsd_mask_round,
1148 __builtin_ia32_scalefss_mask_round): New intrinsics.
1149 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
1150 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
1151 ((match_operand:VF_128 2 "<round_nimm_predicate>"
1152 "<round_constraint>")): Changed to ...
1153 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
1154 "<round_scalar_constraint>")): ... this.
1155 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
1156 %0, %1, %2<round_op3>}"): Changed to ...
1157 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
1158 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
1159 %2<round_scalar_mask_op3>}"): ... this.
1160 * config/i386/subst.md (round_scalar_nimm_predicate): New.
1162 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1164 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
1165 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
1166 (_mm_maskz_sqrt_round_ss): New intrinsics.
1167 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
1168 (__builtin_ia32_sqrtsd_mask_round)
1169 (__builtin_ia32_sqrtss_mask_round): New builtins.
1170 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
1171 (__builtin_ia32_sqrtss_round): Remove.
1172 (__builtin_ia32_sqrtsd_mask_round)
1173 (__builtin_ia32_sqrtss_mask_round): New builtins.
1174 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
1175 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
1176 ((match_operand:VF_128 1 "vector_operand"
1177 "xBm,<round_constraint>")): Changed to ...
1178 ((match_operand:VF_128 1 "vector_operand"
1179 "xBm,<round_scalar_constraint>")): ... this.
1180 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
1181 %0, %2, %<iptr>1<round_op3>}): Changed to ...
1182 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
1183 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
1184 %<iptr>1<round_scalar_mask_op3>}): ... this.
1185 ((set_attr "prefix" "<round_prefix>")): Changed to ...
1186 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
1188 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
1191 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
1192 Cast vec_cmpeq result to correct type.
1193 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
1194 Cast vec_cmpgt result to correct type.
1196 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
1198 * final.c (final_scan_insn_1): Renamed from...
1199 (final_scan_insn): ... this. New wrapper, to recover
1200 seen from the outermost call in recursive ones.
1201 * config/sparc/sparc.c (output_return): Drop seen from call.
1202 (output_sibcall): Likewise.
1203 * config/visium/visium.c (output_branch): Likewise.
1205 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
1207 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
1210 2018-02-10 Alan Modra <amodra@gmail.com>
1213 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
1214 Specify LR as an input.
1216 2018-02-10 Jakub Jelinek <jakub@redhat.com>
1219 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
1220 remove_member_access_dummy_vars): New functions.
1221 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
1222 lower_omp_1, execute_lower_omp): Use them.
1224 PR rtl-optimization/84308
1225 * shrink-wrap.c (spread_components): Release todo vector.
1227 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
1229 PR rtl-optimization/57193
1230 * ira-color.c (struct allocno_color_data): Add member
1231 conflict_allocno_hard_prefs.
1232 (update_conflict_allocno_hard_prefs): New.
1233 (bucket_allocno_compare_func): Add a preference based on
1234 conflict_allocno_hard_prefs.
1235 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
1236 (color_allocnos): Remove a dead code. Initiate
1237 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
1239 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1242 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
1243 constraint from =wa to wa. Avoid a subreg on the output operand,
1244 instead use a pseudo and subreg it in a move.
1245 (p9_xxbrd_<mode>): Changed to ...
1246 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
1247 (p9_xxbrd_v2df): New expander.
1248 (p9_xxbrw_<mode>): Changed to ...
1249 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
1250 (p9_xxbrw_v4sf): New expander.
1252 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1254 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
1256 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
1259 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
1260 multiply in 32-bit mode.
1261 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
1262 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
1265 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1267 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
1268 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
1269 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
1270 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
1272 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
1275 * dwarf2out.c (is_trivial_indirect_ref): New function.
1276 (dwarf2out_late_global_decl): Do not generate a location
1277 attribute for variables that have a non-trivial DECL_VALUE_EXPR
1278 and that are not defined in the current unit.
1280 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1282 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
1283 instead of a libcall for UNORDERED.
1285 2018-02-09 Tamar Christina <tamar.christina@arm.com>
1288 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
1289 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
1291 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1294 * config/s390/s390.c (s390_set_current_function): Invoke
1295 s390_indirect_branch_settings also if fndecl didn't change.
1297 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1299 * config/rs6000/rs6000.md (blockage): Set length to zero.
1301 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1303 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
1305 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1308 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
1309 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
1313 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
1314 PARALLEL incoming that failed vt_get_decl_and_offset check.
1317 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
1318 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
1320 (get_variable_section): For decls in named .bss* sections pass true as
1321 second argument to bss_initializer_p.
1323 2018-02-09 Marek Polacek <polacek@redhat.com>
1324 Jakub Jelinek <jakub@redhat.com>
1327 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
1328 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
1329 Sync some changes from cxx_fold_indirect_ref.
1331 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1333 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
1335 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
1336 (BLOCK_INLINE_ENTRY_LABEL): New.
1337 (dwarf2out_var_location): Disregard inline entry markers.
1338 (inline_entry_data): New struct.
1339 (inline_entry_data_hasher): New hashtable type.
1340 (inline_entry_data_hasher::hash): New.
1341 (inline_entry_data_hasher::equal): New.
1342 (inline_entry_data_table): New variable.
1343 (add_high_low_attributes): Add DW_AT_entry_pc and
1344 DW_AT_GNU_entry_view attributes if a pending entry is found
1345 in inline_entry_data_table. Add old entry_pc attribute only
1346 if debug nonbinding markers are disabled.
1347 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
1348 markers are enabled.
1349 (block_within_block_p, dwarf2out_inline_entry): New.
1350 (dwarf2out_finish): Check that no entries remained in
1351 inline_entry_data_table.
1352 * final.c (reemit_insn_block_notes): Handle inline entry notes.
1353 (final_scan_insn, notice_source_line): Likewise.
1354 (rest_of_clean_state): Skip inline entry markers.
1355 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
1357 * gimple.c (gimple_build_debug_inline_entry): New.
1358 * gimple.h (enum gimple_debug_subcode): Add
1359 GIMPLE_DEBUG_INLINE_ENTRY.
1360 (gimple_build_debug_inline_entry): Declare.
1361 (gimple_debug_inline_entry_p): New.
1362 (gimple_debug_nonbind_marker_p): Adjust.
1363 * insn-notes.def (INLINE_ENTRY): New.
1364 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
1365 inline entry marker notes.
1366 (print_insn): Likewise.
1367 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
1368 (INSN_DEBUG_MARKER_KIND): Likewise.
1369 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
1370 * tree-inline.c (expand_call_inline): Build and insert
1371 debug_inline_entry stmt.
1372 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
1373 inline entry blocks early, if nonbind markers are enabled.
1374 (dump_scope_block): Dump fragment info.
1375 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
1376 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
1377 (gimple_build_debug_inline_entry): New.
1378 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
1379 Enable/disable inline entry points too.
1380 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
1381 (DEBUG_INSN): Describe inline entry markers.
1383 * common.opt (gvariable-location-views): New.
1384 (gvariable-location-views=incompat5): New.
1385 * config.in: Rebuilt.
1386 * configure: Rebuilt.
1387 * configure.ac: Test assembler for view support.
1388 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
1389 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
1390 * dwarf2out.c (var_loc_view): New typedef.
1391 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
1392 (dwarf2out_locviews_in_attribute): New.
1393 (dwarf2out_locviews_in_loclist): New.
1394 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
1395 (enum dw_line_info_opcode): Add LI_adv_address.
1396 (struct dw_line_info_table): Add view.
1397 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
1398 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
1399 (zero_view_p): New variable.
1400 (ZERO_VIEW_P): New macro.
1401 (output_asm_line_debug_info): New.
1402 (struct var_loc_node): Add view.
1403 (add_AT_view_list, AT_loc_list): New.
1404 (add_var_loc_to_decl): Add view param. Test it against last.
1405 (new_loc_list): Add view params. Record them.
1406 (AT_loc_list_ptr): Handle loc and view lists.
1407 (view_list_to_loc_list_val_node): New.
1408 (print_dw_val): Handle dw_val_class_view_list.
1409 (size_of_die): Likewise.
1410 (value_format): Likewise.
1411 (loc_list_has_views): New.
1412 (gen_llsym): Set vl_symbol too.
1413 (maybe_gen_llsym, skip_loc_list_entry): New.
1414 (dwarf2out_maybe_output_loclist_view_pair): New.
1415 (output_loc_list): Output view list or entries too.
1416 (output_view_list_offset): New.
1417 (output_die): Handle dw_val_class_view_list.
1418 (output_dwarf_version): New.
1419 (output_compilation_unit_header): Use it.
1420 (output_skeleton_debug_sections): Likewise.
1421 (output_rnglists, output_line_info): Likewise.
1422 (output_pubnames, output_aranges): Update version comments.
1423 (output_one_line_info_table): Output view numbers in asm comments.
1424 (dw_loc_list): Determine current endview, pass it to new_loc_list.
1425 Call maybe_gen_llsym.
1426 (loc_list_from_tree_1): Adjust.
1427 (add_AT_location_description): Create view list attribute if
1428 needed, check it's absent otherwise.
1429 (convert_cfa_to_fb_loc_list): Adjust.
1430 (maybe_emit_file): Call output_asm_line_debug_info for test.
1431 (dwarf2out_var_location): Reset views as needed. Precompute
1432 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
1433 attribute. Set view.
1434 (new_line_info_table): Reset next view.
1435 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
1436 (dwarf2out_source_line): Likewise. Output view resets and labels to
1437 the assembler, or select appropriate line info opcodes.
1438 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
1439 (optimize_string_length): Catch it. Adjust.
1440 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
1441 dw_val_class_view_list, and remove it if no longer needed.
1442 (hash_loc_list): Hash view numbers.
1443 (loc_list_hasher::equal): Compare them.
1444 (optimize_location_lists): Check whether a view list symbol is
1445 needed, and whether the locview attribute is present, and
1446 whether they match. Remove the locview attribute if no longer
1448 (index_location_lists): Call skip_loc_list_entry for test.
1449 (dwarf2out_finish): Call output_asm_line_debug_info for test.
1450 Use output_dwarf_version.
1451 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
1452 (struct dw_val_node): Add val_view_list.
1453 * final.c (SEEN_NEXT_VIEW): New.
1454 (set_next_view_needed): New.
1455 (clear_next_view_needed): New.
1456 (maybe_output_next_view): New.
1457 (final_start_function): Rename to...
1458 (final_start_function_1): ... this. Take pointer to FIRST,
1459 add SEEN parameter. Emit param bindings in the initial view.
1460 (final_start_function): Reintroduce SEEN-less interface.
1461 (final): Rename to...
1462 (final_1): ... this. Take SEEN parameter. Output final pending
1463 next view at the end.
1464 (final): Reintroduce seen-less interface.
1465 (final_scan_insn): Output pending next view before switching
1466 sections or ending a block. Mark the next view as needed when
1467 outputting variable locations. Notify debug backend of section
1468 changes, and of location view changes.
1469 (rest_of_handle_final): Adjust.
1470 * toplev.c (process_options): Autodetect value for debug variable
1471 location views option. Warn on incompat5 without -gdwarf-5.
1472 * doc/invoke.texi (gvariable-location-views): New.
1473 (gvariable-location-views=incompat5): New.
1474 (gno-variable-location-views): New.
1476 2018-02-08 David Malcolm <dmalcolm@redhat.com>
1478 PR tree-optimization/84136
1479 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
1480 that the result of find_edge is non-NULL.
1482 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1485 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
1486 storing integer register in SImode. Fix cost of 256 and 512
1487 byte aligned SSE register store.
1489 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1491 * config/i386/i386.c (ix86_multiplication_cost): Fix
1492 multiplication cost for TARGET_AVX512DQ.
1494 2018-02-08 Marek Polacek <polacek@redhat.com>
1496 PR tree-optimization/84238
1497 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
1500 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1502 PR tree-optimization/84265
1503 * tree-vect-stmts.c (vectorizable_store): Don't treat
1504 VMAT_CONTIGUOUS accesses as grouped.
1505 (vectorizable_load): Likewise.
1507 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1509 PR tree-optimization/81635
1510 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
1511 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
1512 (test_round_for_mask): New functions.
1513 (wide_int_cc_tests): Call test_round_for_mask.
1514 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
1515 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
1516 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
1517 range returned by get_range_info.
1519 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1522 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
1523 * symtab.c: Include builtins.h
1524 (symtab_node::output_to_lto_symbol_table_p): Move here
1525 from lto-streamer-out.c:output_symbol_p.
1526 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
1527 (output_symbol_p): Move all logic to symtab.c
1528 (produce_symtab): Update.
1530 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1532 * config/s390/s390-opts.h (enum indirect_branch): Define.
1533 * config/s390/s390-protos.h (s390_return_addr_from_memory)
1534 (s390_indirect_branch_via_thunk)
1535 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
1536 (enum s390_indirect_branch_type): Define.
1537 * config/s390/s390.c (struct s390_frame_layout, struct
1538 machine_function): Remove.
1539 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
1540 (indirect_branch_table_label_no, indirect_branch_table_name):
1542 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
1543 (enum s390_indirect_branch_option): Define.
1544 (s390_return_addr_from_memory): New function.
1545 (s390_handle_string_attribute): New function.
1546 (s390_attribute_table): Add new attribute handler.
1547 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
1548 (s390_indirect_branch_via_thunk): New function.
1549 (s390_indirect_branch_via_inline_thunk): New function.
1550 (s390_function_ok_for_sibcall): When jumping via thunk disallow
1551 sibling call optimization for non z10 compiles.
1552 (s390_emit_call): Force indirect branch target to be a single
1553 register. Add r1 clobber for non-z10 compiles.
1554 (s390_emit_epilogue): Emit return jump via return_use expander.
1555 (s390_reorg): Handle JUMP_INSNs as execute targets.
1556 (s390_option_override_internal): Perform validity checks for the
1557 new command line options.
1558 (s390_indirect_branch_attrvalue): New function.
1559 (s390_indirect_branch_settings): New function.
1560 (s390_set_current_function): Invoke s390_indirect_branch_settings.
1561 (s390_output_indirect_thunk_function): New function.
1562 (s390_code_end): Implement target hook.
1563 (s390_case_values_threshold): Implement target hook.
1564 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
1566 * config/s390/s390.h (struct s390_frame_layout)
1567 (struct machine_function): Move here from s390.c.
1568 (TARGET_INDIRECT_BRANCH_NOBP_RET)
1569 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
1570 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
1571 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
1572 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
1573 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
1574 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
1575 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
1576 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
1577 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
1578 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
1579 (mnemonic attribute): Add values which aren't recognized
1581 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
1582 pattern for branch conversion. Fix mnemonic attribute.
1583 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
1584 indirect branch via thunk if requested.
1585 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
1586 ("*indirect_jump"): Disable for branch conversion using out of
1588 ("indirect_jump_via_thunk<mode>_z10")
1589 ("indirect_jump_via_thunk<mode>")
1590 ("indirect_jump_via_inlinethunk<mode>_z10")
1591 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
1592 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
1593 ("casesi_jump_via_inlinethunk<mode>_z10")
1594 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
1595 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
1596 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
1597 ("*indirect2_jump"): Disable for branch conversion.
1598 ("casesi_jump"): Turn into expander and expand patterns for branch
1600 ("return_use"): New expander.
1601 ("*return"): Emit return via thunk and rename it to ...
1602 ("*return<mode>"): ... this one.
1603 * config/s390/s390.opt: Add new options and and enum for the
1606 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1608 * lra-constraints.c (match_reload): Unconditionally use
1609 gen_lowpart_SUBREG, rather than selecting between that
1610 and equivalent gen_rtx_SUBREG code.
1612 2018-02-08 Richard Biener <rguenther@suse.de>
1614 PR tree-optimization/84233
1615 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
1616 changed flag instead of boguously re-using phi_inserted.
1618 2018-02-08 Martin Jambor <mjambor@suse.cz>
1620 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
1621 static local variables.
1623 2018-02-08 Richard Biener <rguenther@suse.de>
1625 PR tree-optimization/84278
1626 * tree-vect-stmts.c (vectorizable_store): When looking for
1627 smaller vector types to perform grouped strided loads/stores
1628 make sure the mode is supported by the target.
1629 (vectorizable_load): Likewise.
1631 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1633 * config/aarch64/aarch64.c (aarch64_components_for_bb):
1634 Increase LDP/STP opportunities by adding adjacent callee-saves.
1636 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1638 PR rtl-optimization/84068
1639 PR rtl-optimization/83459
1640 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
1642 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
1644 PR tree-optimization/84224
1645 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
1646 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
1649 2018-02-07 Iain Sandoe <iain@codesourcery.com>
1652 * config/rs6000/altivec.md (*restore_world): Remove LR use.
1653 * config/rs6000/predicates.md (restore_world_operation): Adjust op
1654 count, remove one USE.
1656 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
1658 * doc/install.texi (Configuration): Document the
1659 --with-long-double-format={ibm,ieee} PowerPC configuration
1663 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
1664 Convert from define_expand to be define_insn_and_split. Rework
1665 float/double/_Float128 conversions to QI/HI/SImode to work with
1666 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
1667 conversions to QI/HImode types did a store and then a load to
1668 truncate the value. For conversions to VSX registers, don't split
1669 the insn, instead emit the code directly. Use the code iterator
1670 any_fix to combine signed and unsigned conversions.
1671 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
1672 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
1673 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
1674 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
1675 (fix_<mode>di2_hw): Likewise.
1676 (fixuns_<mode>di2_hw): Likewise.
1677 (fix_<mode>si2_hw): Likewise.
1678 (fixuns_<mode>si2_hw): Likewise.
1679 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
1680 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
1681 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
1682 fix<uns>_trunc<SFDF:mode>si2_p8.
1683 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
1685 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
1686 (fix<uns>_<mode>_mem): Likewise.
1687 (fctiw<u>z_<mode>_mem): Likewise.
1688 (fix<uns>_<mode>_mem): Likewise.
1689 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
1690 the register allocator from doing a direct move to the GPRs to do
1691 a store, and instead use the ISA 3.0 store byte/half-word from
1692 vector register instruction. For IEEE 128-bit floating point,
1693 also optimize stores of 32-bit ints.
1694 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
1696 2018-02-07 Alan Hayward <alan.hayward@arm.com>
1698 * genextract.c (push_pathstr_operand): New function to support
1700 (walk_rtx): Call push_pathstr_operand.
1701 (print_path): Support [a-zA-Z].
1703 2018-02-07 Richard Biener <rguenther@suse.de>
1705 PR tree-optimization/84037
1706 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
1707 (cse_and_gimplify_to_preheader): Declare.
1708 (vect_get_place_in_interleaving_chain): Likewise.
1709 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
1711 (_loop_vec_info::~_loop_vec_info): Delete it.
1712 (cse_and_gimplify_to_preheader): New function.
1713 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
1714 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
1715 (vectorizable_load): Likewise. For grouped stores always base
1716 the IV on the first element.
1717 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
1718 condition before gimplifying.
1720 2018-02-07 Jakub Jelinek <jakub@redhat.com>
1722 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
1723 *DIV_EXPR and *MOD_EXPR.
1725 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
1728 * config/i386/i386.c (ix86_option_override_internal): Mask out
1729 the CF_SET bit when checking -fcf-protection.
1731 2018-02-07 Tom de Vries <tom@codesourcery.com>
1734 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
1737 2018-02-07 Richard Biener <rguenther@suse.de>
1739 PR tree-optimization/84204
1740 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
1743 PR tree-optimization/84205
1744 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
1745 special-case isl_ast_op_zdiv_r.
1747 PR tree-optimization/84223
1748 * graphite-scop-detection.c (gather_bbs::before_dom_children):
1749 Only add conditions from within the region.
1750 (gather_bbs::after_dom_children): Adjust.
1752 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
1755 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
1756 * config/avr/avr.md: Only post-reload split REG-REG moves if
1757 either register is REGERAL_REG_P.
1759 2018-02-07 Jakub Jelinek <jakub@redhat.com>
1761 PR tree-optimization/84235
1762 * tree-ssa-scopedtables.c
1763 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
1764 if the subtraction is performed in floating point type where NaNs are
1765 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
1766 build 1. Formatting fix.
1768 2018-02-06 Jakub Jelinek <jakub@redhat.com>
1771 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
1772 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
1773 and skip it regardless of bb boundaries. Use CALL_P macro,
1774 don't test INSN_P (insn) together with CALL_P or JUMP_P check
1775 unnecessarily, formatting fix.
1777 2018-02-06 Michael Collison <michael.collison@arm.com>
1779 * config/arm/thumb2.md:
1780 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
1781 (*thumb_mov_notscc): Ditto.
1783 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
1786 * config/rs6000/rs6000.md (su code attribute): Use "u" for
1787 unsigned_fix, not "s".
1789 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1791 * configure.ac (gcc_fn_eh_frame_ro): New function.
1792 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
1793 correct .eh_frame permissions.
1794 * configure: Regenerate.
1796 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
1798 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
1801 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1803 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1804 Display warning message for -mno-speculate-indirect-jumps.
1806 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
1808 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
1810 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
1812 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
1814 PR tree-optimization/84225
1815 * tree-eh.c (find_trapping_overflow): Only call
1816 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
1818 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1821 * config/i386/i386.c: Reimplement the check of possible options
1822 -mibt/-mshstk conbination. Change error messages.
1823 * doc/invoke.texi: Fix a typo: remove extra '='.
1825 2018-02-06 Marek Polacek <polacek@redhat.com>
1827 PR tree-optimization/84228
1828 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
1830 2018-02-06 Tamar Christina <tamar.christina@arm.com>
1833 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
1834 emitted arch directives.
1835 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
1836 __ARM_FEATURE_COPROC before changing architectures.
1838 2018-02-06 Richard Biener <rguenther@suse.de>
1840 * config/i386/i386.c (print_reg): Fix typo.
1841 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
1843 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
1845 * configure: Regenerate.
1847 2018-02-05 Martin Sebor <msebor@redhat.com>
1849 PR tree-optimization/83369
1850 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
1853 2018-02-05 Martin Liska <mliska@suse.cz>
1855 * doc/invoke.texi: Cherry-pick upstream r323995.
1857 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
1859 * ira.c (ira_init_register_move_cost): Adjust comment.
1861 2018-02-05 Martin Liska <mliska@suse.cz>
1863 PR gcov-profile/84137
1864 * doc/gcov.texi: Fix typo in documentation.
1866 2018-02-05 Martin Liska <mliska@suse.cz>
1868 PR gcov-profile/83879
1869 * doc/gcov.texi: Document necessity of --dynamic-list-data when
1870 using dlopen functionality.
1872 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
1874 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
1875 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
1876 _mm_maskz_range_ss, _mm_mask_range_round_ss,
1877 _mm_maskz_range_round_ss): New intrinsics.
1878 (__builtin_ia32_rangesd128_round)
1879 (__builtin_ia32_rangess128_round): Remove.
1880 (__builtin_ia32_rangesd128_mask_round,
1881 __builtin_ia32_rangess128_mask_round): New builtins.
1882 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
1883 __builtin_ia32_rangess128_round): Remove.
1884 (__builtin_ia32_rangesd128_mask_round,
1885 __builtin_ia32_rangess128_mask_round): New builtins.
1886 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
1887 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
1888 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
1889 "<round_saeonly_constraint>")): Changed to ...
1890 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
1891 "<round_saeonly_scalar_constraint>")): ... this.
1892 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
1893 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
1894 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
1895 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
1896 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
1898 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
1900 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
1902 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
1903 Remove all values except native, 8540 and 8548.
1905 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
1907 * config/i386/i386.c (ix86_output_function_return): Pass
1908 INVALID_REGNUM, instead of -1, as invalid register number to
1909 indirect_thunk_name and output_indirect_thunk.
1911 2018-02-02 Julia Koval <julia.koval@intel.com>
1913 * config.gcc: Add -march=icelake.
1914 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
1915 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
1916 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
1917 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
1918 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
1919 (processor_target_table): Add icelake.
1920 (ix86_option_override_internal): Handle new PTAs.
1921 (get_builtin_code_for_version): Handle icelake.
1922 (M_INTEL_COREI7_ICELAKE): New.
1923 (fold_builtin_cpu): Handle icelake.
1924 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
1925 * doc/invoke.texi: Add -march=icelake.
1927 2018-02-02 Julia Koval <julia.koval@intel.com>
1929 * config/i386/i386.c (ix86_option_override_internal): Change flags type
1930 to wide_int_bitmask.
1931 * wide-int-bitmask.h: New.
1933 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1936 * config/i386/i386.md: Replace Pmode with word_mode in
1937 builtin_setjmp_setup and builtin_longjmp to support x32.
1939 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
1943 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
1945 (rs6000_supported_cpu_names): New static variable.
1946 (linux_cpu_translation_table): Likewise.
1947 (elf_platform) <cpu>: Define new static variable and use it.
1948 Translate kernel AT_PLATFORM name to canonical name if needed.
1949 Error if platform name is unknown.
1951 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
1954 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
1956 2018-02-01 Jeff Law <law@redhat.com>
1959 * config/i386/i386.c (release_scratch_register_on_entry): Add new
1960 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
1961 the scratch if RELEASE_VIA_POP is false.
1962 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
1963 If we have to save a temporary register, decrement SIZE appropriately.
1964 Pass new arguments to release_scratch_register_on_entry.
1965 (ix86_adjust_stack_and_probe): Likewise.
1966 (ix86_emit_probe_stack_range): Pass new arguments to
1967 release_scratch_register_on_entry.
1969 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
1971 PR rtl-optimization/84157
1972 * combine.c (change_zero_ext): Use REG_P predicate in
1973 front of HARD_REGISTER_P predicate.
1975 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
1977 * config/avr/avr.c (avr_option_override): Move disabling of
1978 -fdelete-null-pointer-checks to...
1979 * common/config/avr/avr-common.c (avr_option_optimization_table):
1982 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1984 PR tree-optimization/81635
1985 * tree-data-ref.c (split_constant_offset_1): For types that
1986 wrap on overflow, try to use range info to prove that wrapping
1989 2018-02-01 Renlin Li <renlin.li@arm.com>
1992 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
1994 (aarch64_register_move_cost): Likewise.
1995 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
1997 (REG_CLASS_NAMES): Likewise.
1998 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
1999 TAILCALL_ADDR_REGS. Remove IP registers.
2000 * config/aarch64/aarch64.md (Ucs): Update register constraint.
2002 2018-02-01 Richard Biener <rguenther@suse.de>
2004 * domwalk.h (dom_walker::dom_walker): Add additional constructor
2005 for specifying RPO order and allow NULL for that.
2006 * domwalk.c (dom_walker::dom_walker): Likewise.
2007 (dom_walker::walk): Handle NULL RPO order.
2008 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
2010 (rewrite_update_dom_walker): Likewise.
2011 (mark_def_dom_walker): Likewise.
2013 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2015 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
2016 (aarch64_maybe_expand_sve_subreg_move): Declare.
2017 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
2018 * config/aarch64/predicates.md (aarch64_any_register_operand): New
2020 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
2021 that are semantically a reverse operation.
2022 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
2023 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
2024 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
2026 (aarch64_can_change_mode_class): For big-endian, forbid changes
2027 between two SVE modes if they have different element sizes.
2029 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2031 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
2032 the TImode handling for big-endian targets.
2034 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2036 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
2037 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
2039 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
2040 Remove BSWAP handing for big-endian targets and use the form of
2041 LD1RQ appropariate for the mode.
2043 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2045 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
2046 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
2049 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2052 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
2053 check for operands that need to go through aarch64_sve_reload_be.
2055 2018-02-01 Jakub Jelinek <jakub@redhat.com>
2057 PR tree-optimization/81661
2058 PR tree-optimization/84117
2059 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
2060 * tree-eh.c: Include gimplify.h.
2061 (find_trapping_overflow, replace_trapping_overflow,
2062 rewrite_to_non_trapping_overflow): New functions.
2063 * tree-vect-loop.c: Include tree-eh.h.
2064 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
2065 * tree-data-ref.c: Include tree-eh.h.
2066 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
2068 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
2070 PR rtl-optimization/84123
2071 * combine.c (change_zero_ext): Check if hard register satisfies
2072 can_change_dest_mode before calling gen_lowpart_SUBREG.
2074 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
2077 * ira.c (ira_init_register_move_cost): Remove assert.
2079 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2081 PR rtl-optimization/84071
2082 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
2083 * doc/tm.texi: Regenerate.
2085 2018-01-31 Richard Biener <rguenther@suse.de>
2087 PR tree-optimization/84132
2088 * tree-data-ref.c (analyze_miv_subscript): Properly
2089 check whether evolution_function_is_affine_multivariate_p
2090 before calling gcd_of_steps_may_divide_p.
2092 2018-01-31 Julia Koval <julia.koval@intel.com>
2095 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
2096 * config/i386/i386.md (rdpid_rex64) New.
2097 (rdpid): Make 32bit only.
2099 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
2102 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
2103 an IDENTIFIER_NODE for FUNCTION_TYPE's.
2105 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2108 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
2110 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
2112 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2114 PR rtl-optimization/84071
2115 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
2116 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
2118 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2120 * config/arc/arc.c (arc_handle_aux_attribute): New function.
2121 (arc_attribute_table): Add 'aux' attribute.
2122 (arc_in_small_data_p): Consider aux like variables.
2123 (arc_is_aux_reg_p): New function.
2124 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
2125 (arc_get_aux_arg): New function.
2126 (prepare_move_operands): Handle aux-register access.
2127 (arc_handle_aux_attribute): New function.
2128 * doc/extend.texi (ARC Variable attributes): Add subsection.
2130 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2132 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
2133 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
2134 (arc_attribute_table): Add 'uncached' attribute.
2135 (arc_print_operand): Print '.di' flag for uncached memory
2137 (arc_in_small_data_p): Do not consider for small data the uncached
2139 (arc_is_uncached_mem_p): New function.
2140 * config/arc/predicates.md (compact_store_memory_operand): Check
2141 for uncached memory accesses.
2142 (nonvol_nonimm_operand): Likewise.
2143 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
2145 2018-01-31 Jakub Jelinek <jakub@redhat.com>
2148 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
2149 falign-loops=): Add Optimization flag.
2151 2018-01-30 Jeff Law <law@redhat.com>
2154 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
2155 INT_REGISTERS_SAVED. Check it prior to calling
2156 get_scratch_register_on_entry.
2157 (ix86_adjust_stack_and_probe): Similarly.
2158 (ix86_emit_probe_stack_range): Similarly.
2159 (ix86_expand_prologue): Corresponding changes.
2161 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2164 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
2165 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
2167 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
2170 * lra-constraints.c (curr_insn_transform): Process AND in the
2173 2018-01-30 Jakub Jelinek <jakub@redhat.com>
2175 PR rtl-optimization/83986
2176 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
2177 dependence against last_pending_memory_flush in addition to
2180 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
2182 PR tree-optimization/81611
2183 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
2186 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2189 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
2192 2018-01-30 Richard Biener <rguenther@suse.de>
2193 Jakub Jelinek <jakub@redhat.com>
2195 PR tree-optimization/84111
2196 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
2197 inner loops added during recursion, as they don't have up-to-date
2200 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2203 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
2204 (can_inline_edge_by_limits_p): ... here.
2205 (can_early_inline_edge_p, check_callers,
2206 update_caller_keys, update_callee_keys, recursive_inlining,
2207 add_new_edges_to_heap, speculation_useful_p,
2208 inline_small_functions,
2209 inline_small_functions, flatten_function,
2210 inline_to_all_callers_1): Update.
2212 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2214 * profile-count.c (profile_count::combine_with_ipa_count): Handle
2217 2018-01-30 Richard Biener <rguenther@suse.de>
2219 PR tree-optimization/83008
2220 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
2221 invariant and constant vector uses in stmts when they need
2224 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2227 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
2228 * configure: Regenerate.
2230 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2232 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
2234 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
2235 Use gen_rtx_REG rather than gen_lowpart.
2237 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2239 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
2240 rather than 0 when creating partial subregs.
2242 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2244 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
2247 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
2250 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
2251 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
2252 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
2253 flags. This restores the settings used before the 2017-07-24.
2254 Turning off pre increment/decrement/modify allows IVOPTS to
2255 optimize DF/SF loops where the index is an int.
2257 2018-01-29 Richard Biener <rguenther@suse.de>
2258 Kelvin Nilsen <kelvin@gcc.gnu.org>
2261 * tree-vect-stmts.c (vectorizable_call): Don't call
2262 targetm.vectorize_builtin_md_vectorized_function if callee is
2265 2018-01-22 Carl Love <cel@us.ibm.com>
2267 * doc/extend.tex: Fix typo in second arg in
2268 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
2270 2018-01-29 Richard Biener <rguenther@suse.de>
2272 PR tree-optimization/84086
2273 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
2274 (flush_ssaname_freelist): When SSA names were released reset
2275 the SCEV hash table.
2277 2018-01-29 Richard Biener <rguenther@suse.de>
2279 PR tree-optimization/84057
2280 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
2281 removed paths when removing edges.
2283 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
2285 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
2286 -mfunction-return=@var{choice}.
2288 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
2291 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
2293 (layout::print_source_line): Likewise.
2294 (test_get_line_width_without_trailing_whitespace): Add test cases.
2296 2018-01-27 Jakub Jelinek <jakub@redhat.com>
2299 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
2302 2018-01-26 Jim Wilson <jimw@sifive.com>
2304 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
2306 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
2309 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2311 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
2312 and CMP + SUB-immediate -> SUBS.
2314 2018-01-26 Martin Sebor <msebor@redhat.com>
2316 PR tree-optimization/83896
2317 * tree-ssa-strlen.c (get_string_len): Rename...
2318 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
2319 Avoid assuming length is constant.
2320 (handle_char_store): Use HOST_WIDE_INT for string length.
2322 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
2325 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
2326 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
2328 2018-01-26 Richard Biener <rguenther@suse.de>
2330 PR rtl-optimization/84003
2331 * dse.c (record_store): Only record redundant stores when
2332 the earlier store aliases at least all accesses the later one does.
2334 2018-01-26 Jakub Jelinek <jakub@redhat.com>
2336 PR rtl-optimization/83985
2337 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
2338 REG_CFA_RESTORE insns.
2339 (delete_unmarked_insns): Don't ignore separate shrink wrapping
2340 REG_CFA_RESTORE insns here.
2343 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
2344 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
2346 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2348 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
2349 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
2350 (arc_init): Likewise.
2351 (arc_override_options): Likewise.
2352 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
2354 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
2356 * config/arc/arc.h (TARGET_DBNZ): Define.
2357 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
2358 properly set the tune attribute.
2359 (dbnz): Use TARGET_DBNZ guard.
2360 * config/arc/arc.opt (mtune): Add core3 option.
2362 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2364 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
2365 recognize new pic like addresses.
2366 (arc_delegitimize_address): Clean up.
2368 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2370 * config/arc/arc-arches.def: Option mrf16 valid for all
2372 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
2373 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
2374 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
2375 * config/arc/arc-tables.opt: Regenerate.
2376 * config/arc/arc.c (arc_conditional_register_usage): Handle
2377 reduced register file case.
2378 (arc_file_start): Set must have build attributes.
2379 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
2381 * config/arc/arc.opt (mrf16): Add new option.
2382 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
2383 * config/arc/genmultilib.awk: Handle new mrf16 option.
2384 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
2385 * config/arc/t-multilib: Regenerate.
2386 * doc/invoke.texi (ARC Options): Document mrf16 option.
2388 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2390 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
2391 * config/arc/arc.c (arc_handle_secure_attribute): New function.
2392 (arc_attribute_table): Add 'secure_call' attribute.
2393 (arc_print_operand): Print secure call operand.
2394 (arc_function_ok_for_sibcall): Don't optimize tail calls when
2396 (arc_is_secure_call_p): New function. * config/arc/arc.md
2397 (call_i): Add support for sjli instruction.
2398 (call_value_i): Likewise.
2399 * config/arc/constraints.md (Csc): New constraint.
2401 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2402 John Eric Martin <John.Martin@emmicro-us.com>
2404 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
2405 * config/arc/arc.c (_arc_jli_section): New struct.
2406 (arc_jli_section): New type.
2407 (rc_jli_sections): New static variable.
2408 (arc_handle_jli_attribute): New function.
2409 (arc_attribute_table): Add jli_always and jli_fixed attribute.
2410 (arc_file_end): New function.
2411 (TARGET_ASM_FILE_END): Define.
2412 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
2413 (arc_add_jli_section): New function.
2414 (jli_call_scan): Likewise.
2415 (arc_reorg): Call jli_call_scan.
2416 (arc_output_addsi): Remove 'S' from printing asm operand.
2417 (arc_is_jli_call_p): New function.
2418 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
2420 (movhi_insn): Likewise.
2421 (movsi_insn): Likewise.
2422 (movsi_set_cc_insn): Likewise.
2423 (loadqi_update): Likewise.
2424 (load_zeroextendqisi_update): Likewise.
2425 (load_signextendqisi_update): Likewise.
2426 (loadhi_update): Likewise.
2427 (load_zeroextendhisi_update): Likewise.
2428 (load_signextendhisi_update): Likewise.
2429 (loadsi_update): Likewise.
2430 (loadsf_update): Likewise.
2431 (movsicc_insn): Likewise.
2432 (bset_insn): Likewise.
2433 (bxor_insn): Likewise.
2434 (bclr_insn): Likewise.
2435 (bmsk_insn): Likewise.
2436 (bicsi3_insn): Likewise.
2437 (cmpsi_cc_c_insn): Likewise.
2438 (movsi_ne): Likewise.
2439 (movsi_cond_exec): Likewise.
2440 (clrsbsi2): Likewise.
2451 (call_i): Remove 'S' asm letter, add jli instruction.
2452 (call_value_i): Likewise.
2453 * config/arc/arc.op (mjli-always): New option.
2454 * config/arc/constraints.md (Cji): New constraint.
2455 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
2457 (subsf3_fpx): Likewise.
2458 (mulsf3_fpx): Likewise.
2459 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
2461 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
2463 * doc/invoke.texi (ARC): Document mjli-always option.
2465 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
2467 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
2468 avoid addition with 0 and use incw and decw where possible.
2470 2018-01-26 Richard Biener <rguenther@suse.de>
2472 PR tree-optimization/81082
2473 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
2474 association if it requires casting to unsigned.
2475 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
2476 from fold_plusminus_mult_expr to catch important cases late when
2477 range info is available.
2479 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2481 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
2482 * configure.ac (hidden_linkonce): New test.
2483 * configure: Regenerate.
2484 * config.in: Regenerate.
2486 2018-01-26 Julia Koval <julia.koval@intel.com>
2488 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
2489 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
2490 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
2491 _mm_mask_bitshuffle_epi64_mask): Fix type.
2492 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
2493 USI_FTYPE_V4DI_V4DI_USI): Remove.
2494 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
2495 __builtin_ia32_vpshufbitqmb256_mask,
2496 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
2497 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
2498 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
2500 2018-01-26 Alan Modra <amodra@gmail.com>
2503 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
2504 UNSPEC_VBPERMQ. Sort other unspecs.
2506 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
2508 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
2510 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
2513 * predict.c (drop_profile): Do not push/pop cfun; update also
2515 (handle_missing_profiles): Fix logic looking for zero profiles.
2517 2018-01-25 Jakub Jelinek <jakub@redhat.com>
2520 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
2521 on functions with #pragma omp declare simd or functions with simd
2523 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
2524 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
2525 Remove trailing \n from warning_at calls.
2527 2018-01-25 Tom de Vries <tom@codesourcery.com>
2530 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2531 for neutered workers.
2533 2018-01-24 Joseph Myers <joseph@codesourcery.com>
2536 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
2537 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
2539 2018-01-24 Jeff Law <law@redhat.com>
2542 * i386.c (get_probe_interval): Move to earlier point.
2543 (ix86_compute_frame_layout): If -fstack-clash-protection and
2544 the frame is larger than the probe interval, then use pushes
2545 to save registers rather than reg->mem moves.
2546 (ix86_expand_prologue): Remove conditional for int_registers_saved
2549 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
2552 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
2553 min/max for never referenced object.
2555 2018-01-24 Jakub Jelinek <jakub@redhat.com>
2558 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
2560 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
2561 attributes from DECL_ATTRIBUTES (decl) without affecting
2562 DECL_ATTRIBUTES (current_function_decl).
2563 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
2564 functions with non-NULL DECL_ABSTRACT_ORIGIN.
2566 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
2568 PR tree-optimization/83979
2569 * fold-const.c (fold_comparison): Use constant_boolean_node
2570 instead of boolean_{true,false}_node.
2572 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
2574 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
2577 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2579 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
2580 Simplify the clause that sets the length attribute.
2581 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
2582 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
2583 clause that sets the length attribute.
2584 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
2586 2018-01-24 Tom de Vries <tom@codesourcery.com>
2589 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
2590 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
2591 Add strict parameter.
2592 (prevent_branch_around_nothing): Insert dummy insn between branch to
2593 label and label with no ptx insn inbetween.
2594 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
2596 2018-01-24 Tom de Vries <tom@codesourcery.com>
2599 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2600 for neutered threads in warp.
2601 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
2603 2018-01-24 Richard Biener <rguenther@suse.de>
2605 PR tree-optimization/83176
2606 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
2609 2018-01-24 Richard Biener <rguenther@suse.de>
2611 PR tree-optimization/82819
2612 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
2613 code generating pluses that are no-ops in the target precision.
2615 2018-01-24 Richard Biener <rguenther@suse.de>
2618 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
2620 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2622 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
2623 to merge probabilities.
2624 * predict.c (probably_never_executed): Also mark as cold functions
2625 with global 0 profile and guessed local profile.
2626 * profile-count.c (profile_probability::combine_with_count): New
2628 * profile-count.h (profile_probability::operator*,
2629 profile_probability::operator*=, profile_probability::operator/,
2630 profile_probability::operator/=): Reduce precision to adjusted
2631 and set value to guessed on contradictory divisions.
2632 (profile_probability::combine_with_freq): Remove.
2633 (profile_probability::combine_wiht_count): Declare.
2634 (profile_count::force_nonzero):: Set to adjusted.
2635 (profile_count::probability_in):: Set quality to adjusted.
2636 * tree-ssa-tail-merge.c (replace_block_by): Use
2639 2018-01-23 Andrew Waterman <andrew@sifive.com>
2640 Jim Wilson <jimw@sifive.com>
2642 * config/riscv/riscv.c (riscv_stack_boundary): New.
2643 (riscv_option_override): Set riscv_stack_boundary. Handle
2644 riscv_preferred_stack_boundary_arg.
2645 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
2646 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
2647 (STACK_BOUNDARY): Set to riscv_stack_boundary.
2648 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
2649 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
2650 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
2652 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
2655 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
2656 of struct ix86_frame.
2657 (ix86_expand_epilogue): Likewise. Add a local variable for
2658 the reg_save_offset field in struct ix86_frame.
2660 2018-01-23 Bin Cheng <bin.cheng@arm.com>
2662 PR tree-optimization/82604
2663 * tree-loop-distribution.c (enum partition_kind): New enum item
2664 PKIND_PARTIAL_MEMSET.
2665 (partition_builtin_p): Support above new enum item.
2666 (generate_code_for_partition): Ditto.
2667 (compute_access_range): Differentiate cases that equality can be
2668 proven at all loops, the innermost loops or no loops.
2669 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
2670 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
2671 (finalize_partitions, distribute_loop): Don't fuse partition of
2672 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
2673 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
2676 2018-01-23 Martin Liska <mliska@suse.cz>
2678 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
2679 order to ignore the predictor.
2680 (PRED_POLYMORPHIC_CALL): Likewise.
2681 (PRED_RECURSIVE_CALL): Likewise.
2683 2018-01-23 Martin Liska <mliska@suse.cz>
2685 * tree-profile.c (tree_profiling): Print function header to
2686 aware reader which function we are working on.
2687 * value-prof.c (gimple_find_values_to_profile): Do not print
2688 not interesting value histograms.
2690 2018-01-23 Martin Liska <mliska@suse.cz>
2692 * profile-count.h (enum profile_quality): Add
2693 profile_uninitialized as the first value. Do not number values
2694 as they are zero based.
2695 (profile_count::verify): Update sanity check.
2696 (profile_probability::verify): Likewise.
2698 2018-01-23 Nathan Sidwell <nathan@acm.org>
2700 * doc/invoke.texi (ffor-scope): Deprecate.
2702 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2704 PR tree-optimization/83510
2705 * domwalk.c (set_all_edges_as_executable): New function.
2706 (dom_walker::dom_walker): Convert bool param
2707 "skip_unreachable_blocks" to enum reachability. Move setup of
2708 edge flags to set_all_edges_as_executable and only do it when
2709 reachability is REACHABLE_BLOCKS.
2710 * domwalk.h (enum dom_walker::reachability): New enum.
2711 (dom_walker::dom_walker): Convert bool param
2712 "skip_unreachable_blocks" to enum reachability.
2713 (set_all_edges_as_executable): New decl.
2714 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
2715 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
2717 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
2718 but converting true to REACHABLE_BLOCKS.
2719 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
2721 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
2722 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
2723 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
2725 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
2726 if check_all_array_refs will be called.
2728 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2730 * tree.c (selftest::test_location_wrappers): Add more test
2733 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2735 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
2736 (selftest::test_bit_in_range): Likewise.
2738 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
2741 * doc/sourcebuild.texi (vect_float): Say that the selector
2742 only describes the situation when -funsafe-math-optimizations is on.
2743 (vect_float_strict): Document.
2745 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
2747 PR tree-optimization/83965
2748 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
2749 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
2750 instead of checking only for a reduction.
2751 (vect_recog_widen_sum_pattern): Likewise.
2753 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2755 * predict.c (probably_never_executed): Only use precise profile info.
2756 (compute_function_frequency): Skip after inlining hack since we now
2757 have quality checking.
2759 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2761 * profile-count.h (profile_probability::very_unlikely,
2762 profile_probability::unlikely, profile_probability::even): Set
2763 precision to guessed.
2765 2018-01-23 Richard Biener <rguenther@suse.de>
2767 PR tree-optimization/83963
2768 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
2769 Properly terminate dominator walk when crossing the exit edge not
2770 when visiting its source block.
2772 2018-01-23 Jakub Jelinek <jakub@redhat.com>
2775 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
2776 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
2778 2018-01-22 Jakub Jelinek <jakub@redhat.com>
2780 PR tree-optimization/83957
2781 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
2782 semicolon after for body surrounded by braces.
2784 PR tree-optimization/83081
2785 * profile-count.h (profile_probability::split): New method.
2786 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
2787 Use profile_probability::split.
2788 (do_compare_rtx_and_jump): Fix adjustment of probabilities
2789 when splitting a single conditional jump into 2.
2791 2018-01-22 David Malcolm <dmalcolm@redhat.com>
2793 PR tree-optimization/69452
2794 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
2797 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2799 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
2800 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
2801 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
2803 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2805 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
2806 * config/rl78/rl78.md (movdi): New define_expand.
2807 * config/rl78/rl78.c (rl78_split_movdi): New function.
2809 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
2812 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
2814 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
2815 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
2816 128-bit to produce an UNSPEC move to get the double word with the
2817 signbit and then a shift directly to do signbit.
2818 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
2819 implementation with a new version that just does either a direct
2820 move or a regular move. Move memory interface to separate insns.
2821 Move insns so they are next to the expander.
2822 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
2823 with signbit move. Split big and little endian case.
2824 (signbit<mode>2_dm_mem_le): Likewise.
2825 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
2826 (signbit<mode>2_dm2): Likewise.
2828 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2830 * config/rl78/rl78.md (anddi3): New define_expand.
2832 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2834 * config/rl78/rl78.md (umindi3): New define_expand.
2836 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2838 * config/rl78/rl78.md (smindi3): New define_expand.
2840 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2842 * config/rl78/rl78.md (smaxdi3): New define_expand.
2844 2018-01-22 Carl Love <cel@us.ibm.com>
2846 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
2847 LVX_V1TI): Add macro expansion.
2848 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
2849 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
2850 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
2851 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
2852 Change check to determine if the instruction is a byte reversing
2853 entry. Fix typo in comment.
2854 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
2855 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
2856 Add def_builtin calls for new builtins.
2857 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
2858 Add define_insn expansion.
2860 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2862 * config/rl78/rl78.md (umaxdi3): New define_expand.
2864 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2866 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
2867 for non-QImode registers.
2869 2018-01-22 Richard Biener <rguenther@suse.de>
2871 PR tree-optimization/83963
2872 * graphite-scop-detection.c (scop_detection::get_sese): Delay
2873 including the loop exit block.
2874 (scop_detection::merge_sese): Likewise.
2875 (scop_detection::add_scop): Do it here instead.
2877 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2879 * doc/sourcebuild.texi (arm_softfloat): Document.
2881 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
2884 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
2885 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
2886 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
2888 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2889 David Edelsohn <dje.gcc@gmail.com>
2892 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
2893 Change "crset eq" to "crset 2".
2894 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
2895 (*call_indirect_aix<mode>_nospec): Likewise.
2896 (*call_value_indirect_aix<mode>_nospec): Likewise.
2897 (*call_indirect_elfv2<mode>_nospec): Likewise.
2898 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
2899 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
2900 change assembly output from . to $.
2901 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
2902 (indirect_jump<mode>_nospec): Change assembly output from . to $.
2903 (*tablejump<mode>_internal1_nospec): Likewise.
2905 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
2908 * config/sh/sh_optimize_sett_clrt.cc:
2909 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
2911 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
2913 PR tree-optimization/83940
2914 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
2915 offset_dt to vect_constant_def rather than vect_unknown_def_type.
2916 (vect_check_load_store_mask): Add a mask_dt_out parameter and
2917 use it to pass back the definition type.
2918 (vect_check_store_rhs): Likewise rhs_dt_out.
2919 (vect_build_gather_load_calls): Add a mask_dt argument and use
2920 it instead of a call to vect_is_simple_use.
2921 (vectorizable_store): Update calls to vect_check_load_store_mask
2922 and vect_check_store_rhs. Use the dt returned by the latter instead
2923 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
2924 instead of calls to vect_is_simple_use. Pass the scalar rather
2925 than the vector operand to vect_is_simple_use when handling
2926 second and subsequent copies of an rhs value.
2927 (vectorizable_load): Update calls to vect_check_load_store_mask
2928 and vect_build_gather_load_calls. Use the cached mask_dt and
2929 gs_info.offset_dt instead of calls to vect_is_simple_use.
2931 2018-01-20 Jakub Jelinek <jakub@redhat.com>
2934 * tree-emutls.c: Include gimplify.h.
2935 (lower_emutls_2): New function.
2936 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
2937 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
2938 it before further processing.
2941 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
2942 UINTVAL (trueop1) instead of INTVAL (op1).
2944 2018-01-19 Jakub Jelinek <jakub@redhat.com>
2948 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
2949 INCOMING_FRAME_SP_OFFSET if not defined.
2950 (scan_trace): Add ENTRY argument. If true and
2951 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
2952 emit a note to adjust the CFA offset.
2953 (create_cfi_notes): Adjust scan_trace callers.
2954 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
2955 INCOMING_FRAME_SP_OFFSET in the CIE.
2956 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
2957 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
2959 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
2960 * doc/tm.texi: Regenerated.
2962 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2964 PR rtl-optimization/83147
2965 * lra-constraints.c (remove_inheritance_pseudos): Use
2966 lra_substitute_pseudo_within_insn.
2968 2018-01-19 Tom de Vries <tom@codesourcery.com>
2969 Cesar Philippidis <cesar@codesourcery.com>
2972 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
2974 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
2977 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
2978 spaces for function labels.
2980 2018-01-19 Martin Liska <mliska@suse.cz>
2982 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
2983 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
2984 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
2985 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
2986 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
2987 (PRED_CONST_RETURN): Change from 69 to 65.
2988 (PRED_NULL_RETURN): Change from 91 to 71.
2989 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
2990 (PRED_LOOP_GUARD): Change from 66 to 73.
2992 2018-01-19 Martin Liska <mliska@suse.cz>
2994 * predict.c (predict_insn_def): Add new assert.
2995 (struct branch_predictor): Change type to signed integer.
2996 (test_prediction_value_range): Amend test to cover
2998 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
2999 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
3000 (PRED_LOOP_ITERATIONS_MAX): Likewise.
3001 (PRED_LOOP_IV_COMPARE): Likewise.
3002 * predict.h (PROB_UNINITIALIZED): Define new constant.
3004 2018-01-19 Martin Liska <mliska@suse.cz>
3006 * predict.c (dump_prediction): Add new format for
3007 analyze_brprob.py script which is enabled with -details
3009 * profile-count.h (precise_p): New function.
3011 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3013 PR tree-optimization/83922
3014 * tree-vect-loop.c (vect_verify_full_masking): Return false if
3015 there are no statements that need masking.
3016 (vect_active_double_reduction_p): New function.
3017 (vect_analyze_loop_operations): Use it when handling phis that
3018 are not in the loop header.
3020 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3022 PR tree-optimization/83914
3023 * tree-vect-loop.c (vectorizable_induction): Don't convert
3024 init_expr or apply the peeling adjustment for inductions
3025 that are nested within the vectorized loop.
3027 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3029 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
3032 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3036 * function.h (gimplify_parameters): Add gimple_seq * argument.
3037 * function.c: Include gimple.h and options.h.
3038 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
3039 for the added local temporaries if needed.
3040 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
3041 if there are any parameter cleanups, wrap whole body into a
3042 try/finally with the cleanups.
3044 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
3047 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
3048 Use GET_MODE_CLASS for scalar floating point.
3050 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3054 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3055 Fix call of call_cgraph_insertion_hooks.
3057 2018-01-18 Martin Sebor <msebor@redhat.com>
3059 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
3061 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3064 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
3067 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
3070 * common.opt: (-ffile-prefix-map): New option.
3071 * opts.c (common_handle_option): Defer it.
3072 * opts-global.c (handle_common_deferred_options): Handle it.
3073 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
3074 * file-prefix-map.h: New file.
3075 (remap_debug_filename, add_debug_prefix_map): ...here.
3076 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
3077 * final.c (debug_prefix_map, add_debug_prefix_map
3078 remap_debug_filename): Move to...
3079 * file-prefix-map.c: New file.
3080 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
3081 generalize, get rid of alloca(), use strrchr() instead of strchr().
3082 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
3083 Implement in terms of add_prefix_map().
3084 (remap_macro_filename, remap_debug_filename): Implement in term of
3086 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
3087 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
3088 * dbxout.c: Include file-prefix-map.h.
3089 * varasm.c: Likewise.
3090 * vmsdbgout.c: Likewise.
3091 * xcoffout.c: Likewise.
3092 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
3093 * doc/cppopts.texi (-fmacro-prefix-map): Document.
3094 * doc/invoke.texi (-ffile-prefix-map): Document.
3095 (-fdebug-prefix-map): Update description.
3097 2018-01-18 Martin Liska <mliska@suse.cz>
3099 * config/i386/i386.c (indirect_thunk_name): Document that also
3101 (output_indirect_thunk): Document why both instructions
3102 (pause and lfence) are generated.
3104 2018-01-18 Richard Biener <rguenther@suse.de>
3106 PR tree-optimization/83887
3107 * graphite-scop-detection.c
3108 (scop_detection::get_nearest_dom_with_single_entry): Remove.
3109 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
3110 (scop_detection::merge_sese): Re-implement with a flood-fill
3111 algorithm that properly finds a SESE region if it exists.
3113 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3116 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
3117 pointer_diff optimizations use view_convert instead of convert.
3119 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3121 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3122 Generate different code for -mno-speculate-indirect-jumps.
3123 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3124 (*call_indirect_aix<mode>): Disable for
3125 -mno-speculate-indirect-jumps.
3126 (*call_indirect_aix<mode>_nospec): New define_insn.
3127 (*call_value_indirect_aix<mode>): Disable for
3128 -mno-speculate-indirect-jumps.
3129 (*call_value_indirect_aix<mode>_nospec): New define_insn.
3130 (*sibcall_nonlocal_sysv<mode>): Generate different code for
3131 -mno-speculate-indirect-jumps.
3132 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3134 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3136 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
3137 long double type, set the flags for noting the default long double
3138 type, even if we don't pass or return a long double type.
3140 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
3143 * ipa-inline.c (flatten_function): Do not overwrite final inlining
3146 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
3148 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
3149 support for merge[hl].
3150 (fold_mergehl_helper): New helper function.
3151 (tree-vector-builder.h): New #include for tree_vector_builder usage.
3152 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
3153 (altivec_vmrglw_direct): Add xxmrglw insn.
3155 2018-01-17 Andrew Waterman <andrew@sifive.com>
3157 * config/riscv/riscv.c (riscv_conditional_register_usage): If
3158 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
3160 2018-01-17 David Malcolm <dmalcolm@redhat.com>
3163 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
3164 call the lto_location_cache before reading the
3165 DECL_SOURCE_LOCATION of the types.
3167 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3168 Richard Sandiford <richard.sandiford@linaro.org>
3170 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
3171 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
3172 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
3173 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
3174 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
3176 * config/aarch64/constraints.md (aarch64_movti_operand):
3178 * config/aarch64/predicates.md (Uti): Add new constraint.
3180 2018-01-17 Carl Love <cel@us.ibm.com>
3182 * config/rs6000/vsx.md (define_expand xl_len_r,
3183 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
3184 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
3186 (define_expand, define_insn): Move the shift left from the
3187 define_insn to the define_expand for lxvl and stxvl instructions.
3188 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
3189 and XL_LEN_R definitions to PURE.
3191 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
3193 * config/i386/i386.c (indirect_thunk_name): Declare regno
3194 as unsigned int. Compare regno with INVALID_REGNUM.
3195 (output_indirect_thunk): Ditto.
3196 (output_indirect_thunk_function): Ditto.
3197 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
3198 in the call to output_indirect_thunk_function.
3200 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
3203 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
3204 rather than the size of inner_type to determine the stack slot size
3205 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
3207 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
3210 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
3213 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
3215 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
3216 endian Linux systems to optionally enable multilibs for selecting
3217 the long double type if the user configured an explicit type.
3218 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
3219 have no long double multilibs if not defined.
3220 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
3221 warn if the user used -mabi={ieee,ibm}longdouble and we built
3222 multilibs for long double.
3223 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
3224 appropriate multilib option.
3225 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
3227 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
3228 for building long double multilibs.
3229 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
3231 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
3233 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
3236 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
3238 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
3241 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
3244 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
3247 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3249 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
3250 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
3252 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3254 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
3255 different rtl trees depending on TARGET_64BIT.
3256 (rs6000_gen_lvx): Likewise.
3258 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3260 * config/visium/visium.md (nop): Tweak comment.
3261 (hazard_nop): Likewise.
3263 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3265 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
3266 -mspeculate-indirect-jumps.
3267 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
3268 for -mno-speculate-indirect-jumps.
3269 (*call_indirect_elfv2<mode>_nospec): New define_insn.
3270 (*call_value_indirect_elfv2<mode>): Disable for
3271 -mno-speculate-indirect-jumps.
3272 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
3273 (indirect_jump): Emit different RTL for
3274 -mno-speculate-indirect-jumps.
3275 (*indirect_jump<mode>): Disable for
3276 -mno-speculate-indirect-jumps.
3277 (*indirect_jump<mode>_nospec): New define_insn.
3278 (tablejump): Emit different RTL for
3279 -mno-speculate-indirect-jumps.
3280 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
3281 (tablejumpsi_nospec): New define_expand.
3282 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
3283 (tablejumpdi_nospec): New define_expand.
3284 (*tablejump<mode>_internal1): Disable for
3285 -mno-speculate-indirect-jumps.
3286 (*tablejump<mode>_internal1_nospec): New define_insn.
3287 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
3290 2018-01-16 Artyom Skrobov tyomitch@gmail.com
3292 * caller-save.c (insert_save): Drop unnecessary parameter. All
3295 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3296 Richard Biener <rguenth@suse.de>
3299 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
3300 return early, inline manually is_gimple_sizepos. Make sure if we
3301 call gimplify_expr we don't end up with a gimple constant.
3302 * tree.c (variably_modified_type_p): Don't return true for
3303 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
3304 * gimplify.h (is_gimple_sizepos): Remove.
3306 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3308 PR tree-optimization/83857
3309 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
3310 vectorizable_live_operation for pure SLP statements.
3311 (vectorizable_live_operation): Handle PHIs.
3313 2018-01-16 Richard Biener <rguenther@suse.de>
3315 PR tree-optimization/83867
3316 * tree-vect-stmts.c (vect_transform_stmt): Precompute
3317 nested_in_vect_loop_p since the scalar stmt may get invalidated.
3319 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3322 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
3323 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
3324 If off is not INTEGER_CST, issue a may not be aligned warning
3325 rather than isn't aligned. Use isn%'t rather than isn't.
3326 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
3328 <case MULT_EXPR>: Improve the case when bottom and one of the
3329 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
3330 operand, in that case check if the other operand is multiple of
3331 bottom divided by the INTEGER_CST operand.
3333 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3336 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
3337 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
3338 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
3339 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
3340 * config/pa/pa.c (pa_function_arg_advance): Likewise.
3341 (pa_function_arg, pa_arg_partial_bytes): Likewise.
3342 (pa_function_arg_size): New function.
3344 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3346 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
3347 in a separate statement.
3349 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3351 PR tree-optimization/83847
3352 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
3353 group gathers and scatters.
3355 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3357 PR rtl-optimization/86620
3358 * params.def (max-sched-ready-insns): Bump minimum value to 1.
3360 PR rtl-optimization/83213
3361 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
3362 to last if both are JUMP_INSNs.
3364 PR tree-optimization/83843
3365 * gimple-ssa-store-merging.c
3366 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
3367 store_immediate_info for bswap/nop orig_stores.
3369 2018-01-15 Andrew Waterman <andrew@sifive.com>
3371 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
3373 <UDIV>: Increase cost if !TARGET_DIV.
3375 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
3377 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
3378 (define_attr "cr_logical_3op"): New.
3379 (cceq_ior_compare): Adjust.
3380 (cceq_ior_compare_complement): Adjust.
3381 (*cceq_rev_compare): Adjust.
3382 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
3383 (is_cracked_insn): Adjust.
3384 (insn_must_be_first_in_group): Adjust.
3385 * config/rs6000/40x.md: Adjust.
3386 * config/rs6000/440.md: Adjust.
3387 * config/rs6000/476.md: Adjust.
3388 * config/rs6000/601.md: Adjust.
3389 * config/rs6000/603.md: Adjust.
3390 * config/rs6000/6xx.md: Adjust.
3391 * config/rs6000/7450.md: Adjust.
3392 * config/rs6000/7xx.md: Adjust.
3393 * config/rs6000/8540.md: Adjust.
3394 * config/rs6000/cell.md: Adjust.
3395 * config/rs6000/e300c2c3.md: Adjust.
3396 * config/rs6000/e500mc.md: Adjust.
3397 * config/rs6000/e500mc64.md: Adjust.
3398 * config/rs6000/e5500.md: Adjust.
3399 * config/rs6000/e6500.md: Adjust.
3400 * config/rs6000/mpc.md: Adjust.
3401 * config/rs6000/power4.md: Adjust.
3402 * config/rs6000/power5.md: Adjust.
3403 * config/rs6000/power6.md: Adjust.
3404 * config/rs6000/power7.md: Adjust.
3405 * config/rs6000/power8.md: Adjust.
3406 * config/rs6000/power9.md: Adjust.
3407 * config/rs6000/rs64.md: Adjust.
3408 * config/rs6000/titan.md: Adjust.
3410 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3412 * config/i386/predicates.md (indirect_branch_operand): Rewrite
3413 ix86_indirect_branch_register logic.
3415 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3417 * config/i386/constraints.md (Bs): Update
3418 ix86_indirect_branch_register check. Don't check
3419 ix86_indirect_branch_register with GOT_memory_operand.
3421 * config/i386/predicates.md (GOT_memory_operand): Don't check
3422 ix86_indirect_branch_register here.
3423 (GOT32_symbol_operand): Likewise.
3425 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3427 * config/i386/predicates.md (constant_call_address_operand):
3428 Rewrite ix86_indirect_branch_register logic.
3429 (sibcall_insn_operand): Likewise.
3431 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3433 * config/i386/constraints.md (Bs): Replace
3434 ix86_indirect_branch_thunk_register with
3435 ix86_indirect_branch_register.
3437 * config/i386/i386.md (indirect_jump): Likewise.
3438 (tablejump): Likewise.
3439 (*sibcall_memory): Likewise.
3440 (*sibcall_value_memory): Likewise.
3441 Peepholes of indirect call and jump via memory: Likewise.
3442 * config/i386/i386.opt: Likewise.
3443 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3444 (GOT_memory_operand): Likewise.
3445 (call_insn_operand): Likewise.
3446 (sibcall_insn_operand): Likewise.
3447 (GOT32_symbol_operand): Likewise.
3449 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3452 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
3453 type rather than type addr's type points to.
3454 (expand_omp_atomic_mutex): Likewise.
3455 (expand_omp_atomic): Likewise.
3457 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3460 * config/i386/i386.c (output_indirect_thunk_function): Use
3461 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
3462 for __x86_return_thunk.
3464 2018-01-15 Richard Biener <rguenther@suse.de>
3467 * expmed.c (extract_bit_field_1): Fix typo.
3469 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3472 * config/arm/iterators.md (VF): New mode iterator.
3473 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
3474 Remove integer-related logic from pattern.
3475 (neon_vabd<mode>_3): Likewise.
3477 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3480 * common.opt (fstrict-overflow): No longer an alias.
3481 (fwrapv-pointer): New option.
3482 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
3483 also for pointer types based on flag_wrapv_pointer.
3484 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
3485 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
3486 opts->x_flag_wrapv got set.
3487 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
3488 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
3489 POINTER_TYPE_OVERFLOW_UNDEFINED.
3490 * match.pd: Likewise in address comparison pattern.
3491 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
3493 2018-01-15 Richard Biener <rguenther@suse.de>
3496 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
3497 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
3498 Reset type names to their identifier if their TYPE_DECL doesn't
3499 have linkage (and thus is used for ODR and devirt).
3500 (save_debug_info_for_decl): Remove.
3501 (save_debug_info_for_type): Likewise.
3502 (add_tree_to_fld_list): Adjust.
3503 * tree-pretty-print.c (dump_generic_node): Make dumping of
3504 type names more robust.
3506 2018-01-15 Richard Biener <rguenther@suse.de>
3508 * BASE-VER: Bump to 8.0.1.
3510 2018-01-14 Martin Sebor <msebor@redhat.com>
3513 * builtins.c (check_access): Avoid warning when the no-warning bit
3516 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
3518 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
3519 * ira-color (allocno_hard_regs_compare): Likewise.
3521 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
3524 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
3525 Use .pushsection/.popsection.
3527 2018-01-14 Martin Sebor <msebor@redhat.com>
3530 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
3532 2018-01-14 Jakub Jelinek <jakub@redhat.com>
3534 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
3535 entry from extra_headers.
3536 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
3537 extra_headers, make the list bitwise identical to the i?86-*-* one.
3539 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3541 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
3542 -mcmodel=large with -mindirect-branch=thunk,
3543 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
3544 -mfunction-return=thunk-extern.
3545 * doc/invoke.texi: Document -mcmodel=large is incompatible with
3546 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
3547 -mfunction-return=thunk and -mfunction-return=thunk-extern.
3549 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3551 * config/i386/i386.c (print_reg): Print the name of the full
3552 integer register without '%'.
3553 (ix86_print_operand): Handle 'V'.
3554 * doc/extend.texi: Document 'V' modifier.
3556 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3558 * config/i386/constraints.md (Bs): Disallow memory operand for
3559 -mindirect-branch-register.
3561 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3562 (GOT_memory_operand): Likewise.
3563 (call_insn_operand): Likewise.
3564 (sibcall_insn_operand): Likewise.
3565 (GOT32_symbol_operand): Likewise.
3566 * config/i386/i386.md (indirect_jump): Call convert_memory_address
3567 for -mindirect-branch-register.
3568 (tablejump): Likewise.
3569 (*sibcall_memory): Likewise.
3570 (*sibcall_value_memory): Likewise.
3571 Disallow peepholes of indirect call and jump via memory for
3572 -mindirect-branch-register.
3573 (*call_pop): Replace m with Bw.
3574 (*call_value_pop): Likewise.
3575 (*sibcall_pop_memory): Replace m with Bs.
3576 * config/i386/i386.opt (mindirect-branch-register): New option.
3577 * doc/invoke.texi: Document -mindirect-branch-register option.
3579 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3581 * config/i386/i386-protos.h (ix86_output_function_return): New.
3582 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
3583 set function_return_type.
3584 (indirect_thunk_name): Add ret_p to indicate thunk for function
3586 (output_indirect_thunk_function): Pass false to
3587 indirect_thunk_name.
3588 (ix86_output_indirect_branch_via_reg): Likewise.
3589 (ix86_output_indirect_branch_via_push): Likewise.
3590 (output_indirect_thunk_function): Create alias for function
3591 return thunk if regno < 0.
3592 (ix86_output_function_return): New function.
3593 (ix86_handle_fndecl_attribute): Handle function_return.
3594 (ix86_attribute_table): Add function_return.
3595 * config/i386/i386.h (machine_function): Add
3596 function_return_type.
3597 * config/i386/i386.md (simple_return_internal): Use
3598 ix86_output_function_return.
3599 (simple_return_internal_long): Likewise.
3600 * config/i386/i386.opt (mfunction-return=): New option.
3601 (indirect_branch): Mention -mfunction-return=.
3602 * doc/extend.texi: Document function_return function attribute.
3603 * doc/invoke.texi: Document -mfunction-return= option.
3605 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3607 * config/i386/i386-opts.h (indirect_branch): New.
3608 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
3609 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
3610 with local indirect jump when converting indirect call and jump.
3611 (ix86_set_indirect_branch_type): New.
3612 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
3613 (indirectlabelno): New.
3614 (indirect_thunk_needed): Likewise.
3615 (indirect_thunk_bnd_needed): Likewise.
3616 (indirect_thunks_used): Likewise.
3617 (indirect_thunks_bnd_used): Likewise.
3618 (INDIRECT_LABEL): Likewise.
3619 (indirect_thunk_name): Likewise.
3620 (output_indirect_thunk): Likewise.
3621 (output_indirect_thunk_function): Likewise.
3622 (ix86_output_indirect_branch_via_reg): Likewise.
3623 (ix86_output_indirect_branch_via_push): Likewise.
3624 (ix86_output_indirect_branch): Likewise.
3625 (ix86_output_indirect_jmp): Likewise.
3626 (ix86_code_end): Call output_indirect_thunk_function if needed.
3627 (ix86_output_call_insn): Call ix86_output_indirect_branch if
3629 (ix86_handle_fndecl_attribute): Handle indirect_branch.
3630 (ix86_attribute_table): Add indirect_branch.
3631 * config/i386/i386.h (machine_function): Add indirect_branch_type
3632 and has_local_indirect_jump.
3633 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
3635 (tablejump): Likewise.
3636 (*indirect_jump): Use ix86_output_indirect_jmp.
3637 (*tablejump_1): Likewise.
3638 (simple_return_indirect_internal): Likewise.
3639 * config/i386/i386.opt (mindirect-branch=): New option.
3640 (indirect_branch): New.
3643 (thunk-inline): Likewise.
3644 (thunk-extern): Likewise.
3645 * doc/extend.texi: Document indirect_branch function attribute.
3646 * doc/invoke.texi: Document -mindirect-branch= option.
3648 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
3651 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
3653 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
3655 * ipa-inline.c (want_inline_small_function_p): Return false if
3656 inlining has already failed with CIF_FINAL_ERROR.
3657 (update_caller_keys): Call want_inline_small_function_p before
3659 (update_callee_keys): Likewise.
3661 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3663 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
3665 (rs6000_quadword_masked_address_p): Likewise.
3666 (quad_aligned_load_p): Likewise.
3667 (quad_aligned_store_p): Likewise.
3668 (const_load_sequence_p): Add comment to describe the outer-most loop.
3669 (mimic_memory_attributes_and_flags): New function.
3670 (rs6000_gen_stvx): Likewise.
3671 (replace_swapped_aligned_store): Likewise.
3672 (rs6000_gen_lvx): Likewise.
3673 (replace_swapped_aligned_load): Likewise.
3674 (replace_swapped_load_constant): Capitalize argument name in
3675 comment describing this function.
3676 (rs6000_analyze_swaps): Add a third pass to search for vector loads
3677 and stores that access quad-word aligned addresses and replace
3678 with stvx or lvx instructions when appropriate.
3679 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
3680 New function prototype.
3681 (rs6000_quadword_masked_address_p): Likewise.
3682 (rs6000_gen_lvx): Likewise.
3683 (rs6000_gen_stvx): Likewise.
3684 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
3685 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
3686 when memory address is aligned.
3687 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
3688 this split to select lvx instruction when memory address is aligned.
3689 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
3690 instruction when memory address is aligned.
3691 (*vsx_le_perm_load_v16qi): Likewise.
3692 (four unnamed splitters): Modify to select the stvx instruction
3693 when memory is aligned.
3695 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
3697 * predict.c (determine_unlikely_bbs): Handle correctly BBs
3698 which appears in the queue multiple times.
3700 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3701 Alan Hayward <alan.hayward@arm.com>
3702 David Sherwood <david.sherwood@arm.com>
3704 * tree-vectorizer.h (vec_lower_bound): New structure.
3705 (_loop_vec_info): Add check_nonzero and lower_bounds.
3706 (LOOP_VINFO_CHECK_NONZERO): New macro.
3707 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
3708 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
3709 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
3710 fields. Make seg_len the distance travelled, not including the
3712 (dr_direction_indicator): Declare.
3713 (dr_zero_step_indicator): Likewise.
3714 (dr_known_forward_stride_p): Likewise.
3715 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
3717 (runtime_alias_check_p): Allow runtime alias checks with
3719 (operator ==): Compare access_size and align.
3720 (prune_runtime_alias_test_list): Rework for new distinction between
3721 the access_size and seg_len.
3722 (create_intersect_range_checks_index): Likewise. Cope with polynomial
3724 (get_segment_min_max): New function.
3725 (create_intersect_range_checks): Use it.
3726 (dr_step_indicator): New function.
3727 (dr_direction_indicator): Likewise.
3728 (dr_zero_step_indicator): Likewise.
3729 (dr_known_forward_stride_p): Likewise.
3730 * tree-loop-distribution.c (data_ref_segment_size): Return
3731 DR_STEP * (niters - 1).
3732 (compute_alias_check_pairs): Update call to the dr_with_seg_len
3734 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
3735 (vect_preserves_scalar_order_p): New function, split out from...
3736 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
3737 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
3738 (vect_vfa_access_size): New function.
3739 (vect_vfa_align): Likewise.
3740 (vect_compile_time_alias): Take access_size_a and access_b arguments.
3741 (dump_lower_bound): New function.
3742 (vect_check_lower_bound): Likewise.
3743 (vect_small_gap_p): Likewise.
3744 (vectorizable_with_step_bound_p): Likewise.
3745 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
3746 depencies if the vectorization factor is 1. Convert the checks
3747 for nonzero steps into checks on the bounds of DR_STEP. Try using
3748 a bunds check for variable steps if the minimum required step is
3749 relatively small. Update calls to the dr_with_seg_len
3750 constructor and to vect_compile_time_alias.
3751 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
3753 (vect_loop_versioning): Call it.
3754 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
3756 (vect_estimate_min_profitable_iters): Account for any bounds checks.
3758 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3759 Alan Hayward <alan.hayward@arm.com>
3760 David Sherwood <david.sherwood@arm.com>
3762 * doc/sourcebuild.texi (vect_scatter_store): Document.
3763 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
3765 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
3767 * genopinit.c (main): Add supports_vec_scatter_store and
3768 supports_vec_scatter_store_cached to target_optabs.
3769 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
3770 IFN_MASK_SCATTER_STORE.
3771 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
3773 * internal-fn.h (internal_store_fn_p): Declare.
3774 (internal_fn_stored_value_index): Likewise.
3775 * internal-fn.c (scatter_store_direct): New macro.
3776 (expand_scatter_store_optab_fn): New function.
3777 (direct_scatter_store_optab_supported_p): New macro.
3778 (internal_store_fn_p): New function.
3779 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
3780 IFN_MASK_SCATTER_STORE.
3781 (internal_fn_mask_index): Likewise.
3782 (internal_fn_stored_value_index): New function.
3783 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
3785 * optabs-query.h (supports_vec_scatter_store_p): Declare.
3786 * optabs-query.c (supports_vec_scatter_store_p): New function.
3787 * tree-vectorizer.h (vect_get_store_rhs): Declare.
3788 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
3789 true for scatter stores.
3790 (vect_gather_scatter_fn_p): Handle scatter stores too.
3791 (vect_check_gather_scatter): Consider using scatter stores if
3792 supports_vec_scatter_store_p.
3793 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
3795 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
3796 internal_fn_stored_value_index.
3797 (check_load_store_masking): Handle scatter stores too.
3798 (vect_get_store_rhs): Make public.
3799 (vectorizable_call): Use internal_store_fn_p.
3800 (vectorizable_store): Handle scatter store internal functions.
3801 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
3802 when deciding whether the end of the group has been reached.
3803 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
3804 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
3805 (mask_scatter_store<mode>): New insns.
3807 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3808 Alan Hayward <alan.hayward@arm.com>
3809 David Sherwood <david.sherwood@arm.com>
3811 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
3812 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
3813 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
3815 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
3816 Use vect_truncate_gather_scatter_offset if we can't treat the
3817 operation as a normal gather load or scatter store.
3818 (get_group_load_store_type): Take the gather_scatter_info
3819 as argument. Try using a gather load or scatter store for
3820 single-element groups.
3821 (get_load_store_type): Update calls to get_group_load_store_type
3822 and vect_use_strided_gather_scatters_p.
3824 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3825 Alan Hayward <alan.hayward@arm.com>
3826 David Sherwood <david.sherwood@arm.com>
3828 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
3829 optional tree argument.
3830 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
3832 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
3833 but continue to use the current value as a fallback.
3834 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
3835 to compare the updates.
3836 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
3837 (get_load_store_type): Use it when handling a strided access.
3838 (vect_get_strided_load_store_ops): New function.
3839 (vect_get_data_ptr_increment): Likewise.
3840 (vectorizable_load): Handle strided gather loads. Always pass
3841 a step to vect_create_data_ref_ptr and bump_vector_ptr.
3843 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3844 Alan Hayward <alan.hayward@arm.com>
3845 David Sherwood <david.sherwood@arm.com>
3847 * doc/md.texi (gather_load@var{m}): Document.
3848 (mask_gather_load@var{m}): Likewise.
3849 * genopinit.c (main): Add supports_vec_gather_load and
3850 supports_vec_gather_load_cached to target_optabs.
3851 * optabs-tree.c (init_tree_optimization_optabs): Use
3852 ggc_cleared_alloc to allocate target_optabs.
3853 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
3854 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
3856 * internal-fn.h (internal_load_fn_p): Declare.
3857 (internal_gather_scatter_fn_p): Likewise.
3858 (internal_fn_mask_index): Likewise.
3859 (internal_gather_scatter_fn_supported_p): Likewise.
3860 * internal-fn.c (gather_load_direct): New macro.
3861 (expand_gather_load_optab_fn): New function.
3862 (direct_gather_load_optab_supported_p): New macro.
3863 (direct_internal_fn_optab): New function.
3864 (internal_load_fn_p): Likewise.
3865 (internal_gather_scatter_fn_p): Likewise.
3866 (internal_fn_mask_index): Likewise.
3867 (internal_gather_scatter_fn_supported_p): Likewise.
3868 * optabs-query.c (supports_at_least_one_mode_p): New function.
3869 (supports_vec_gather_load_p): Likewise.
3870 * optabs-query.h (supports_vec_gather_load_p): Declare.
3871 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
3872 and memory_type field.
3873 (NUM_PATTERNS): Bump to 15.
3874 * tree-vect-data-refs.c: Include internal-fn.h.
3875 (vect_gather_scatter_fn_p): New function.
3876 (vect_describe_gather_scatter_call): Likewise.
3877 (vect_check_gather_scatter): Try using internal functions for
3878 gather loads. Recognize existing calls to a gather load function.
3879 (vect_analyze_data_refs): Consider using gather loads if
3880 supports_vec_gather_load_p.
3881 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
3882 (vect_get_gather_scatter_offset_type): Likewise.
3883 (vect_convert_mask_for_vectype): Likewise.
3884 (vect_add_conversion_to_patterm): Likewise.
3885 (vect_try_gather_scatter_pattern): Likewise.
3886 (vect_recog_gather_scatter_pattern): New pattern recognizer.
3887 (vect_vect_recog_func_ptrs): Add it.
3888 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
3889 internal_fn_mask_index and internal_gather_scatter_fn_p.
3890 (check_load_store_masking): Take the gather_scatter_info as an
3891 argument and handle gather loads.
3892 (vect_get_gather_scatter_ops): New function.
3893 (vectorizable_call): Check internal_load_fn_p.
3894 (vectorizable_load): Likewise. Handle gather load internal
3896 (vectorizable_store): Update call to check_load_store_masking.
3897 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
3898 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
3899 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
3900 (aarch64_gather_scale_operand_d): New predicates.
3901 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
3902 (mask_gather_load<mode>): New insns.
3904 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3905 Alan Hayward <alan.hayward@arm.com>
3906 David Sherwood <david.sherwood@arm.com>
3908 * optabs.def (fold_left_plus_optab): New optab.
3909 * doc/md.texi (fold_left_plus_@var{m}): Document.
3910 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
3911 * internal-fn.c (fold_left_direct): Define.
3912 (expand_fold_left_optab_fn): Likewise.
3913 (direct_fold_left_optab_supported_p): Likewise.
3914 * fold-const-call.c (fold_const_fold_left): New function.
3915 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
3916 * tree-parloops.c (valid_reduction_p): New function.
3917 (gather_scalar_reductions): Use it.
3918 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
3919 (vect_finish_replace_stmt): Declare.
3920 * tree-vect-loop.c (fold_left_reduction_fn): New function.
3921 (needs_fold_left_reduction_p): New function, split out from...
3922 (vect_is_simple_reduction): ...here. Accept reductions that
3923 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
3924 (vect_force_simple_reduction): Also store the reduction type in
3925 the assignment's STMT_VINFO_REDUC_TYPE.
3926 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
3927 (merge_with_identity): New function.
3928 (vect_expand_fold_left): Likewise.
3929 (vectorize_fold_left_reduction): Likewise.
3930 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
3931 scalar phi in place for it. Check for target support and reject
3932 cases that would reassociate the operation. Defer the transform
3933 phase to vectorize_fold_left_reduction.
3934 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
3935 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
3936 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
3938 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3940 * tree-if-conv.c (predicate_mem_writes): Remove redundant
3941 call to ifc_temp_var.
3943 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3944 Alan Hayward <alan.hayward@arm.com>
3945 David Sherwood <david.sherwood@arm.com>
3947 * target.def (legitimize_address_displacement): Take the original
3948 offset as a poly_int.
3949 * targhooks.h (default_legitimize_address_displacement): Update
3951 * targhooks.c (default_legitimize_address_displacement): Likewise.
3952 * doc/tm.texi: Regenerate.
3953 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
3954 as an argument, moving assert of ad->disp == ad->disp_term to...
3955 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
3956 Try calling targetm.legitimize_address_displacement before expanding
3957 the address rather than afterwards, and adjust for the new interface.
3958 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
3959 Match the new hook interface. Handle SVE addresses.
3960 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
3963 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3965 * Makefile.in (OBJS): Add early-remat.o.
3966 * target.def (select_early_remat_modes): New hook.
3967 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
3968 * doc/tm.texi: Regenerate.
3969 * targhooks.h (default_select_early_remat_modes): Declare.
3970 * targhooks.c (default_select_early_remat_modes): New function.
3971 * timevar.def (TV_EARLY_REMAT): New timevar.
3972 * passes.def (pass_early_remat): New pass.
3973 * tree-pass.h (make_pass_early_remat): Declare.
3974 * early-remat.c: New file.
3975 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
3977 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
3979 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3980 Alan Hayward <alan.hayward@arm.com>
3981 David Sherwood <david.sherwood@arm.com>
3983 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
3984 vfm1 with a bound_epilog parameter.
3985 (vect_do_peeling): Update calls accordingly, and move the prologue
3986 call earlier in the function. Treat the base bound_epilog as 0 for
3987 fully-masked loops and retain vf - 1 for other loops. Add 1 to
3988 this base when peeling for gaps.
3989 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
3990 with fully-masked loops.
3991 (vect_estimate_min_profitable_iters): Handle the single peeled
3992 iteration in that case.
3994 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3995 Alan Hayward <alan.hayward@arm.com>
3996 David Sherwood <david.sherwood@arm.com>
3998 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
3999 single-element interleaving even if the size is not a power of 2.
4000 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
4001 accesses for single-element interleaving if the group size is
4004 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4005 Alan Hayward <alan.hayward@arm.com>
4006 David Sherwood <david.sherwood@arm.com>
4008 * doc/md.texi (fold_extract_last_@var{m}): Document.
4009 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
4010 * optabs.def (fold_extract_last_optab): New optab.
4011 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
4012 * internal-fn.c (fold_extract_direct): New macro.
4013 (expand_fold_extract_optab_fn): Likewise.
4014 (direct_fold_extract_optab_supported_p): Likewise.
4015 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
4016 * tree-vect-loop.c (vect_model_reduction_cost): Handle
4017 EXTRACT_LAST_REDUCTION.
4018 (get_initial_def_for_reduction): Do not create an initial vector
4019 for EXTRACT_LAST_REDUCTION reductions.
4020 (vectorizable_reduction): Leave the scalar phi in place for
4021 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
4022 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
4023 epilogue code for EXTRACT_LAST_REDUCTION and defer the
4024 transform phase to vectorizable_condition.
4025 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
4027 (vect_finish_stmt_generation): ...here.
4028 (vect_finish_replace_stmt): New function.
4029 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
4030 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
4032 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
4034 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4035 Alan Hayward <alan.hayward@arm.com>
4036 David Sherwood <david.sherwood@arm.com>
4038 * doc/md.texi (extract_last_@var{m}): Document.
4039 * optabs.def (extract_last_optab): New optab.
4040 * internal-fn.def (EXTRACT_LAST): New internal function.
4041 * internal-fn.c (cond_unary_direct): New macro.
4042 (expand_cond_unary_optab_fn): Likewise.
4043 (direct_cond_unary_optab_supported_p): Likewise.
4044 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
4045 loops using EXTRACT_LAST.
4046 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
4047 (extract_last_<mode>): ...this optab.
4048 (vec_extract<mode><Vel>): Update accordingly.
4050 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4051 Alan Hayward <alan.hayward@arm.com>
4052 David Sherwood <david.sherwood@arm.com>
4054 * target.def (empty_mask_is_expensive): New hook.
4055 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
4056 * doc/tm.texi: Regenerate.
4057 * targhooks.h (default_empty_mask_is_expensive): Declare.
4058 * targhooks.c (default_empty_mask_is_expensive): New function.
4059 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
4060 if the target says that empty masks are expensive.
4061 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
4063 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
4065 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4066 Alan Hayward <alan.hayward@arm.com>
4067 David Sherwood <david.sherwood@arm.com>
4069 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
4070 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
4071 (vect_use_loop_mask_for_alignment_p): New function.
4072 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
4073 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
4074 niters_skip argument. Make sure that the first niters_skip elements
4075 of the first iteration are inactive.
4076 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
4077 Update call to vect_set_loop_masks_directly.
4078 (get_misalign_in_elems): New function, split out from...
4079 (vect_gen_prolog_loop_niters): ...here.
4080 (vect_update_init_of_dr): Take a code argument that specifies whether
4081 the adjustment should be added or subtracted.
4082 (vect_update_init_of_drs): Likewise.
4083 (vect_prepare_for_masked_peels): New function.
4084 (vect_do_peeling): Skip prologue peeling if we're using a mask
4085 instead. Update call to vect_update_inits_of_drs.
4086 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4088 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
4089 alignment. Do not include the number of peeled iterations in
4090 the minimum threshold in that case.
4091 (vectorizable_induction): Adjust the start value down by
4092 LOOP_VINFO_MASK_SKIP_NITERS iterations.
4093 (vect_transform_loop): Call vect_prepare_for_masked_peels.
4094 Take the number of skipped iterations into account when calculating
4096 * tree-vect-stmts.c (vect_gen_while_not): New function.
4098 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4099 Alan Hayward <alan.hayward@arm.com>
4100 David Sherwood <david.sherwood@arm.com>
4102 * doc/sourcebuild.texi (vect_fully_masked): Document.
4103 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
4105 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
4107 (vect_analyze_loop_2): ...here. Don't check the vectorization
4108 factor against the number of loop iterations if the loop is
4111 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4112 Alan Hayward <alan.hayward@arm.com>
4113 David Sherwood <david.sherwood@arm.com>
4115 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
4116 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
4117 (dump_groups): Update accordingly.
4118 (iv_use::mem_type): New member variable.
4119 (address_p): New function.
4120 (record_use): Add a mem_type argument and initialize the new
4122 (record_group_use): Add a mem_type argument. Use address_p.
4123 Remove obsolete null checks of base_object. Update call to record_use.
4124 (find_interesting_uses_op): Update call to record_group_use.
4125 (find_interesting_uses_cond): Likewise.
4126 (find_interesting_uses_address): Likewise.
4127 (get_mem_type_for_internal_fn): New function.
4128 (find_address_like_use): Likewise.
4129 (find_interesting_uses_stmt): Try find_address_like_use before
4130 calling find_interesting_uses_op.
4131 (addr_offset_valid_p): Use the iv mem_type field as the type
4132 of the addressed memory.
4133 (add_autoinc_candidates): Likewise.
4134 (get_address_cost): Likewise.
4135 (split_small_address_groups_p): Use address_p.
4136 (split_address_groups): Likewise.
4137 (add_iv_candidate_for_use): Likewise.
4138 (autoinc_possible_for_pair): Likewise.
4139 (rewrite_groups): Likewise.
4140 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
4141 (determine_group_iv_cost): Update after split of USE_ADDRESS.
4142 (get_alias_ptr_type_for_ptr_address): New function.
4143 (rewrite_use_address): Rewrite address uses in calls that were
4144 identified by find_address_like_use.
4146 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4147 Alan Hayward <alan.hayward@arm.com>
4148 David Sherwood <david.sherwood@arm.com>
4150 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
4152 * gimple-expr.h (is_gimple_addressable: Likewise.
4153 * gimple-expr.c (is_gimple_address): Likewise.
4154 * internal-fn.c (expand_call_mem_ref): New function.
4155 (expand_mask_load_optab_fn): Use it.
4156 (expand_mask_store_optab_fn): Likewise.
4158 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4159 Alan Hayward <alan.hayward@arm.com>
4160 David Sherwood <david.sherwood@arm.com>
4162 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
4163 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
4164 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
4165 (cond_umax@var{mode}): Document.
4166 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
4167 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
4168 (cond_umin_optab, cond_umax_optab): New optabs.
4169 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
4170 (COND_IOR, COND_XOR): New internal functions.
4171 * internal-fn.h (get_conditional_internal_fn): Declare.
4172 * internal-fn.c (cond_binary_direct): New macro.
4173 (expand_cond_binary_optab_fn): Likewise.
4174 (direct_cond_binary_optab_supported_p): Likewise.
4175 (get_conditional_internal_fn): New function.
4176 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
4177 Cope with reduction statements that are vectorized as calls rather
4179 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
4180 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
4181 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
4182 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4183 (UNSPEC_COND_EOR): New unspecs.
4184 (optab): Add mappings for them.
4185 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
4186 (sve_int_op, sve_fp_op): New int attributes.
4188 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4189 Alan Hayward <alan.hayward@arm.com>
4190 David Sherwood <david.sherwood@arm.com>
4192 * optabs.def (while_ult_optab): New optab.
4193 * doc/md.texi (while_ult@var{m}@var{n}): Document.
4194 * internal-fn.def (WHILE_ULT): New internal function.
4195 * internal-fn.h (direct_internal_fn_supported_p): New override
4196 that takes two types as argument.
4197 * internal-fn.c (while_direct): New macro.
4198 (expand_while_optab_fn): New function.
4199 (convert_optab_supported_p): Likewise.
4200 (direct_while_optab_supported_p): New macro.
4201 * wide-int.h (wi::udiv_ceil): New function.
4202 * tree-vectorizer.h (rgroup_masks): New structure.
4203 (vec_loop_masks): New typedef.
4204 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
4206 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
4207 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
4208 (vect_max_vf): New function.
4209 (slpeel_make_loop_iterate_ntimes): Delete.
4210 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
4211 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
4212 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
4213 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
4214 internal-fn.h, stor-layout.h and optabs-query.h.
4215 (vect_set_loop_mask): New function.
4216 (add_preheader_seq): Likewise.
4217 (add_header_seq): Likewise.
4218 (interleave_supported_p): Likewise.
4219 (vect_maybe_permute_loop_masks): Likewise.
4220 (vect_set_loop_masks_directly): Likewise.
4221 (vect_set_loop_condition_masked): Likewise.
4222 (vect_set_loop_condition_unmasked): New function, split out from
4223 slpeel_make_loop_iterate_ntimes.
4224 (slpeel_make_loop_iterate_ntimes): Rename to..
4225 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
4226 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
4227 (vect_do_peeling): Update call accordingly.
4228 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
4230 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4231 mask_compare_type, can_fully_mask_p and fully_masked_p.
4232 (release_vec_loop_masks): New function.
4233 (_loop_vec_info): Use it to free the loop masks.
4234 (can_produce_all_loop_masks_p): New function.
4235 (vect_get_max_nscalars_per_iter): Likewise.
4236 (vect_verify_full_masking): Likewise.
4237 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
4238 retries, and free the mask rgroups before retrying. Check loop-wide
4239 reasons for disallowing fully-masked loops. Make the final decision
4240 about whether use a fully-masked loop or not.
4241 (vect_estimate_min_profitable_iters): Do not assume that peeling
4242 for the number of iterations will be needed for fully-masked loops.
4243 (vectorizable_reduction): Disable fully-masked loops.
4244 (vectorizable_live_operation): Likewise.
4245 (vect_halve_mask_nunits): New function.
4246 (vect_double_mask_nunits): Likewise.
4247 (vect_record_loop_mask): Likewise.
4248 (vect_get_loop_mask): Likewise.
4249 (vect_transform_loop): Handle the case in which the final loop
4250 iteration might handle a partial vector. Call vect_set_loop_condition
4251 instead of slpeel_make_loop_iterate_ntimes.
4252 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
4253 (check_load_store_masking): New function.
4254 (prepare_load_store_mask): Likewise.
4255 (vectorizable_store): Handle fully-masked loops.
4256 (vectorizable_load): Likewise.
4257 (supportable_widening_operation): Use vect_halve_mask_nunits for
4259 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
4260 (vect_gen_while): New function.
4261 * config/aarch64/aarch64.md (umax<mode>3): New expander.
4262 (aarch64_uqdec<mode>): New insn.
4264 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4265 Alan Hayward <alan.hayward@arm.com>
4266 David Sherwood <david.sherwood@arm.com>
4268 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
4269 (reduc_xor_scal_optab): New optabs.
4270 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
4271 (reduc_xor_scal_@var{m}): Document.
4272 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
4273 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
4275 * fold-const-call.c (fold_const_call): Handle them.
4276 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
4277 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
4278 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
4279 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
4280 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
4281 (UNSPEC_XORV): New unspecs.
4282 (optab): Add entries for them.
4283 (BITWISEV): New int iterator.
4284 (bit_reduc_op): New int attributes.
4286 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4287 Alan Hayward <alan.hayward@arm.com>
4288 David Sherwood <david.sherwood@arm.com>
4290 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
4291 * internal-fn.def (VEC_SHL_INSERT): New internal function.
4292 * optabs.def (vec_shl_insert_optab): New optab.
4293 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
4294 (duplicate_and_interleave): Likewise.
4295 * tree-vect-loop.c: Include internal-fn.h.
4296 (neutral_op_for_slp_reduction): New function, split out from
4297 get_initial_defs_for_reduction.
4298 (get_initial_def_for_reduction): Handle option 2 for variable-length
4299 vectors by loading the neutral value into a vector and then shifting
4300 the initial value into element 0.
4301 (get_initial_defs_for_reduction): Replace the code argument with
4302 the neutral value calculated by neutral_op_for_slp_reduction.
4303 Use gimple_build_vector for constant-length vectors.
4304 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
4305 but the first group_size elements have a neutral value.
4306 Use duplicate_and_interleave otherwise.
4307 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
4308 Update call to get_initial_defs_for_reduction. Handle SLP
4309 reductions for variable-length vectors by creating one vector
4310 result for each scalar result, with the elements associated
4311 with other scalar results stubbed out with the neutral value.
4312 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
4313 Require IFN_VEC_SHL_INSERT for double reductions on
4314 variable-length vectors, or SLP reductions that have
4315 a neutral value. Require can_duplicate_and_interleave_p
4316 support for variable-length unchained SLP reductions if there
4317 is no neutral value, such as for MIN/MAX reductions. Also require
4318 the number of vector elements to be a multiple of the number of
4319 SLP statements when doing variable-length unchained SLP reductions.
4320 Update call to vect_create_epilog_for_reduction.
4321 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
4322 and remove initial values.
4323 (duplicate_and_interleave): Make public.
4324 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
4325 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
4327 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4328 Alan Hayward <alan.hayward@arm.com>
4329 David Sherwood <david.sherwood@arm.com>
4331 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
4332 (can_duplicate_and_interleave_p): New function.
4333 (vect_get_and_check_slp_defs): Take the vector of statements
4334 rather than just the current one. Remove excess parentheses.
4335 Restriction rejectinon of vect_constant_def and vect_external_def
4336 for variable-length vectors to boolean types, or types for which
4337 can_duplicate_and_interleave_p is false.
4338 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
4339 (duplicate_and_interleave): New function.
4340 (vect_get_constant_vectors): Use gimple_build_vector for
4341 constant-length vectors and suitable variable-length constant
4342 vectors. Use duplicate_and_interleave for other variable-length
4343 vectors. Don't defer the update when inserting new statements.
4345 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4346 Alan Hayward <alan.hayward@arm.com>
4347 David Sherwood <david.sherwood@arm.com>
4349 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
4350 min_profitable_iters doesn't go negative.
4352 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4353 Alan Hayward <alan.hayward@arm.com>
4354 David Sherwood <david.sherwood@arm.com>
4356 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
4357 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
4358 * optabs.def (vec_mask_load_lanes_optab): New optab.
4359 (vec_mask_store_lanes_optab): Likewise.
4360 * internal-fn.def (MASK_LOAD_LANES): New internal function.
4361 (MASK_STORE_LANES): Likewise.
4362 * internal-fn.c (mask_load_lanes_direct): New macro.
4363 (mask_store_lanes_direct): Likewise.
4364 (expand_mask_load_optab_fn): Handle masked operations.
4365 (expand_mask_load_lanes_optab_fn): New macro.
4366 (expand_mask_store_optab_fn): Handle masked operations.
4367 (expand_mask_store_lanes_optab_fn): New macro.
4368 (direct_mask_load_lanes_optab_supported_p): Likewise.
4369 (direct_mask_store_lanes_optab_supported_p): Likewise.
4370 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
4372 (vect_load_lanes_supported): Likewise.
4373 * tree-vect-data-refs.c (strip_conversion): New function.
4374 (can_group_stmts_p): Likewise.
4375 (vect_analyze_data_ref_accesses): Use it instead of checking
4376 for a pair of assignments.
4377 (vect_store_lanes_supported): Take a masked_p parameter.
4378 (vect_load_lanes_supported): Likewise.
4379 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
4380 vect_store_lanes_supported and vect_load_lanes_supported.
4381 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4382 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
4383 parameter. Don't allow gaps for masked accesses.
4384 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
4385 and vect_load_lanes_supported.
4386 (get_load_store_type): Take a masked_p parameter and update
4387 call to get_group_load_store_type.
4388 (vectorizable_store): Update call to get_load_store_type.
4389 Handle IFN_MASK_STORE_LANES.
4390 (vectorizable_load): Update call to get_load_store_type.
4391 Handle IFN_MASK_LOAD_LANES.
4393 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4394 Alan Hayward <alan.hayward@arm.com>
4395 David Sherwood <david.sherwood@arm.com>
4397 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
4399 * config/aarch64/aarch64-protos.h
4400 (aarch64_sve_struct_memory_operand_p): Declare.
4401 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
4402 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
4403 (VPRED, vpred): Handle SVE structure modes.
4404 * config/aarch64/constraints.md (Utx): New constraint.
4405 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
4406 (aarch64_sve_struct_nonimmediate_operand): New predicates.
4407 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
4408 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
4409 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
4410 structure modes. Split into pieces after RA.
4411 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
4412 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
4414 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
4415 SVE structure modes.
4416 (aarch64_classify_address): Likewise.
4417 (sizetochar): Move earlier in file.
4418 (aarch64_print_operand): Handle SVE register lists.
4419 (aarch64_array_mode): New function.
4420 (aarch64_sve_struct_memory_operand_p): Likewise.
4421 (TARGET_ARRAY_MODE): Redefine.
4423 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4424 Alan Hayward <alan.hayward@arm.com>
4425 David Sherwood <david.sherwood@arm.com>
4427 * target.def (array_mode): New target hook.
4428 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
4429 * doc/tm.texi: Regenerate.
4430 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
4431 * hooks.c (hook_optmode_mode_uhwi_none): New function.
4432 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
4434 * stor-layout.c (mode_for_array): Likewise. Support polynomial
4437 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4438 Alan Hayward <alan.hayward@arm.com>
4439 David Sherwood <david.sherwood@arm.com>
4441 * fold-const.c (fold_binary_loc): Check the argument types
4442 rather than the result type when testing for a vector operation.
4444 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4446 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
4447 * doc/tm.texi: Regenerate.
4449 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4450 Alan Hayward <alan.hayward@arm.com>
4451 David Sherwood <david.sherwood@arm.com>
4453 * doc/invoke.texi (-msve-vector-bits=): Document new option.
4454 (sve): Document new AArch64 extension.
4455 * doc/md.texi (w): Extend the description of the AArch64
4456 constraint to include SVE vectors.
4457 (Upl, Upa): Document new AArch64 predicate constraints.
4458 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
4460 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
4461 (msve-vector-bits=): New option.
4462 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
4463 SVE when these are disabled.
4464 (sve): New extension.
4465 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
4466 modes. Adjust their number of units based on aarch64_sve_vg.
4467 (MAX_BITSIZE_MODE_ANY_MODE): Define.
4468 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
4469 aarch64_addr_query_type.
4470 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
4471 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
4472 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
4473 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
4474 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
4475 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
4476 (aarch64_simd_imm_zero_p): Delete.
4477 (aarch64_check_zero_based_sve_index_immediate): Declare.
4478 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4479 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4480 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4481 (aarch64_sve_float_mul_immediate_p): Likewise.
4482 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4484 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
4485 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
4486 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
4487 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
4488 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
4489 (aarch64_regmode_natural_size): Likewise.
4490 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
4491 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
4493 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
4494 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
4495 for VG and the SVE predicate registers.
4496 (V_ALIASES): Add a "z"-prefixed alias.
4497 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
4498 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
4499 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
4500 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
4501 (REG_CLASS_NAMES): Add entries for them.
4502 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
4503 and the predicate registers.
4504 (aarch64_sve_vg): Declare.
4505 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
4506 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
4507 (REGMODE_NATURAL_SIZE): Define.
4508 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
4510 * config/aarch64/aarch64.c: Include cfgrtl.h.
4511 (simd_immediate_info): Add a constructor for series vectors,
4512 and an associated step field.
4513 (aarch64_sve_vg): New variable.
4514 (aarch64_dbx_register_number): Handle VG and the predicate registers.
4515 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
4516 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
4517 (VEC_ANY_DATA, VEC_STRUCT): New constants.
4518 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
4519 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
4520 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
4521 (aarch64_get_mask_mode): New functions.
4522 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
4523 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
4524 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
4525 predicate modes and predicate registers. Explicitly restrict
4526 GPRs to modes of 16 bytes or smaller. Only allow FP registers
4527 to store a vector mode if it is recognized by
4528 aarch64_classify_vector_mode.
4529 (aarch64_regmode_natural_size): New function.
4530 (aarch64_hard_regno_caller_save_mode): Return the original mode
4532 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
4533 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
4534 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
4535 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
4537 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
4538 does not overlap dest if the function is frame-related. Handle
4540 (aarch64_split_add_offset): New function.
4541 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
4542 them aarch64_add_offset.
4543 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
4544 and update call to aarch64_sub_sp.
4545 (aarch64_add_cfa_expression): New function.
4546 (aarch64_expand_prologue): Pass extra temporary registers to the
4547 functions above. Handle the case in which we need to emit new
4548 DW_CFA_expressions for registers that were originally saved
4549 relative to the stack pointer, but now have to be expressed
4550 relative to the frame pointer.
4551 (aarch64_output_mi_thunk): Pass extra temporary registers to the
4553 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
4554 IP0 and IP1 values for SVE frames.
4555 (aarch64_expand_vec_series): New function.
4556 (aarch64_expand_sve_widened_duplicate): Likewise.
4557 (aarch64_expand_sve_const_vector): Likewise.
4558 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
4559 Handle SVE constants. Use emit_move_insn to move a force_const_mem
4560 into the register, rather than emitting a SET directly.
4561 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
4562 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
4563 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
4564 (offset_9bit_signed_scaled_p): New functions.
4565 (aarch64_replicate_bitmask_imm): New function.
4566 (aarch64_bitmask_imm): Use it.
4567 (aarch64_cannot_force_const_mem): Reject expressions involving
4568 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
4569 (aarch64_classify_index): Handle SVE indices, by requiring
4570 a plain register index with a scale that matches the element size.
4571 (aarch64_classify_address): Handle SVE addresses. Assert that
4572 the mode of the address is VOIDmode or an integer mode.
4573 Update call to aarch64_classify_symbol.
4574 (aarch64_classify_symbolic_expression): Update call to
4575 aarch64_classify_symbol.
4576 (aarch64_const_vec_all_in_range_p): New function.
4577 (aarch64_print_vector_float_operand): Likewise.
4578 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
4579 "vN" for FP registers with SVE modes. Handle (const ...) vectors
4580 and the FP immediates 1.0 and 0.5.
4581 (aarch64_print_address_internal): Handle SVE addresses.
4582 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
4583 (aarch64_regno_regclass): Handle predicate registers.
4584 (aarch64_secondary_reload): Handle big-endian reloads of SVE
4586 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
4587 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
4588 (aarch64_convert_sve_vector_bits): New function.
4589 (aarch64_override_options): Use it to handle -msve-vector-bits=.
4590 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4592 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
4593 Handle SVE vector and predicate modes. Accept VL-based constants
4594 that need only one temporary register, and VL offsets that require
4595 no temporary registers.
4596 (aarch64_conditional_register_usage): Mark the predicate registers
4597 as fixed if SVE isn't available.
4598 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
4599 Return true for SVE vector and predicate modes.
4600 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
4601 rather than an unsigned int. Handle SVE modes.
4602 (aarch64_preferred_simd_mode): Update call accordingly. Handle
4604 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
4606 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4607 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4608 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4609 (aarch64_sve_float_mul_immediate_p): New functions.
4610 (aarch64_sve_valid_immediate): New function.
4611 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
4612 Explicitly reject structure modes. Check for INDEX constants.
4613 Handle PTRUE and PFALSE constants.
4614 (aarch64_check_zero_based_sve_index_immediate): New function.
4615 (aarch64_simd_imm_zero_p): Delete.
4616 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
4617 vector modes. Accept constants in the range of CNT[BHWD].
4618 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
4619 ask for an Advanced SIMD mode.
4620 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
4621 (aarch64_simd_vector_alignment): Handle SVE predicates.
4622 (aarch64_vectorize_preferred_vector_alignment): New function.
4623 (aarch64_simd_vector_alignment_reachable): Use it instead of
4625 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
4626 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
4628 (MAX_VECT_LEN): Delete.
4629 (expand_vec_perm_d): Add a vec_flags field.
4630 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
4631 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
4632 (aarch64_evpc_ext): Don't apply a big-endian lane correction
4634 (aarch64_evpc_rev): Rename to...
4635 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
4636 (aarch64_evpc_rev_global): New function.
4637 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
4638 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
4640 (aarch64_evpc_sve_tbl): New function.
4641 (aarch64_expand_vec_perm_const_1): Update after rename of
4642 aarch64_evpc_rev. Handle SVE permutes too, trying
4643 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
4644 than aarch64_evpc_tbl.
4645 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
4646 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
4647 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
4648 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
4649 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
4650 (aarch64_expand_sve_vcond): New functions.
4651 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
4652 of aarch64_vector_mode_p.
4653 (aarch64_dwarf_poly_indeterminate_value): New function.
4654 (aarch64_compute_pressure_classes): Likewise.
4655 (aarch64_can_change_mode_class): Likewise.
4656 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
4657 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
4658 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
4659 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
4660 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
4661 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
4662 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
4663 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
4665 (Dn, Dl, Dr): Accept const as well as const_vector.
4666 (Dz): Likewise. Compare against CONST0_RTX.
4667 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
4668 of "vector" where appropriate.
4669 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
4670 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
4671 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
4672 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
4673 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
4674 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
4675 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
4676 (v_int_equiv): Extend to SVE modes.
4677 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
4679 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
4680 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
4681 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
4682 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
4683 (SVE_COND_FP_CMP): New int iterators.
4684 (perm_hilo): Handle the new unpack unspecs.
4685 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
4687 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
4688 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
4689 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
4690 (aarch64_equality_operator, aarch64_constant_vector_operand)
4691 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
4692 (aarch64_sve_nonimmediate_operand): Likewise.
4693 (aarch64_sve_general_operand): Likewise.
4694 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
4695 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
4696 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
4697 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
4698 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
4699 (aarch64_sve_float_arith_immediate): Likewise.
4700 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
4701 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
4702 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
4703 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
4704 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
4705 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
4706 (aarch64_sve_float_arith_operand): Likewise.
4707 (aarch64_sve_float_arith_with_sub_operand): Likewise.
4708 (aarch64_sve_float_mul_operand): Likewise.
4709 (aarch64_sve_vec_perm_operand): Likewise.
4710 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
4711 (aarch64_mov_operand): Accept const_poly_int and const_vector.
4712 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
4713 as well as const_vector.
4714 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
4715 in file. Use CONST0_RTX and CONSTM1_RTX.
4716 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
4717 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
4718 Use aarch64_simd_imm_zero.
4719 * config/aarch64/aarch64-sve.md: New file.
4720 * config/aarch64/aarch64.md: Include it.
4721 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
4722 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
4723 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
4724 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
4725 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
4726 (sve): New attribute.
4727 (enabled): Disable instructions with the sve attribute unless
4729 (movqi, movhi): Pass CONST_POLY_INT operaneds through
4730 aarch64_expand_mov_immediate.
4731 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
4732 CNT[BHSD] immediates.
4733 (movti): Split CONST_POLY_INT moves into two halves.
4734 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
4735 Split additions that need a temporary here if the destination
4736 is the stack pointer.
4737 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
4738 (*add<mode>3_poly_1): New instruction.
4739 (set_clobber_cc): New expander.
4741 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4743 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
4744 parameter and use it instead of GET_MODE_SIZE (innermode). Use
4745 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
4746 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
4747 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
4748 Change innermode from fixed_mode_size to machine_mode.
4749 (simplify_subreg): Update call accordingly. Handle a constant-sized
4750 subreg of a variable-length CONST_VECTOR.
4752 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4753 Alan Hayward <alan.hayward@arm.com>
4754 David Sherwood <david.sherwood@arm.com>
4756 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
4757 (add_offset_to_base): New function, split out from...
4758 (create_mem_ref): ...here. When handling a scale other than 1,
4759 check first whether the address is valid without the offset.
4760 Add it into the base if so, leaving the index and scale as-is.
4762 2018-01-12 Jakub Jelinek <jakub@redhat.com>
4765 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
4766 fold_for_warn before checking if arg2 is INTEGER_CST.
4768 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
4770 * config/rs6000/predicates.md (load_multiple_operation): Delete.
4771 (store_multiple_operation): Delete.
4772 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
4773 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
4774 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
4775 guarded by TARGET_STRING.
4776 (rs6000_output_load_multiple): Delete.
4777 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
4778 OPTION_MASK_STRING / TARGET_STRING handling.
4779 (print_operand) <'N', 'O'>: Add comment that these are unused now.
4780 (const rs6000_opt_masks) <"string">: Change mask to 0.
4781 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
4782 (MASK_STRING): Delete.
4783 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
4785 (load_multiple): Delete.
4792 (store_multiple): Delete.
4799 (movmemsi_8reg): Delete.
4800 (corresponding unnamed define_insn): Delete.
4801 (movmemsi_6reg): Delete.
4802 (corresponding unnamed define_insn): Delete.
4803 (movmemsi_4reg): Delete.
4804 (corresponding unnamed define_insn): Delete.
4805 (movmemsi_2reg): Delete.
4806 (corresponding unnamed define_insn): Delete.
4807 (movmemsi_1reg): Delete.
4808 (corresponding unnamed define_insn): Delete.
4809 * config/rs6000/rs6000.opt (mno-string): New.
4810 (mstring): Replace by deprecation warning stub.
4811 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
4813 2018-01-12 Jakub Jelinek <jakub@redhat.com>
4815 * regrename.c (regrename_do_replace): If replacing the same
4816 reg multiple times, try to reuse last created gen_raw_REG.
4819 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
4820 main to workaround a bug in GDB.
4822 2018-01-12 Tom de Vries <tom@codesourcery.com>
4825 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
4827 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
4829 PR rtl-optimization/80481
4830 * ira-color.c (get_cap_member): New function.
4831 (allocnos_conflict_by_live_ranges_p): Use it.
4832 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
4833 (setup_slot_coalesced_allocno_live_ranges): Ditto.
4835 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
4838 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
4839 (*saddl_se_1): Ditto.
4841 (*ssubl_se_1): Ditto.
4843 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
4845 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
4846 rather than wi::to_widest for DR_INITs.
4847 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
4848 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
4849 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
4851 (vect_analyze_group_access_1): Note that here.
4853 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
4855 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
4856 polynomial type sizes.
4858 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
4860 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
4861 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
4862 (gimple_add_tmp_var): Likewise.
4864 2018-01-12 Martin Liska <mliska@suse.cz>
4866 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
4867 (gimple_alloc_sizes): Likewise.
4868 (dump_gimple_statistics): Use PRIu64 in printf format.
4869 * gimple.h: Change uint64_t to int.
4871 2018-01-12 Martin Liska <mliska@suse.cz>
4873 * tree-core.h: Use uint64_t instead of int.
4874 * tree.c (tree_node_counts): Likewise.
4875 (tree_node_sizes): Likewise.
4876 (dump_tree_statistics): Use PRIu64 in printf format.
4878 2018-01-12 Martin Liska <mliska@suse.cz>
4880 * Makefile.in: As qsort_chk is implemented in vec.c, add
4881 vec.o to linkage of gencfn-macros.
4882 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
4883 passing the info to record_node_allocation_statistics.
4884 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
4886 * ggc-common.c (struct ggc_usage): Add operator== and use
4887 it in operator< and compare function.
4888 * mem-stats.h (struct mem_usage): Likewise.
4889 * vec.c (struct vec_usage): Remove operator< and compare
4890 function. Can be simply inherited.
4892 2018-01-12 Martin Jambor <mjambor@suse.cz>
4895 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
4896 * tree-ssa-math-opts.c: Include domwalk.h.
4897 (convert_mult_to_fma_1): New function.
4898 (fma_transformation_info): New type.
4899 (fma_deferring_state): Likewise.
4900 (cancel_fma_deferring): New function.
4901 (result_of_phi): Likewise.
4902 (last_fma_candidate_feeds_initial_phi): Likewise.
4903 (convert_mult_to_fma): Added deferring logic, split actual
4904 transformation to convert_mult_to_fma_1.
4905 (math_opts_dom_walker): New type.
4906 (math_opts_dom_walker::after_dom_children): New method, body moved
4907 here from pass_optimize_widening_mul::execute, added deferring logic
4909 (pass_optimize_widening_mul::execute): Moved most of code to
4910 math_opts_dom_walker::after_dom_children.
4911 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
4912 * config/i386/i386.c (ix86_option_override_internal): Added
4913 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
4915 2018-01-12 Richard Biener <rguenther@suse.de>
4918 * dwarf2out.c (gen_variable_die): Do not reset old_die for
4919 inline instance vars.
4921 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
4924 * config/rx/rx.c (rx_is_restricted_memory_address):
4927 2018-01-12 Richard Biener <rguenther@suse.de>
4929 PR tree-optimization/80846
4930 * target.def (split_reduction): New target hook.
4931 * targhooks.c (default_split_reduction): New function.
4932 * targhooks.h (default_split_reduction): Declare.
4933 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
4934 target requests first reduce vectors by combining low and high
4936 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
4937 (get_vectype_for_scalar_type_and_size): Export.
4938 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
4939 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
4940 * doc/tm.texi: Regenerate.
4941 * config/i386/i386.c (ix86_split_reduction): Implement
4942 TARGET_VECTORIZE_SPLIT_REDUCTION.
4944 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4947 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
4948 in PIC mode except for TARGET_VXWORKS_RTP.
4949 * config/sparc/sparc.c: Include cfgrtl.h.
4950 (TARGET_INIT_PIC_REG): Define.
4951 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
4952 (sparc_pic_register_p): New predicate.
4953 (sparc_legitimate_address_p): Use it.
4954 (sparc_legitimize_pic_address): Likewise.
4955 (sparc_delegitimize_address): Likewise.
4956 (sparc_mode_dependent_address_p): Likewise.
4957 (gen_load_pcrel_sym): Remove 4th parameter.
4958 (load_got_register): Adjust call to above. Remove obsolete stuff.
4959 (sparc_expand_prologue): Do not call load_got_register here.
4960 (sparc_flat_expand_prologue): Likewise.
4961 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
4962 (sparc_use_pseudo_pic_reg): New function.
4963 (sparc_init_pic_reg): Likewise.
4964 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
4965 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
4967 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
4969 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
4970 Add item for branch_cost.
4972 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4974 PR rtl-optimization/83565
4975 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
4976 not extend the result to a larger mode for rotate operations.
4977 (num_sign_bit_copies1): Likewise.
4979 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4982 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
4984 Use values-Xc.o for -pedantic.
4985 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
4987 2018-01-12 Martin Liska <mliska@suse.cz>
4990 * ipa-devirt.c (final_warning_record::grow_type_warnings):
4992 (possible_polymorphic_call_targets): Use it.
4993 (ipa_devirt): Likewise.
4995 2018-01-12 Martin Liska <mliska@suse.cz>
4997 * profile-count.h (enum profile_quality): Use 0 as invalid
4998 enum value of profile_quality.
5000 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
5002 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
5003 -mext-string options.
5005 2018-01-12 Richard Biener <rguenther@suse.de>
5007 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
5008 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
5009 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
5011 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
5013 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5015 * configure.ac (--with-long-double-format): Add support for the
5016 configuration option to change the default long double format on
5018 * config.gcc (powerpc*-linux*-*): Likewise.
5019 * configure: Regenerate.
5020 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
5021 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
5022 used without modification.
5024 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5026 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
5027 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
5028 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
5029 MISC_BUILTIN_SPEC_BARRIER.
5030 (rs6000_init_builtins): Likewise.
5031 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
5033 (speculation_barrier): New define_insn.
5034 * doc/extend.texi: Document __builtin_speculation_barrier.
5036 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5039 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
5040 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
5041 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
5043 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
5044 integral modes instead of "ss" and "sd".
5045 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
5046 vectors with 32-bit and 64-bit elements.
5047 (vecdupssescalarmodesuffix): New mode attribute.
5048 (vec_dup<mode>): Use it.
5050 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
5053 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
5054 frame if argument is passed on stack.
5056 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5059 * ree.c (combine_reaching_defs): Optimize also
5060 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
5061 reg2=any_extend(exp); reg1=reg2;, formatting fix.
5063 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5066 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
5068 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5071 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
5072 after they are computed.
5074 2018-01-11 Bin Cheng <bin.cheng@arm.com>
5076 PR tree-optimization/83695
5077 * gimple-loop-linterchange.cc
5078 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
5079 reset cached scev information after interchange.
5080 (pass_linterchange::execute): Remove call to scev_reset_htab.
5082 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5084 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
5085 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
5086 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
5087 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
5088 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
5089 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
5090 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
5091 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
5092 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
5093 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
5094 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
5095 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
5096 (V_lane_reg): Likewise.
5097 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
5099 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
5100 (vfmal_lane_low<mode>_intrinsic,
5101 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
5102 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
5103 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
5104 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
5105 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
5106 vfmsl_lane_high<mode>_intrinsic): New define_insns.
5108 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5110 * config/arm/arm-cpus.in (fp16fml): New feature.
5111 (ALL_SIMD): Add fp16fml.
5112 (armv8.2-a): Add fp16fml as an option.
5113 (armv8.3-a): Likewise.
5114 (armv8.4-a): Add fp16fml as part of fp16.
5115 * config/arm/arm.h (TARGET_FP16FML): Define.
5116 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
5118 * config/arm/arm-modes.def (V2HF): Define.
5119 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
5120 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
5121 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
5122 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
5123 vfmsl_low, vfmsl_high): New set of builtins.
5124 * config/arm/iterators.md (PLUSMINUS): New code iterator.
5125 (vfml_op): New code attribute.
5126 (VFMLHALVES): New int iterator.
5127 (VFML, VFMLSEL): New mode attributes.
5128 (V_reg): Define mapping for V2HF.
5129 (V_hi, V_lo): New mode attributes.
5130 (VF_constraint): Likewise.
5131 (vfml_half, vfml_half_selector): New int attributes.
5132 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
5134 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
5135 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
5137 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
5138 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
5139 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
5140 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
5142 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
5143 Document new effective target and option set.
5145 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5147 * config/arm/arm-cpus.in (armv8_4): New feature.
5148 (ARMv8_4a): New fgroup.
5149 (armv8.4-a): New arch.
5150 * config/arm/arm-tables.opt: Regenerate.
5151 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
5152 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
5153 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
5154 Add matching rules for -march=armv8.4-a and extensions.
5155 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
5157 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
5160 * config/rx/rx.md (BW): New mode attribute.
5161 (sync_lock_test_and_setsi): Add mode suffix to insn output.
5163 2018-01-11 Richard Biener <rguenther@suse.de>
5165 PR tree-optimization/83435
5166 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
5167 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
5168 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
5170 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5171 Alan Hayward <alan.hayward@arm.com>
5172 David Sherwood <david.sherwood@arm.com>
5174 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
5176 (aarch64_classify_address): Initialize it. Track polynomial offsets.
5177 (aarch64_print_address_internal): Use it to check for a zero offset.
5179 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5180 Alan Hayward <alan.hayward@arm.com>
5181 David Sherwood <david.sherwood@arm.com>
5183 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
5184 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
5185 Return a poly_int64 rather than a HOST_WIDE_INT.
5186 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
5187 rather than a HOST_WIDE_INT.
5188 * config/aarch64/aarch64.h (aarch64_frame): Protect with
5189 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
5190 hard_fp_offset, frame_size, initial_adjust, callee_offset and
5191 final_offset from HOST_WIDE_INT to poly_int64.
5192 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
5193 to_constant when getting the number of units in an Advanced SIMD
5195 (aarch64_builtin_vectorized_function): Check for a constant number
5197 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
5199 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
5200 attribute instead of GET_MODE_NUNITS.
5201 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
5202 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
5203 GET_MODE_SIZE for fixed-size registers.
5204 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
5205 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
5206 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
5207 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
5208 (aarch64_print_operand, aarch64_print_address_internal)
5209 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
5210 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
5211 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
5212 Handle polynomial GET_MODE_SIZE.
5213 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
5214 wider than SImode without modification.
5215 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
5216 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
5217 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
5218 passing and returning SVE modes.
5219 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
5220 rather than GEN_INT.
5221 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
5222 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
5223 (aarch64_allocate_and_probe_stack_space): Likewise.
5224 (aarch64_layout_frame): Cope with polynomial offsets.
5225 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
5226 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
5228 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
5229 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
5230 poly_int64 rather than a HOST_WIDE_INT.
5231 (aarch64_get_separate_components, aarch64_process_components)
5232 (aarch64_expand_prologue, aarch64_expand_epilogue)
5233 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
5234 (aarch64_anchor_offset): New function, split out from...
5235 (aarch64_legitimize_address): ...here.
5236 (aarch64_builtin_vectorization_cost): Handle polynomial
5237 TYPE_VECTOR_SUBPARTS.
5238 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
5240 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
5241 number of elements from the PARALLEL rather than the mode.
5242 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
5243 rather than GET_MODE_BITSIZE.
5244 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
5245 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
5246 (aarch64_expand_vec_perm_const_1): Handle polynomial
5247 d->perm.length () and d->perm elements.
5248 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
5249 Apply to_constant to d->perm elements.
5250 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
5251 polynomial CONST_VECTOR_NUNITS.
5252 (aarch64_move_pointer): Take amount as a poly_int64 rather
5254 (aarch64_progress_pointer): Avoid temporary variable.
5255 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
5256 the mode attribute instead of GET_MODE.
5258 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5259 Alan Hayward <alan.hayward@arm.com>
5260 David Sherwood <david.sherwood@arm.com>
5262 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
5263 x exists before using it.
5264 (aarch64_add_constant_internal): Rename to...
5265 (aarch64_add_offset_1): ...this. Replace regnum with separate
5266 src and dest rtxes. Handle the case in which they're different,
5267 including when the offset is zero. Replace scratchreg with an rtx.
5268 Use 2 additions if there is no spare register into which we can
5269 move a 16-bit constant.
5270 (aarch64_add_constant): Delete.
5271 (aarch64_add_offset): Replace reg with separate src and dest
5272 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
5273 Use aarch64_add_offset_1.
5274 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
5275 an rtx rather than an int. Take the delta as a poly_int64
5276 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
5277 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
5278 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
5279 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
5280 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
5282 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
5283 aarch64_add_constant.
5285 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5287 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
5288 Use scalar_float_mode.
5290 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5292 * config/aarch64/aarch64-simd.md
5293 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
5294 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
5295 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
5296 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
5297 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
5298 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
5299 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
5300 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
5301 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
5302 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
5304 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5307 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
5308 targ_options->x_arm_arch_string is non NULL.
5310 2018-01-11 Tamar Christina <tamar.christina@arm.com>
5312 * config/aarch64/aarch64.h
5313 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
5315 2018-01-11 Sudakshina Das <sudi.das@arm.com>
5318 * expmed.c (emit_store_flag_force): Swap if const op0
5319 and change VOIDmode to mode of op0.
5321 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5323 PR rtl-optimization/83761
5324 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
5325 than bytes to mode_for_size.
5327 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5330 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
5331 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
5334 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5337 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
5338 when in layout mode.
5339 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
5340 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
5343 2018-01-10 Michael Collison <michael.collison@arm.com>
5345 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
5346 * config/aarch64/aarch64-option-extension.def: Add
5347 AARCH64_OPT_EXTENSION of 'fp16fml'.
5348 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5349 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
5350 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
5351 * config/aarch64/constraints.md (Ui7): New constraint.
5352 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
5353 (VFMLA_SEL_W): Ditto.
5356 (VFMLA16_LOW): New int iterator.
5357 (VFMLA16_HIGH): Ditto.
5358 (UNSPEC_FMLAL): New unspec.
5359 (UNSPEC_FMLSL): Ditto.
5360 (UNSPEC_FMLAL2): Ditto.
5361 (UNSPEC_FMLSL2): Ditto.
5362 (f16mac): New code attribute.
5363 * config/aarch64/aarch64-simd-builtins.def
5364 (aarch64_fmlal_lowv2sf): Ditto.
5365 (aarch64_fmlsl_lowv2sf): Ditto.
5366 (aarch64_fmlalq_lowv4sf): Ditto.
5367 (aarch64_fmlslq_lowv4sf): Ditto.
5368 (aarch64_fmlal_highv2sf): Ditto.
5369 (aarch64_fmlsl_highv2sf): Ditto.
5370 (aarch64_fmlalq_highv4sf): Ditto.
5371 (aarch64_fmlslq_highv4sf): Ditto.
5372 (aarch64_fmlal_lane_lowv2sf): Ditto.
5373 (aarch64_fmlsl_lane_lowv2sf): Ditto.
5374 (aarch64_fmlal_laneq_lowv2sf): Ditto.
5375 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
5376 (aarch64_fmlalq_lane_lowv4sf): Ditto.
5377 (aarch64_fmlsl_lane_lowv4sf): Ditto.
5378 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
5379 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
5380 (aarch64_fmlal_lane_highv2sf): Ditto.
5381 (aarch64_fmlsl_lane_highv2sf): Ditto.
5382 (aarch64_fmlal_laneq_highv2sf): Ditto.
5383 (aarch64_fmlsl_laneq_highv2sf): Ditto.
5384 (aarch64_fmlalq_lane_highv4sf): Ditto.
5385 (aarch64_fmlsl_lane_highv4sf): Ditto.
5386 (aarch64_fmlalq_laneq_highv4sf): Ditto.
5387 (aarch64_fmlsl_laneq_highv4sf): Ditto.
5388 * config/aarch64/aarch64-simd.md:
5389 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
5390 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5391 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
5392 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5393 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
5394 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
5395 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
5396 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
5397 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
5398 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
5399 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
5400 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
5401 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
5402 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
5403 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
5404 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
5405 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
5406 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
5407 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
5408 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
5409 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
5410 (vfmlsl_low_u32): Ditto.
5411 (vfmlalq_low_u32): Ditto.
5412 (vfmlslq_low_u32): Ditto.
5413 (vfmlal_high_u32): Ditto.
5414 (vfmlsl_high_u32): Ditto.
5415 (vfmlalq_high_u32): Ditto.
5416 (vfmlslq_high_u32): Ditto.
5417 (vfmlal_lane_low_u32): Ditto.
5418 (vfmlsl_lane_low_u32): Ditto.
5419 (vfmlal_laneq_low_u32): Ditto.
5420 (vfmlsl_laneq_low_u32): Ditto.
5421 (vfmlalq_lane_low_u32): Ditto.
5422 (vfmlslq_lane_low_u32): Ditto.
5423 (vfmlalq_laneq_low_u32): Ditto.
5424 (vfmlslq_laneq_low_u32): Ditto.
5425 (vfmlal_lane_high_u32): Ditto.
5426 (vfmlsl_lane_high_u32): Ditto.
5427 (vfmlal_laneq_high_u32): Ditto.
5428 (vfmlsl_laneq_high_u32): Ditto.
5429 (vfmlalq_lane_high_u32): Ditto.
5430 (vfmlslq_lane_high_u32): Ditto.
5431 (vfmlalq_laneq_high_u32): Ditto.
5432 (vfmlslq_laneq_high_u32): Ditto.
5433 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
5434 (AARCH64_FL_FOR_ARCH8_4): New.
5435 (AARCH64_ISA_F16FML): New ISA flag.
5436 (TARGET_F16FML): New feature flag for fp16fml.
5437 (doc/invoke.texi): Document new fp16fml option.
5439 2018-01-10 Michael Collison <michael.collison@arm.com>
5441 * config/aarch64/aarch64-builtins.c:
5442 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
5443 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5444 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
5445 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
5446 (AARCH64_ISA_SHA3): New ISA flag.
5447 (TARGET_SHA3): New feature flag for sha3.
5448 * config/aarch64/iterators.md (sha512_op): New int attribute.
5449 (CRYPTO_SHA512): New int iterator.
5450 (UNSPEC_SHA512H): New unspec.
5451 (UNSPEC_SHA512H2): Ditto.
5452 (UNSPEC_SHA512SU0): Ditto.
5453 (UNSPEC_SHA512SU1): Ditto.
5454 * config/aarch64/aarch64-simd-builtins.def
5455 (aarch64_crypto_sha512hqv2di): New builtin.
5456 (aarch64_crypto_sha512h2qv2di): Ditto.
5457 (aarch64_crypto_sha512su0qv2di): Ditto.
5458 (aarch64_crypto_sha512su1qv2di): Ditto.
5459 (aarch64_eor3qv8hi): Ditto.
5460 (aarch64_rax1qv2di): Ditto.
5461 (aarch64_xarqv2di): Ditto.
5462 (aarch64_bcaxqv8hi): Ditto.
5463 * config/aarch64/aarch64-simd.md:
5464 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
5465 (aarch64_crypto_sha512su0qv2di): Ditto.
5466 (aarch64_crypto_sha512su1qv2di): Ditto.
5467 (aarch64_eor3qv8hi): Ditto.
5468 (aarch64_rax1qv2di): Ditto.
5469 (aarch64_xarqv2di): Ditto.
5470 (aarch64_bcaxqv8hi): Ditto.
5471 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
5472 (vsha512h2q_u64): Ditto.
5473 (vsha512su0q_u64): Ditto.
5474 (vsha512su1q_u64): Ditto.
5475 (veor3q_u16): Ditto.
5476 (vrax1q_u64): Ditto.
5478 (vbcaxq_u16): Ditto.
5479 * config/arm/types.md (crypto_sha512): New type attribute.
5480 (crypto_sha3): Ditto.
5481 (doc/invoke.texi): Document new sha3 option.
5483 2018-01-10 Michael Collison <michael.collison@arm.com>
5485 * config/aarch64/aarch64-builtins.c:
5486 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
5487 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5488 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
5489 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
5490 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
5491 (AARCH64_ISA_SM4): New ISA flag.
5492 (TARGET_SM4): New feature flag for sm4.
5493 * config/aarch64/aarch64-simd-builtins.def
5494 (aarch64_sm3ss1qv4si): Ditto.
5495 (aarch64_sm3tt1aq4si): Ditto.
5496 (aarch64_sm3tt1bq4si): Ditto.
5497 (aarch64_sm3tt2aq4si): Ditto.
5498 (aarch64_sm3tt2bq4si): Ditto.
5499 (aarch64_sm3partw1qv4si): Ditto.
5500 (aarch64_sm3partw2qv4si): Ditto.
5501 (aarch64_sm4eqv4si): Ditto.
5502 (aarch64_sm4ekeyqv4si): Ditto.
5503 * config/aarch64/aarch64-simd.md:
5504 (aarch64_sm3ss1qv4si): Ditto.
5505 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
5506 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
5507 (aarch64_sm4eqv4si): Ditto.
5508 (aarch64_sm4ekeyqv4si): Ditto.
5509 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
5510 (sm3part_op): Ditto.
5511 (CRYPTO_SM3TT): Ditto.
5512 (CRYPTO_SM3PART): Ditto.
5513 (UNSPEC_SM3SS1): New unspec.
5514 (UNSPEC_SM3TT1A): Ditto.
5515 (UNSPEC_SM3TT1B): Ditto.
5516 (UNSPEC_SM3TT2A): Ditto.
5517 (UNSPEC_SM3TT2B): Ditto.
5518 (UNSPEC_SM3PARTW1): Ditto.
5519 (UNSPEC_SM3PARTW2): Ditto.
5520 (UNSPEC_SM4E): Ditto.
5521 (UNSPEC_SM4EKEY): Ditto.
5522 * config/aarch64/constraints.md (Ui2): New constraint.
5523 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
5524 * config/arm/types.md (crypto_sm3): New type attribute.
5525 (crypto_sm4): Ditto.
5526 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
5527 (vsm3tt1aq_u32): Ditto.
5528 (vsm3tt1bq_u32): Ditto.
5529 (vsm3tt2aq_u32): Ditto.
5530 (vsm3tt2bq_u32): Ditto.
5531 (vsm3partw1q_u32): Ditto.
5532 (vsm3partw2q_u32): Ditto.
5533 (vsm4eq_u32): Ditto.
5534 (vsm4ekeyq_u32): Ditto.
5535 (doc/invoke.texi): Document new sm4 option.
5537 2018-01-10 Michael Collison <michael.collison@arm.com>
5539 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
5540 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
5541 (AARCH64_FL_FOR_ARCH8_4): New.
5542 (AARCH64_FL_V8_4): New flag.
5543 (doc/invoke.texi): Document new armv8.4-a option.
5545 2018-01-10 Michael Collison <michael.collison@arm.com>
5547 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5548 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
5549 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
5550 * config/aarch64/aarch64-option-extension.def: Add
5551 AARCH64_OPT_EXTENSION of 'sha2'.
5552 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
5553 (crypto): Disable sha2 and aes if crypto disabled.
5554 (crypto): Enable aes and sha2 if enabled.
5555 (simd): Disable sha2 and aes if simd disabled.
5556 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
5558 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
5559 (TARGET_SHA2): New feature flag for sha2.
5560 (TARGET_AES): New feature flag for aes.
5561 * config/aarch64/aarch64-simd.md:
5562 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
5563 conditional on TARGET_AES.
5564 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
5565 (aarch64_crypto_sha1hsi): Make pattern conditional
5567 (aarch64_crypto_sha1hv4si): Ditto.
5568 (aarch64_be_crypto_sha1hv4si): Ditto.
5569 (aarch64_crypto_sha1su1v4si): Ditto.
5570 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
5571 (aarch64_crypto_sha1su0v4si): Ditto.
5572 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
5573 (aarch64_crypto_sha256su0v4si): Ditto.
5574 (aarch64_crypto_sha256su1v4si): Ditto.
5575 (doc/invoke.texi): Document new aes and sha2 options.
5577 2018-01-10 Martin Sebor <msebor@redhat.com>
5579 PR tree-optimization/83781
5580 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
5583 2018-01-11 Martin Sebor <msebor@gmail.com>
5584 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5586 PR tree-optimization/83501
5587 PR tree-optimization/81703
5589 * tree-ssa-strlen.c (get_string_cst): Rename...
5590 (get_string_len): ...to this. Handle global constants.
5591 (handle_char_store): Adjust.
5593 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
5594 Jim Wilson <jimw@sifive.com>
5596 * config/riscv/riscv-protos.h (riscv_output_return): New.
5597 * config/riscv/riscv.c (struct machine_function): New naked_p field.
5598 (riscv_attribute_table, riscv_output_return),
5599 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
5600 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
5601 (riscv_compute_frame_info): Only compute frame->mask if not a naked
5603 (riscv_expand_prologue): Add early return for naked function.
5604 (riscv_expand_epilogue): Likewise.
5605 (riscv_function_ok_for_sibcall): Return false for naked function.
5606 (riscv_set_current_function): New.
5607 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
5608 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
5609 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
5610 * doc/extend.texi (RISC-V Function Attributes): New.
5612 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
5614 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
5615 check for 128-bit long double before checking TCmode.
5616 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
5617 128-bit long doubles before checking TFmode or TCmode.
5618 (FLOAT128_IBM_P): Likewise.
5620 2018-01-10 Martin Sebor <msebor@redhat.com>
5622 PR tree-optimization/83671
5623 * builtins.c (c_strlen): Unconditionally return zero for the empty
5625 Use -Warray-bounds for warnings.
5626 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
5627 for non-constant array indices with COMPONENT_REF, arrays of
5628 arrays, and pointers to arrays.
5629 (gimple_fold_builtin_strlen): Determine and set length range for
5630 non-constant character arrays.
5632 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
5635 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
5638 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
5640 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
5642 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5645 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
5646 VECTOR_MEM_ALTIVEC_OR_VSX_P.
5647 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
5648 indexed_or_indirect_operand predicate.
5649 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
5650 (*vsx_le_perm_load_v8hi): Likewise.
5651 (*vsx_le_perm_load_v16qi): Likewise.
5652 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
5653 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
5654 (*vsx_le_perm_store_v8hi): Likewise.
5655 (*vsx_le_perm_store_v16qi): Likewise.
5656 (eight unnamed splitters): Likewise.
5658 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5660 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
5661 * config/rs6000/emmintrin.h: Likewise.
5662 * config/rs6000/mmintrin.h: Likewise.
5663 * config/rs6000/xmmintrin.h: Likewise.
5665 2018-01-10 David Malcolm <dmalcolm@redhat.com>
5668 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
5670 * tree.c (tree_nop_conversion): Return true for location wrapper
5672 (maybe_wrap_with_location): New function.
5673 (selftest::check_strip_nops): New function.
5674 (selftest::test_location_wrappers): New function.
5675 (selftest::tree_c_tests): Call it.
5676 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
5677 (maybe_wrap_with_location): New decl.
5678 (EXPR_LOCATION_WRAPPER_P): New macro.
5679 (location_wrapper_p): New inline function.
5680 (tree_strip_any_location_wrapper): New inline function.
5682 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
5685 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
5686 stack_realign_offset for the largest alignment of stack slot
5688 (ix86_find_max_used_stack_alignment): New function.
5689 (ix86_finalize_stack_frame_flags): Use it. Set
5690 max_used_stack_alignment if we don't realign stack.
5691 * config/i386/i386.h (machine_function): Add
5692 max_used_stack_alignment.
5694 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
5696 * config/arm/arm.opt (-mbranch-cost): New option.
5697 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
5700 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
5703 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
5704 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
5706 2018-01-10 Richard Biener <rguenther@suse.de>
5709 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
5710 early out so it also covers the case where we have a non-NULL
5713 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
5715 PR tree-optimization/83753
5716 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
5717 for non-strided grouped accesses if the number of elements is 1.
5719 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5722 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
5723 * i386.h (TARGET_USE_GATHER): Define.
5724 * x86-tune.def (X86_TUNE_USE_GATHER): New.
5726 2018-01-10 Martin Liska <mliska@suse.cz>
5729 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
5730 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
5732 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
5733 CLEANUP_NO_PARTITIONING is not set.
5735 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
5737 * doc/rtl.texi: Remove documentation of (const ...) wrappers
5738 for vectors, as a partial revert of r254296.
5739 * rtl.h (const_vec_p): Delete.
5740 (const_vec_duplicate_p): Don't test for vector CONSTs.
5741 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
5742 * expmed.c (make_tree): Likewise.
5745 * common.md (E, F): Use CONSTANT_P instead of checking for
5747 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
5748 checking for CONST_VECTOR.
5750 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5753 * predict.c (force_edge_cold): Handle in more sane way edges
5756 2018-01-09 Carl Love <cel@us.ibm.com>
5758 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
5760 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
5761 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
5762 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
5763 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
5764 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
5765 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
5766 * config/rs6000/rs6000-protos.h: Add extern defition for
5767 rs6000_generate_float2_double_code.
5768 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
5770 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
5771 (float2_v2df): Add define_expand.
5773 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
5776 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
5777 op_mode in the force_to_mode call.
5779 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5781 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
5782 instead of checking each element individually.
5783 (aarch64_evpc_uzp): Likewise.
5784 (aarch64_evpc_zip): Likewise.
5785 (aarch64_evpc_ext): Likewise.
5786 (aarch64_evpc_rev): Likewise.
5787 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
5788 instead of checking each element individually. Return true without
5790 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
5791 whether all selected elements come from the same input, instead of
5792 checking each element individually. Remove calls to gen_rtx_REG,
5793 start_sequence and end_sequence and instead assert that no rtl is
5796 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5798 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
5799 order of HIGH and CONST checks.
5801 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5803 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
5804 if the destination isn't an SSA_NAME.
5806 2018-01-09 Richard Biener <rguenther@suse.de>
5808 PR tree-optimization/83668
5809 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
5811 (canonicalize_loop_form): ... here, renamed from ...
5812 (canonicalize_loop_closed_ssa_form): ... this and amended to
5813 swap successor edges for loop exit blocks to make us use
5814 the RPO order we need for initial schedule generation.
5816 2018-01-09 Joseph Myers <joseph@codesourcery.com>
5818 PR tree-optimization/64811
5819 * match.pd: When optimizing comparisons with Inf, avoid
5820 introducing or losing exceptions from comparisons with NaN.
5822 2018-01-09 Martin Liska <mliska@suse.cz>
5825 * asan.c (shadow_mem_size): Add gcc_assert.
5827 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
5829 Don't save registers in main().
5832 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
5833 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
5834 * config/avr/avr.c (avr_set_current_function): Don't error if
5835 naked, OS_task or OS_main are specified at the same time.
5836 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
5838 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
5840 * common/config/avr/avr-common.c (avr_option_optimization_table):
5841 Switch on -mmain-is-OS_task for optimizing compilations.
5843 2018-01-09 Richard Biener <rguenther@suse.de>
5845 PR tree-optimization/83572
5846 * graphite.c: Include cfganal.h.
5847 (graphite_transform_loops): Connect infinite loops to exit
5848 and remove fake edges at the end.
5850 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5852 * ipa-inline.c (edge_badness): Revert accidental checkin.
5854 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5857 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
5858 symbols; not inline clones.
5860 2018-01-09 Jakub Jelinek <jakub@redhat.com>
5863 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
5864 hard registers. Formatting fixes.
5866 PR preprocessor/83722
5867 * gcc.c (try_generate_repro): Pass
5868 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
5869 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
5872 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
5873 Kito Cheng <kito.cheng@gmail.com>
5875 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
5876 (riscv_leaf_function_p): Delete.
5877 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
5879 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5881 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
5883 (do_ifelse): New function.
5884 (do_isel): New function.
5885 (do_sub3): New function.
5886 (do_add3): New function.
5887 (do_load_mask_compare): New function.
5888 (do_overlap_load_compare): New function.
5889 (expand_compare_loop): New function.
5890 (expand_block_compare): Call expand_compare_loop() when appropriate.
5891 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
5893 (-mblock-compare-inline-loop-limit): New option.
5895 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5898 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
5899 Reverse order of second and third operands in first alternative.
5900 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
5901 of first and second elements in UNSPEC_VPERMR vector.
5902 (altivec_expand_vec_perm_le): Likewise.
5904 2018-01-08 Jeff Law <law@redhat.com>
5906 PR rtl-optimizatin/81308
5907 * tree-switch-conversion.c (cfg_altered): New file scoped static.
5908 (process_switch): If group_case_labels makes a change, then set
5910 (pass_convert_switch::execute): If a switch is converted, then
5911 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
5913 PR rtl-optimization/81308
5914 * recog.c (split_all_insns): Conditionally cleanup the CFG after
5917 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
5919 PR target/83663 - Revert r255946
5920 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
5921 generation for cases where splatting a value is not useful.
5922 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
5923 across a vec_duplicate and a paradoxical subreg forming a vector
5924 mode to a vec_concat.
5926 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5928 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
5929 -march=armv8.3-a variants.
5930 * config/arm/t-multilib: Likewise.
5931 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
5933 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5935 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
5937 (cceq_ior_compare_complement): Give it a name so I can use it, and
5938 change boolean_or_operator predicate to boolean_operator so it can
5939 be used to generate a crand.
5940 (eqne): New code iterator.
5941 (bd/bd_neg): New code_attrs.
5942 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
5943 a single define_insn.
5944 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
5945 decrement (bdnzt/bdnzf/bdzt/bdzf).
5946 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
5947 with the new names of the branch decrement patterns, and added the
5948 names of the branch decrement conditional patterns.
5950 2018-01-08 Richard Biener <rguenther@suse.de>
5952 PR tree-optimization/83563
5953 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
5956 2018-01-08 Richard Biener <rguenther@suse.de>
5959 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
5961 2018-01-08 Richard Biener <rguenther@suse.de>
5963 PR tree-optimization/83685
5964 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
5965 references to abnormals.
5967 2018-01-08 Richard Biener <rguenther@suse.de>
5970 * dwarf2out.c (output_indirect_strings): Handle empty
5971 skeleton_debug_str_hash.
5972 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
5974 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
5976 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
5977 (emit_store_direct): Likewise.
5978 (arc_trampoline_adjust_address): Likewise.
5979 (arc_asm_trampoline_template): New function.
5980 (arc_initialize_trampoline): Use asm_trampoline_template.
5981 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
5982 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
5983 * config/arc/arc.md (flush_icache): Delete pattern.
5985 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
5987 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
5988 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
5991 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
5994 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
5995 by not USED_FOR_TARGET.
5996 (make_pass_resolve_sw_modes): Likewise.
5998 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6000 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
6003 2018-01-08 Richard Biener <rguenther@suse.de>
6006 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
6008 2018-01-08 Richard Biener <rguenther@suse.de>
6011 * match.pd ((t * 2) / 2) -> t): Add missing :c.
6013 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
6016 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
6017 basic blocks with a small number of successors.
6018 (convert_control_dep_chain_into_preds): Improve handling of
6020 (dump_predicates): Split apart into...
6021 (dump_pred_chain): ...here...
6022 (dump_pred_info): ...and here.
6023 (can_one_predicate_be_invalidated_p): Add debugging printfs.
6024 (can_chain_union_be_invalidated_p): Improve check for invalidation
6026 (uninit_uses_cannot_happen): Avoid unnecessary if
6027 convert_control_dep_chain_into_preds yielded nothing.
6029 2018-01-06 Martin Sebor <msebor@redhat.com>
6031 PR tree-optimization/83640
6032 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
6033 subtracting negative offset from size.
6034 (builtin_access::overlap): Adjust offset bounds of the access to fall
6035 within the size of the object if possible.
6037 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
6039 PR rtl-optimization/83699
6040 * expmed.c (extract_bit_field_1): Restrict the vector usage of
6041 extract_bit_field_as_subreg to cases in which the extracted
6042 value is also a vector.
6044 * lra-constraints.c (process_alt_operands): Test for the equivalence
6045 substitutions when detecting a possible reload cycle.
6047 2018-01-06 Jakub Jelinek <jakub@redhat.com>
6050 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
6051 by default if flag_selective_schedling{,2}. Formatting fixes.
6053 PR rtl-optimization/83682
6054 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
6055 if it has non-VECTOR_MODE element mode.
6056 (vec_duplicate_p): Likewise.
6059 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
6060 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
6062 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6065 * config/i386/i386-builtin.def
6066 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
6067 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
6068 Require also OPTION_MASK_ISA_AVX512F in addition to
6069 OPTION_MASK_ISA_GFNI.
6070 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
6071 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
6072 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
6073 to OPTION_MASK_ISA_GFNI.
6074 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
6075 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
6076 OPTION_MASK_ISA_AVX512BW.
6077 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
6078 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
6079 addition to OPTION_MASK_ISA_GFNI.
6080 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
6081 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
6082 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
6083 to OPTION_MASK_ISA_GFNI.
6084 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
6085 a requirement for all ISAs rather than any of them with a few
6087 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
6089 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
6090 bitmasks to be enabled with 3 exceptions, instead of requiring any
6091 enabled ISA with lots of exceptions.
6092 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
6093 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
6094 Change avx512bw in isa attribute to avx512f.
6095 * config/i386/sgxintrin.h: Add license boilerplate.
6096 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
6097 to __AVX512F__ and __AVX512VL to __AVX512VL__.
6098 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
6099 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
6101 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
6102 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
6103 temporarily sse2 rather than sse if not enabled already.
6106 * config/i386/sse.md (VI248_VLBW): Rename to ...
6107 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
6108 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
6109 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
6110 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
6111 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
6112 mode iterator instead of VI248_VLBW.
6114 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
6116 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
6117 (record_modified): Skip clobbers; add debug output.
6118 (param_change_prob): Use sreal frequencies.
6120 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6122 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
6123 punt for user-aligned variables.
6125 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6127 * tree-chrec.c (chrec_contains_symbols): Return true for
6130 2018-01-05 Sudakshina Das <sudi.das@arm.com>
6133 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
6134 of (x|y) == x for BICS pattern.
6136 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6138 PR tree-optimization/83605
6139 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
6140 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
6143 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
6145 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
6146 * config/epiphany/rtems.h: New file.
6148 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6149 Uros Bizjak <ubizjak@gmail.com>
6152 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
6153 QIreg_operand instead of register_operand predicate.
6154 * config/i386/i386.c (ix86_rop_should_change_byte_p,
6155 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
6156 comments instead of -fmitigate[-_]rop.
6158 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6161 * cgraphunit.c (symbol_table::compile): Switch to text_section
6162 before calling assembly_start debug hook.
6163 * run-rtl-passes.c (run_rtl_passes): Likewise.
6166 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6168 * tree-vrp.c (extract_range_from_binary_expr_1): Check
6169 range_int_cst_p rather than !symbolic_range_p before calling
6170 extract_range_from_multiplicative_op_1.
6172 2018-01-04 Jeff Law <law@redhat.com>
6174 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
6175 redundant test in assertion.
6177 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6179 * doc/rtl.texi: Document machine_mode wrapper classes.
6181 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6183 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
6186 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6188 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
6189 the VEC_PERM_EXPR fold to fail.
6191 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6194 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
6195 to switched_sections.
6197 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6200 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
6203 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
6206 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
6207 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
6209 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6212 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
6213 is BLKmode and bitpos not zero or mode change is needed.
6215 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6218 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
6221 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
6224 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
6225 instead of MULT rtx. Update all corresponding splitters.
6227 (*ssub<modesuffix>): Ditto.
6229 (*cmp_sadd_di): Update split patterns.
6230 (*cmp_sadd_si): Ditto.
6231 (*cmp_sadd_sidi): Ditto.
6232 (*cmp_ssub_di): Ditto.
6233 (*cmp_ssub_si): Ditto.
6234 (*cmp_ssub_sidi): Ditto.
6235 * config/alpha/predicates.md (const23_operand): New predicate.
6236 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
6237 Look for ASHIFT, not MULT inner operand.
6238 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
6240 2018-01-04 Martin Liska <mliska@suse.cz>
6242 PR gcov-profile/83669
6243 * gcov.c (output_intermediate_file): Add version to intermediate
6245 * doc/gcov.texi: Document new field 'version' in intermediate
6246 file format. Fix location of '-k' option of gcov command.
6248 2018-01-04 Martin Liska <mliska@suse.cz>
6251 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
6253 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6255 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
6257 2018-01-03 Martin Sebor <msebor@redhat.com>
6259 PR tree-optimization/83655
6260 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
6261 checking calls with invalid arguments.
6263 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6265 * tree-vect-stmts.c (vect_get_store_rhs): New function.
6266 (vectorizable_mask_load_store): Delete.
6267 (vectorizable_call): Return false for masked loads and stores.
6268 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
6269 instead of gimple_assign_rhs1.
6270 (vectorizable_load): Handle IFN_MASK_LOAD.
6271 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
6273 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6275 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
6277 (vectorizable_mask_load_store): ...here.
6278 (vectorizable_load): ...and here.
6280 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6282 * tree-vect-stmts.c (vect_build_all_ones_mask)
6283 (vect_build_zero_merge_argument): New functions, split out from...
6284 (vectorizable_load): ...here.
6286 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6288 * tree-vect-stmts.c (vect_check_store_rhs): New function,
6290 (vectorizable_mask_load_store): ...here.
6291 (vectorizable_store): ...and here.
6293 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6295 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
6297 (vectorizable_mask_load_store): ...here.
6299 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6301 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
6302 (vect_model_store_cost): Take a vec_load_store_type instead of a
6304 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
6305 (vect_model_store_cost): Take a vec_load_store_type instead of a
6307 (vectorizable_mask_load_store): Update accordingly.
6308 (vectorizable_store): Likewise.
6309 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
6311 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6313 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
6314 IFN_MASK_LOAD calls here rather than...
6315 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
6317 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6318 Alan Hayward <alan.hayward@arm.com>
6319 David Sherwood <david.sherwood@arm.com>
6321 * expmed.c (extract_bit_field_1): For vector extracts,
6322 fall back to extract_bit_field_as_subreg if vec_extract
6325 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6326 Alan Hayward <alan.hayward@arm.com>
6327 David Sherwood <david.sherwood@arm.com>
6329 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
6330 they are variable or constant sized.
6331 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
6332 slots for constant-sized data.
6334 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6335 Alan Hayward <alan.hayward@arm.com>
6336 David Sherwood <david.sherwood@arm.com>
6338 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
6339 handling COND_EXPRs with boolean comparisons, try to find a better
6340 basis for the mask type than the boolean itself.
6342 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6344 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
6345 is calculated and how it can be overridden.
6346 * genmodes.c (max_bitsize_mode_any_mode): New variable.
6347 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
6349 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
6352 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6353 Alan Hayward <alan.hayward@arm.com>
6354 David Sherwood <david.sherwood@arm.com>
6356 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
6357 Remove the mode argument.
6358 (aarch64_simd_valid_immediate): Remove the mode and inverse
6360 * config/aarch64/iterators.md (bitsize): New iterator.
6361 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
6362 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
6363 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
6364 aarch64_simd_valid_immediate.
6365 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
6366 (aarch64_reg_or_bic_imm): Likewise.
6367 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
6368 with an insn_type enum and msl with a modifier_type enum.
6369 Replace element_width with a scalar_mode. Change the shift
6370 to unsigned int. Add constructors for scalar_float_mode and
6371 scalar_int_mode elements.
6372 (aarch64_vect_float_const_representable_p): Delete.
6373 (aarch64_can_const_movi_rtx_p)
6374 (aarch64_simd_scalar_immediate_valid_for_move)
6375 (aarch64_simd_make_constant): Update call to
6376 aarch64_simd_valid_immediate.
6377 (aarch64_advsimd_valid_immediate_hs): New function.
6378 (aarch64_advsimd_valid_immediate): Likewise.
6379 (aarch64_simd_valid_immediate): Remove mode and inverse
6380 arguments. Rewrite to use the above. Use const_vec_duplicate_p
6381 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
6382 and aarch64_float_const_representable_p on the result.
6383 (aarch64_output_simd_mov_immediate): Remove mode argument.
6384 Update call to aarch64_simd_valid_immediate and use of
6385 simd_immediate_info.
6386 (aarch64_output_scalar_simd_mov_immediate): Update call
6389 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6390 Alan Hayward <alan.hayward@arm.com>
6391 David Sherwood <david.sherwood@arm.com>
6393 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
6394 (mode_nunits): Likewise CONST_MODE_NUNITS.
6395 * machmode.def (ADJUST_NUNITS): Document.
6396 * genmodes.c (mode_data::need_nunits_adj): New field.
6397 (blank_mode): Update accordingly.
6398 (adj_nunits): New variable.
6399 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
6401 (emit_mode_size_inline): Set need_bytesize_adj for all modes
6402 listed in adj_nunits.
6403 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
6404 listed in adj_nunits. Don't emit case statements for such modes.
6405 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
6406 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
6407 nothing if adj_nunits is nonnull.
6408 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
6409 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
6410 (emit_mode_fbit): Update use of print_maybe_const_decl.
6411 (emit_move_size): Likewise. Treat the array as non-const
6413 (emit_mode_adjustments): Handle adj_nunits.
6415 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6417 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
6418 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
6419 (VECTOR_MODES): Use it.
6420 (make_vector_modes): Take the prefix as an argument.
6422 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6423 Alan Hayward <alan.hayward@arm.com>
6424 David Sherwood <david.sherwood@arm.com>
6426 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
6427 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
6428 for MODE_VECTOR_BOOL.
6429 * machmode.def (VECTOR_BOOL_MODE): Document.
6430 * genmodes.c (VECTOR_BOOL_MODE): New macro.
6431 (make_vector_bool_mode): New function.
6432 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
6434 * lto-streamer-in.c (lto_input_mode_table): Likewise.
6435 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
6437 * stor-layout.c (int_mode_for_mode): Likewise.
6438 * tree.c (build_vector_type_for_mode): Likewise.
6439 * varasm.c (output_constant_pool_2): Likewise.
6440 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
6441 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
6442 for MODE_VECTOR_BOOL.
6443 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
6444 of mode class checks.
6445 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
6446 instead of a list of mode class checks.
6447 (expand_vector_scalar_condition): Likewise.
6448 (type_for_widest_vector_mode): Handle BImode as an inner mode.
6450 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6451 Alan Hayward <alan.hayward@arm.com>
6452 David Sherwood <david.sherwood@arm.com>
6454 * machmode.h (mode_size): Change from unsigned short to
6456 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
6457 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6458 or if measurement_type is not polynomial.
6459 (fixed_size_mode::includes_p): Check for constant-sized modes.
6460 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
6461 return a poly_uint16 rather than an unsigned short.
6462 (emit_mode_size): Change the type of mode_size from unsigned short
6463 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
6464 (emit_mode_adjustments): Cope with polynomial vector sizes.
6465 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6467 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6469 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
6470 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
6471 * caller-save.c (setup_save_areas): Likewise.
6472 (replace_reg_with_saved_mem): Likewise.
6473 * calls.c (emit_library_call_value_1): Likewise.
6474 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
6475 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
6476 (gen_lowpart_for_combine): Likewise.
6477 * convert.c (convert_to_integer_1): Likewise.
6478 * cse.c (equiv_constant, cse_insn): Likewise.
6479 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
6480 (cselib_subst_to_values): Likewise.
6481 * dce.c (word_dce_process_block): Likewise.
6482 * df-problems.c (df_word_lr_mark_ref): Likewise.
6483 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
6484 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
6485 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
6486 (rtl_for_decl_location): Likewise.
6487 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
6488 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
6489 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
6490 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
6491 (expand_expr_real_1): Likewise.
6492 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
6493 (pad_below): Likewise.
6494 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6495 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
6496 * ira.c (get_subreg_tracking_sizes): Likewise.
6497 * ira-build.c (ira_create_allocno_objects): Likewise.
6498 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
6499 (ira_sort_regnos_for_alter_reg): Likewise.
6500 * ira-costs.c (record_operand_costs): Likewise.
6501 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
6502 (resolve_simple_move): Likewise.
6503 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
6504 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
6505 (lra_constraints): Likewise.
6506 (CONST_POOL_OK_P): Reject variable-sized modes.
6507 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
6508 (add_pseudo_to_slot, lra_spill): Likewise.
6509 * omp-low.c (omp_clause_aligned_alignment): Likewise.
6510 * optabs-query.c (get_best_extraction_insn): Likewise.
6511 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6512 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
6513 (expand_mult_highpart, valid_multiword_target_p): Likewise.
6514 * recog.c (offsettable_address_addr_space_p): Likewise.
6515 * regcprop.c (maybe_mode_change): Likewise.
6516 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
6517 * regrename.c (build_def_use): Likewise.
6518 * regstat.c (dump_reg_info): Likewise.
6519 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
6520 (find_reloads, find_reloads_subreg_address): Likewise.
6521 * reload1.c (eliminate_regs_1): Likewise.
6522 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
6523 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
6524 (simplify_binary_operation_1, simplify_subreg): Likewise.
6525 * targhooks.c (default_function_arg_padding): Likewise.
6526 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
6527 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
6528 (verify_gimple_assign_ternary): Likewise.
6529 * tree-inline.c (estimate_move_cost): Likewise.
6530 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6531 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
6532 (get_address_cost_ainc): Likewise.
6533 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
6534 (vect_supportable_dr_alignment): Likewise.
6535 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6536 (vectorizable_reduction): Likewise.
6537 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
6538 (vectorizable_operation, vectorizable_load): Likewise.
6539 * tree.c (build_same_sized_truth_vector_type): Likewise.
6540 * valtrack.c (cleanup_auto_inc_dec): Likewise.
6541 * var-tracking.c (emit_note_insn_var_location): Likewise.
6542 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
6543 (ADDR_VEC_ALIGN): Likewise.
6545 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6546 Alan Hayward <alan.hayward@arm.com>
6547 David Sherwood <david.sherwood@arm.com>
6549 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
6551 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6552 or if measurement_type is polynomial.
6553 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
6554 * combine.c (make_extraction): Likewise.
6555 * dse.c (find_shift_sequence): Likewise.
6556 * dwarf2out.c (mem_loc_descriptor): Likewise.
6557 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
6558 (extract_bit_field, extract_low_bits): Likewise.
6559 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
6560 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
6561 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
6562 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
6563 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6564 * reload.c (find_reloads): Likewise.
6565 * reload1.c (alter_reg): Likewise.
6566 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
6567 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
6568 * tree-if-conv.c (predicate_mem_writes): Likewise.
6569 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
6570 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
6571 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
6572 * valtrack.c (dead_debug_insert_temp): Likewise.
6573 * varasm.c (mergeable_constant_section): Likewise.
6574 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
6576 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6577 Alan Hayward <alan.hayward@arm.com>
6578 David Sherwood <david.sherwood@arm.com>
6580 * expr.c (expand_assignment): Cope with polynomial mode sizes
6581 when assigning to a CONCAT.
6583 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6584 Alan Hayward <alan.hayward@arm.com>
6585 David Sherwood <david.sherwood@arm.com>
6587 * machmode.h (mode_precision): Change from unsigned short to
6589 (mode_to_precision): Return a poly_uint16 rather than an unsigned
6591 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
6592 or if measurement_type is not polynomial.
6593 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
6594 in which the mode is already known to be a scalar_int_mode.
6595 * genmodes.c (emit_mode_precision): Change the type of mode_precision
6596 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
6598 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6599 for GET_MODE_PRECISION.
6600 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6601 for GET_MODE_PRECISION.
6602 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
6604 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
6605 (expand_field_assignment, make_extraction): Likewise.
6606 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
6607 (get_last_value): Likewise.
6608 * convert.c (convert_to_integer_1): Likewise.
6609 * cse.c (cse_insn): Likewise.
6610 * expr.c (expand_expr_real_1): Likewise.
6611 * lra-constraints.c (simplify_operand_subreg): Likewise.
6612 * optabs-query.c (can_atomic_load_p): Likewise.
6613 * optabs.c (expand_atomic_load): Likewise.
6614 (expand_atomic_store): Likewise.
6615 * ree.c (combine_reaching_defs): Likewise.
6616 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
6617 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
6618 * tree.h (type_has_mode_precision_p): Likewise.
6619 * ubsan.c (instrument_si_overflow): Likewise.
6621 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6622 Alan Hayward <alan.hayward@arm.com>
6623 David Sherwood <david.sherwood@arm.com>
6625 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
6626 polynomial numbers of units.
6627 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
6628 (valid_vector_subparts_p): New function.
6629 (build_vector_type): Remove temporary shim and take the number
6630 of units as a poly_uint64 rather than an int.
6631 (build_opaque_vector_type): Take the number of units as a
6632 poly_uint64 rather than an int.
6633 * tree.c (build_vector_from_ctor): Handle polynomial
6634 TYPE_VECTOR_SUBPARTS.
6635 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
6636 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
6637 (build_vector_from_val): If the number of units is variable,
6638 use build_vec_duplicate_cst for constant operands and
6639 VEC_DUPLICATE_EXPR otherwise.
6640 (make_vector_type): Remove temporary is_constant ().
6641 (build_vector_type, build_opaque_vector_type): Take the number of
6642 units as a poly_uint64 rather than an int.
6643 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
6645 * cfgexpand.c (expand_debug_expr): Likewise.
6646 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
6647 (store_constructor, expand_expr_real_1): Likewise.
6648 (const_scalar_mask_from_tree): Likewise.
6649 * fold-const-call.c (fold_const_reduction): Likewise.
6650 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
6651 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
6652 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
6653 (fold_relational_const): Likewise.
6654 (native_interpret_vector): Likewise. Change the size from an
6655 int to an unsigned int.
6656 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
6657 TYPE_VECTOR_SUBPARTS.
6658 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
6659 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
6660 duplicating a non-constant operand into a variable-length vector.
6661 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
6662 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
6663 * ipa-icf.c (sem_variable::equals): Likewise.
6664 * match.pd: Likewise.
6665 * omp-simd-clone.c (simd_clone_subparts): Likewise.
6666 * print-tree.c (print_node): Likewise.
6667 * stor-layout.c (layout_type): Likewise.
6668 * targhooks.c (default_builtin_vectorization_cost): Likewise.
6669 * tree-cfg.c (verify_gimple_comparison): Likewise.
6670 (verify_gimple_assign_binary): Likewise.
6671 (verify_gimple_assign_ternary): Likewise.
6672 (verify_gimple_assign_single): Likewise.
6673 * tree-pretty-print.c (dump_generic_node): Likewise.
6674 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6675 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
6676 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
6677 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
6678 (vect_shift_permute_load_chain): Likewise.
6679 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
6680 (expand_vector_condition, optimize_vector_constructor): Likewise.
6681 (lower_vec_perm, get_compute_type): Likewise.
6682 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6683 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
6684 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
6685 (vect_recog_mask_conversion_pattern): Likewise.
6686 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
6687 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
6688 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
6689 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
6690 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
6691 (vectorizable_shift, vectorizable_operation, vectorizable_store)
6692 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
6693 (supportable_widening_operation): Likewise.
6694 (supportable_narrowing_operation): Likewise.
6695 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
6697 * varasm.c (output_constant): Likewise.
6699 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6700 Alan Hayward <alan.hayward@arm.com>
6701 David Sherwood <david.sherwood@arm.com>
6703 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
6704 so that both the length == 3 and length != 3 cases set up their
6705 own permute vectors. Add comments explaining why we know the
6706 number of elements is constant.
6707 (vect_permute_load_chain): Likewise.
6709 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6710 Alan Hayward <alan.hayward@arm.com>
6711 David Sherwood <david.sherwood@arm.com>
6713 * machmode.h (mode_nunits): Change from unsigned char to
6715 (ONLY_FIXED_SIZE_MODES): New macro.
6716 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
6717 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
6718 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
6720 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
6721 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
6722 or if measurement_type is not polynomial.
6723 * genmodes.c (ZERO_COEFFS): New macro.
6724 (emit_mode_nunits_inline): Make mode_nunits_inline return a
6726 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
6727 Use ZERO_COEFFS when emitting initializers.
6728 * data-streamer.h (bp_pack_poly_value): New function.
6729 (bp_unpack_poly_value): Likewise.
6730 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6731 for GET_MODE_NUNITS.
6732 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6733 for GET_MODE_NUNITS.
6734 * tree.c (make_vector_type): Remove temporary shim and make
6735 the real function take the number of units as a poly_uint64
6737 (build_vector_type_for_mode): Handle polynomial nunits.
6738 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
6739 * emit-rtl.c (const_vec_series_p_1): Likewise.
6740 (gen_rtx_CONST_VECTOR): Likewise.
6741 * fold-const.c (test_vec_duplicate_folding): Likewise.
6742 * genrecog.c (validate_pattern): Likewise.
6743 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
6744 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6745 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
6746 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
6747 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
6748 * rtlanal.c (subreg_get_info): Likewise.
6749 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
6750 (vect_grouped_load_supported): Likewise.
6751 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
6752 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
6753 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
6754 (simplify_const_unary_operation, simplify_binary_operation_1)
6755 (simplify_const_binary_operation, simplify_ternary_operation)
6756 (test_vector_ops_duplicate, test_vector_ops): Likewise.
6757 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
6758 instead of CONST_VECTOR_NUNITS.
6759 * varasm.c (output_constant_pool_2): Likewise.
6760 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
6761 explicit-encoded elements in the XVEC for variable-length vectors.
6763 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6765 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
6767 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6768 Alan Hayward <alan.hayward@arm.com>
6769 David Sherwood <david.sherwood@arm.com>
6771 * coretypes.h (fixed_size_mode): Declare.
6772 (fixed_size_mode_pod): New typedef.
6773 * builtins.h (target_builtins::x_apply_args_mode)
6774 (target_builtins::x_apply_result_mode): Change type to
6775 fixed_size_mode_pod.
6776 * builtins.c (apply_args_size, apply_result_size, result_vector)
6777 (expand_builtin_apply_args_1, expand_builtin_apply)
6778 (expand_builtin_return): Update accordingly.
6780 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6782 * cse.c (hash_rtx_cb): Hash only the encoded elements.
6783 * cselib.c (cselib_hash_rtx): Likewise.
6784 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
6785 CONST_VECTOR encoding.
6787 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6788 Jeff Law <law@redhat.com>
6791 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
6792 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
6793 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
6794 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
6797 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
6798 explicitly probe *sp in a noreturn function if there were any callee
6799 register saves or frame pointer is needed.
6801 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6804 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
6805 BLKmode for ternary, binary or unary expressions.
6808 * var-tracking.c (delete_vta_debug_insn): New inline function.
6809 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
6810 insns from get_insns () to NULL instead of each bb separately.
6811 Use delete_vta_debug_insn. No longer static.
6812 (vt_debug_insns_local, variable_tracking_main_1): Adjust
6813 delete_vta_debug_insns callers.
6814 * rtl.h (delete_vta_debug_insns): Declare.
6815 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
6816 instead of variable_tracking_main.
6818 2018-01-03 Martin Sebor <msebor@redhat.com>
6820 PR tree-optimization/83603
6821 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
6822 arguments past the endof the argument list in functions declared
6823 without a prototype.
6824 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
6825 Avoid checking when arguments are null.
6827 2018-01-03 Martin Sebor <msebor@redhat.com>
6830 * doc/extend.texi (attribute const): Fix a typo.
6831 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
6832 issuing -Wsuggest-attribute for void functions.
6834 2018-01-03 Martin Sebor <msebor@redhat.com>
6836 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
6837 offset_int::from instead of wide_int::to_shwi.
6838 (maybe_diag_overlap): Remove assertion.
6839 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
6840 * gimple-ssa-sprintf.c (format_directive): Same.
6841 (parse_directive): Same.
6842 (sprintf_dom_walker::compute_format_length): Same.
6843 (try_substitute_return_value): Same.
6845 2018-01-03 Jeff Law <law@redhat.com>
6848 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
6849 non-constant residual for zero at runtime and avoid probing in
6850 that case. Reorganize code for trailing problem to mirror handling
6853 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6855 PR tree-optimization/83501
6856 * tree-ssa-strlen.c (get_string_cst): New.
6857 (handle_char_store): Call get_string_cst.
6859 2018-01-03 Martin Liska <mliska@suse.cz>
6861 PR tree-optimization/83593
6862 * tree-ssa-strlen.c: Include tree-cfg.h.
6863 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
6864 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
6865 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
6867 (strlen_dom_walker::before_dom_children): Call
6868 gimple_purge_dead_eh_edges. Dump tranformation with details
6870 (strlen_dom_walker::before_dom_children): Update call by adding
6871 new argument cleanup_eh.
6872 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
6874 2018-01-03 Martin Liska <mliska@suse.cz>
6877 * cif-code.def (VARIADIC_THUNK): New enum value.
6878 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
6881 2018-01-03 Jan Beulich <jbeulich@suse.com>
6883 * sse.md (mov<mode>_internal): Tighten condition for when to use
6884 vmovdqu<ssescalarsize> for TI and OI modes.
6886 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6888 Update copyright years.
6890 2018-01-03 Martin Liska <mliska@suse.cz>
6893 * ipa-visibility.c (function_and_variable_visibility): Skip
6894 functions with noipa attribure.
6896 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6898 * gcc.c (process_command): Update copyright notice dates.
6899 * gcov-dump.c (print_version): Ditto.
6900 * gcov.c (print_version): Ditto.
6901 * gcov-tool.c (print_version): Ditto.
6902 * gengtype.c (create_file): Ditto.
6903 * doc/cpp.texi: Bump @copying's copyright year.
6904 * doc/cppinternals.texi: Ditto.
6905 * doc/gcc.texi: Ditto.
6906 * doc/gccint.texi: Ditto.
6907 * doc/gcov.texi: Ditto.
6908 * doc/install.texi: Ditto.
6909 * doc/invoke.texi: Ditto.
6911 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6913 * vector-builder.h (vector_builder::m_full_nelts): Change from
6914 unsigned int to poly_uint64.
6915 (vector_builder::full_nelts): Update prototype accordingly.
6916 (vector_builder::new_vector): Likewise.
6917 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
6918 (vector_builder::operator ==): Likewise.
6919 (vector_builder::finalize): Likewise.
6920 * int-vector-builder.h (int_vector_builder::int_vector_builder):
6921 Take the number of elements as a poly_uint64 rather than an
6923 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
6924 from unsigned int to poly_uint64.
6925 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
6926 (vec_perm_indices::new_vector): Likewise.
6927 (vec_perm_indices::length): Likewise.
6928 (vec_perm_indices::nelts_per_input): Likewise.
6929 (vec_perm_indices::input_nelts): Likewise.
6930 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
6931 number of elements per input as a poly_uint64 rather than an
6932 unsigned int. Use the original encoding for variable-length
6933 vectors, rather than clamping each individual element.
6934 For the second and subsequent elements in each pattern,
6935 clamp the step and base before clamping their sum.
6936 (vec_perm_indices::series_p): Handle polynomial element counts.
6937 (vec_perm_indices::all_in_range_p): Likewise.
6938 (vec_perm_indices_to_tree): Likewise.
6939 (vec_perm_indices_to_rtx): Likewise.
6940 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
6941 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
6942 (tree_vector_builder::new_binary_operation): Handle polynomial
6943 element counts. Return false if we need to know the number
6944 of elements at compile time.
6945 * fold-const.c (fold_vec_perm): Punt if the number of elements
6946 isn't known at compile time.
6948 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6950 * vec-perm-indices.h (vec_perm_builder): Change element type
6951 from HOST_WIDE_INT to poly_int64.
6952 (vec_perm_indices::element_type): Update accordingly.
6953 (vec_perm_indices::clamp): Handle polynomial element_types.
6954 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
6955 (vec_perm_indices::all_in_range_p): Likewise.
6956 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
6958 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
6959 polynomial vec_perm_indices element types.
6960 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
6961 * fold-const.c (fold_vec_perm): Likewise.
6962 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
6963 * tree-vect-generic.c (lower_vec_perm): Likewise.
6964 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
6965 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
6966 element type to HOST_WIDE_INT.
6968 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6969 Alan Hayward <alan.hayward@arm.com>
6970 David Sherwood <david.sherwood@arm.com>
6972 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
6973 rather than an int. Use plus_constant.
6974 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
6975 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
6977 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6978 Alan Hayward <alan.hayward@arm.com>
6979 David Sherwood <david.sherwood@arm.com>
6981 * calls.c (emit_call_1, expand_call): Change struct_value_size from
6982 a HOST_WIDE_INT to a poly_int64.
6984 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6985 Alan Hayward <alan.hayward@arm.com>
6986 David Sherwood <david.sherwood@arm.com>
6988 * calls.c (load_register_parameters): Cope with polynomial
6989 mode sizes. Require a constant size for BLKmode parameters
6990 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
6991 forces a parameter to be padded at the lsb end in order to
6992 fill a complete number of words, require the parameter size
6993 to be ordered wrt UNITS_PER_WORD.
6995 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6996 Alan Hayward <alan.hayward@arm.com>
6997 David Sherwood <david.sherwood@arm.com>
6999 * reload1.c (spill_stack_slot_width): Change element type
7000 from unsigned int to poly_uint64_pod.
7001 (alter_reg): Treat mode sizes as polynomial.
7003 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7004 Alan Hayward <alan.hayward@arm.com>
7005 David Sherwood <david.sherwood@arm.com>
7007 * reload.c (complex_word_subreg_p): New function.
7008 (reload_inner_reg_of_subreg, push_reload): Use it.
7010 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7011 Alan Hayward <alan.hayward@arm.com>
7012 David Sherwood <david.sherwood@arm.com>
7014 * lra-constraints.c (process_alt_operands): Reject matched
7015 operands whose sizes aren't ordered.
7016 (match_reload): Refer to this check here.
7018 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7019 Alan Hayward <alan.hayward@arm.com>
7020 David Sherwood <david.sherwood@arm.com>
7022 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
7023 that the mode size is in the set {1, 2, 4, 8, 16}.
7025 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7026 Alan Hayward <alan.hayward@arm.com>
7027 David Sherwood <david.sherwood@arm.com>
7029 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
7030 Use plus_constant instead of gen_rtx_PLUS.
7032 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7033 Alan Hayward <alan.hayward@arm.com>
7034 David Sherwood <david.sherwood@arm.com>
7036 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
7037 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
7038 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
7039 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
7040 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
7041 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
7042 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
7043 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
7044 * config/i386/i386.c (ix86_push_rounding): ...this new function.
7045 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
7047 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
7048 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
7049 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
7050 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
7051 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
7052 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
7053 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
7054 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
7055 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
7056 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
7058 * expr.c (emit_move_resolve_push): Treat the input and result
7059 of PUSH_ROUNDING as a poly_int64.
7060 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
7061 (emit_push_insn): Likewise.
7062 * lra-eliminations.c (mark_not_eliminable): Likewise.
7063 * recog.c (push_operand): Likewise.
7064 * reload1.c (elimination_effects): Likewise.
7065 * rtlanal.c (nonzero_bits1): Likewise.
7066 * calls.c (store_one_arg): Likewise. Require the padding to be
7067 known at compile time.
7069 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7070 Alan Hayward <alan.hayward@arm.com>
7071 David Sherwood <david.sherwood@arm.com>
7073 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
7074 Use plus_constant instead of gen_rtx_PLUS.
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 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
7083 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7084 Alan Hayward <alan.hayward@arm.com>
7085 David Sherwood <david.sherwood@arm.com>
7087 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
7088 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
7089 via stack temporaries. Treat the mode size as polynomial too.
7091 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7092 Alan Hayward <alan.hayward@arm.com>
7093 David Sherwood <david.sherwood@arm.com>
7095 * expr.c (expand_expr_real_2): When handling conversions involving
7096 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
7097 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
7098 as a poly_uint64 too.
7100 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7101 Alan Hayward <alan.hayward@arm.com>
7102 David Sherwood <david.sherwood@arm.com>
7104 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
7106 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7107 Alan Hayward <alan.hayward@arm.com>
7108 David Sherwood <david.sherwood@arm.com>
7110 * combine.c (can_change_dest_mode): Handle polynomial
7111 REGMODE_NATURAL_SIZE.
7112 * expmed.c (store_bit_field_1): Likewise.
7113 * expr.c (store_constructor): Likewise.
7114 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
7115 and polynomial REGMODE_NATURAL_SIZE.
7116 (gen_lowpart_common): Likewise.
7117 * reginfo.c (record_subregs_of_mode): Likewise.
7118 * rtlanal.c (read_modify_subreg_p): Likewise.
7120 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7121 Alan Hayward <alan.hayward@arm.com>
7122 David Sherwood <david.sherwood@arm.com>
7124 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
7125 numbers of elements.
7127 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7128 Alan Hayward <alan.hayward@arm.com>
7129 David Sherwood <david.sherwood@arm.com>
7131 * match.pd: Cope with polynomial numbers of vector elements.
7133 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7134 Alan Hayward <alan.hayward@arm.com>
7135 David Sherwood <david.sherwood@arm.com>
7137 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
7138 in a POINTER_PLUS_EXPR.
7140 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7141 Alan Hayward <alan.hayward@arm.com>
7142 David Sherwood <david.sherwood@arm.com>
7144 * omp-simd-clone.c (simd_clone_subparts): New function.
7145 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
7146 (ipa_simd_modify_function_body): Likewise.
7148 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7149 Alan Hayward <alan.hayward@arm.com>
7150 David Sherwood <david.sherwood@arm.com>
7152 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
7153 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
7154 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
7155 (expand_vector_condition, vector_element): Likewise.
7156 (subparts_gt): New function.
7157 (get_compute_type): Use subparts_gt.
7158 (count_type_subparts): Delete.
7159 (expand_vector_operations_1): Use subparts_gt instead of
7160 count_type_subparts.
7162 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7163 Alan Hayward <alan.hayward@arm.com>
7164 David Sherwood <david.sherwood@arm.com>
7166 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
7167 (vect_compile_time_alias): ...this new function. Do the calculation
7168 on poly_ints rather than trees.
7169 (vect_prune_runtime_alias_test_list): Update call accordingly.
7171 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7172 Alan Hayward <alan.hayward@arm.com>
7173 David Sherwood <david.sherwood@arm.com>
7175 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
7177 (vect_schedule_slp_instance): Likewise.
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-slp.c (vect_get_and_check_slp_defs): Reject
7184 constant and extern definitions for variable-length vectors.
7185 (vect_get_constant_vectors): Note that the number of units
7186 is known to be constant.
7188 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7189 Alan Hayward <alan.hayward@arm.com>
7190 David Sherwood <david.sherwood@arm.com>
7192 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
7193 of units as polynomial. Choose between WIDE and NARROW based
7196 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7197 Alan Hayward <alan.hayward@arm.com>
7198 David Sherwood <david.sherwood@arm.com>
7200 * tree-vect-stmts.c (simd_clone_subparts): New function.
7201 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
7203 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7204 Alan Hayward <alan.hayward@arm.com>
7205 David Sherwood <david.sherwood@arm.com>
7207 * tree-vect-stmts.c (vectorizable_call): Treat the number of
7208 vectors as polynomial. Use build_index_vector for
7211 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7212 Alan Hayward <alan.hayward@arm.com>
7213 David Sherwood <david.sherwood@arm.com>
7215 * tree-vect-stmts.c (get_load_store_type): Treat the number of
7216 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
7217 for variable-length vectors.
7218 (vectorizable_mask_load_store): Treat the number of units as
7219 polynomial, asserting that it is constant if the condition has
7220 already been enforced.
7221 (vectorizable_store, vectorizable_load): Likewise.
7223 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7224 Alan Hayward <alan.hayward@arm.com>
7225 David Sherwood <david.sherwood@arm.com>
7227 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
7228 of units as polynomial. Punt if we can't tell at compile time
7229 which vector contains the final result.
7231 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7232 Alan Hayward <alan.hayward@arm.com>
7233 David Sherwood <david.sherwood@arm.com>
7235 * tree-vect-loop.c (vectorizable_induction): Treat the number
7236 of units as polynomial. Punt on SLP inductions. Use an integer
7237 VEC_SERIES_EXPR for variable-length integer reductions. Use a
7238 cast of such a series for variable-length floating-point
7241 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7242 Alan Hayward <alan.hayward@arm.com>
7243 David Sherwood <david.sherwood@arm.com>
7245 * tree.h (build_index_vector): Declare.
7246 * tree.c (build_index_vector): New function.
7247 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
7248 of units as polynomial, forcibly converting it to a constant if
7249 vectorizable_reduction has already enforced the condition.
7250 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
7251 to create a {1,2,3,...} vector.
7252 (vectorizable_reduction): Treat the number of units as polynomial.
7253 Choose vectype_in based on the largest scalar element size rather
7254 than the smallest number of units. Enforce the restrictions
7257 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7258 Alan Hayward <alan.hayward@arm.com>
7259 David Sherwood <david.sherwood@arm.com>
7261 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
7262 number of units as polynomial.
7264 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7265 Alan Hayward <alan.hayward@arm.com>
7266 David Sherwood <david.sherwood@arm.com>
7268 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
7269 * target.def (autovectorize_vector_sizes): Return the vector sizes
7270 by pointer, using vector_sizes rather than a bitmask.
7271 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
7272 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
7273 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
7275 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
7276 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
7277 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
7278 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
7279 * omp-general.c (omp_max_vf): Likewise.
7280 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7281 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
7282 * tree-vect-loop.c (vect_analyze_loop): Likewise.
7283 * tree-vect-slp.c (vect_slp_bb): Likewise.
7284 * doc/tm.texi: Regenerate.
7285 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
7287 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
7288 the vector size as a poly_uint64 rather than an unsigned int.
7289 (current_vector_size): Change from an unsigned int to a poly_uint64.
7290 (get_vectype_for_scalar_type): Update accordingly.
7291 * tree.h (build_truth_vector_type): Take the size and number of
7292 units as a poly_uint64 rather than an unsigned int.
7293 (build_vector_type): Add a temporary overload that takes
7294 the number of units as a poly_uint64 rather than an unsigned int.
7295 * tree.c (make_vector_type): Likewise.
7296 (build_truth_vector_type): Take the number of units as a poly_uint64
7297 rather than an unsigned int.
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 * target.def (get_mask_mode): Take the number of units and length
7304 as poly_uint64s rather than unsigned ints.
7305 * targhooks.h (default_get_mask_mode): Update accordingly.
7306 * targhooks.c (default_get_mask_mode): Likewise.
7307 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
7308 * doc/tm.texi: Regenerate.
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 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
7315 * omp-general.c (omp_max_vf): Likewise.
7316 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
7317 (expand_omp_simd): Handle polynomial safelen.
7318 * omp-low.c (omplow_simd_context): Add a default constructor.
7319 (omplow_simd_context::max_vf): Change from int to poly_uint64.
7320 (lower_rec_simd_input_clauses): Update accordingly.
7321 (lower_rec_input_clauses): Likewise.
7323 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7324 Alan Hayward <alan.hayward@arm.com>
7325 David Sherwood <david.sherwood@arm.com>
7327 * tree-vectorizer.h (vect_nunits_for_cost): New function.
7328 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
7329 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
7330 (vect_analyze_slp_cost): Likewise.
7331 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
7332 (vect_model_load_cost): Likewise.
7334 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7335 Alan Hayward <alan.hayward@arm.com>
7336 David Sherwood <david.sherwood@arm.com>
7338 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
7339 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
7340 from an unsigned int * to a poly_uint64_pod *.
7341 (calculate_unrolling_factor): New function.
7342 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
7344 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7345 Alan Hayward <alan.hayward@arm.com>
7346 David Sherwood <david.sherwood@arm.com>
7348 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
7349 from an unsigned int to a poly_uint64.
7350 (_loop_vec_info::slp_unrolling_factor): Likewise.
7351 (_loop_vec_info::vectorization_factor): Change from an int
7353 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
7354 (vect_get_num_vectors): New function.
7355 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
7356 (vect_get_num_copies): Use vect_get_num_vectors.
7357 (vect_analyze_data_ref_dependences): Change max_vf from an int *
7358 to an unsigned int *.
7359 (vect_analyze_data_refs): Change min_vf from an int * to a
7361 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7362 than an unsigned HOST_WIDE_INT.
7363 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7364 (vect_analyze_data_ref_dependence): Change max_vf from an int *
7365 to an unsigned int *.
7366 (vect_analyze_data_ref_dependences): Likewise.
7367 (vect_compute_data_ref_alignment): Handle polynomial vf.
7368 (vect_enhance_data_refs_alignment): Likewise.
7369 (vect_prune_runtime_alias_test_list): Likewise.
7370 (vect_shift_permute_load_chain): Likewise.
7371 (vect_supportable_dr_alignment): Likewise.
7372 (dependence_distance_ge_vf): Take the vectorization factor as a
7373 poly_uint64 rather than an unsigned HOST_WIDE_INT.
7374 (vect_analyze_data_refs): Change min_vf from an int * to a
7376 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
7377 vfm1 as a poly_uint64 rather than an int. Make the same change
7378 for the returned bound_scalar.
7379 (vect_gen_vector_loop_niters): Handle polynomial vf.
7380 (vect_do_peeling): Likewise. Update call to
7381 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
7382 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
7384 * tree-vect-loop.c (vect_determine_vectorization_factor)
7385 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
7386 (vect_get_known_peeling_cost): Likewise.
7387 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
7388 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
7389 (vect_transform_loop): Likewise. Use the lowest possible VF when
7390 updating the upper bounds of the loop.
7391 (vect_min_worthwhile_factor): Make static. Return an unsigned int
7393 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
7394 polynomial unroll factors.
7395 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
7396 (vect_make_slp_decision): Likewise.
7397 (vect_supported_load_permutation_p): Likewise, and polynomial
7399 (vect_analyze_slp_cost): Handle polynomial vf.
7400 (vect_slp_analyze_node_operations): Likewise.
7401 (vect_slp_analyze_bb_1): Likewise.
7402 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7403 than an unsigned HOST_WIDE_INT.
7404 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
7405 (vectorizable_load): Handle polynomial vf.
7406 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
7408 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
7410 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7411 Alan Hayward <alan.hayward@arm.com>
7412 David Sherwood <david.sherwood@arm.com>
7414 * match.pd: Handle bit operations involving three constants
7415 and try to fold one pair.
7417 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7419 * tree-vect-loop-manip.c: Include gimple-fold.h.
7420 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
7421 niters_maybe_zero parameters. Handle other cases besides a step of 1.
7422 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
7423 Add a path that uses a step of VF instead of 1, but disable it
7425 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
7426 and niters_no_overflow parameters. Update calls to
7427 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
7428 Create a new SSA name if the latter choses to use a ste other
7429 than zero, and return it via niters_vector_mult_vf_var.
7430 * tree-vect-loop.c (vect_transform_loop): Update calls to
7431 vect_do_peeling, vect_gen_vector_loop_niters and
7432 slpeel_make_loop_iterate_ntimes.
7433 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
7434 (vect_gen_vector_loop_niters): Update declarations after above changes.
7436 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
7438 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
7439 128-bit round to integer instructions.
7440 (ceil<mode>2): Likewise.
7441 (btrunc<mode>2): Likewise.
7442 (round<mode>2): Likewise.
7444 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7446 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
7447 unaligned VSX load/store on P8/P9.
7448 (expand_block_clear): Allow the use of unaligned VSX
7449 load/store on P8/P9.
7451 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7453 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
7455 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
7456 swap associated with both a load and a store.
7458 2018-01-02 Andrew Waterman <andrew@sifive.com>
7460 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
7461 * config/riscv/riscv.md (clear_cache): Use it.
7463 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
7465 * web.c: Remove out-of-date comment.
7467 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7469 * expr.c (fixup_args_size_notes): Check that any existing
7470 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
7471 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
7472 (emit_single_push_insn): ...here.
7474 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7476 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
7477 (const_vector_encoded_nelts): New function.
7478 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
7479 (const_vector_int_elt, const_vector_elt): Declare.
7480 * emit-rtl.c (const_vector_int_elt_1): New function.
7481 (const_vector_elt): Likewise.
7482 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
7483 of CONST_VECTOR_ELT.
7485 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7487 * expr.c: Include rtx-vector-builder.h.
7488 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
7489 directly on the tree encoding.
7490 (const_vector_from_tree): Likewise.
7491 * optabs.c: Include rtx-vector-builder.h.
7492 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
7493 sequence of "u" values.
7494 * vec-perm-indices.c: Include rtx-vector-builder.h.
7495 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
7496 directly on the vec_perm_indices encoding.
7498 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7500 * doc/rtl.texi (const_vector): Describe new encoding scheme.
7501 * Makefile.in (OBJS): Add rtx-vector-builder.o.
7502 * rtx-vector-builder.h: New file.
7503 * rtx-vector-builder.c: Likewise.
7504 * rtl.h (rtx_def::u2): Add a const_vector field.
7505 (CONST_VECTOR_NPATTERNS): New macro.
7506 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
7507 (CONST_VECTOR_DUPLICATE_P): Likewise.
7508 (CONST_VECTOR_STEPPED_P): Likewise.
7509 (CONST_VECTOR_ENCODED_ELT): Likewise.
7510 (const_vec_duplicate_p): Check for a duplicated vector encoding.
7511 (unwrap_const_vec_duplicate): Likewise.
7512 (const_vec_series_p): Check for a non-duplicated vector encoding.
7513 Say that the function only returns true for integer vectors.
7514 * emit-rtl.c: Include rtx-vector-builder.h.
7515 (gen_const_vec_duplicate_1): Delete.
7516 (gen_const_vector): Call gen_const_vec_duplicate instead of
7517 gen_const_vec_duplicate_1.
7518 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
7519 (gen_const_vec_duplicate): Use rtx_vector_builder.
7520 (gen_const_vec_series): Likewise.
7521 (gen_rtx_CONST_VECTOR): Likewise.
7522 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
7523 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7524 Build a new vector rather than modifying a CONST_VECTOR in-place.
7525 (handle_special_swappables): Update call accordingly.
7526 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
7527 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7528 Build a new vector rather than modifying a CONST_VECTOR in-place.
7529 (handle_special_swappables): Update call accordingly.
7531 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7533 * simplify-rtx.c (simplify_const_binary_operation): Use
7534 CONST_VECTOR_ELT instead of XVECEXP.
7536 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7538 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
7539 the selector elements to be different from the data elements
7540 if the selector is a VECTOR_CST.
7541 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
7542 ssizetype for the selector.
7544 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7546 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
7547 before testing each element individually.
7548 * tree-vect-generic.c (lower_vec_perm): Likewise.
7550 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7552 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
7553 * selftest-run-tests.c (selftest::run_tests): Call it.
7554 * vector-builder.h (vector_builder::operator ==): New function.
7555 (vector_builder::operator !=): Likewise.
7556 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
7557 (vec_perm_indices::all_from_input_p): New function.
7558 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7559 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
7560 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
7561 instead of reading the VECTOR_CST directly. Detect whether both
7562 vector inputs are the same before constructing the vec_perm_indices,
7563 and update the number of inputs argument accordingly. Use the
7564 utility functions added above. Only construct sel2 if we need to.
7566 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7568 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
7569 the broadcast of the low byte.
7570 (expand_mult_highpart): Use an explicit encoding for the permutes.
7571 * optabs-query.c (can_mult_highpart_p): Likewise.
7572 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
7573 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7574 (vectorizable_bswap): Likewise.
7575 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
7576 explicit encoding for the power-of-2 permutes.
7577 (vect_permute_store_chain): Likewise.
7578 (vect_grouped_load_supported): Likewise.
7579 (vect_permute_load_chain): Likewise.
7581 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7583 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
7584 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
7585 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
7586 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7587 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
7588 (vect_gen_perm_mask_any): Likewise.
7590 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7592 * int-vector-builder.h: New file.
7593 * vec-perm-indices.h: Include int-vector-builder.h.
7594 (vec_perm_indices): Redefine as an int_vector_builder.
7595 (auto_vec_perm_indices): Delete.
7596 (vec_perm_builder): Redefine as a stand-alone class.
7597 (vec_perm_indices::vec_perm_indices): New function.
7598 (vec_perm_indices::clamp): Likewise.
7599 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
7600 (vec_perm_indices::new_vector): New function.
7601 (vec_perm_indices::new_expanded_vector): Update for new
7602 vec_perm_indices class.
7603 (vec_perm_indices::rotate_inputs): New function.
7604 (vec_perm_indices::all_in_range_p): Operate directly on the
7605 encoded form, without computing elided elements.
7606 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
7607 encoding. Update for new vec_perm_indices class.
7608 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
7609 the given vec_perm_builder.
7610 (expand_vec_perm_var): Update vec_perm_builder constructor.
7611 (expand_mult_highpart): Use vec_perm_builder instead of
7612 auto_vec_perm_indices.
7613 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
7614 vec_perm_indices instead of auto_vec_perm_indices. Use a single
7615 or double series encoding as appropriate.
7616 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
7617 vec_perm_indices instead of auto_vec_perm_indices.
7618 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7619 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7620 (vect_permute_store_chain): Likewise.
7621 (vect_grouped_load_supported): Likewise.
7622 (vect_permute_load_chain): Likewise.
7623 (vect_shift_permute_load_chain): Likewise.
7624 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
7625 (vect_transform_slp_perm_load): Likewise.
7626 (vect_schedule_slp_instance): Likewise.
7627 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7628 (vectorizable_mask_load_store): Likewise.
7629 (vectorizable_bswap): Likewise.
7630 (vectorizable_store): Likewise.
7631 (vectorizable_load): Likewise.
7632 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
7633 vec_perm_indices instead of auto_vec_perm_indices. Use
7634 tree_to_vec_perm_builder to read the vector from a tree.
7635 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
7636 vec_perm_builder instead of a vec_perm_indices.
7637 (have_whole_vector_shift): Use vec_perm_builder and
7638 vec_perm_indices instead of auto_vec_perm_indices. Leave the
7639 truncation to calc_vec_perm_mask_for_shift.
7640 (vect_create_epilog_for_reduction): Likewise.
7641 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
7642 from auto_vec_perm_indices to vec_perm_indices.
7643 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7644 instead of changing individual elements.
7645 (aarch64_vectorize_vec_perm_const): Use new_vector to install
7646 the vector in d.perm.
7647 * config/arm/arm.c (expand_vec_perm_d::perm): Change
7648 from auto_vec_perm_indices to vec_perm_indices.
7649 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7650 instead of changing individual elements.
7651 (arm_vectorize_vec_perm_const): Use new_vector to install
7652 the vector in d.perm.
7653 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
7654 Update vec_perm_builder constructor.
7655 (rs6000_expand_interleave): Likewise.
7656 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
7657 (rs6000_expand_interleave): Likewise.
7659 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7661 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
7662 to qimode could truncate the indices.
7663 * optabs.c (expand_vec_perm_var): Likewise.
7665 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7667 * Makefile.in (OBJS): Add vec-perm-indices.o.
7668 * vec-perm-indices.h: New file.
7669 * vec-perm-indices.c: Likewise.
7670 * target.h (vec_perm_indices): Replace with a forward class
7672 (auto_vec_perm_indices): Move to vec-perm-indices.h.
7673 * optabs.h: Include vec-perm-indices.h.
7674 (expand_vec_perm): Delete.
7675 (selector_fits_mode_p, expand_vec_perm_var): Declare.
7676 (expand_vec_perm_const): Declare.
7677 * target.def (vec_perm_const_ok): Replace with...
7678 (vec_perm_const): ...this new hook.
7679 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
7680 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
7681 * doc/tm.texi: Regenerate.
7682 * optabs.def (vec_perm_const): Delete.
7683 * doc/md.texi (vec_perm_const): Likewise.
7684 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
7685 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
7686 expand_vec_perm for constant permutation vectors. Assert that
7687 the mode of variable permutation vectors is the integer equivalent
7688 of the mode that is being permuted.
7689 * optabs-query.h (selector_fits_mode_p): Declare.
7690 * optabs-query.c: Include vec-perm-indices.h.
7691 (selector_fits_mode_p): New function.
7692 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
7693 is defined, instead of checking whether the vec_perm_const_optab
7694 exists. Use targetm.vectorize.vec_perm_const instead of
7695 targetm.vectorize.vec_perm_const_ok. Check whether the indices
7696 fit in the vector mode before using a variable permute.
7697 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
7698 vec_perm_indices instead of an rtx.
7699 (expand_vec_perm): Replace with...
7700 (expand_vec_perm_const): ...this new function. Take the selector
7701 as a vec_perm_indices rather than an rtx. Also take the mode of
7702 the selector. Update call to shift_amt_for_vec_perm_mask.
7703 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
7704 Use vec_perm_indices::new_expanded_vector to expand the original
7705 selector into bytes. Check whether the indices fit in the vector
7706 mode before using a variable permute.
7707 (expand_vec_perm_var): Make global.
7708 (expand_mult_highpart): Use expand_vec_perm_const.
7709 * fold-const.c: Includes vec-perm-indices.h.
7710 * tree-ssa-forwprop.c: Likewise.
7711 * tree-vect-data-refs.c: Likewise.
7712 * tree-vect-generic.c: Likewise.
7713 * tree-vect-loop.c: Likewise.
7714 * tree-vect-slp.c: Likewise.
7715 * tree-vect-stmts.c: Likewise.
7716 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
7718 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
7719 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
7720 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
7721 (aarch64_vectorize_vec_perm_const): ...this new function.
7722 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7723 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7724 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
7725 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
7726 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7727 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7728 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
7730 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
7731 check for NEON modes.
7732 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
7733 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
7734 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
7735 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
7737 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
7738 the old VEC_PERM_CONST conditions.
7739 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
7740 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
7741 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
7742 (ia64_vectorize_vec_perm_const_ok): Merge into...
7743 (ia64_vectorize_vec_perm_const): ...this new function.
7744 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
7745 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
7746 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
7747 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
7748 * config/mips/mips.c (mips_expand_vec_perm_const)
7749 (mips_vectorize_vec_perm_const_ok): Merge into...
7750 (mips_vectorize_vec_perm_const): ...this new function.
7751 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
7752 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
7753 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
7754 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
7755 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
7756 (rs6000_expand_vec_perm_const): Delete.
7757 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
7759 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7760 (altivec_expand_vec_perm_const_le): Take each operand individually.
7761 Operate on constant selectors rather than rtxes.
7762 (altivec_expand_vec_perm_const): Likewise. Update call to
7763 altivec_expand_vec_perm_const_le.
7764 (rs6000_expand_vec_perm_const): Delete.
7765 (rs6000_vectorize_vec_perm_const_ok): Delete.
7766 (rs6000_vectorize_vec_perm_const): New function.
7767 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
7768 an element count and rtx array.
7769 (rs6000_expand_extract_even): Update call accordingly.
7770 (rs6000_expand_interleave): Likewise.
7771 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
7772 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
7773 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
7774 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
7775 (rs6000_expand_vec_perm_const): Delete.
7776 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7777 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7778 (altivec_expand_vec_perm_const_le): Take each operand individually.
7779 Operate on constant selectors rather than rtxes.
7780 (altivec_expand_vec_perm_const): Likewise. Update call to
7781 altivec_expand_vec_perm_const_le.
7782 (rs6000_expand_vec_perm_const): Delete.
7783 (rs6000_vectorize_vec_perm_const_ok): Delete.
7784 (rs6000_vectorize_vec_perm_const): New function. Remove stray
7785 reference to the SPE evmerge intructions.
7786 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
7787 an element count and rtx array.
7788 (rs6000_expand_extract_even): Update call accordingly.
7789 (rs6000_expand_interleave): Likewise.
7790 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
7791 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
7793 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7795 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7797 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
7798 vector mode and that that mode matches the mode of the data
7800 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
7801 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
7802 directly using expand_vec_perm_1 when forcing selectors into
7804 (expand_vec_perm_var): New function, split out from expand_vec_perm.
7806 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7808 * optabs-query.h (can_vec_perm_p): Delete.
7809 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
7810 * optabs-query.c (can_vec_perm_p): Split into...
7811 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
7812 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
7813 particular selector is valid.
7814 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7815 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7816 (vect_grouped_load_supported): Likewise.
7817 (vect_shift_permute_load_chain): Likewise.
7818 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
7819 (vect_transform_slp_perm_load): Likewise.
7820 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7821 (vectorizable_bswap): Likewise.
7822 (vect_gen_perm_mask_checked): Likewise.
7823 * fold-const.c (fold_ternary_loc): Likewise. Don't take
7824 implementations of variable permutation vectors into account
7825 when deciding which selector to use.
7826 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
7827 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
7828 with a false third argument.
7829 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
7830 to test whether the constant selector is valid and can_vec_perm_var_p
7831 to test whether a variable selector is valid.
7833 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7835 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
7836 * optabs-query.c (can_vec_perm_p): Likewise.
7837 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
7838 instead of vec_perm_indices.
7839 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
7840 (vect_gen_perm_mask_checked): Likewise,
7841 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
7842 (vect_gen_perm_mask_checked): Likewise,
7844 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7846 * optabs-query.h (qimode_for_vec_perm): Declare.
7847 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
7848 (qimode_for_vec_perm): ...this new function.
7849 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
7851 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7853 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
7854 does not have a conditional at the top.
7856 2018-01-02 Richard Biener <rguenther@suse.de>
7858 * ipa-inline.c (big_speedup_p): Fix expression.
7860 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
7863 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
7866 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
7870 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
7871 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
7872 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
7873 cond_taken_branch_cost 3->4.
7875 2018-01-01 Jakub Jelinek <jakub@redhat.com>
7877 PR tree-optimization/83581
7878 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
7879 TODO_cleanup_cfg if any changes have been made.
7882 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
7883 convert_modes if target mode has the right side, but different mode
7887 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
7888 last argument when extracting from CONCAT. If either from_real or
7889 from_imag is NULL, use expansion through memory. If result is not
7890 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
7891 the parts directly to inner mode, if even that fails, use expansion
7895 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
7896 check for bswap in mode rather than HImode and use that in expand_unop
7899 Copyright (C) 2018 Free Software Foundation, Inc.
7901 Copying and distribution of this file, with or without modification,
7902 are permitted in any medium without royalty provided the copyright
7903 notice and this notice are preserved.