1 2018-03-01 Matthew Fortune <mfortune@gmail.com>
3 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5 (mips_final_postscan_insn): Likewise.
7 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
9 PR rtl-optimization/84528
10 * alias.c (init_alias_target): Add commentary.
11 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
12 a unique base value if the frame pointer is not eliminated
15 2018-03-01 Tom de Vries <tom@codesourcery.com>
17 PR rtl-optimization/83327
18 * lra-int.h (hard_regs_spilled_into): Declare.
19 * lra.c (hard_regs_spilled_into): Define.
20 (init_reg_info): Init hard_regs_spilled_into.
21 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
22 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
23 (process_bb_lives): Handle hard_regs_spilled_into.
24 (lra_create_live_ranges_1): Before doing liveness propagation, clear
25 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
27 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
29 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
30 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
31 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
32 * config/rs6000/aix72.h: New file.
34 2018-02-28 Jakub Jelinek <jakub@redhat.com>
36 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
37 instead of warning_at with conditional singular and plural messages
41 * stor-layout.c (update_alignment_for_field): For
42 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
43 && !DECL_PACKED (field), do the alignment update, just use
44 only desired_align instead of MAX (type_align, desired_align)
46 (place_field): Don't do known_align < desired_align handling
47 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
48 is non-NULL, instead do it after rli->prev_field handling and
49 only if not within a bitfield word. For DECL_PACKED (field)
50 use type_align of BITS_PER_UNIT.
52 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
54 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
55 superfluous parentheses and trailing spaces.
57 2018-02-28 Richard Biener <rguenther@suse.de>
59 PR tree-optimization/84584
60 * graphite-scop-detection.c (scop_detection::add_scop): Discard
61 SCoPs with fake exit edge.
63 2018-02-28 Martin Liska <mliska@suse.cz>
66 * timevar.c (timer::print): Fix format to properly print 100%
69 2018-02-28 Richard Biener <rguenther@suse.de>
72 * genmatch.c (capture_info::walk_match): Do not mark
73 captured expressions without operands as expr_p given
74 they act more like predicates and should be subject to
75 "lost tail" side-effect preserving.
77 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
79 PR rtl-optimization/81611
80 * auto-inc-dec.c (attempt_change): Move dead note from
81 mem_insn if it's the next use of regno
82 (find_address): Take address use of reg holding
83 non-incremented value. Add parm to limit search to the named
85 (merge_in_block): Attempt to use a mem insn that is the next
86 use of the original regno.
88 2018-02-27 Martin Sebor <msebor@redhat.com>
91 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
92 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
94 2018-02-27 Martin Sebor <msebor@redhat.com>
97 * diagnostic-core.h (warning_n, error_n, inform_n): Change
98 n argument to unsigned HOST_WIDE_INT.
99 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
100 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
101 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
102 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
104 2018-02-27 Richard Biener <rguenther@suse.de>
106 PR tree-optimization/84512
107 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
108 Do not use the estimate returned from record_stmt_cost for
109 the scalar iteration cost but sum properly using add_stmt_cost.
111 2018-02-27 Richard Biener <rguenther@suse.de>
113 PR tree-optimization/84466
114 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
115 Adjust last change to less strictly validate use operands.
117 2018-02-27 Martin Liska <mliska@suse.cz>
119 PR gcov-profile/84548
120 * gcov.c (process_file): Allow partial overlap and consider it
121 also as group functions.
122 (output_lines): Properly calculate range of lines for a group.
124 2018-02-27 Martin Liska <mliska@suse.cz>
126 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
127 'ggc' suffixes. Change first column width.
128 (timer::print): Fix formatting of the column.
130 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
132 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
133 preserve inline entry blocks for the sake of debug inline
134 entry point markers alone.
135 (remove_unused_locals): Suggest in comments a better place to
136 force the preservation of inline entry blocks that are
137 otherwise unused, but do not preserve them.
139 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
141 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
143 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
146 * config/i386/constraints.md (Bs): Replace
147 ix86_indirect_branch_register with
148 TARGET_INDIRECT_BRANCH_REGISTER.
150 * config/i386/i386.md (indirect_jump): Likewise.
151 (tablejump): Likewise.
152 (*sibcall_memory): Likewise.
153 (*sibcall_value_memory): Likewise.
154 Peepholes of indirect call and jump via memory: Likewise.
155 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
156 (*sibcall_value_GOT_32): Likewise.
157 * config/i386/predicates.md (indirect_branch_operand): Likewise.
158 (GOT_memory_operand): Likewise.
159 (call_insn_operand): Likewise.
160 (sibcall_insn_operand): Likewise.
161 (GOT32_symbol_operand): Likewise.
162 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
164 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
166 PR rtl-optimization/83496
167 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
168 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
169 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
170 redundant insn, if any.
171 (relax_delay_slots): Likewise.
172 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
174 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
176 PR tree-optimization/83965
177 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
178 that grouped statements are part of a reduction chain. Return
179 true if the statement is not marked as a reduction itself but
181 (vect_recog_dot_prod_pattern): Don't check whether the statement
182 is part of a group here.
183 (vect_recog_sad_pattern): Likewise.
184 (vect_recog_widen_sum_pattern): Likewise.
186 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
189 * final.c (rest_of_clean_state): Also look for calls inside sequences.
191 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
194 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
196 (ix86_output_indirect_function_return): New prototype.
197 (ix86_split_simple_return_pop_internal): Likewise.
198 * config/i386/i386.c (indirect_return_via_cx): New.
199 (indirect_return_via_cx_bnd): Likewise.
200 (indirect_thunk_name): Handle return va CX_REG.
201 (output_indirect_thunk_function): Create alias for
202 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
203 (ix86_output_indirect_jmp): Remove the bool argument.
204 (ix86_output_indirect_function_return): New function.
205 (ix86_split_simple_return_pop_internal): Likewise.
206 * config/i386/i386.md (*indirect_jump): Don't pass false
207 to ix86_output_indirect_jmp.
208 (*tablejump_1): Likewise.
209 (simple_return_pop_internal): Change it to define_insn_and_split.
210 Call ix86_split_simple_return_pop_internal to split it for
212 (simple_return_indirect_internal): Call
213 ix86_output_indirect_function_return instead of
214 ix86_output_indirect_jmp.
216 2018-02-26 Jakub Jelinek <jakub@redhat.com>
219 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
220 memset and value initialization afterwards.
222 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
224 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
226 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
229 * common/config/aarch64/aarch64-common.c
230 (aarch_option_optimization_table[]): Switch
231 off fomit-frame-pointer
233 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
234 Chung-Ju Wu <jasonwucj@gmail.com>
236 * config/nds32/nds32-multiple.md (load_multiple): Disallow
238 (store_multiple): Ditto.
240 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
242 * config.gcc: Add --with-cpu support for nds32 target.
243 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
244 * config/nds32/nds32.opt: Add -mcpu= option.
246 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
248 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
249 isel=yes): Warn for these deprecated options.
251 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
253 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
254 ISA_2_5_MASKS_EMBEDDED.
256 2018-02-23 Jakub Jelinek <jakub@redhat.com>
258 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
259 p->max as pointers rather than using iterative_hash_expr.
261 2018-02-23 Carl Love <cel@us.ibm.com>
263 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
264 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
266 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
267 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
268 P8V_BUILTIN_VEC_VUNSIGNED2.
270 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
273 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
274 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
276 * lra-constraints.c (curr_insn_transform): Set up
277 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
279 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
280 * lra-eliminations.c (spill_pseudos): Ditto.
281 (process_insn_for_elimination): Ditto.
282 * lra-lives.c (reg_early_clobber_p): Use the new macros.
283 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
284 LRA_NON_CLOBBERED_ALT.
286 2018-02-22 Martin Sebor <msebor@redhat.com>
288 PR tree-optimization/84480
289 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
290 to maybe_diag_stxncpy_trunc. Call it.
291 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
292 from gimple_fold_builtin_strcpy. Print inlining stack.
293 (handle_builtin_stxncpy): Print inlining stack.
294 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
296 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
299 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
300 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
301 and -fcheck-pointer-bounds are used together.
302 (indirect_thunk_prefix): New enum.
303 (indirect_thunk_need_prefix): New function.
304 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
305 "_nt" instead of "_bnd" for NOTRACK prefix.
306 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
307 (output_indirect_thunk_function): Likewise.
309 (ix86_code_end): Update output_indirect_thunk_function calls.
310 (ix86_output_indirect_branch_via_reg): Replace
311 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
312 (ix86_output_indirect_branch_via_push): Likewise.
313 (ix86_output_function_return): Likewise.
314 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
315 incompatible with -fcf-protection=branch and
316 -fcheck-pointer-bounds.
318 2018-02-22 Steve Ellcey <sellcey@cavium.com>
321 * config/aarch64/aarch64.c (aarch64_print_address_internal):
322 Change gcc_assert call to output_operand_lossage.
324 2018-02-22 Steve Ellcey <sellcey@cavium.com>
326 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
328 2018-02-22 DJ Delorie <dj@redhat.com>
329 Sebastian Perta <sebastian.perta@renesas.com>
330 Oleg Endo <olegendo@gcc.gnu.org>
332 * config/rx/rx.c (rx_rtx_costs): New function.
333 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
335 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
337 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
339 2018-02-22 Martin Liska <mliska@suse.cz>
342 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
343 Add "native" as a possible value.
345 2018-02-22 Martin Liska <mliska@suse.cz>
348 * config/i386/i386.c (ix86_option_override_internal):
349 Add "native" as a possible value for -march and -mtune.
351 2018-02-22 Jakub Jelinek <jakub@redhat.com>
354 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
355 to all type variants.
357 PR tree-optimization/84503
358 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
359 width as info->bitpos + info->bitsize - start.
360 (merged_store_group::merge_overlapping): Simplify width computation.
361 (check_no_overlap): New function.
362 (imm_store_chain_info::try_coalesce_bswap): Compute expected
363 start + width and last_order of the group, fail if check_no_overlap
365 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
366 to group if check_no_overlap fails.
368 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
370 * config/rs6000/altivec.md: Delete contraint arguments to
371 define_expand, define_split, and define_peephole2, and in
372 define_insn_and_split if always unused.
373 * config/rs6000/darwin.md: Ditto.
374 * config/rs6000/dfp.md: Ditto.
375 * config/rs6000/rs6000.md: Ditto.
376 * config/rs6000/sync.md: Ditto.
377 * config/rs6000/vector.md: Ditto.
378 * config/rs6000/vsx.md: Ditto.
380 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
382 * config/rs6000/altivec.md: Write output control strings as braced
383 blocks instead of double-quoted strings.
384 * config/rs6000/darwin.md: Ditto.
385 * config/rs6000/rs6000.md: Ditto.
386 * config/rs6000/vector.md: Ditto.
387 * config/rs6000/vsx.md: Ditto.
389 2018-02-21 Jason Merrill <jason@redhat.com>
391 PR c++/84314 - ICE with templates and fastcall attribute.
392 * attribs.c (build_type_attribute_qual_variant): Remove assert.
394 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
396 * ipa-cp.c (determine_versionability): Fix comment typos.
398 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
401 * ipa-cp.c (determine_versionability): Do not version functions caling
404 2018-02-21 Martin Liska <mliska@suse.cz>
407 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
408 Add "native" as a possible value.
409 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
410 the macro when native cpu detection is available.
412 2018-02-21 Martin Liska <mliska@suse.cz>
415 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
416 Add "native" as a possible value.
417 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
418 when native cpu detection is available.
420 2018-02-21 Jakub Jelinek <jakub@redhat.com>
421 Martin Sebor <msebor@redhat.com>
423 PR tree-optimization/84478
424 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
426 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
427 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
428 support which is conservatively correct, for 2 only stay conservative
429 for maxlen. Formatting and comment capitalization fixes. Add STRICT
430 argument to the 2 argument get_range_strlen, adjust 6 arg
431 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
433 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
434 (gimple_fold_builtin_strlen): Pass true as last argument to
437 2018-02-20 Martin Sebor <msebor@redhat.com>
440 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
441 (builtin_memref::set_base_and_offset): Same. Handle inner references.
442 (builtin_memref::builtin_memref): Factor out parts into
443 set_base_and_offset and call it.
445 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
448 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
449 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
450 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
451 search at the associated MODE_INT.
453 2018-02-20 Jeff Law <law@redhat.com>
456 PR tree-optimization/81592
458 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
459 for range data rather than using global data.
460 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
461 range data rather than using global data.
462 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
463 pass it to children as needed.
464 (struct directive::fmtresult): Similarly.
465 (struct directive::set_width): Similarly.
466 (struct directive::set_precision): Similarly.
467 (format_integer, format_directive, parse_directive): Similarly.
468 (format_none): Accept unnamed vr_values parameter.
469 (format_percent, format_floating, format_character): Similarly.
470 (format_string, format_plain): Similarly.
471 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
472 the EVRP range analyzer for range data rather than using global data.
473 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
474 gimple-ssa-evrp-analyze.h
475 (class sprintf_dom_walker): Add after_dom_children member function.
476 Add evrp_range_analyzer member.
477 (sprintf_dom_walker::before_dom_children): Call into the EVRP
478 range analyzer as needed.
479 (sprintf_dom_walker::after_dom_children): New member function.
480 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
482 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
483 (evrp_range_analyzer::pop_to_marker): Likewise.
485 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
487 PR tree-optimization/84419
488 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
489 with the required type if its current type is compatible but
492 2018-02-20 Jakub Jelinek <jakub@redhat.com>
495 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
498 2018-02-20 Martin Liska <mliska@suse.cz>
501 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
502 possible values if we don't have a hint.
504 2018-02-20 Martin Liska <mliska@suse.cz>
508 * final.c (shorten_branches): Build align_tab array with one
510 * opts.c (finish_options): Add alignment option limit check.
511 (MAX_CODE_ALIGN): Likewise.
512 (MAX_CODE_ALIGN_VALUE): Likewise.
513 * doc/invoke.texi: Document maximum allowed option value for
514 all -falign-* options.
516 2018-02-19 Jakub Jelinek <jakub@redhat.com>
519 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
520 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
521 * var-tracking.c (emit_note_insn_var_location): Remove all references
522 to NOTE_INSN_CALL_ARG_LOCATION.
523 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
524 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
525 Use copy_rtx_if_shared.
526 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
527 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
528 (dwarf2out_var_location): Remove handling of
529 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
531 * final.c (final_scan_insn): Remove all references to
532 NOTE_INSN_CALL_ARG_LOCATION.
533 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
534 before dumping final insns.
535 * except.c (emit_note_eh_region_end): Remove all references to
536 NOTE_INSN_CALL_ARG_LOCATION.
537 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
538 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
539 * config/arc/arc.c (hwloop_optimize): Likewise.
540 * config/arm/arm.c (create_fix_barrier): Likewise.
541 * config/s390/s390.c (s390_chunkify_start): Likewise.
542 * config/sh/sh.c (find_barrier): Likewise.
543 * config/i386/i386.c (rest_of_insert_endbranch,
544 ix86_seh_fixup_eh_fallthru): Likewise.
545 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
546 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
547 * config/frv/frv.c (frv_function_prologue): Likewise.
548 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
550 (note_outside_basic_block_p): Remove all references to
551 NOTE_INSN_CALL_ARG_LOCATION.
552 * gengtype.c (adjust_field_rtx_def): Likewise.
553 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
555 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
556 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
559 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
562 PR tree-optimization/84452
563 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
564 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
567 2018-02-19 Martin Liska <mliska@suse.cz>
570 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
572 2018-02-19 Martin Liska <mliska@suse.cz>
573 Richard Sandiford <richard.sandiford@linaro.org>
575 PR tree-optimization/82491
576 * gimple-fold.c (get_base_constructor): Make earlier bail out
579 2018-02-19 Carl Love <cel@us.ibm.com>
581 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
582 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
584 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
587 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
589 * config/rl78/rl78.md (movdf): New define expand.
591 2018-02-19 Martin Liska <mliska@suse.cz>
594 * doc/invoke.texi: Fix typo.
595 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
597 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
599 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
600 handle rs6000_single_float and rs6000_double_float specially for
603 2018-02-16 Jeff Law <law@redhat.com>
605 * config/rx/rx.c (add_pop_cfi_notes): New function.;
608 2018-02-16 Jakub Jelinek <jakub@redhat.com>
611 * ipa-inline.c (inline_small_functions): Fix a typo.
613 2018-02-16 Nathan Sidwell <nathan@acm.org>
615 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
617 2018-02-16 Carl Love <cel@us.ibm.com>
619 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
620 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
621 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
622 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
623 expansion to P8V_BUILTIN_VEC_FLOAT2.
625 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
627 PR rtl-optimization/70023
628 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
629 src_regno into account.
631 2018-02-16 Carl Love <cel@us.ibm.com>
633 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
634 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
635 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
636 * config/rs6000/rs6000.c: Remove case statements for
637 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
638 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
639 and P9V_BUILTIN_VEC_VINSERT4B.
640 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
641 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
642 * config/rs6000/vsx.md:
643 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
646 2018-02-16 Carl Love <cel@us.ibm.com>
648 * config/rs6000/altivec.h: Add builtin names vec_extract4b
650 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
652 * config/rs6000/rs6000-c.c: Add the definitions for
653 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
654 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
655 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
656 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
657 definition for insert4b and define insn *insert3b_internal.
658 * doc/extend.texi: Add documentation for vec_extract4b.
660 2018-02-16 Nathan Sidwell <nathan@acm.org>
662 * doc/extend.texi (Backwards Compatibility): Mention friend
663 injection. Note for-scope is deprecated.
664 * doc/invoke.texi (-ffriend-injection): Deprecate.
666 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
668 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
669 that moved to I2, also allow destinations that are a paradoxical
670 subreg (instead of a normal reg).
672 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
675 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
678 2018-02-16 Richard Biener <rguenther@suse.de>
680 PR tree-optimization/84037
681 PR tree-optimization/84016
683 * config/i386/i386.c (ix86_builtin_vectorization_cost):
684 Adjust vec_construct for the fact we need additional higher latency
685 128bit inserts for AVX256 and AVX512 vector builds.
686 (ix86_add_stmt_cost): Scale vector construction cost for
689 2018-02-16 Richard Biener <rguenther@suse.de>
691 PR tree-optimization/84417
692 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
693 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
694 (non_rewritable_lvalue_p): Likewise, use poly-ints.
696 2018-02-16 Martin Liska <mliska@suse.cz>
699 * internal-fn.def (ASAN_CHECK): Set proper flags.
700 (ASAN_MARK): Likewise.
702 2018-02-16 Julia Koval <julia.koval@intel.com>
704 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
707 2018-02-16 Jakub Jelinek <jakub@redhat.com>
710 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
711 Use ++iter rather than iter++ for std::list iterators.
712 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
713 defer deleting them until all nodes in the forest are processed. Do
714 free even leaf nodes. Change to_process into auto_vec.
717 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
718 * vec.h (vec_default_construct): Use memset instead of placement new
719 if BROKEN_VALUE_INITIALIZATION is defined.
720 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
721 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
724 PR rtl-optimization/83723
725 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
726 * lra.c (lra_substitute_pseudo): Likewise. If true, use
727 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
729 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
731 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
733 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
735 PR rtl-optimization/81443
736 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
737 from inner REGs to paradoxical SUBREGs.
739 2018-02-16 Richard Biener <rguenther@suse.de>
741 PR tree-optimization/84399
742 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
743 For operands we can analyze at their definition make sure we can
744 analyze them at each use as well.
746 2018-02-16 Richard Biener <rguenther@suse.de>
748 PR tree-optimization/84190
749 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
750 volatile accesses if the decl isn't volatile.
752 2018-02-15 Jason Merrill <jason@redhat.com>
754 PR c++/84314 - ICE with templates and fastcall attribute.
755 * attribs.c (build_type_attribute_qual_variant): Don't clobber
756 TYPE_CANONICAL on an existing type.
758 2018-02-15 Jakub Jelinek <jakub@redhat.com>
760 PR tree-optimization/84383
761 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
762 dstoff nor call operand_equal_p if dstbase is NULL.
764 PR tree-optimization/84334
765 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
766 also a CONSTANT_CLASS_P, punt.
768 2018-02-14 Jim Wilson <jimw@sifive.com>
770 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
771 first SMALL_OPERAND check. New local min_second_step. Move assert
772 to where locals are set. Add TARGET_RVC support.
773 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
775 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
777 * doc/invoke.texi: Correct -Wformat-overflow code sample.
779 2018-02-14 Martin Sebor <msebor@redhat.com>
781 PR tree-optimization/83698
782 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
783 arrays constrain the offset range to their bounds.
784 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
785 (builtin_access::overlap): Avoid setting the size of overlap if it's
787 (maybe_diag_overlap): Also consider arrays when deciding what values
788 of offsets to include in diagnostics.
790 2018-02-14 Martin Sebor <msebor@redhat.com>
793 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
794 that correspond to the kind of a declaration.
796 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
799 * config/pa/pa.md: Load address of PIC label using the linkage table
800 if the label is nonlocal.
802 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
804 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
805 warning message if user requests -maltivec=be.
806 * doc/invoke.texi: Document deprecation of -maltivec=be.
808 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
811 * config/rs6000/rs6000-c.c: Update definitions for
812 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
813 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
815 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
818 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
819 add _get_ssp intrinsics. Remove argument from
820 __builtin_ia32_rdssp[d|q].
821 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
822 * config/i386/i386-builtin.def: Remove argument from
823 __builtin_ia32_rdssp[d|q].
824 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
825 ix86_expand_special_args_builtin for _rdssp[d|q].
826 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
827 Clear register before usage.
828 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
829 Add documentation for new _get_ssp and _inc_ssp intrinsics.
831 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
833 PR tree-optimization/84357
834 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
835 operand 1 of an ARRAY_REF too.
837 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
840 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
841 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
843 (set_of_reg): New struct.
844 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
845 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
846 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
848 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
849 Split into bitclr, bitset, bitinvert patterns if appropriate.
850 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
851 use rx_fuse_in_memory_bitop.
852 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
853 to named insn, correct maximum insn length.
855 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
858 * machmode.def: Define a complex mode for PARTIAL_INT.
859 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
861 * doc/rtl.texi: Document CSPImode.
862 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
864 (msp430_hard_regno_nregs_with_padding): Likewise.
866 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
869 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
871 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
873 PR rtl-optimization/84169
874 * combine.c (try_combine): New variable split_i2i3. Set it to true if
875 we generated a parallel as new i3 and we split that to new i2 and i3
876 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
877 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
878 those to i2, not i1. Partially rewrite this scan code.
880 2018-02-13 Jakub Jelinek <jakub@redhat.com>
883 * stor-layout.c (place_field): For variable length fields, adjust
884 offset_align afterwards not just based on the field's alignment,
885 but also on the size.
888 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
889 of exps and logs in the use_exp2 case.
891 2018-02-13 Jeff Law <law@redhat.com>
893 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
896 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
899 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
903 * common.opt (gas-loc-support, gas-locview-support): New.
904 (ginline-points, ginternal-reset-location-views): New.
905 * doc/invoke.texi: Document them. Use @itemx where intended.
906 (gvariable-location-views): Adjust.
907 * target.def (reset_location_view): New.
908 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
909 (TARGET_RESET_LOCATION_VIEW): New.
910 * doc/tm.texi: Rebuilt.
911 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
912 (dwarf2out_default_as_locview_support): New.
913 (output_asm_line_debug_info): Use option variables.
914 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
915 (output_loc_list): Likewise.
916 (add_high_low_attributes): Check option variables.
917 Don't output entry view attribute in strict mode.
918 (gen_inlined_subroutine_die): Check option variables.
919 (dwarf2out_inline_entry): Likewise.
920 (init_sections_and_labels): Likewise.
921 (dwarf2out_early_finish): Likewise.
922 (maybe_reset_location_view): New, from...
923 (dwarf2out_var_location): ... here. Call it.
924 * debug.h (dwarf2out_default_as_loc_support): Declare.
925 (dwarf2out_default_as_locview_support): Declare.
926 * hooks.c (hook_int_rtx_insn_0): New.
927 * hooks.h (hook_int_rtx_insn_0): Declare.
928 * toplev.c (process_options): Take -gas-loc-support and
929 -gas-locview-support from dwarf2out. Enable
930 -gvariable-location-views by default only with locview
931 assembler support. Enable -ginternal-reset-location-views by
932 default only if the target defines the corresponding hook.
933 Enable -ginline-points by default if location views are
934 enabled; force it disabled if statement frontiers are
936 * tree-inline.c (expand_call_inline): Check option variables.
937 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
939 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
941 PR tree-optimization/84321
942 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
943 handling. Also check whether the anti-range contains any values
944 that satisfy the mask; switch to a VR_RANGE if not.
946 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
949 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
951 2018-02-13 Martin Jambor <mjambor@suse.cz>
954 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
955 of call statements, also set location of a load to a temporary.
957 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
959 * config/rl78/rl78.c (add_vector_labels): New function.
960 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
961 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
962 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
963 which checks that no arguments are passed.
964 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
965 * doc/extend.texi: Documentation for the new attribute.
967 2018-02-13 Andreas Schwab <schwab@suse.de>
969 * config/riscv/linux.h (CPP_SPEC): Define.
971 2018-02-13 Jakub Jelinek <jakub@redhat.com>
974 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
975 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
976 OPTION_MASK_ISA_AES as first argument to def_builtin_const
977 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
978 instead of OPTION_MASK_ISA_PCLMUL as first argument to
979 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
980 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
981 temporarily for AES and PCLMUL builtins.
983 PR tree-optimization/84339
984 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
985 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
989 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
990 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
991 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
993 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
995 * omp-simd-clone.h: New file.
996 * omp-simd-clone.c: Include omp-simd-clone.h.
997 (expand_simd_clones): No longer static.
998 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
999 cgraph.h and omp-simd-clone.h.
1000 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
1001 (vect_recog_widen_shift_pattern): Formatting fix.
1002 (vect_pattern_recog_1): Don't check optab for calls.
1005 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
1006 operands[2] into a REG before using gen_lowpart on it.
1008 2018-02-12 Jeff Law <law@redhat.com>
1011 * config/sh/sh.c (find_barrier): Consider a sibling call
1014 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
1015 successfully back substituting a reg.
1017 2018-02-12 Richard Biener <rguenther@suse.de>
1019 PR tree-optimization/84037
1020 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
1021 parameter, move visited init to caller.
1022 (vect_slp_analyze_operations): Separate cost from validity
1023 check, initialize visited once for all instances.
1024 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
1026 * tree-vect-stmts.c (vect_model_simple_cost): Make early
1028 (vect_model_promotion_demotion_cost): Likewise.
1029 (vectorizable_bswap): Guard cost modeling with !slp_node
1030 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
1032 (vectorizable_call): Likewise.
1033 (vectorizable_conversion): Likewise.
1034 (vectorizable_assignment): Likewise.
1035 (vectorizable_shift): Likewise.
1036 (vectorizable_operation): Likewise.
1037 (vectorizable_store): Likewise.
1038 (vectorizable_load): Likewise.
1039 (vectorizable_condition): Likewise.
1040 (vectorizable_comparison): Likewise.
1042 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
1045 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
1046 (ASAN_MARK): Fix fnspec to account for return value, change pointer
1047 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
1049 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1052 * params.def (inline-min-speedup): Increase from 8 to 15.
1053 (max-inline-insns-auto): Decrease from 40 to 30.
1054 * ipa-split.c (consider_split): Add some buffer for function to
1055 be considered inlining candidate.
1056 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
1059 2018-02-12 Richard Biener <rguenther@suse.de>
1061 PR tree-optimization/84037
1062 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
1063 matched stmts if we cannot swap the non-matched ones.
1065 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1067 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
1068 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
1069 _mm_maskz_scalef_round_ss): New intrinsics.
1070 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
1071 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
1072 __builtin_ia32_scalefss_round): Remove.
1073 (__builtin_ia32_scalefsd_mask_round,
1074 __builtin_ia32_scalefss_mask_round): New intrinsics.
1075 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
1076 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
1077 ((match_operand:VF_128 2 "<round_nimm_predicate>"
1078 "<round_constraint>")): Changed to ...
1079 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
1080 "<round_scalar_constraint>")): ... this.
1081 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
1082 %0, %1, %2<round_op3>}"): Changed to ...
1083 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
1084 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
1085 %2<round_scalar_mask_op3>}"): ... this.
1086 * config/i386/subst.md (round_scalar_nimm_predicate): New.
1088 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1090 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
1091 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
1092 (_mm_maskz_sqrt_round_ss): New intrinsics.
1093 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
1094 (__builtin_ia32_sqrtsd_mask_round)
1095 (__builtin_ia32_sqrtss_mask_round): New builtins.
1096 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
1097 (__builtin_ia32_sqrtss_round): Remove.
1098 (__builtin_ia32_sqrtsd_mask_round)
1099 (__builtin_ia32_sqrtss_mask_round): New builtins.
1100 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
1101 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
1102 ((match_operand:VF_128 1 "vector_operand"
1103 "xBm,<round_constraint>")): Changed to ...
1104 ((match_operand:VF_128 1 "vector_operand"
1105 "xBm,<round_scalar_constraint>")): ... this.
1106 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
1107 %0, %2, %<iptr>1<round_op3>}): Changed to ...
1108 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
1109 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
1110 %<iptr>1<round_scalar_mask_op3>}): ... this.
1111 ((set_attr "prefix" "<round_prefix>")): Changed to ...
1112 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
1114 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
1117 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
1118 Cast vec_cmpeq result to correct type.
1119 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
1120 Cast vec_cmpgt result to correct type.
1122 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
1124 * final.c (final_scan_insn_1): Renamed from...
1125 (final_scan_insn): ... this. New wrapper, to recover
1126 seen from the outermost call in recursive ones.
1127 * config/sparc/sparc.c (output_return): Drop seen from call.
1128 (output_sibcall): Likewise.
1129 * config/visium/visium.c (output_branch): Likewise.
1131 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
1133 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
1136 2018-02-10 Alan Modra <amodra@gmail.com>
1139 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
1140 Specify LR as an input.
1142 2018-02-10 Jakub Jelinek <jakub@redhat.com>
1145 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
1146 remove_member_access_dummy_vars): New functions.
1147 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
1148 lower_omp_1, execute_lower_omp): Use them.
1150 PR rtl-optimization/84308
1151 * shrink-wrap.c (spread_components): Release todo vector.
1153 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
1155 PR rtl-optimization/57193
1156 * ira-color.c (struct allocno_color_data): Add member
1157 conflict_allocno_hard_prefs.
1158 (update_conflict_allocno_hard_prefs): New.
1159 (bucket_allocno_compare_func): Add a preference based on
1160 conflict_allocno_hard_prefs.
1161 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
1162 (color_allocnos): Remove a dead code. Initiate
1163 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
1165 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1168 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
1169 constraint from =wa to wa. Avoid a subreg on the output operand,
1170 instead use a pseudo and subreg it in a move.
1171 (p9_xxbrd_<mode>): Changed to ...
1172 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
1173 (p9_xxbrd_v2df): New expander.
1174 (p9_xxbrw_<mode>): Changed to ...
1175 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
1176 (p9_xxbrw_v4sf): New expander.
1178 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1180 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
1182 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
1185 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
1186 multiply in 32-bit mode.
1187 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
1188 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
1191 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1193 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
1194 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
1195 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
1196 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
1198 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
1201 * dwarf2out.c (is_trivial_indirect_ref): New function.
1202 (dwarf2out_late_global_decl): Do not generate a location
1203 attribute for variables that have a non-trivial DECL_VALUE_EXPR
1204 and that are not defined in the current unit.
1206 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1208 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
1209 instead of a libcall for UNORDERED.
1211 2018-02-09 Tamar Christina <tamar.christina@arm.com>
1214 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
1215 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
1217 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1220 * config/s390/s390.c (s390_set_current_function): Invoke
1221 s390_indirect_branch_settings also if fndecl didn't change.
1223 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1225 * config/rs6000/rs6000.md (blockage): Set length to zero.
1227 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1229 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
1231 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1234 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
1235 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
1239 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
1240 PARALLEL incoming that failed vt_get_decl_and_offset check.
1243 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
1244 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
1246 (get_variable_section): For decls in named .bss* sections pass true as
1247 second argument to bss_initializer_p.
1249 2018-02-09 Marek Polacek <polacek@redhat.com>
1250 Jakub Jelinek <jakub@redhat.com>
1253 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
1254 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
1255 Sync some changes from cxx_fold_indirect_ref.
1257 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1259 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
1261 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
1262 (BLOCK_INLINE_ENTRY_LABEL): New.
1263 (dwarf2out_var_location): Disregard inline entry markers.
1264 (inline_entry_data): New struct.
1265 (inline_entry_data_hasher): New hashtable type.
1266 (inline_entry_data_hasher::hash): New.
1267 (inline_entry_data_hasher::equal): New.
1268 (inline_entry_data_table): New variable.
1269 (add_high_low_attributes): Add DW_AT_entry_pc and
1270 DW_AT_GNU_entry_view attributes if a pending entry is found
1271 in inline_entry_data_table. Add old entry_pc attribute only
1272 if debug nonbinding markers are disabled.
1273 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
1274 markers are enabled.
1275 (block_within_block_p, dwarf2out_inline_entry): New.
1276 (dwarf2out_finish): Check that no entries remained in
1277 inline_entry_data_table.
1278 * final.c (reemit_insn_block_notes): Handle inline entry notes.
1279 (final_scan_insn, notice_source_line): Likewise.
1280 (rest_of_clean_state): Skip inline entry markers.
1281 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
1283 * gimple.c (gimple_build_debug_inline_entry): New.
1284 * gimple.h (enum gimple_debug_subcode): Add
1285 GIMPLE_DEBUG_INLINE_ENTRY.
1286 (gimple_build_debug_inline_entry): Declare.
1287 (gimple_debug_inline_entry_p): New.
1288 (gimple_debug_nonbind_marker_p): Adjust.
1289 * insn-notes.def (INLINE_ENTRY): New.
1290 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
1291 inline entry marker notes.
1292 (print_insn): Likewise.
1293 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
1294 (INSN_DEBUG_MARKER_KIND): Likewise.
1295 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
1296 * tree-inline.c (expand_call_inline): Build and insert
1297 debug_inline_entry stmt.
1298 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
1299 inline entry blocks early, if nonbind markers are enabled.
1300 (dump_scope_block): Dump fragment info.
1301 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
1302 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
1303 (gimple_build_debug_inline_entry): New.
1304 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
1305 Enable/disable inline entry points too.
1306 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
1307 (DEBUG_INSN): Describe inline entry markers.
1309 * common.opt (gvariable-location-views): New.
1310 (gvariable-location-views=incompat5): New.
1311 * config.in: Rebuilt.
1312 * configure: Rebuilt.
1313 * configure.ac: Test assembler for view support.
1314 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
1315 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
1316 * dwarf2out.c (var_loc_view): New typedef.
1317 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
1318 (dwarf2out_locviews_in_attribute): New.
1319 (dwarf2out_locviews_in_loclist): New.
1320 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
1321 (enum dw_line_info_opcode): Add LI_adv_address.
1322 (struct dw_line_info_table): Add view.
1323 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
1324 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
1325 (zero_view_p): New variable.
1326 (ZERO_VIEW_P): New macro.
1327 (output_asm_line_debug_info): New.
1328 (struct var_loc_node): Add view.
1329 (add_AT_view_list, AT_loc_list): New.
1330 (add_var_loc_to_decl): Add view param. Test it against last.
1331 (new_loc_list): Add view params. Record them.
1332 (AT_loc_list_ptr): Handle loc and view lists.
1333 (view_list_to_loc_list_val_node): New.
1334 (print_dw_val): Handle dw_val_class_view_list.
1335 (size_of_die): Likewise.
1336 (value_format): Likewise.
1337 (loc_list_has_views): New.
1338 (gen_llsym): Set vl_symbol too.
1339 (maybe_gen_llsym, skip_loc_list_entry): New.
1340 (dwarf2out_maybe_output_loclist_view_pair): New.
1341 (output_loc_list): Output view list or entries too.
1342 (output_view_list_offset): New.
1343 (output_die): Handle dw_val_class_view_list.
1344 (output_dwarf_version): New.
1345 (output_compilation_unit_header): Use it.
1346 (output_skeleton_debug_sections): Likewise.
1347 (output_rnglists, output_line_info): Likewise.
1348 (output_pubnames, output_aranges): Update version comments.
1349 (output_one_line_info_table): Output view numbers in asm comments.
1350 (dw_loc_list): Determine current endview, pass it to new_loc_list.
1351 Call maybe_gen_llsym.
1352 (loc_list_from_tree_1): Adjust.
1353 (add_AT_location_description): Create view list attribute if
1354 needed, check it's absent otherwise.
1355 (convert_cfa_to_fb_loc_list): Adjust.
1356 (maybe_emit_file): Call output_asm_line_debug_info for test.
1357 (dwarf2out_var_location): Reset views as needed. Precompute
1358 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
1359 attribute. Set view.
1360 (new_line_info_table): Reset next view.
1361 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
1362 (dwarf2out_source_line): Likewise. Output view resets and labels to
1363 the assembler, or select appropriate line info opcodes.
1364 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
1365 (optimize_string_length): Catch it. Adjust.
1366 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
1367 dw_val_class_view_list, and remove it if no longer needed.
1368 (hash_loc_list): Hash view numbers.
1369 (loc_list_hasher::equal): Compare them.
1370 (optimize_location_lists): Check whether a view list symbol is
1371 needed, and whether the locview attribute is present, and
1372 whether they match. Remove the locview attribute if no longer
1374 (index_location_lists): Call skip_loc_list_entry for test.
1375 (dwarf2out_finish): Call output_asm_line_debug_info for test.
1376 Use output_dwarf_version.
1377 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
1378 (struct dw_val_node): Add val_view_list.
1379 * final.c (SEEN_NEXT_VIEW): New.
1380 (set_next_view_needed): New.
1381 (clear_next_view_needed): New.
1382 (maybe_output_next_view): New.
1383 (final_start_function): Rename to...
1384 (final_start_function_1): ... this. Take pointer to FIRST,
1385 add SEEN parameter. Emit param bindings in the initial view.
1386 (final_start_function): Reintroduce SEEN-less interface.
1387 (final): Rename to...
1388 (final_1): ... this. Take SEEN parameter. Output final pending
1389 next view at the end.
1390 (final): Reintroduce seen-less interface.
1391 (final_scan_insn): Output pending next view before switching
1392 sections or ending a block. Mark the next view as needed when
1393 outputting variable locations. Notify debug backend of section
1394 changes, and of location view changes.
1395 (rest_of_handle_final): Adjust.
1396 * toplev.c (process_options): Autodetect value for debug variable
1397 location views option. Warn on incompat5 without -gdwarf-5.
1398 * doc/invoke.texi (gvariable-location-views): New.
1399 (gvariable-location-views=incompat5): New.
1400 (gno-variable-location-views): New.
1402 2018-02-08 David Malcolm <dmalcolm@redhat.com>
1404 PR tree-optimization/84136
1405 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
1406 that the result of find_edge is non-NULL.
1408 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1411 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
1412 storing integer register in SImode. Fix cost of 256 and 512
1413 byte aligned SSE register store.
1415 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1417 * config/i386/i386.c (ix86_multiplication_cost): Fix
1418 multiplication cost for TARGET_AVX512DQ.
1420 2018-02-08 Marek Polacek <polacek@redhat.com>
1422 PR tree-optimization/84238
1423 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
1426 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1428 PR tree-optimization/84265
1429 * tree-vect-stmts.c (vectorizable_store): Don't treat
1430 VMAT_CONTIGUOUS accesses as grouped.
1431 (vectorizable_load): Likewise.
1433 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1435 PR tree-optimization/81635
1436 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
1437 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
1438 (test_round_for_mask): New functions.
1439 (wide_int_cc_tests): Call test_round_for_mask.
1440 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
1441 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
1442 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
1443 range returned by get_range_info.
1445 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1448 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
1449 * symtab.c: Include builtins.h
1450 (symtab_node::output_to_lto_symbol_table_p): Move here
1451 from lto-streamer-out.c:output_symbol_p.
1452 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
1453 (output_symbol_p): Move all logic to symtab.c
1454 (produce_symtab): Update.
1456 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1458 * config/s390/s390-opts.h (enum indirect_branch): Define.
1459 * config/s390/s390-protos.h (s390_return_addr_from_memory)
1460 (s390_indirect_branch_via_thunk)
1461 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
1462 (enum s390_indirect_branch_type): Define.
1463 * config/s390/s390.c (struct s390_frame_layout, struct
1464 machine_function): Remove.
1465 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
1466 (indirect_branch_table_label_no, indirect_branch_table_name):
1468 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
1469 (enum s390_indirect_branch_option): Define.
1470 (s390_return_addr_from_memory): New function.
1471 (s390_handle_string_attribute): New function.
1472 (s390_attribute_table): Add new attribute handler.
1473 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
1474 (s390_indirect_branch_via_thunk): New function.
1475 (s390_indirect_branch_via_inline_thunk): New function.
1476 (s390_function_ok_for_sibcall): When jumping via thunk disallow
1477 sibling call optimization for non z10 compiles.
1478 (s390_emit_call): Force indirect branch target to be a single
1479 register. Add r1 clobber for non-z10 compiles.
1480 (s390_emit_epilogue): Emit return jump via return_use expander.
1481 (s390_reorg): Handle JUMP_INSNs as execute targets.
1482 (s390_option_override_internal): Perform validity checks for the
1483 new command line options.
1484 (s390_indirect_branch_attrvalue): New function.
1485 (s390_indirect_branch_settings): New function.
1486 (s390_set_current_function): Invoke s390_indirect_branch_settings.
1487 (s390_output_indirect_thunk_function): New function.
1488 (s390_code_end): Implement target hook.
1489 (s390_case_values_threshold): Implement target hook.
1490 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
1492 * config/s390/s390.h (struct s390_frame_layout)
1493 (struct machine_function): Move here from s390.c.
1494 (TARGET_INDIRECT_BRANCH_NOBP_RET)
1495 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
1496 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
1497 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
1498 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
1499 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
1500 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
1501 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
1502 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
1503 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
1504 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
1505 (mnemonic attribute): Add values which aren't recognized
1507 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
1508 pattern for branch conversion. Fix mnemonic attribute.
1509 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
1510 indirect branch via thunk if requested.
1511 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
1512 ("*indirect_jump"): Disable for branch conversion using out of
1514 ("indirect_jump_via_thunk<mode>_z10")
1515 ("indirect_jump_via_thunk<mode>")
1516 ("indirect_jump_via_inlinethunk<mode>_z10")
1517 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
1518 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
1519 ("casesi_jump_via_inlinethunk<mode>_z10")
1520 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
1521 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
1522 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
1523 ("*indirect2_jump"): Disable for branch conversion.
1524 ("casesi_jump"): Turn into expander and expand patterns for branch
1526 ("return_use"): New expander.
1527 ("*return"): Emit return via thunk and rename it to ...
1528 ("*return<mode>"): ... this one.
1529 * config/s390/s390.opt: Add new options and and enum for the
1532 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1534 * lra-constraints.c (match_reload): Unconditionally use
1535 gen_lowpart_SUBREG, rather than selecting between that
1536 and equivalent gen_rtx_SUBREG code.
1538 2018-02-08 Richard Biener <rguenther@suse.de>
1540 PR tree-optimization/84233
1541 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
1542 changed flag instead of boguously re-using phi_inserted.
1544 2018-02-08 Martin Jambor <mjambor@suse.cz>
1546 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
1547 static local variables.
1549 2018-02-08 Richard Biener <rguenther@suse.de>
1551 PR tree-optimization/84278
1552 * tree-vect-stmts.c (vectorizable_store): When looking for
1553 smaller vector types to perform grouped strided loads/stores
1554 make sure the mode is supported by the target.
1555 (vectorizable_load): Likewise.
1557 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1559 * config/aarch64/aarch64.c (aarch64_components_for_bb):
1560 Increase LDP/STP opportunities by adding adjacent callee-saves.
1562 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1564 PR rtl-optimization/84068
1565 PR rtl-optimization/83459
1566 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
1568 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
1570 PR tree-optimization/84224
1571 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
1572 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
1575 2018-02-07 Iain Sandoe <iain@codesourcery.com>
1578 * config/rs6000/altivec.md (*restore_world): Remove LR use.
1579 * config/rs6000/predicates.md (restore_world_operation): Adjust op
1580 count, remove one USE.
1582 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
1584 * doc/install.texi (Configuration): Document the
1585 --with-long-double-format={ibm,ieee} PowerPC configuration
1589 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
1590 Convert from define_expand to be define_insn_and_split. Rework
1591 float/double/_Float128 conversions to QI/HI/SImode to work with
1592 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
1593 conversions to QI/HImode types did a store and then a load to
1594 truncate the value. For conversions to VSX registers, don't split
1595 the insn, instead emit the code directly. Use the code iterator
1596 any_fix to combine signed and unsigned conversions.
1597 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
1598 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
1599 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
1600 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
1601 (fix_<mode>di2_hw): Likewise.
1602 (fixuns_<mode>di2_hw): Likewise.
1603 (fix_<mode>si2_hw): Likewise.
1604 (fixuns_<mode>si2_hw): Likewise.
1605 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
1606 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
1607 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
1608 fix<uns>_trunc<SFDF:mode>si2_p8.
1609 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
1611 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
1612 (fix<uns>_<mode>_mem): Likewise.
1613 (fctiw<u>z_<mode>_mem): Likewise.
1614 (fix<uns>_<mode>_mem): Likewise.
1615 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
1616 the register allocator from doing a direct move to the GPRs to do
1617 a store, and instead use the ISA 3.0 store byte/half-word from
1618 vector register instruction. For IEEE 128-bit floating point,
1619 also optimize stores of 32-bit ints.
1620 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
1622 2018-02-07 Alan Hayward <alan.hayward@arm.com>
1624 * genextract.c (push_pathstr_operand): New function to support
1626 (walk_rtx): Call push_pathstr_operand.
1627 (print_path): Support [a-zA-Z].
1629 2018-02-07 Richard Biener <rguenther@suse.de>
1631 PR tree-optimization/84037
1632 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
1633 (cse_and_gimplify_to_preheader): Declare.
1634 (vect_get_place_in_interleaving_chain): Likewise.
1635 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
1637 (_loop_vec_info::~_loop_vec_info): Delete it.
1638 (cse_and_gimplify_to_preheader): New function.
1639 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
1640 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
1641 (vectorizable_load): Likewise. For grouped stores always base
1642 the IV on the first element.
1643 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
1644 condition before gimplifying.
1646 2018-02-07 Jakub Jelinek <jakub@redhat.com>
1648 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
1649 *DIV_EXPR and *MOD_EXPR.
1651 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
1654 * config/i386/i386.c (ix86_option_override_internal): Mask out
1655 the CF_SET bit when checking -fcf-protection.
1657 2018-02-07 Tom de Vries <tom@codesourcery.com>
1660 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
1663 2018-02-07 Richard Biener <rguenther@suse.de>
1665 PR tree-optimization/84204
1666 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
1669 PR tree-optimization/84205
1670 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
1671 special-case isl_ast_op_zdiv_r.
1673 PR tree-optimization/84223
1674 * graphite-scop-detection.c (gather_bbs::before_dom_children):
1675 Only add conditions from within the region.
1676 (gather_bbs::after_dom_children): Adjust.
1678 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
1681 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
1682 * config/avr/avr.md: Only post-reload split REG-REG moves if
1683 either register is REGERAL_REG_P.
1685 2018-02-07 Jakub Jelinek <jakub@redhat.com>
1687 PR tree-optimization/84235
1688 * tree-ssa-scopedtables.c
1689 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
1690 if the subtraction is performed in floating point type where NaNs are
1691 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
1692 build 1. Formatting fix.
1694 2018-02-06 Jakub Jelinek <jakub@redhat.com>
1697 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
1698 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
1699 and skip it regardless of bb boundaries. Use CALL_P macro,
1700 don't test INSN_P (insn) together with CALL_P or JUMP_P check
1701 unnecessarily, formatting fix.
1703 2018-02-06 Michael Collison <michael.collison@arm.com>
1705 * config/arm/thumb2.md:
1706 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
1707 (*thumb_mov_notscc): Ditto.
1709 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
1712 * config/rs6000/rs6000.md (su code attribute): Use "u" for
1713 unsigned_fix, not "s".
1715 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1717 * configure.ac (gcc_fn_eh_frame_ro): New function.
1718 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
1719 correct .eh_frame permissions.
1720 * configure: Regenerate.
1722 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
1724 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
1727 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1729 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1730 Display warning message for -mno-speculate-indirect-jumps.
1732 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
1734 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
1736 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
1738 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
1740 PR tree-optimization/84225
1741 * tree-eh.c (find_trapping_overflow): Only call
1742 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
1744 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1747 * config/i386/i386.c: Reimplement the check of possible options
1748 -mibt/-mshstk conbination. Change error messages.
1749 * doc/invoke.texi: Fix a typo: remove extra '='.
1751 2018-02-06 Marek Polacek <polacek@redhat.com>
1753 PR tree-optimization/84228
1754 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
1756 2018-02-06 Tamar Christina <tamar.christina@arm.com>
1759 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
1760 emitted arch directives.
1761 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
1762 __ARM_FEATURE_COPROC before changing architectures.
1764 2018-02-06 Richard Biener <rguenther@suse.de>
1766 * config/i386/i386.c (print_reg): Fix typo.
1767 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
1769 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
1771 * configure: Regenerate.
1773 2018-02-05 Martin Sebor <msebor@redhat.com>
1775 PR tree-optimization/83369
1776 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
1779 2018-02-05 Martin Liska <mliska@suse.cz>
1781 * doc/invoke.texi: Cherry-pick upstream r323995.
1783 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
1785 * ira.c (ira_init_register_move_cost): Adjust comment.
1787 2018-02-05 Martin Liska <mliska@suse.cz>
1789 PR gcov-profile/84137
1790 * doc/gcov.texi: Fix typo in documentation.
1792 2018-02-05 Martin Liska <mliska@suse.cz>
1794 PR gcov-profile/83879
1795 * doc/gcov.texi: Document necessity of --dynamic-list-data when
1796 using dlopen functionality.
1798 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
1800 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
1801 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
1802 _mm_maskz_range_ss, _mm_mask_range_round_ss,
1803 _mm_maskz_range_round_ss): New intrinsics.
1804 (__builtin_ia32_rangesd128_round)
1805 (__builtin_ia32_rangess128_round): Remove.
1806 (__builtin_ia32_rangesd128_mask_round,
1807 __builtin_ia32_rangess128_mask_round): New builtins.
1808 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
1809 __builtin_ia32_rangess128_round): Remove.
1810 (__builtin_ia32_rangesd128_mask_round,
1811 __builtin_ia32_rangess128_mask_round): New builtins.
1812 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
1813 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
1814 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
1815 "<round_saeonly_constraint>")): Changed to ...
1816 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
1817 "<round_saeonly_scalar_constraint>")): ... this.
1818 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
1819 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
1820 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
1821 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
1822 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
1824 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
1826 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
1828 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
1829 Remove all values except native, 8540 and 8548.
1831 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
1833 * config/i386/i386.c (ix86_output_function_return): Pass
1834 INVALID_REGNUM, instead of -1, as invalid register number to
1835 indirect_thunk_name and output_indirect_thunk.
1837 2018-02-02 Julia Koval <julia.koval@intel.com>
1839 * config.gcc: Add -march=icelake.
1840 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
1841 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
1842 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
1843 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
1844 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
1845 (processor_target_table): Add icelake.
1846 (ix86_option_override_internal): Handle new PTAs.
1847 (get_builtin_code_for_version): Handle icelake.
1848 (M_INTEL_COREI7_ICELAKE): New.
1849 (fold_builtin_cpu): Handle icelake.
1850 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
1851 * doc/invoke.texi: Add -march=icelake.
1853 2018-02-02 Julia Koval <julia.koval@intel.com>
1855 * config/i386/i386.c (ix86_option_override_internal): Change flags type
1856 to wide_int_bitmask.
1857 * wide-int-bitmask.h: New.
1859 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1862 * config/i386/i386.md: Replace Pmode with word_mode in
1863 builtin_setjmp_setup and builtin_longjmp to support x32.
1865 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
1869 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
1871 (rs6000_supported_cpu_names): New static variable.
1872 (linux_cpu_translation_table): Likewise.
1873 (elf_platform) <cpu>: Define new static variable and use it.
1874 Translate kernel AT_PLATFORM name to canonical name if needed.
1875 Error if platform name is unknown.
1877 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
1880 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
1882 2018-02-01 Jeff Law <law@redhat.com>
1885 * config/i386/i386.c (release_scratch_register_on_entry): Add new
1886 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
1887 the scratch if RELEASE_VIA_POP is false.
1888 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
1889 If we have to save a temporary register, decrement SIZE appropriately.
1890 Pass new arguments to release_scratch_register_on_entry.
1891 (ix86_adjust_stack_and_probe): Likewise.
1892 (ix86_emit_probe_stack_range): Pass new arguments to
1893 release_scratch_register_on_entry.
1895 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
1897 PR rtl-optimization/84157
1898 * combine.c (change_zero_ext): Use REG_P predicate in
1899 front of HARD_REGISTER_P predicate.
1901 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
1903 * config/avr/avr.c (avr_option_override): Move disabling of
1904 -fdelete-null-pointer-checks to...
1905 * common/config/avr/avr-common.c (avr_option_optimization_table):
1908 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1910 PR tree-optimization/81635
1911 * tree-data-ref.c (split_constant_offset_1): For types that
1912 wrap on overflow, try to use range info to prove that wrapping
1915 2018-02-01 Renlin Li <renlin.li@arm.com>
1918 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
1920 (aarch64_register_move_cost): Likewise.
1921 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
1923 (REG_CLASS_NAMES): Likewise.
1924 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
1925 TAILCALL_ADDR_REGS. Remove IP registers.
1926 * config/aarch64/aarch64.md (Ucs): Update register constraint.
1928 2018-02-01 Richard Biener <rguenther@suse.de>
1930 * domwalk.h (dom_walker::dom_walker): Add additional constructor
1931 for specifying RPO order and allow NULL for that.
1932 * domwalk.c (dom_walker::dom_walker): Likewise.
1933 (dom_walker::walk): Handle NULL RPO order.
1934 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
1936 (rewrite_update_dom_walker): Likewise.
1937 (mark_def_dom_walker): Likewise.
1939 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1941 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
1942 (aarch64_maybe_expand_sve_subreg_move): Declare.
1943 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
1944 * config/aarch64/predicates.md (aarch64_any_register_operand): New
1946 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
1947 that are semantically a reverse operation.
1948 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
1949 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
1950 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
1952 (aarch64_can_change_mode_class): For big-endian, forbid changes
1953 between two SVE modes if they have different element sizes.
1955 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1957 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
1958 the TImode handling for big-endian targets.
1960 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1962 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
1963 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
1965 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
1966 Remove BSWAP handing for big-endian targets and use the form of
1967 LD1RQ appropariate for the mode.
1969 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1971 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
1972 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
1975 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1978 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
1979 check for operands that need to go through aarch64_sve_reload_be.
1981 2018-02-01 Jakub Jelinek <jakub@redhat.com>
1983 PR tree-optimization/81661
1984 PR tree-optimization/84117
1985 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
1986 * tree-eh.c: Include gimplify.h.
1987 (find_trapping_overflow, replace_trapping_overflow,
1988 rewrite_to_non_trapping_overflow): New functions.
1989 * tree-vect-loop.c: Include tree-eh.h.
1990 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
1991 * tree-data-ref.c: Include tree-eh.h.
1992 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
1994 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
1996 PR rtl-optimization/84123
1997 * combine.c (change_zero_ext): Check if hard register satisfies
1998 can_change_dest_mode before calling gen_lowpart_SUBREG.
2000 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
2003 * ira.c (ira_init_register_move_cost): Remove assert.
2005 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2007 PR rtl-optimization/84071
2008 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
2009 * doc/tm.texi: Regenerate.
2011 2018-01-31 Richard Biener <rguenther@suse.de>
2013 PR tree-optimization/84132
2014 * tree-data-ref.c (analyze_miv_subscript): Properly
2015 check whether evolution_function_is_affine_multivariate_p
2016 before calling gcd_of_steps_may_divide_p.
2018 2018-01-31 Julia Koval <julia.koval@intel.com>
2021 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
2022 * config/i386/i386.md (rdpid_rex64) New.
2023 (rdpid): Make 32bit only.
2025 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
2028 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
2029 an IDENTIFIER_NODE for FUNCTION_TYPE's.
2031 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2034 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
2036 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
2038 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2040 PR rtl-optimization/84071
2041 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
2042 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
2044 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2046 * config/arc/arc.c (arc_handle_aux_attribute): New function.
2047 (arc_attribute_table): Add 'aux' attribute.
2048 (arc_in_small_data_p): Consider aux like variables.
2049 (arc_is_aux_reg_p): New function.
2050 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
2051 (arc_get_aux_arg): New function.
2052 (prepare_move_operands): Handle aux-register access.
2053 (arc_handle_aux_attribute): New function.
2054 * doc/extend.texi (ARC Variable attributes): Add subsection.
2056 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2058 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
2059 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
2060 (arc_attribute_table): Add 'uncached' attribute.
2061 (arc_print_operand): Print '.di' flag for uncached memory
2063 (arc_in_small_data_p): Do not consider for small data the uncached
2065 (arc_is_uncached_mem_p): New function.
2066 * config/arc/predicates.md (compact_store_memory_operand): Check
2067 for uncached memory accesses.
2068 (nonvol_nonimm_operand): Likewise.
2069 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
2071 2018-01-31 Jakub Jelinek <jakub@redhat.com>
2074 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
2075 falign-loops=): Add Optimization flag.
2077 2018-01-30 Jeff Law <law@redhat.com>
2080 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
2081 INT_REGISTERS_SAVED. Check it prior to calling
2082 get_scratch_register_on_entry.
2083 (ix86_adjust_stack_and_probe): Similarly.
2084 (ix86_emit_probe_stack_range): Similarly.
2085 (ix86_expand_prologue): Corresponding changes.
2087 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2090 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
2091 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
2093 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
2096 * lra-constraints.c (curr_insn_transform): Process AND in the
2099 2018-01-30 Jakub Jelinek <jakub@redhat.com>
2101 PR rtl-optimization/83986
2102 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
2103 dependence against last_pending_memory_flush in addition to
2106 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
2108 PR tree-optimization/81611
2109 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
2112 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2115 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
2118 2018-01-30 Richard Biener <rguenther@suse.de>
2119 Jakub Jelinek <jakub@redhat.com>
2121 PR tree-optimization/84111
2122 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
2123 inner loops added during recursion, as they don't have up-to-date
2126 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2129 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
2130 (can_inline_edge_by_limits_p): ... here.
2131 (can_early_inline_edge_p, check_callers,
2132 update_caller_keys, update_callee_keys, recursive_inlining,
2133 add_new_edges_to_heap, speculation_useful_p,
2134 inline_small_functions,
2135 inline_small_functions, flatten_function,
2136 inline_to_all_callers_1): Update.
2138 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2140 * profile-count.c (profile_count::combine_with_ipa_count): Handle
2143 2018-01-30 Richard Biener <rguenther@suse.de>
2145 PR tree-optimization/83008
2146 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
2147 invariant and constant vector uses in stmts when they need
2150 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2153 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
2154 * configure: Regenerate.
2156 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2158 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
2160 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
2161 Use gen_rtx_REG rather than gen_lowpart.
2163 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2165 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
2166 rather than 0 when creating partial subregs.
2168 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2170 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
2173 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
2176 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
2177 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
2178 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
2179 flags. This restores the settings used before the 2017-07-24.
2180 Turning off pre increment/decrement/modify allows IVOPTS to
2181 optimize DF/SF loops where the index is an int.
2183 2018-01-29 Richard Biener <rguenther@suse.de>
2184 Kelvin Nilsen <kelvin@gcc.gnu.org>
2187 * tree-vect-stmts.c (vectorizable_call): Don't call
2188 targetm.vectorize_builtin_md_vectorized_function if callee is
2191 2018-01-22 Carl Love <cel@us.ibm.com>
2193 * doc/extend.tex: Fix typo in second arg in
2194 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
2196 2018-01-29 Richard Biener <rguenther@suse.de>
2198 PR tree-optimization/84086
2199 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
2200 (flush_ssaname_freelist): When SSA names were released reset
2201 the SCEV hash table.
2203 2018-01-29 Richard Biener <rguenther@suse.de>
2205 PR tree-optimization/84057
2206 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
2207 removed paths when removing edges.
2209 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
2211 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
2212 -mfunction-return=@var{choice}.
2214 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
2217 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
2219 (layout::print_source_line): Likewise.
2220 (test_get_line_width_without_trailing_whitespace): Add test cases.
2222 2018-01-27 Jakub Jelinek <jakub@redhat.com>
2225 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
2228 2018-01-26 Jim Wilson <jimw@sifive.com>
2230 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
2232 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
2235 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2237 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
2238 and CMP + SUB-immediate -> SUBS.
2240 2018-01-26 Martin Sebor <msebor@redhat.com>
2242 PR tree-optimization/83896
2243 * tree-ssa-strlen.c (get_string_len): Rename...
2244 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
2245 Avoid assuming length is constant.
2246 (handle_char_store): Use HOST_WIDE_INT for string length.
2248 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
2251 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
2252 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
2254 2018-01-26 Richard Biener <rguenther@suse.de>
2256 PR rtl-optimization/84003
2257 * dse.c (record_store): Only record redundant stores when
2258 the earlier store aliases at least all accesses the later one does.
2260 2018-01-26 Jakub Jelinek <jakub@redhat.com>
2262 PR rtl-optimization/83985
2263 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
2264 REG_CFA_RESTORE insns.
2265 (delete_unmarked_insns): Don't ignore separate shrink wrapping
2266 REG_CFA_RESTORE insns here.
2269 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
2270 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
2272 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2274 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
2275 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
2276 (arc_init): Likewise.
2277 (arc_override_options): Likewise.
2278 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
2280 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
2282 * config/arc/arc.h (TARGET_DBNZ): Define.
2283 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
2284 properly set the tune attribute.
2285 (dbnz): Use TARGET_DBNZ guard.
2286 * config/arc/arc.opt (mtune): Add core3 option.
2288 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2290 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
2291 recognize new pic like addresses.
2292 (arc_delegitimize_address): Clean up.
2294 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2296 * config/arc/arc-arches.def: Option mrf16 valid for all
2298 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
2299 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
2300 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
2301 * config/arc/arc-tables.opt: Regenerate.
2302 * config/arc/arc.c (arc_conditional_register_usage): Handle
2303 reduced register file case.
2304 (arc_file_start): Set must have build attributes.
2305 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
2307 * config/arc/arc.opt (mrf16): Add new option.
2308 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
2309 * config/arc/genmultilib.awk: Handle new mrf16 option.
2310 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
2311 * config/arc/t-multilib: Regenerate.
2312 * doc/invoke.texi (ARC Options): Document mrf16 option.
2314 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2316 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
2317 * config/arc/arc.c (arc_handle_secure_attribute): New function.
2318 (arc_attribute_table): Add 'secure_call' attribute.
2319 (arc_print_operand): Print secure call operand.
2320 (arc_function_ok_for_sibcall): Don't optimize tail calls when
2322 (arc_is_secure_call_p): New function. * config/arc/arc.md
2323 (call_i): Add support for sjli instruction.
2324 (call_value_i): Likewise.
2325 * config/arc/constraints.md (Csc): New constraint.
2327 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2328 John Eric Martin <John.Martin@emmicro-us.com>
2330 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
2331 * config/arc/arc.c (_arc_jli_section): New struct.
2332 (arc_jli_section): New type.
2333 (rc_jli_sections): New static variable.
2334 (arc_handle_jli_attribute): New function.
2335 (arc_attribute_table): Add jli_always and jli_fixed attribute.
2336 (arc_file_end): New function.
2337 (TARGET_ASM_FILE_END): Define.
2338 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
2339 (arc_add_jli_section): New function.
2340 (jli_call_scan): Likewise.
2341 (arc_reorg): Call jli_call_scan.
2342 (arc_output_addsi): Remove 'S' from printing asm operand.
2343 (arc_is_jli_call_p): New function.
2344 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
2346 (movhi_insn): Likewise.
2347 (movsi_insn): Likewise.
2348 (movsi_set_cc_insn): Likewise.
2349 (loadqi_update): Likewise.
2350 (load_zeroextendqisi_update): Likewise.
2351 (load_signextendqisi_update): Likewise.
2352 (loadhi_update): Likewise.
2353 (load_zeroextendhisi_update): Likewise.
2354 (load_signextendhisi_update): Likewise.
2355 (loadsi_update): Likewise.
2356 (loadsf_update): Likewise.
2357 (movsicc_insn): Likewise.
2358 (bset_insn): Likewise.
2359 (bxor_insn): Likewise.
2360 (bclr_insn): Likewise.
2361 (bmsk_insn): Likewise.
2362 (bicsi3_insn): Likewise.
2363 (cmpsi_cc_c_insn): Likewise.
2364 (movsi_ne): Likewise.
2365 (movsi_cond_exec): Likewise.
2366 (clrsbsi2): Likewise.
2377 (call_i): Remove 'S' asm letter, add jli instruction.
2378 (call_value_i): Likewise.
2379 * config/arc/arc.op (mjli-always): New option.
2380 * config/arc/constraints.md (Cji): New constraint.
2381 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
2383 (subsf3_fpx): Likewise.
2384 (mulsf3_fpx): Likewise.
2385 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
2387 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
2389 * doc/invoke.texi (ARC): Document mjli-always option.
2391 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
2393 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
2394 avoid addition with 0 and use incw and decw where possible.
2396 2018-01-26 Richard Biener <rguenther@suse.de>
2398 PR tree-optimization/81082
2399 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
2400 association if it requires casting to unsigned.
2401 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
2402 from fold_plusminus_mult_expr to catch important cases late when
2403 range info is available.
2405 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2407 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
2408 * configure.ac (hidden_linkonce): New test.
2409 * configure: Regenerate.
2410 * config.in: Regenerate.
2412 2018-01-26 Julia Koval <julia.koval@intel.com>
2414 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
2415 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
2416 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
2417 _mm_mask_bitshuffle_epi64_mask): Fix type.
2418 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
2419 USI_FTYPE_V4DI_V4DI_USI): Remove.
2420 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
2421 __builtin_ia32_vpshufbitqmb256_mask,
2422 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
2423 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
2424 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
2426 2018-01-26 Alan Modra <amodra@gmail.com>
2429 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
2430 UNSPEC_VBPERMQ. Sort other unspecs.
2432 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
2434 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
2436 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
2439 * predict.c (drop_profile): Do not push/pop cfun; update also
2441 (handle_missing_profiles): Fix logic looking for zero profiles.
2443 2018-01-25 Jakub Jelinek <jakub@redhat.com>
2446 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
2447 on functions with #pragma omp declare simd or functions with simd
2449 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
2450 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
2451 Remove trailing \n from warning_at calls.
2453 2018-01-25 Tom de Vries <tom@codesourcery.com>
2456 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2457 for neutered workers.
2459 2018-01-24 Joseph Myers <joseph@codesourcery.com>
2462 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
2463 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
2465 2018-01-24 Jeff Law <law@redhat.com>
2468 * i386.c (get_probe_interval): Move to earlier point.
2469 (ix86_compute_frame_layout): If -fstack-clash-protection and
2470 the frame is larger than the probe interval, then use pushes
2471 to save registers rather than reg->mem moves.
2472 (ix86_expand_prologue): Remove conditional for int_registers_saved
2475 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
2478 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
2479 min/max for never referenced object.
2481 2018-01-24 Jakub Jelinek <jakub@redhat.com>
2484 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
2486 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
2487 attributes from DECL_ATTRIBUTES (decl) without affecting
2488 DECL_ATTRIBUTES (current_function_decl).
2489 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
2490 functions with non-NULL DECL_ABSTRACT_ORIGIN.
2492 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
2494 PR tree-optimization/83979
2495 * fold-const.c (fold_comparison): Use constant_boolean_node
2496 instead of boolean_{true,false}_node.
2498 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
2500 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
2503 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2505 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
2506 Simplify the clause that sets the length attribute.
2507 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
2508 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
2509 clause that sets the length attribute.
2510 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
2512 2018-01-24 Tom de Vries <tom@codesourcery.com>
2515 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
2516 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
2517 Add strict parameter.
2518 (prevent_branch_around_nothing): Insert dummy insn between branch to
2519 label and label with no ptx insn inbetween.
2520 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
2522 2018-01-24 Tom de Vries <tom@codesourcery.com>
2525 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2526 for neutered threads in warp.
2527 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
2529 2018-01-24 Richard Biener <rguenther@suse.de>
2531 PR tree-optimization/83176
2532 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
2535 2018-01-24 Richard Biener <rguenther@suse.de>
2537 PR tree-optimization/82819
2538 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
2539 code generating pluses that are no-ops in the target precision.
2541 2018-01-24 Richard Biener <rguenther@suse.de>
2544 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
2546 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2548 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
2549 to merge probabilities.
2550 * predict.c (probably_never_executed): Also mark as cold functions
2551 with global 0 profile and guessed local profile.
2552 * profile-count.c (profile_probability::combine_with_count): New
2554 * profile-count.h (profile_probability::operator*,
2555 profile_probability::operator*=, profile_probability::operator/,
2556 profile_probability::operator/=): Reduce precision to adjusted
2557 and set value to guessed on contradictory divisions.
2558 (profile_probability::combine_with_freq): Remove.
2559 (profile_probability::combine_wiht_count): Declare.
2560 (profile_count::force_nonzero):: Set to adjusted.
2561 (profile_count::probability_in):: Set quality to adjusted.
2562 * tree-ssa-tail-merge.c (replace_block_by): Use
2565 2018-01-23 Andrew Waterman <andrew@sifive.com>
2566 Jim Wilson <jimw@sifive.com>
2568 * config/riscv/riscv.c (riscv_stack_boundary): New.
2569 (riscv_option_override): Set riscv_stack_boundary. Handle
2570 riscv_preferred_stack_boundary_arg.
2571 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
2572 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
2573 (STACK_BOUNDARY): Set to riscv_stack_boundary.
2574 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
2575 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
2576 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
2578 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
2581 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
2582 of struct ix86_frame.
2583 (ix86_expand_epilogue): Likewise. Add a local variable for
2584 the reg_save_offset field in struct ix86_frame.
2586 2018-01-23 Bin Cheng <bin.cheng@arm.com>
2588 PR tree-optimization/82604
2589 * tree-loop-distribution.c (enum partition_kind): New enum item
2590 PKIND_PARTIAL_MEMSET.
2591 (partition_builtin_p): Support above new enum item.
2592 (generate_code_for_partition): Ditto.
2593 (compute_access_range): Differentiate cases that equality can be
2594 proven at all loops, the innermost loops or no loops.
2595 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
2596 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
2597 (finalize_partitions, distribute_loop): Don't fuse partition of
2598 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
2599 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
2602 2018-01-23 Martin Liska <mliska@suse.cz>
2604 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
2605 order to ignore the predictor.
2606 (PRED_POLYMORPHIC_CALL): Likewise.
2607 (PRED_RECURSIVE_CALL): Likewise.
2609 2018-01-23 Martin Liska <mliska@suse.cz>
2611 * tree-profile.c (tree_profiling): Print function header to
2612 aware reader which function we are working on.
2613 * value-prof.c (gimple_find_values_to_profile): Do not print
2614 not interesting value histograms.
2616 2018-01-23 Martin Liska <mliska@suse.cz>
2618 * profile-count.h (enum profile_quality): Add
2619 profile_uninitialized as the first value. Do not number values
2620 as they are zero based.
2621 (profile_count::verify): Update sanity check.
2622 (profile_probability::verify): Likewise.
2624 2018-01-23 Nathan Sidwell <nathan@acm.org>
2626 * doc/invoke.texi (ffor-scope): Deprecate.
2628 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2630 PR tree-optimization/83510
2631 * domwalk.c (set_all_edges_as_executable): New function.
2632 (dom_walker::dom_walker): Convert bool param
2633 "skip_unreachable_blocks" to enum reachability. Move setup of
2634 edge flags to set_all_edges_as_executable and only do it when
2635 reachability is REACHABLE_BLOCKS.
2636 * domwalk.h (enum dom_walker::reachability): New enum.
2637 (dom_walker::dom_walker): Convert bool param
2638 "skip_unreachable_blocks" to enum reachability.
2639 (set_all_edges_as_executable): New decl.
2640 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
2641 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
2643 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
2644 but converting true to REACHABLE_BLOCKS.
2645 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
2647 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
2648 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
2649 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
2651 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
2652 if check_all_array_refs will be called.
2654 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2656 * tree.c (selftest::test_location_wrappers): Add more test
2659 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2661 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
2662 (selftest::test_bit_in_range): Likewise.
2664 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
2667 * doc/sourcebuild.texi (vect_float): Say that the selector
2668 only describes the situation when -funsafe-math-optimizations is on.
2669 (vect_float_strict): Document.
2671 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
2673 PR tree-optimization/83965
2674 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
2675 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
2676 instead of checking only for a reduction.
2677 (vect_recog_widen_sum_pattern): Likewise.
2679 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2681 * predict.c (probably_never_executed): Only use precise profile info.
2682 (compute_function_frequency): Skip after inlining hack since we now
2683 have quality checking.
2685 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2687 * profile-count.h (profile_probability::very_unlikely,
2688 profile_probability::unlikely, profile_probability::even): Set
2689 precision to guessed.
2691 2018-01-23 Richard Biener <rguenther@suse.de>
2693 PR tree-optimization/83963
2694 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
2695 Properly terminate dominator walk when crossing the exit edge not
2696 when visiting its source block.
2698 2018-01-23 Jakub Jelinek <jakub@redhat.com>
2701 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
2702 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
2704 2018-01-22 Jakub Jelinek <jakub@redhat.com>
2706 PR tree-optimization/83957
2707 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
2708 semicolon after for body surrounded by braces.
2710 PR tree-optimization/83081
2711 * profile-count.h (profile_probability::split): New method.
2712 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
2713 Use profile_probability::split.
2714 (do_compare_rtx_and_jump): Fix adjustment of probabilities
2715 when splitting a single conditional jump into 2.
2717 2018-01-22 David Malcolm <dmalcolm@redhat.com>
2719 PR tree-optimization/69452
2720 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
2723 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2725 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
2726 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
2727 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
2729 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2731 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
2732 * config/rl78/rl78.md (movdi): New define_expand.
2733 * config/rl78/rl78.c (rl78_split_movdi): New function.
2735 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
2738 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
2740 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
2741 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
2742 128-bit to produce an UNSPEC move to get the double word with the
2743 signbit and then a shift directly to do signbit.
2744 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
2745 implementation with a new version that just does either a direct
2746 move or a regular move. Move memory interface to separate insns.
2747 Move insns so they are next to the expander.
2748 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
2749 with signbit move. Split big and little endian case.
2750 (signbit<mode>2_dm_mem_le): Likewise.
2751 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
2752 (signbit<mode>2_dm2): Likewise.
2754 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2756 * config/rl78/rl78.md (anddi3): New define_expand.
2758 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2760 * config/rl78/rl78.md (umindi3): New define_expand.
2762 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2764 * config/rl78/rl78.md (smindi3): New define_expand.
2766 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2768 * config/rl78/rl78.md (smaxdi3): New define_expand.
2770 2018-01-22 Carl Love <cel@us.ibm.com>
2772 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
2773 LVX_V1TI): Add macro expansion.
2774 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
2775 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
2776 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
2777 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
2778 Change check to determine if the instruction is a byte reversing
2779 entry. Fix typo in comment.
2780 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
2781 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
2782 Add def_builtin calls for new builtins.
2783 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
2784 Add define_insn expansion.
2786 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2788 * config/rl78/rl78.md (umaxdi3): New define_expand.
2790 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2792 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
2793 for non-QImode registers.
2795 2018-01-22 Richard Biener <rguenther@suse.de>
2797 PR tree-optimization/83963
2798 * graphite-scop-detection.c (scop_detection::get_sese): Delay
2799 including the loop exit block.
2800 (scop_detection::merge_sese): Likewise.
2801 (scop_detection::add_scop): Do it here instead.
2803 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2805 * doc/sourcebuild.texi (arm_softfloat): Document.
2807 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
2810 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
2811 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
2812 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
2814 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2815 David Edelsohn <dje.gcc@gmail.com>
2818 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
2819 Change "crset eq" to "crset 2".
2820 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
2821 (*call_indirect_aix<mode>_nospec): Likewise.
2822 (*call_value_indirect_aix<mode>_nospec): Likewise.
2823 (*call_indirect_elfv2<mode>_nospec): Likewise.
2824 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
2825 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
2826 change assembly output from . to $.
2827 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
2828 (indirect_jump<mode>_nospec): Change assembly output from . to $.
2829 (*tablejump<mode>_internal1_nospec): Likewise.
2831 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
2834 * config/sh/sh_optimize_sett_clrt.cc:
2835 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
2837 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
2839 PR tree-optimization/83940
2840 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
2841 offset_dt to vect_constant_def rather than vect_unknown_def_type.
2842 (vect_check_load_store_mask): Add a mask_dt_out parameter and
2843 use it to pass back the definition type.
2844 (vect_check_store_rhs): Likewise rhs_dt_out.
2845 (vect_build_gather_load_calls): Add a mask_dt argument and use
2846 it instead of a call to vect_is_simple_use.
2847 (vectorizable_store): Update calls to vect_check_load_store_mask
2848 and vect_check_store_rhs. Use the dt returned by the latter instead
2849 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
2850 instead of calls to vect_is_simple_use. Pass the scalar rather
2851 than the vector operand to vect_is_simple_use when handling
2852 second and subsequent copies of an rhs value.
2853 (vectorizable_load): Update calls to vect_check_load_store_mask
2854 and vect_build_gather_load_calls. Use the cached mask_dt and
2855 gs_info.offset_dt instead of calls to vect_is_simple_use.
2857 2018-01-20 Jakub Jelinek <jakub@redhat.com>
2860 * tree-emutls.c: Include gimplify.h.
2861 (lower_emutls_2): New function.
2862 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
2863 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
2864 it before further processing.
2867 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
2868 UINTVAL (trueop1) instead of INTVAL (op1).
2870 2018-01-19 Jakub Jelinek <jakub@redhat.com>
2874 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
2875 INCOMING_FRAME_SP_OFFSET if not defined.
2876 (scan_trace): Add ENTRY argument. If true and
2877 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
2878 emit a note to adjust the CFA offset.
2879 (create_cfi_notes): Adjust scan_trace callers.
2880 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
2881 INCOMING_FRAME_SP_OFFSET in the CIE.
2882 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
2883 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
2885 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
2886 * doc/tm.texi: Regenerated.
2888 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2890 PR rtl-optimization/83147
2891 * lra-constraints.c (remove_inheritance_pseudos): Use
2892 lra_substitute_pseudo_within_insn.
2894 2018-01-19 Tom de Vries <tom@codesourcery.com>
2895 Cesar Philippidis <cesar@codesourcery.com>
2898 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
2900 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
2903 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
2904 spaces for function labels.
2906 2018-01-19 Martin Liska <mliska@suse.cz>
2908 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
2909 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
2910 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
2911 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
2912 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
2913 (PRED_CONST_RETURN): Change from 69 to 65.
2914 (PRED_NULL_RETURN): Change from 91 to 71.
2915 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
2916 (PRED_LOOP_GUARD): Change from 66 to 73.
2918 2018-01-19 Martin Liska <mliska@suse.cz>
2920 * predict.c (predict_insn_def): Add new assert.
2921 (struct branch_predictor): Change type to signed integer.
2922 (test_prediction_value_range): Amend test to cover
2924 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
2925 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
2926 (PRED_LOOP_ITERATIONS_MAX): Likewise.
2927 (PRED_LOOP_IV_COMPARE): Likewise.
2928 * predict.h (PROB_UNINITIALIZED): Define new constant.
2930 2018-01-19 Martin Liska <mliska@suse.cz>
2932 * predict.c (dump_prediction): Add new format for
2933 analyze_brprob.py script which is enabled with -details
2935 * profile-count.h (precise_p): New function.
2937 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
2939 PR tree-optimization/83922
2940 * tree-vect-loop.c (vect_verify_full_masking): Return false if
2941 there are no statements that need masking.
2942 (vect_active_double_reduction_p): New function.
2943 (vect_analyze_loop_operations): Use it when handling phis that
2944 are not in the loop header.
2946 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
2948 PR tree-optimization/83914
2949 * tree-vect-loop.c (vectorizable_induction): Don't convert
2950 init_expr or apply the peeling adjustment for inductions
2951 that are nested within the vectorized loop.
2953 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2955 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
2958 2018-01-18 Jakub Jelinek <jakub@redhat.com>
2962 * function.h (gimplify_parameters): Add gimple_seq * argument.
2963 * function.c: Include gimple.h and options.h.
2964 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
2965 for the added local temporaries if needed.
2966 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
2967 if there are any parameter cleanups, wrap whole body into a
2968 try/finally with the cleanups.
2970 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
2973 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
2974 Use GET_MODE_CLASS for scalar floating point.
2976 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
2980 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
2981 Fix call of call_cgraph_insertion_hooks.
2983 2018-01-18 Martin Sebor <msebor@redhat.com>
2985 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
2987 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
2990 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
2993 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
2996 * common.opt: (-ffile-prefix-map): New option.
2997 * opts.c (common_handle_option): Defer it.
2998 * opts-global.c (handle_common_deferred_options): Handle it.
2999 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
3000 * file-prefix-map.h: New file.
3001 (remap_debug_filename, add_debug_prefix_map): ...here.
3002 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
3003 * final.c (debug_prefix_map, add_debug_prefix_map
3004 remap_debug_filename): Move to...
3005 * file-prefix-map.c: New file.
3006 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
3007 generalize, get rid of alloca(), use strrchr() instead of strchr().
3008 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
3009 Implement in terms of add_prefix_map().
3010 (remap_macro_filename, remap_debug_filename): Implement in term of
3012 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
3013 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
3014 * dbxout.c: Include file-prefix-map.h.
3015 * varasm.c: Likewise.
3016 * vmsdbgout.c: Likewise.
3017 * xcoffout.c: Likewise.
3018 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
3019 * doc/cppopts.texi (-fmacro-prefix-map): Document.
3020 * doc/invoke.texi (-ffile-prefix-map): Document.
3021 (-fdebug-prefix-map): Update description.
3023 2018-01-18 Martin Liska <mliska@suse.cz>
3025 * config/i386/i386.c (indirect_thunk_name): Document that also
3027 (output_indirect_thunk): Document why both instructions
3028 (pause and lfence) are generated.
3030 2018-01-18 Richard Biener <rguenther@suse.de>
3032 PR tree-optimization/83887
3033 * graphite-scop-detection.c
3034 (scop_detection::get_nearest_dom_with_single_entry): Remove.
3035 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
3036 (scop_detection::merge_sese): Re-implement with a flood-fill
3037 algorithm that properly finds a SESE region if it exists.
3039 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3042 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
3043 pointer_diff optimizations use view_convert instead of convert.
3045 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3047 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3048 Generate different code for -mno-speculate-indirect-jumps.
3049 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3050 (*call_indirect_aix<mode>): Disable for
3051 -mno-speculate-indirect-jumps.
3052 (*call_indirect_aix<mode>_nospec): New define_insn.
3053 (*call_value_indirect_aix<mode>): Disable for
3054 -mno-speculate-indirect-jumps.
3055 (*call_value_indirect_aix<mode>_nospec): New define_insn.
3056 (*sibcall_nonlocal_sysv<mode>): Generate different code for
3057 -mno-speculate-indirect-jumps.
3058 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3060 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3062 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
3063 long double type, set the flags for noting the default long double
3064 type, even if we don't pass or return a long double type.
3066 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
3069 * ipa-inline.c (flatten_function): Do not overwrite final inlining
3072 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
3074 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
3075 support for merge[hl].
3076 (fold_mergehl_helper): New helper function.
3077 (tree-vector-builder.h): New #include for tree_vector_builder usage.
3078 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
3079 (altivec_vmrglw_direct): Add xxmrglw insn.
3081 2018-01-17 Andrew Waterman <andrew@sifive.com>
3083 * config/riscv/riscv.c (riscv_conditional_register_usage): If
3084 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
3086 2018-01-17 David Malcolm <dmalcolm@redhat.com>
3089 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
3090 call the lto_location_cache before reading the
3091 DECL_SOURCE_LOCATION of the types.
3093 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3094 Richard Sandiford <richard.sandiford@linaro.org>
3096 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
3097 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
3098 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
3099 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
3100 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
3102 * config/aarch64/constraints.md (aarch64_movti_operand):
3104 * config/aarch64/predicates.md (Uti): Add new constraint.
3106 2018-01-17 Carl Love <cel@us.ibm.com>
3108 * config/rs6000/vsx.md (define_expand xl_len_r,
3109 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
3110 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
3112 (define_expand, define_insn): Move the shift left from the
3113 define_insn to the define_expand for lxvl and stxvl instructions.
3114 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
3115 and XL_LEN_R definitions to PURE.
3117 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
3119 * config/i386/i386.c (indirect_thunk_name): Declare regno
3120 as unsigned int. Compare regno with INVALID_REGNUM.
3121 (output_indirect_thunk): Ditto.
3122 (output_indirect_thunk_function): Ditto.
3123 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
3124 in the call to output_indirect_thunk_function.
3126 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
3129 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
3130 rather than the size of inner_type to determine the stack slot size
3131 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
3133 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
3136 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
3139 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
3141 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
3142 endian Linux systems to optionally enable multilibs for selecting
3143 the long double type if the user configured an explicit type.
3144 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
3145 have no long double multilibs if not defined.
3146 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
3147 warn if the user used -mabi={ieee,ibm}longdouble and we built
3148 multilibs for long double.
3149 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
3150 appropriate multilib option.
3151 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
3153 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
3154 for building long double multilibs.
3155 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
3157 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
3159 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
3162 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
3164 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
3167 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
3170 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
3173 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3175 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
3176 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
3178 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3180 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
3181 different rtl trees depending on TARGET_64BIT.
3182 (rs6000_gen_lvx): Likewise.
3184 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3186 * config/visium/visium.md (nop): Tweak comment.
3187 (hazard_nop): Likewise.
3189 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3191 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
3192 -mspeculate-indirect-jumps.
3193 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
3194 for -mno-speculate-indirect-jumps.
3195 (*call_indirect_elfv2<mode>_nospec): New define_insn.
3196 (*call_value_indirect_elfv2<mode>): Disable for
3197 -mno-speculate-indirect-jumps.
3198 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
3199 (indirect_jump): Emit different RTL for
3200 -mno-speculate-indirect-jumps.
3201 (*indirect_jump<mode>): Disable for
3202 -mno-speculate-indirect-jumps.
3203 (*indirect_jump<mode>_nospec): New define_insn.
3204 (tablejump): Emit different RTL for
3205 -mno-speculate-indirect-jumps.
3206 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
3207 (tablejumpsi_nospec): New define_expand.
3208 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
3209 (tablejumpdi_nospec): New define_expand.
3210 (*tablejump<mode>_internal1): Disable for
3211 -mno-speculate-indirect-jumps.
3212 (*tablejump<mode>_internal1_nospec): New define_insn.
3213 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
3216 2018-01-16 Artyom Skrobov tyomitch@gmail.com
3218 * caller-save.c (insert_save): Drop unnecessary parameter. All
3221 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3222 Richard Biener <rguenth@suse.de>
3225 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
3226 return early, inline manually is_gimple_sizepos. Make sure if we
3227 call gimplify_expr we don't end up with a gimple constant.
3228 * tree.c (variably_modified_type_p): Don't return true for
3229 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
3230 * gimplify.h (is_gimple_sizepos): Remove.
3232 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3234 PR tree-optimization/83857
3235 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
3236 vectorizable_live_operation for pure SLP statements.
3237 (vectorizable_live_operation): Handle PHIs.
3239 2018-01-16 Richard Biener <rguenther@suse.de>
3241 PR tree-optimization/83867
3242 * tree-vect-stmts.c (vect_transform_stmt): Precompute
3243 nested_in_vect_loop_p since the scalar stmt may get invalidated.
3245 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3248 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
3249 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
3250 If off is not INTEGER_CST, issue a may not be aligned warning
3251 rather than isn't aligned. Use isn%'t rather than isn't.
3252 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
3254 <case MULT_EXPR>: Improve the case when bottom and one of the
3255 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
3256 operand, in that case check if the other operand is multiple of
3257 bottom divided by the INTEGER_CST operand.
3259 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3262 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
3263 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
3264 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
3265 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
3266 * config/pa/pa.c (pa_function_arg_advance): Likewise.
3267 (pa_function_arg, pa_arg_partial_bytes): Likewise.
3268 (pa_function_arg_size): New function.
3270 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3272 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
3273 in a separate statement.
3275 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3277 PR tree-optimization/83847
3278 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
3279 group gathers and scatters.
3281 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3283 PR rtl-optimization/86620
3284 * params.def (max-sched-ready-insns): Bump minimum value to 1.
3286 PR rtl-optimization/83213
3287 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
3288 to last if both are JUMP_INSNs.
3290 PR tree-optimization/83843
3291 * gimple-ssa-store-merging.c
3292 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
3293 store_immediate_info for bswap/nop orig_stores.
3295 2018-01-15 Andrew Waterman <andrew@sifive.com>
3297 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
3299 <UDIV>: Increase cost if !TARGET_DIV.
3301 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
3303 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
3304 (define_attr "cr_logical_3op"): New.
3305 (cceq_ior_compare): Adjust.
3306 (cceq_ior_compare_complement): Adjust.
3307 (*cceq_rev_compare): Adjust.
3308 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
3309 (is_cracked_insn): Adjust.
3310 (insn_must_be_first_in_group): Adjust.
3311 * config/rs6000/40x.md: Adjust.
3312 * config/rs6000/440.md: Adjust.
3313 * config/rs6000/476.md: Adjust.
3314 * config/rs6000/601.md: Adjust.
3315 * config/rs6000/603.md: Adjust.
3316 * config/rs6000/6xx.md: Adjust.
3317 * config/rs6000/7450.md: Adjust.
3318 * config/rs6000/7xx.md: Adjust.
3319 * config/rs6000/8540.md: Adjust.
3320 * config/rs6000/cell.md: Adjust.
3321 * config/rs6000/e300c2c3.md: Adjust.
3322 * config/rs6000/e500mc.md: Adjust.
3323 * config/rs6000/e500mc64.md: Adjust.
3324 * config/rs6000/e5500.md: Adjust.
3325 * config/rs6000/e6500.md: Adjust.
3326 * config/rs6000/mpc.md: Adjust.
3327 * config/rs6000/power4.md: Adjust.
3328 * config/rs6000/power5.md: Adjust.
3329 * config/rs6000/power6.md: Adjust.
3330 * config/rs6000/power7.md: Adjust.
3331 * config/rs6000/power8.md: Adjust.
3332 * config/rs6000/power9.md: Adjust.
3333 * config/rs6000/rs64.md: Adjust.
3334 * config/rs6000/titan.md: Adjust.
3336 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3338 * config/i386/predicates.md (indirect_branch_operand): Rewrite
3339 ix86_indirect_branch_register logic.
3341 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3343 * config/i386/constraints.md (Bs): Update
3344 ix86_indirect_branch_register check. Don't check
3345 ix86_indirect_branch_register with GOT_memory_operand.
3347 * config/i386/predicates.md (GOT_memory_operand): Don't check
3348 ix86_indirect_branch_register here.
3349 (GOT32_symbol_operand): Likewise.
3351 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3353 * config/i386/predicates.md (constant_call_address_operand):
3354 Rewrite ix86_indirect_branch_register logic.
3355 (sibcall_insn_operand): Likewise.
3357 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3359 * config/i386/constraints.md (Bs): Replace
3360 ix86_indirect_branch_thunk_register with
3361 ix86_indirect_branch_register.
3363 * config/i386/i386.md (indirect_jump): Likewise.
3364 (tablejump): Likewise.
3365 (*sibcall_memory): Likewise.
3366 (*sibcall_value_memory): Likewise.
3367 Peepholes of indirect call and jump via memory: Likewise.
3368 * config/i386/i386.opt: Likewise.
3369 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3370 (GOT_memory_operand): Likewise.
3371 (call_insn_operand): Likewise.
3372 (sibcall_insn_operand): Likewise.
3373 (GOT32_symbol_operand): Likewise.
3375 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3378 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
3379 type rather than type addr's type points to.
3380 (expand_omp_atomic_mutex): Likewise.
3381 (expand_omp_atomic): Likewise.
3383 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3386 * config/i386/i386.c (output_indirect_thunk_function): Use
3387 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
3388 for __x86_return_thunk.
3390 2018-01-15 Richard Biener <rguenther@suse.de>
3393 * expmed.c (extract_bit_field_1): Fix typo.
3395 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3398 * config/arm/iterators.md (VF): New mode iterator.
3399 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
3400 Remove integer-related logic from pattern.
3401 (neon_vabd<mode>_3): Likewise.
3403 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3406 * common.opt (fstrict-overflow): No longer an alias.
3407 (fwrapv-pointer): New option.
3408 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
3409 also for pointer types based on flag_wrapv_pointer.
3410 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
3411 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
3412 opts->x_flag_wrapv got set.
3413 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
3414 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
3415 POINTER_TYPE_OVERFLOW_UNDEFINED.
3416 * match.pd: Likewise in address comparison pattern.
3417 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
3419 2018-01-15 Richard Biener <rguenther@suse.de>
3422 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
3423 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
3424 Reset type names to their identifier if their TYPE_DECL doesn't
3425 have linkage (and thus is used for ODR and devirt).
3426 (save_debug_info_for_decl): Remove.
3427 (save_debug_info_for_type): Likewise.
3428 (add_tree_to_fld_list): Adjust.
3429 * tree-pretty-print.c (dump_generic_node): Make dumping of
3430 type names more robust.
3432 2018-01-15 Richard Biener <rguenther@suse.de>
3434 * BASE-VER: Bump to 8.0.1.
3436 2018-01-14 Martin Sebor <msebor@redhat.com>
3439 * builtins.c (check_access): Avoid warning when the no-warning bit
3442 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
3444 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
3445 * ira-color (allocno_hard_regs_compare): Likewise.
3447 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
3450 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
3451 Use .pushsection/.popsection.
3453 2018-01-14 Martin Sebor <msebor@redhat.com>
3456 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
3458 2018-01-14 Jakub Jelinek <jakub@redhat.com>
3460 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
3461 entry from extra_headers.
3462 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
3463 extra_headers, make the list bitwise identical to the i?86-*-* one.
3465 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3467 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
3468 -mcmodel=large with -mindirect-branch=thunk,
3469 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
3470 -mfunction-return=thunk-extern.
3471 * doc/invoke.texi: Document -mcmodel=large is incompatible with
3472 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
3473 -mfunction-return=thunk and -mfunction-return=thunk-extern.
3475 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3477 * config/i386/i386.c (print_reg): Print the name of the full
3478 integer register without '%'.
3479 (ix86_print_operand): Handle 'V'.
3480 * doc/extend.texi: Document 'V' modifier.
3482 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3484 * config/i386/constraints.md (Bs): Disallow memory operand for
3485 -mindirect-branch-register.
3487 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3488 (GOT_memory_operand): Likewise.
3489 (call_insn_operand): Likewise.
3490 (sibcall_insn_operand): Likewise.
3491 (GOT32_symbol_operand): Likewise.
3492 * config/i386/i386.md (indirect_jump): Call convert_memory_address
3493 for -mindirect-branch-register.
3494 (tablejump): Likewise.
3495 (*sibcall_memory): Likewise.
3496 (*sibcall_value_memory): Likewise.
3497 Disallow peepholes of indirect call and jump via memory for
3498 -mindirect-branch-register.
3499 (*call_pop): Replace m with Bw.
3500 (*call_value_pop): Likewise.
3501 (*sibcall_pop_memory): Replace m with Bs.
3502 * config/i386/i386.opt (mindirect-branch-register): New option.
3503 * doc/invoke.texi: Document -mindirect-branch-register option.
3505 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3507 * config/i386/i386-protos.h (ix86_output_function_return): New.
3508 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
3509 set function_return_type.
3510 (indirect_thunk_name): Add ret_p to indicate thunk for function
3512 (output_indirect_thunk_function): Pass false to
3513 indirect_thunk_name.
3514 (ix86_output_indirect_branch_via_reg): Likewise.
3515 (ix86_output_indirect_branch_via_push): Likewise.
3516 (output_indirect_thunk_function): Create alias for function
3517 return thunk if regno < 0.
3518 (ix86_output_function_return): New function.
3519 (ix86_handle_fndecl_attribute): Handle function_return.
3520 (ix86_attribute_table): Add function_return.
3521 * config/i386/i386.h (machine_function): Add
3522 function_return_type.
3523 * config/i386/i386.md (simple_return_internal): Use
3524 ix86_output_function_return.
3525 (simple_return_internal_long): Likewise.
3526 * config/i386/i386.opt (mfunction-return=): New option.
3527 (indirect_branch): Mention -mfunction-return=.
3528 * doc/extend.texi: Document function_return function attribute.
3529 * doc/invoke.texi: Document -mfunction-return= option.
3531 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3533 * config/i386/i386-opts.h (indirect_branch): New.
3534 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
3535 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
3536 with local indirect jump when converting indirect call and jump.
3537 (ix86_set_indirect_branch_type): New.
3538 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
3539 (indirectlabelno): New.
3540 (indirect_thunk_needed): Likewise.
3541 (indirect_thunk_bnd_needed): Likewise.
3542 (indirect_thunks_used): Likewise.
3543 (indirect_thunks_bnd_used): Likewise.
3544 (INDIRECT_LABEL): Likewise.
3545 (indirect_thunk_name): Likewise.
3546 (output_indirect_thunk): Likewise.
3547 (output_indirect_thunk_function): Likewise.
3548 (ix86_output_indirect_branch_via_reg): Likewise.
3549 (ix86_output_indirect_branch_via_push): Likewise.
3550 (ix86_output_indirect_branch): Likewise.
3551 (ix86_output_indirect_jmp): Likewise.
3552 (ix86_code_end): Call output_indirect_thunk_function if needed.
3553 (ix86_output_call_insn): Call ix86_output_indirect_branch if
3555 (ix86_handle_fndecl_attribute): Handle indirect_branch.
3556 (ix86_attribute_table): Add indirect_branch.
3557 * config/i386/i386.h (machine_function): Add indirect_branch_type
3558 and has_local_indirect_jump.
3559 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
3561 (tablejump): Likewise.
3562 (*indirect_jump): Use ix86_output_indirect_jmp.
3563 (*tablejump_1): Likewise.
3564 (simple_return_indirect_internal): Likewise.
3565 * config/i386/i386.opt (mindirect-branch=): New option.
3566 (indirect_branch): New.
3569 (thunk-inline): Likewise.
3570 (thunk-extern): Likewise.
3571 * doc/extend.texi: Document indirect_branch function attribute.
3572 * doc/invoke.texi: Document -mindirect-branch= option.
3574 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
3577 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
3579 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
3581 * ipa-inline.c (want_inline_small_function_p): Return false if
3582 inlining has already failed with CIF_FINAL_ERROR.
3583 (update_caller_keys): Call want_inline_small_function_p before
3585 (update_callee_keys): Likewise.
3587 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3589 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
3591 (rs6000_quadword_masked_address_p): Likewise.
3592 (quad_aligned_load_p): Likewise.
3593 (quad_aligned_store_p): Likewise.
3594 (const_load_sequence_p): Add comment to describe the outer-most loop.
3595 (mimic_memory_attributes_and_flags): New function.
3596 (rs6000_gen_stvx): Likewise.
3597 (replace_swapped_aligned_store): Likewise.
3598 (rs6000_gen_lvx): Likewise.
3599 (replace_swapped_aligned_load): Likewise.
3600 (replace_swapped_load_constant): Capitalize argument name in
3601 comment describing this function.
3602 (rs6000_analyze_swaps): Add a third pass to search for vector loads
3603 and stores that access quad-word aligned addresses and replace
3604 with stvx or lvx instructions when appropriate.
3605 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
3606 New function prototype.
3607 (rs6000_quadword_masked_address_p): Likewise.
3608 (rs6000_gen_lvx): Likewise.
3609 (rs6000_gen_stvx): Likewise.
3610 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
3611 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
3612 when memory address is aligned.
3613 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
3614 this split to select lvx instruction when memory address is aligned.
3615 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
3616 instruction when memory address is aligned.
3617 (*vsx_le_perm_load_v16qi): Likewise.
3618 (four unnamed splitters): Modify to select the stvx instruction
3619 when memory is aligned.
3621 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
3623 * predict.c (determine_unlikely_bbs): Handle correctly BBs
3624 which appears in the queue multiple times.
3626 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3627 Alan Hayward <alan.hayward@arm.com>
3628 David Sherwood <david.sherwood@arm.com>
3630 * tree-vectorizer.h (vec_lower_bound): New structure.
3631 (_loop_vec_info): Add check_nonzero and lower_bounds.
3632 (LOOP_VINFO_CHECK_NONZERO): New macro.
3633 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
3634 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
3635 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
3636 fields. Make seg_len the distance travelled, not including the
3638 (dr_direction_indicator): Declare.
3639 (dr_zero_step_indicator): Likewise.
3640 (dr_known_forward_stride_p): Likewise.
3641 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
3643 (runtime_alias_check_p): Allow runtime alias checks with
3645 (operator ==): Compare access_size and align.
3646 (prune_runtime_alias_test_list): Rework for new distinction between
3647 the access_size and seg_len.
3648 (create_intersect_range_checks_index): Likewise. Cope with polynomial
3650 (get_segment_min_max): New function.
3651 (create_intersect_range_checks): Use it.
3652 (dr_step_indicator): New function.
3653 (dr_direction_indicator): Likewise.
3654 (dr_zero_step_indicator): Likewise.
3655 (dr_known_forward_stride_p): Likewise.
3656 * tree-loop-distribution.c (data_ref_segment_size): Return
3657 DR_STEP * (niters - 1).
3658 (compute_alias_check_pairs): Update call to the dr_with_seg_len
3660 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
3661 (vect_preserves_scalar_order_p): New function, split out from...
3662 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
3663 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
3664 (vect_vfa_access_size): New function.
3665 (vect_vfa_align): Likewise.
3666 (vect_compile_time_alias): Take access_size_a and access_b arguments.
3667 (dump_lower_bound): New function.
3668 (vect_check_lower_bound): Likewise.
3669 (vect_small_gap_p): Likewise.
3670 (vectorizable_with_step_bound_p): Likewise.
3671 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
3672 depencies if the vectorization factor is 1. Convert the checks
3673 for nonzero steps into checks on the bounds of DR_STEP. Try using
3674 a bunds check for variable steps if the minimum required step is
3675 relatively small. Update calls to the dr_with_seg_len
3676 constructor and to vect_compile_time_alias.
3677 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
3679 (vect_loop_versioning): Call it.
3680 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
3682 (vect_estimate_min_profitable_iters): Account for any bounds checks.
3684 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3685 Alan Hayward <alan.hayward@arm.com>
3686 David Sherwood <david.sherwood@arm.com>
3688 * doc/sourcebuild.texi (vect_scatter_store): Document.
3689 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
3691 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
3693 * genopinit.c (main): Add supports_vec_scatter_store and
3694 supports_vec_scatter_store_cached to target_optabs.
3695 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
3696 IFN_MASK_SCATTER_STORE.
3697 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
3699 * internal-fn.h (internal_store_fn_p): Declare.
3700 (internal_fn_stored_value_index): Likewise.
3701 * internal-fn.c (scatter_store_direct): New macro.
3702 (expand_scatter_store_optab_fn): New function.
3703 (direct_scatter_store_optab_supported_p): New macro.
3704 (internal_store_fn_p): New function.
3705 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
3706 IFN_MASK_SCATTER_STORE.
3707 (internal_fn_mask_index): Likewise.
3708 (internal_fn_stored_value_index): New function.
3709 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
3711 * optabs-query.h (supports_vec_scatter_store_p): Declare.
3712 * optabs-query.c (supports_vec_scatter_store_p): New function.
3713 * tree-vectorizer.h (vect_get_store_rhs): Declare.
3714 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
3715 true for scatter stores.
3716 (vect_gather_scatter_fn_p): Handle scatter stores too.
3717 (vect_check_gather_scatter): Consider using scatter stores if
3718 supports_vec_scatter_store_p.
3719 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
3721 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
3722 internal_fn_stored_value_index.
3723 (check_load_store_masking): Handle scatter stores too.
3724 (vect_get_store_rhs): Make public.
3725 (vectorizable_call): Use internal_store_fn_p.
3726 (vectorizable_store): Handle scatter store internal functions.
3727 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
3728 when deciding whether the end of the group has been reached.
3729 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
3730 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
3731 (mask_scatter_store<mode>): New insns.
3733 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3734 Alan Hayward <alan.hayward@arm.com>
3735 David Sherwood <david.sherwood@arm.com>
3737 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
3738 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
3739 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
3741 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
3742 Use vect_truncate_gather_scatter_offset if we can't treat the
3743 operation as a normal gather load or scatter store.
3744 (get_group_load_store_type): Take the gather_scatter_info
3745 as argument. Try using a gather load or scatter store for
3746 single-element groups.
3747 (get_load_store_type): Update calls to get_group_load_store_type
3748 and vect_use_strided_gather_scatters_p.
3750 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3751 Alan Hayward <alan.hayward@arm.com>
3752 David Sherwood <david.sherwood@arm.com>
3754 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
3755 optional tree argument.
3756 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
3758 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
3759 but continue to use the current value as a fallback.
3760 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
3761 to compare the updates.
3762 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
3763 (get_load_store_type): Use it when handling a strided access.
3764 (vect_get_strided_load_store_ops): New function.
3765 (vect_get_data_ptr_increment): Likewise.
3766 (vectorizable_load): Handle strided gather loads. Always pass
3767 a step to vect_create_data_ref_ptr and bump_vector_ptr.
3769 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3770 Alan Hayward <alan.hayward@arm.com>
3771 David Sherwood <david.sherwood@arm.com>
3773 * doc/md.texi (gather_load@var{m}): Document.
3774 (mask_gather_load@var{m}): Likewise.
3775 * genopinit.c (main): Add supports_vec_gather_load and
3776 supports_vec_gather_load_cached to target_optabs.
3777 * optabs-tree.c (init_tree_optimization_optabs): Use
3778 ggc_cleared_alloc to allocate target_optabs.
3779 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
3780 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
3782 * internal-fn.h (internal_load_fn_p): Declare.
3783 (internal_gather_scatter_fn_p): Likewise.
3784 (internal_fn_mask_index): Likewise.
3785 (internal_gather_scatter_fn_supported_p): Likewise.
3786 * internal-fn.c (gather_load_direct): New macro.
3787 (expand_gather_load_optab_fn): New function.
3788 (direct_gather_load_optab_supported_p): New macro.
3789 (direct_internal_fn_optab): New function.
3790 (internal_load_fn_p): Likewise.
3791 (internal_gather_scatter_fn_p): Likewise.
3792 (internal_fn_mask_index): Likewise.
3793 (internal_gather_scatter_fn_supported_p): Likewise.
3794 * optabs-query.c (supports_at_least_one_mode_p): New function.
3795 (supports_vec_gather_load_p): Likewise.
3796 * optabs-query.h (supports_vec_gather_load_p): Declare.
3797 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
3798 and memory_type field.
3799 (NUM_PATTERNS): Bump to 15.
3800 * tree-vect-data-refs.c: Include internal-fn.h.
3801 (vect_gather_scatter_fn_p): New function.
3802 (vect_describe_gather_scatter_call): Likewise.
3803 (vect_check_gather_scatter): Try using internal functions for
3804 gather loads. Recognize existing calls to a gather load function.
3805 (vect_analyze_data_refs): Consider using gather loads if
3806 supports_vec_gather_load_p.
3807 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
3808 (vect_get_gather_scatter_offset_type): Likewise.
3809 (vect_convert_mask_for_vectype): Likewise.
3810 (vect_add_conversion_to_patterm): Likewise.
3811 (vect_try_gather_scatter_pattern): Likewise.
3812 (vect_recog_gather_scatter_pattern): New pattern recognizer.
3813 (vect_vect_recog_func_ptrs): Add it.
3814 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
3815 internal_fn_mask_index and internal_gather_scatter_fn_p.
3816 (check_load_store_masking): Take the gather_scatter_info as an
3817 argument and handle gather loads.
3818 (vect_get_gather_scatter_ops): New function.
3819 (vectorizable_call): Check internal_load_fn_p.
3820 (vectorizable_load): Likewise. Handle gather load internal
3822 (vectorizable_store): Update call to check_load_store_masking.
3823 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
3824 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
3825 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
3826 (aarch64_gather_scale_operand_d): New predicates.
3827 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
3828 (mask_gather_load<mode>): New insns.
3830 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3831 Alan Hayward <alan.hayward@arm.com>
3832 David Sherwood <david.sherwood@arm.com>
3834 * optabs.def (fold_left_plus_optab): New optab.
3835 * doc/md.texi (fold_left_plus_@var{m}): Document.
3836 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
3837 * internal-fn.c (fold_left_direct): Define.
3838 (expand_fold_left_optab_fn): Likewise.
3839 (direct_fold_left_optab_supported_p): Likewise.
3840 * fold-const-call.c (fold_const_fold_left): New function.
3841 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
3842 * tree-parloops.c (valid_reduction_p): New function.
3843 (gather_scalar_reductions): Use it.
3844 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
3845 (vect_finish_replace_stmt): Declare.
3846 * tree-vect-loop.c (fold_left_reduction_fn): New function.
3847 (needs_fold_left_reduction_p): New function, split out from...
3848 (vect_is_simple_reduction): ...here. Accept reductions that
3849 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
3850 (vect_force_simple_reduction): Also store the reduction type in
3851 the assignment's STMT_VINFO_REDUC_TYPE.
3852 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
3853 (merge_with_identity): New function.
3854 (vect_expand_fold_left): Likewise.
3855 (vectorize_fold_left_reduction): Likewise.
3856 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
3857 scalar phi in place for it. Check for target support and reject
3858 cases that would reassociate the operation. Defer the transform
3859 phase to vectorize_fold_left_reduction.
3860 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
3861 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
3862 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
3864 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3866 * tree-if-conv.c (predicate_mem_writes): Remove redundant
3867 call to ifc_temp_var.
3869 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3870 Alan Hayward <alan.hayward@arm.com>
3871 David Sherwood <david.sherwood@arm.com>
3873 * target.def (legitimize_address_displacement): Take the original
3874 offset as a poly_int.
3875 * targhooks.h (default_legitimize_address_displacement): Update
3877 * targhooks.c (default_legitimize_address_displacement): Likewise.
3878 * doc/tm.texi: Regenerate.
3879 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
3880 as an argument, moving assert of ad->disp == ad->disp_term to...
3881 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
3882 Try calling targetm.legitimize_address_displacement before expanding
3883 the address rather than afterwards, and adjust for the new interface.
3884 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
3885 Match the new hook interface. Handle SVE addresses.
3886 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
3889 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3891 * Makefile.in (OBJS): Add early-remat.o.
3892 * target.def (select_early_remat_modes): New hook.
3893 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
3894 * doc/tm.texi: Regenerate.
3895 * targhooks.h (default_select_early_remat_modes): Declare.
3896 * targhooks.c (default_select_early_remat_modes): New function.
3897 * timevar.def (TV_EARLY_REMAT): New timevar.
3898 * passes.def (pass_early_remat): New pass.
3899 * tree-pass.h (make_pass_early_remat): Declare.
3900 * early-remat.c: New file.
3901 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
3903 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
3905 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3906 Alan Hayward <alan.hayward@arm.com>
3907 David Sherwood <david.sherwood@arm.com>
3909 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
3910 vfm1 with a bound_epilog parameter.
3911 (vect_do_peeling): Update calls accordingly, and move the prologue
3912 call earlier in the function. Treat the base bound_epilog as 0 for
3913 fully-masked loops and retain vf - 1 for other loops. Add 1 to
3914 this base when peeling for gaps.
3915 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
3916 with fully-masked loops.
3917 (vect_estimate_min_profitable_iters): Handle the single peeled
3918 iteration in that case.
3920 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3921 Alan Hayward <alan.hayward@arm.com>
3922 David Sherwood <david.sherwood@arm.com>
3924 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
3925 single-element interleaving even if the size is not a power of 2.
3926 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
3927 accesses for single-element interleaving if the group size is
3930 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3931 Alan Hayward <alan.hayward@arm.com>
3932 David Sherwood <david.sherwood@arm.com>
3934 * doc/md.texi (fold_extract_last_@var{m}): Document.
3935 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
3936 * optabs.def (fold_extract_last_optab): New optab.
3937 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
3938 * internal-fn.c (fold_extract_direct): New macro.
3939 (expand_fold_extract_optab_fn): Likewise.
3940 (direct_fold_extract_optab_supported_p): Likewise.
3941 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
3942 * tree-vect-loop.c (vect_model_reduction_cost): Handle
3943 EXTRACT_LAST_REDUCTION.
3944 (get_initial_def_for_reduction): Do not create an initial vector
3945 for EXTRACT_LAST_REDUCTION reductions.
3946 (vectorizable_reduction): Leave the scalar phi in place for
3947 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
3948 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
3949 epilogue code for EXTRACT_LAST_REDUCTION and defer the
3950 transform phase to vectorizable_condition.
3951 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
3953 (vect_finish_stmt_generation): ...here.
3954 (vect_finish_replace_stmt): New function.
3955 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
3956 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
3958 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
3960 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3961 Alan Hayward <alan.hayward@arm.com>
3962 David Sherwood <david.sherwood@arm.com>
3964 * doc/md.texi (extract_last_@var{m}): Document.
3965 * optabs.def (extract_last_optab): New optab.
3966 * internal-fn.def (EXTRACT_LAST): New internal function.
3967 * internal-fn.c (cond_unary_direct): New macro.
3968 (expand_cond_unary_optab_fn): Likewise.
3969 (direct_cond_unary_optab_supported_p): Likewise.
3970 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
3971 loops using EXTRACT_LAST.
3972 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
3973 (extract_last_<mode>): ...this optab.
3974 (vec_extract<mode><Vel>): Update accordingly.
3976 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3977 Alan Hayward <alan.hayward@arm.com>
3978 David Sherwood <david.sherwood@arm.com>
3980 * target.def (empty_mask_is_expensive): New hook.
3981 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
3982 * doc/tm.texi: Regenerate.
3983 * targhooks.h (default_empty_mask_is_expensive): Declare.
3984 * targhooks.c (default_empty_mask_is_expensive): New function.
3985 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
3986 if the target says that empty masks are expensive.
3987 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
3989 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
3991 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3992 Alan Hayward <alan.hayward@arm.com>
3993 David Sherwood <david.sherwood@arm.com>
3995 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
3996 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
3997 (vect_use_loop_mask_for_alignment_p): New function.
3998 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
3999 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
4000 niters_skip argument. Make sure that the first niters_skip elements
4001 of the first iteration are inactive.
4002 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
4003 Update call to vect_set_loop_masks_directly.
4004 (get_misalign_in_elems): New function, split out from...
4005 (vect_gen_prolog_loop_niters): ...here.
4006 (vect_update_init_of_dr): Take a code argument that specifies whether
4007 the adjustment should be added or subtracted.
4008 (vect_update_init_of_drs): Likewise.
4009 (vect_prepare_for_masked_peels): New function.
4010 (vect_do_peeling): Skip prologue peeling if we're using a mask
4011 instead. Update call to vect_update_inits_of_drs.
4012 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4014 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
4015 alignment. Do not include the number of peeled iterations in
4016 the minimum threshold in that case.
4017 (vectorizable_induction): Adjust the start value down by
4018 LOOP_VINFO_MASK_SKIP_NITERS iterations.
4019 (vect_transform_loop): Call vect_prepare_for_masked_peels.
4020 Take the number of skipped iterations into account when calculating
4022 * tree-vect-stmts.c (vect_gen_while_not): New function.
4024 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4025 Alan Hayward <alan.hayward@arm.com>
4026 David Sherwood <david.sherwood@arm.com>
4028 * doc/sourcebuild.texi (vect_fully_masked): Document.
4029 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
4031 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
4033 (vect_analyze_loop_2): ...here. Don't check the vectorization
4034 factor against the number of loop iterations if the loop is
4037 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4038 Alan Hayward <alan.hayward@arm.com>
4039 David Sherwood <david.sherwood@arm.com>
4041 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
4042 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
4043 (dump_groups): Update accordingly.
4044 (iv_use::mem_type): New member variable.
4045 (address_p): New function.
4046 (record_use): Add a mem_type argument and initialize the new
4048 (record_group_use): Add a mem_type argument. Use address_p.
4049 Remove obsolete null checks of base_object. Update call to record_use.
4050 (find_interesting_uses_op): Update call to record_group_use.
4051 (find_interesting_uses_cond): Likewise.
4052 (find_interesting_uses_address): Likewise.
4053 (get_mem_type_for_internal_fn): New function.
4054 (find_address_like_use): Likewise.
4055 (find_interesting_uses_stmt): Try find_address_like_use before
4056 calling find_interesting_uses_op.
4057 (addr_offset_valid_p): Use the iv mem_type field as the type
4058 of the addressed memory.
4059 (add_autoinc_candidates): Likewise.
4060 (get_address_cost): Likewise.
4061 (split_small_address_groups_p): Use address_p.
4062 (split_address_groups): Likewise.
4063 (add_iv_candidate_for_use): Likewise.
4064 (autoinc_possible_for_pair): Likewise.
4065 (rewrite_groups): Likewise.
4066 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
4067 (determine_group_iv_cost): Update after split of USE_ADDRESS.
4068 (get_alias_ptr_type_for_ptr_address): New function.
4069 (rewrite_use_address): Rewrite address uses in calls that were
4070 identified by find_address_like_use.
4072 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4073 Alan Hayward <alan.hayward@arm.com>
4074 David Sherwood <david.sherwood@arm.com>
4076 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
4078 * gimple-expr.h (is_gimple_addressable: Likewise.
4079 * gimple-expr.c (is_gimple_address): Likewise.
4080 * internal-fn.c (expand_call_mem_ref): New function.
4081 (expand_mask_load_optab_fn): Use it.
4082 (expand_mask_store_optab_fn): Likewise.
4084 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4085 Alan Hayward <alan.hayward@arm.com>
4086 David Sherwood <david.sherwood@arm.com>
4088 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
4089 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
4090 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
4091 (cond_umax@var{mode}): Document.
4092 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
4093 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
4094 (cond_umin_optab, cond_umax_optab): New optabs.
4095 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
4096 (COND_IOR, COND_XOR): New internal functions.
4097 * internal-fn.h (get_conditional_internal_fn): Declare.
4098 * internal-fn.c (cond_binary_direct): New macro.
4099 (expand_cond_binary_optab_fn): Likewise.
4100 (direct_cond_binary_optab_supported_p): Likewise.
4101 (get_conditional_internal_fn): New function.
4102 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
4103 Cope with reduction statements that are vectorized as calls rather
4105 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
4106 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
4107 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
4108 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4109 (UNSPEC_COND_EOR): New unspecs.
4110 (optab): Add mappings for them.
4111 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
4112 (sve_int_op, sve_fp_op): New int attributes.
4114 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4115 Alan Hayward <alan.hayward@arm.com>
4116 David Sherwood <david.sherwood@arm.com>
4118 * optabs.def (while_ult_optab): New optab.
4119 * doc/md.texi (while_ult@var{m}@var{n}): Document.
4120 * internal-fn.def (WHILE_ULT): New internal function.
4121 * internal-fn.h (direct_internal_fn_supported_p): New override
4122 that takes two types as argument.
4123 * internal-fn.c (while_direct): New macro.
4124 (expand_while_optab_fn): New function.
4125 (convert_optab_supported_p): Likewise.
4126 (direct_while_optab_supported_p): New macro.
4127 * wide-int.h (wi::udiv_ceil): New function.
4128 * tree-vectorizer.h (rgroup_masks): New structure.
4129 (vec_loop_masks): New typedef.
4130 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
4132 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
4133 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
4134 (vect_max_vf): New function.
4135 (slpeel_make_loop_iterate_ntimes): Delete.
4136 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
4137 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
4138 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
4139 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
4140 internal-fn.h, stor-layout.h and optabs-query.h.
4141 (vect_set_loop_mask): New function.
4142 (add_preheader_seq): Likewise.
4143 (add_header_seq): Likewise.
4144 (interleave_supported_p): Likewise.
4145 (vect_maybe_permute_loop_masks): Likewise.
4146 (vect_set_loop_masks_directly): Likewise.
4147 (vect_set_loop_condition_masked): Likewise.
4148 (vect_set_loop_condition_unmasked): New function, split out from
4149 slpeel_make_loop_iterate_ntimes.
4150 (slpeel_make_loop_iterate_ntimes): Rename to..
4151 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
4152 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
4153 (vect_do_peeling): Update call accordingly.
4154 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
4156 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4157 mask_compare_type, can_fully_mask_p and fully_masked_p.
4158 (release_vec_loop_masks): New function.
4159 (_loop_vec_info): Use it to free the loop masks.
4160 (can_produce_all_loop_masks_p): New function.
4161 (vect_get_max_nscalars_per_iter): Likewise.
4162 (vect_verify_full_masking): Likewise.
4163 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
4164 retries, and free the mask rgroups before retrying. Check loop-wide
4165 reasons for disallowing fully-masked loops. Make the final decision
4166 about whether use a fully-masked loop or not.
4167 (vect_estimate_min_profitable_iters): Do not assume that peeling
4168 for the number of iterations will be needed for fully-masked loops.
4169 (vectorizable_reduction): Disable fully-masked loops.
4170 (vectorizable_live_operation): Likewise.
4171 (vect_halve_mask_nunits): New function.
4172 (vect_double_mask_nunits): Likewise.
4173 (vect_record_loop_mask): Likewise.
4174 (vect_get_loop_mask): Likewise.
4175 (vect_transform_loop): Handle the case in which the final loop
4176 iteration might handle a partial vector. Call vect_set_loop_condition
4177 instead of slpeel_make_loop_iterate_ntimes.
4178 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
4179 (check_load_store_masking): New function.
4180 (prepare_load_store_mask): Likewise.
4181 (vectorizable_store): Handle fully-masked loops.
4182 (vectorizable_load): Likewise.
4183 (supportable_widening_operation): Use vect_halve_mask_nunits for
4185 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
4186 (vect_gen_while): New function.
4187 * config/aarch64/aarch64.md (umax<mode>3): New expander.
4188 (aarch64_uqdec<mode>): New insn.
4190 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4191 Alan Hayward <alan.hayward@arm.com>
4192 David Sherwood <david.sherwood@arm.com>
4194 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
4195 (reduc_xor_scal_optab): New optabs.
4196 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
4197 (reduc_xor_scal_@var{m}): Document.
4198 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
4199 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
4201 * fold-const-call.c (fold_const_call): Handle them.
4202 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
4203 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
4204 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
4205 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
4206 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
4207 (UNSPEC_XORV): New unspecs.
4208 (optab): Add entries for them.
4209 (BITWISEV): New int iterator.
4210 (bit_reduc_op): New int attributes.
4212 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4213 Alan Hayward <alan.hayward@arm.com>
4214 David Sherwood <david.sherwood@arm.com>
4216 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
4217 * internal-fn.def (VEC_SHL_INSERT): New internal function.
4218 * optabs.def (vec_shl_insert_optab): New optab.
4219 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
4220 (duplicate_and_interleave): Likewise.
4221 * tree-vect-loop.c: Include internal-fn.h.
4222 (neutral_op_for_slp_reduction): New function, split out from
4223 get_initial_defs_for_reduction.
4224 (get_initial_def_for_reduction): Handle option 2 for variable-length
4225 vectors by loading the neutral value into a vector and then shifting
4226 the initial value into element 0.
4227 (get_initial_defs_for_reduction): Replace the code argument with
4228 the neutral value calculated by neutral_op_for_slp_reduction.
4229 Use gimple_build_vector for constant-length vectors.
4230 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
4231 but the first group_size elements have a neutral value.
4232 Use duplicate_and_interleave otherwise.
4233 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
4234 Update call to get_initial_defs_for_reduction. Handle SLP
4235 reductions for variable-length vectors by creating one vector
4236 result for each scalar result, with the elements associated
4237 with other scalar results stubbed out with the neutral value.
4238 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
4239 Require IFN_VEC_SHL_INSERT for double reductions on
4240 variable-length vectors, or SLP reductions that have
4241 a neutral value. Require can_duplicate_and_interleave_p
4242 support for variable-length unchained SLP reductions if there
4243 is no neutral value, such as for MIN/MAX reductions. Also require
4244 the number of vector elements to be a multiple of the number of
4245 SLP statements when doing variable-length unchained SLP reductions.
4246 Update call to vect_create_epilog_for_reduction.
4247 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
4248 and remove initial values.
4249 (duplicate_and_interleave): Make public.
4250 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
4251 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
4253 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4254 Alan Hayward <alan.hayward@arm.com>
4255 David Sherwood <david.sherwood@arm.com>
4257 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
4258 (can_duplicate_and_interleave_p): New function.
4259 (vect_get_and_check_slp_defs): Take the vector of statements
4260 rather than just the current one. Remove excess parentheses.
4261 Restriction rejectinon of vect_constant_def and vect_external_def
4262 for variable-length vectors to boolean types, or types for which
4263 can_duplicate_and_interleave_p is false.
4264 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
4265 (duplicate_and_interleave): New function.
4266 (vect_get_constant_vectors): Use gimple_build_vector for
4267 constant-length vectors and suitable variable-length constant
4268 vectors. Use duplicate_and_interleave for other variable-length
4269 vectors. Don't defer the update when inserting new statements.
4271 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4272 Alan Hayward <alan.hayward@arm.com>
4273 David Sherwood <david.sherwood@arm.com>
4275 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
4276 min_profitable_iters doesn't go negative.
4278 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4279 Alan Hayward <alan.hayward@arm.com>
4280 David Sherwood <david.sherwood@arm.com>
4282 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
4283 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
4284 * optabs.def (vec_mask_load_lanes_optab): New optab.
4285 (vec_mask_store_lanes_optab): Likewise.
4286 * internal-fn.def (MASK_LOAD_LANES): New internal function.
4287 (MASK_STORE_LANES): Likewise.
4288 * internal-fn.c (mask_load_lanes_direct): New macro.
4289 (mask_store_lanes_direct): Likewise.
4290 (expand_mask_load_optab_fn): Handle masked operations.
4291 (expand_mask_load_lanes_optab_fn): New macro.
4292 (expand_mask_store_optab_fn): Handle masked operations.
4293 (expand_mask_store_lanes_optab_fn): New macro.
4294 (direct_mask_load_lanes_optab_supported_p): Likewise.
4295 (direct_mask_store_lanes_optab_supported_p): Likewise.
4296 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
4298 (vect_load_lanes_supported): Likewise.
4299 * tree-vect-data-refs.c (strip_conversion): New function.
4300 (can_group_stmts_p): Likewise.
4301 (vect_analyze_data_ref_accesses): Use it instead of checking
4302 for a pair of assignments.
4303 (vect_store_lanes_supported): Take a masked_p parameter.
4304 (vect_load_lanes_supported): Likewise.
4305 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
4306 vect_store_lanes_supported and vect_load_lanes_supported.
4307 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4308 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
4309 parameter. Don't allow gaps for masked accesses.
4310 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
4311 and vect_load_lanes_supported.
4312 (get_load_store_type): Take a masked_p parameter and update
4313 call to get_group_load_store_type.
4314 (vectorizable_store): Update call to get_load_store_type.
4315 Handle IFN_MASK_STORE_LANES.
4316 (vectorizable_load): Update call to get_load_store_type.
4317 Handle IFN_MASK_LOAD_LANES.
4319 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4320 Alan Hayward <alan.hayward@arm.com>
4321 David Sherwood <david.sherwood@arm.com>
4323 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
4325 * config/aarch64/aarch64-protos.h
4326 (aarch64_sve_struct_memory_operand_p): Declare.
4327 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
4328 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
4329 (VPRED, vpred): Handle SVE structure modes.
4330 * config/aarch64/constraints.md (Utx): New constraint.
4331 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
4332 (aarch64_sve_struct_nonimmediate_operand): New predicates.
4333 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
4334 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
4335 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
4336 structure modes. Split into pieces after RA.
4337 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
4338 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
4340 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
4341 SVE structure modes.
4342 (aarch64_classify_address): Likewise.
4343 (sizetochar): Move earlier in file.
4344 (aarch64_print_operand): Handle SVE register lists.
4345 (aarch64_array_mode): New function.
4346 (aarch64_sve_struct_memory_operand_p): Likewise.
4347 (TARGET_ARRAY_MODE): Redefine.
4349 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4350 Alan Hayward <alan.hayward@arm.com>
4351 David Sherwood <david.sherwood@arm.com>
4353 * target.def (array_mode): New target hook.
4354 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
4355 * doc/tm.texi: Regenerate.
4356 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
4357 * hooks.c (hook_optmode_mode_uhwi_none): New function.
4358 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
4360 * stor-layout.c (mode_for_array): Likewise. Support polynomial
4363 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4364 Alan Hayward <alan.hayward@arm.com>
4365 David Sherwood <david.sherwood@arm.com>
4367 * fold-const.c (fold_binary_loc): Check the argument types
4368 rather than the result type when testing for a vector operation.
4370 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4372 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
4373 * doc/tm.texi: Regenerate.
4375 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4376 Alan Hayward <alan.hayward@arm.com>
4377 David Sherwood <david.sherwood@arm.com>
4379 * doc/invoke.texi (-msve-vector-bits=): Document new option.
4380 (sve): Document new AArch64 extension.
4381 * doc/md.texi (w): Extend the description of the AArch64
4382 constraint to include SVE vectors.
4383 (Upl, Upa): Document new AArch64 predicate constraints.
4384 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
4386 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
4387 (msve-vector-bits=): New option.
4388 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
4389 SVE when these are disabled.
4390 (sve): New extension.
4391 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
4392 modes. Adjust their number of units based on aarch64_sve_vg.
4393 (MAX_BITSIZE_MODE_ANY_MODE): Define.
4394 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
4395 aarch64_addr_query_type.
4396 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
4397 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
4398 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
4399 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
4400 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
4401 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
4402 (aarch64_simd_imm_zero_p): Delete.
4403 (aarch64_check_zero_based_sve_index_immediate): Declare.
4404 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4405 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4406 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4407 (aarch64_sve_float_mul_immediate_p): Likewise.
4408 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4410 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
4411 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
4412 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
4413 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
4414 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
4415 (aarch64_regmode_natural_size): Likewise.
4416 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
4417 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
4419 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
4420 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
4421 for VG and the SVE predicate registers.
4422 (V_ALIASES): Add a "z"-prefixed alias.
4423 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
4424 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
4425 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
4426 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
4427 (REG_CLASS_NAMES): Add entries for them.
4428 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
4429 and the predicate registers.
4430 (aarch64_sve_vg): Declare.
4431 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
4432 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
4433 (REGMODE_NATURAL_SIZE): Define.
4434 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
4436 * config/aarch64/aarch64.c: Include cfgrtl.h.
4437 (simd_immediate_info): Add a constructor for series vectors,
4438 and an associated step field.
4439 (aarch64_sve_vg): New variable.
4440 (aarch64_dbx_register_number): Handle VG and the predicate registers.
4441 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
4442 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
4443 (VEC_ANY_DATA, VEC_STRUCT): New constants.
4444 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
4445 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
4446 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
4447 (aarch64_get_mask_mode): New functions.
4448 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
4449 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
4450 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
4451 predicate modes and predicate registers. Explicitly restrict
4452 GPRs to modes of 16 bytes or smaller. Only allow FP registers
4453 to store a vector mode if it is recognized by
4454 aarch64_classify_vector_mode.
4455 (aarch64_regmode_natural_size): New function.
4456 (aarch64_hard_regno_caller_save_mode): Return the original mode
4458 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
4459 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
4460 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
4461 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
4463 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
4464 does not overlap dest if the function is frame-related. Handle
4466 (aarch64_split_add_offset): New function.
4467 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
4468 them aarch64_add_offset.
4469 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
4470 and update call to aarch64_sub_sp.
4471 (aarch64_add_cfa_expression): New function.
4472 (aarch64_expand_prologue): Pass extra temporary registers to the
4473 functions above. Handle the case in which we need to emit new
4474 DW_CFA_expressions for registers that were originally saved
4475 relative to the stack pointer, but now have to be expressed
4476 relative to the frame pointer.
4477 (aarch64_output_mi_thunk): Pass extra temporary registers to the
4479 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
4480 IP0 and IP1 values for SVE frames.
4481 (aarch64_expand_vec_series): New function.
4482 (aarch64_expand_sve_widened_duplicate): Likewise.
4483 (aarch64_expand_sve_const_vector): Likewise.
4484 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
4485 Handle SVE constants. Use emit_move_insn to move a force_const_mem
4486 into the register, rather than emitting a SET directly.
4487 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
4488 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
4489 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
4490 (offset_9bit_signed_scaled_p): New functions.
4491 (aarch64_replicate_bitmask_imm): New function.
4492 (aarch64_bitmask_imm): Use it.
4493 (aarch64_cannot_force_const_mem): Reject expressions involving
4494 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
4495 (aarch64_classify_index): Handle SVE indices, by requiring
4496 a plain register index with a scale that matches the element size.
4497 (aarch64_classify_address): Handle SVE addresses. Assert that
4498 the mode of the address is VOIDmode or an integer mode.
4499 Update call to aarch64_classify_symbol.
4500 (aarch64_classify_symbolic_expression): Update call to
4501 aarch64_classify_symbol.
4502 (aarch64_const_vec_all_in_range_p): New function.
4503 (aarch64_print_vector_float_operand): Likewise.
4504 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
4505 "vN" for FP registers with SVE modes. Handle (const ...) vectors
4506 and the FP immediates 1.0 and 0.5.
4507 (aarch64_print_address_internal): Handle SVE addresses.
4508 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
4509 (aarch64_regno_regclass): Handle predicate registers.
4510 (aarch64_secondary_reload): Handle big-endian reloads of SVE
4512 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
4513 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
4514 (aarch64_convert_sve_vector_bits): New function.
4515 (aarch64_override_options): Use it to handle -msve-vector-bits=.
4516 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4518 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
4519 Handle SVE vector and predicate modes. Accept VL-based constants
4520 that need only one temporary register, and VL offsets that require
4521 no temporary registers.
4522 (aarch64_conditional_register_usage): Mark the predicate registers
4523 as fixed if SVE isn't available.
4524 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
4525 Return true for SVE vector and predicate modes.
4526 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
4527 rather than an unsigned int. Handle SVE modes.
4528 (aarch64_preferred_simd_mode): Update call accordingly. Handle
4530 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
4532 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4533 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4534 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4535 (aarch64_sve_float_mul_immediate_p): New functions.
4536 (aarch64_sve_valid_immediate): New function.
4537 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
4538 Explicitly reject structure modes. Check for INDEX constants.
4539 Handle PTRUE and PFALSE constants.
4540 (aarch64_check_zero_based_sve_index_immediate): New function.
4541 (aarch64_simd_imm_zero_p): Delete.
4542 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
4543 vector modes. Accept constants in the range of CNT[BHWD].
4544 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
4545 ask for an Advanced SIMD mode.
4546 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
4547 (aarch64_simd_vector_alignment): Handle SVE predicates.
4548 (aarch64_vectorize_preferred_vector_alignment): New function.
4549 (aarch64_simd_vector_alignment_reachable): Use it instead of
4551 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
4552 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
4554 (MAX_VECT_LEN): Delete.
4555 (expand_vec_perm_d): Add a vec_flags field.
4556 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
4557 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
4558 (aarch64_evpc_ext): Don't apply a big-endian lane correction
4560 (aarch64_evpc_rev): Rename to...
4561 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
4562 (aarch64_evpc_rev_global): New function.
4563 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
4564 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
4566 (aarch64_evpc_sve_tbl): New function.
4567 (aarch64_expand_vec_perm_const_1): Update after rename of
4568 aarch64_evpc_rev. Handle SVE permutes too, trying
4569 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
4570 than aarch64_evpc_tbl.
4571 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
4572 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
4573 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
4574 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
4575 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
4576 (aarch64_expand_sve_vcond): New functions.
4577 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
4578 of aarch64_vector_mode_p.
4579 (aarch64_dwarf_poly_indeterminate_value): New function.
4580 (aarch64_compute_pressure_classes): Likewise.
4581 (aarch64_can_change_mode_class): Likewise.
4582 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
4583 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
4584 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
4585 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
4586 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
4587 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
4588 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
4589 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
4591 (Dn, Dl, Dr): Accept const as well as const_vector.
4592 (Dz): Likewise. Compare against CONST0_RTX.
4593 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
4594 of "vector" where appropriate.
4595 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
4596 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
4597 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
4598 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
4599 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
4600 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
4601 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
4602 (v_int_equiv): Extend to SVE modes.
4603 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
4605 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
4606 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
4607 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
4608 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
4609 (SVE_COND_FP_CMP): New int iterators.
4610 (perm_hilo): Handle the new unpack unspecs.
4611 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
4613 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
4614 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
4615 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
4616 (aarch64_equality_operator, aarch64_constant_vector_operand)
4617 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
4618 (aarch64_sve_nonimmediate_operand): Likewise.
4619 (aarch64_sve_general_operand): Likewise.
4620 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
4621 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
4622 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
4623 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
4624 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
4625 (aarch64_sve_float_arith_immediate): Likewise.
4626 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
4627 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
4628 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
4629 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
4630 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
4631 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
4632 (aarch64_sve_float_arith_operand): Likewise.
4633 (aarch64_sve_float_arith_with_sub_operand): Likewise.
4634 (aarch64_sve_float_mul_operand): Likewise.
4635 (aarch64_sve_vec_perm_operand): Likewise.
4636 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
4637 (aarch64_mov_operand): Accept const_poly_int and const_vector.
4638 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
4639 as well as const_vector.
4640 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
4641 in file. Use CONST0_RTX and CONSTM1_RTX.
4642 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
4643 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
4644 Use aarch64_simd_imm_zero.
4645 * config/aarch64/aarch64-sve.md: New file.
4646 * config/aarch64/aarch64.md: Include it.
4647 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
4648 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
4649 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
4650 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
4651 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
4652 (sve): New attribute.
4653 (enabled): Disable instructions with the sve attribute unless
4655 (movqi, movhi): Pass CONST_POLY_INT operaneds through
4656 aarch64_expand_mov_immediate.
4657 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
4658 CNT[BHSD] immediates.
4659 (movti): Split CONST_POLY_INT moves into two halves.
4660 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
4661 Split additions that need a temporary here if the destination
4662 is the stack pointer.
4663 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
4664 (*add<mode>3_poly_1): New instruction.
4665 (set_clobber_cc): New expander.
4667 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4669 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
4670 parameter and use it instead of GET_MODE_SIZE (innermode). Use
4671 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
4672 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
4673 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
4674 Change innermode from fixed_mode_size to machine_mode.
4675 (simplify_subreg): Update call accordingly. Handle a constant-sized
4676 subreg of a variable-length CONST_VECTOR.
4678 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4679 Alan Hayward <alan.hayward@arm.com>
4680 David Sherwood <david.sherwood@arm.com>
4682 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
4683 (add_offset_to_base): New function, split out from...
4684 (create_mem_ref): ...here. When handling a scale other than 1,
4685 check first whether the address is valid without the offset.
4686 Add it into the base if so, leaving the index and scale as-is.
4688 2018-01-12 Jakub Jelinek <jakub@redhat.com>
4691 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
4692 fold_for_warn before checking if arg2 is INTEGER_CST.
4694 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
4696 * config/rs6000/predicates.md (load_multiple_operation): Delete.
4697 (store_multiple_operation): Delete.
4698 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
4699 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
4700 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
4701 guarded by TARGET_STRING.
4702 (rs6000_output_load_multiple): Delete.
4703 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
4704 OPTION_MASK_STRING / TARGET_STRING handling.
4705 (print_operand) <'N', 'O'>: Add comment that these are unused now.
4706 (const rs6000_opt_masks) <"string">: Change mask to 0.
4707 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
4708 (MASK_STRING): Delete.
4709 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
4711 (load_multiple): Delete.
4718 (store_multiple): Delete.
4725 (movmemsi_8reg): Delete.
4726 (corresponding unnamed define_insn): Delete.
4727 (movmemsi_6reg): Delete.
4728 (corresponding unnamed define_insn): Delete.
4729 (movmemsi_4reg): Delete.
4730 (corresponding unnamed define_insn): Delete.
4731 (movmemsi_2reg): Delete.
4732 (corresponding unnamed define_insn): Delete.
4733 (movmemsi_1reg): Delete.
4734 (corresponding unnamed define_insn): Delete.
4735 * config/rs6000/rs6000.opt (mno-string): New.
4736 (mstring): Replace by deprecation warning stub.
4737 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
4739 2018-01-12 Jakub Jelinek <jakub@redhat.com>
4741 * regrename.c (regrename_do_replace): If replacing the same
4742 reg multiple times, try to reuse last created gen_raw_REG.
4745 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
4746 main to workaround a bug in GDB.
4748 2018-01-12 Tom de Vries <tom@codesourcery.com>
4751 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
4753 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
4755 PR rtl-optimization/80481
4756 * ira-color.c (get_cap_member): New function.
4757 (allocnos_conflict_by_live_ranges_p): Use it.
4758 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
4759 (setup_slot_coalesced_allocno_live_ranges): Ditto.
4761 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
4764 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
4765 (*saddl_se_1): Ditto.
4767 (*ssubl_se_1): Ditto.
4769 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
4771 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
4772 rather than wi::to_widest for DR_INITs.
4773 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
4774 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
4775 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
4777 (vect_analyze_group_access_1): Note that here.
4779 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
4781 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
4782 polynomial type sizes.
4784 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
4786 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
4787 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
4788 (gimple_add_tmp_var): Likewise.
4790 2018-01-12 Martin Liska <mliska@suse.cz>
4792 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
4793 (gimple_alloc_sizes): Likewise.
4794 (dump_gimple_statistics): Use PRIu64 in printf format.
4795 * gimple.h: Change uint64_t to int.
4797 2018-01-12 Martin Liska <mliska@suse.cz>
4799 * tree-core.h: Use uint64_t instead of int.
4800 * tree.c (tree_node_counts): Likewise.
4801 (tree_node_sizes): Likewise.
4802 (dump_tree_statistics): Use PRIu64 in printf format.
4804 2018-01-12 Martin Liska <mliska@suse.cz>
4806 * Makefile.in: As qsort_chk is implemented in vec.c, add
4807 vec.o to linkage of gencfn-macros.
4808 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
4809 passing the info to record_node_allocation_statistics.
4810 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
4812 * ggc-common.c (struct ggc_usage): Add operator== and use
4813 it in operator< and compare function.
4814 * mem-stats.h (struct mem_usage): Likewise.
4815 * vec.c (struct vec_usage): Remove operator< and compare
4816 function. Can be simply inherited.
4818 2018-01-12 Martin Jambor <mjambor@suse.cz>
4821 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
4822 * tree-ssa-math-opts.c: Include domwalk.h.
4823 (convert_mult_to_fma_1): New function.
4824 (fma_transformation_info): New type.
4825 (fma_deferring_state): Likewise.
4826 (cancel_fma_deferring): New function.
4827 (result_of_phi): Likewise.
4828 (last_fma_candidate_feeds_initial_phi): Likewise.
4829 (convert_mult_to_fma): Added deferring logic, split actual
4830 transformation to convert_mult_to_fma_1.
4831 (math_opts_dom_walker): New type.
4832 (math_opts_dom_walker::after_dom_children): New method, body moved
4833 here from pass_optimize_widening_mul::execute, added deferring logic
4835 (pass_optimize_widening_mul::execute): Moved most of code to
4836 math_opts_dom_walker::after_dom_children.
4837 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
4838 * config/i386/i386.c (ix86_option_override_internal): Added
4839 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
4841 2018-01-12 Richard Biener <rguenther@suse.de>
4844 * dwarf2out.c (gen_variable_die): Do not reset old_die for
4845 inline instance vars.
4847 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
4850 * config/rx/rx.c (rx_is_restricted_memory_address):
4853 2018-01-12 Richard Biener <rguenther@suse.de>
4855 PR tree-optimization/80846
4856 * target.def (split_reduction): New target hook.
4857 * targhooks.c (default_split_reduction): New function.
4858 * targhooks.h (default_split_reduction): Declare.
4859 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
4860 target requests first reduce vectors by combining low and high
4862 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
4863 (get_vectype_for_scalar_type_and_size): Export.
4864 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
4865 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
4866 * doc/tm.texi: Regenerate.
4867 * config/i386/i386.c (ix86_split_reduction): Implement
4868 TARGET_VECTORIZE_SPLIT_REDUCTION.
4870 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4873 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
4874 in PIC mode except for TARGET_VXWORKS_RTP.
4875 * config/sparc/sparc.c: Include cfgrtl.h.
4876 (TARGET_INIT_PIC_REG): Define.
4877 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
4878 (sparc_pic_register_p): New predicate.
4879 (sparc_legitimate_address_p): Use it.
4880 (sparc_legitimize_pic_address): Likewise.
4881 (sparc_delegitimize_address): Likewise.
4882 (sparc_mode_dependent_address_p): Likewise.
4883 (gen_load_pcrel_sym): Remove 4th parameter.
4884 (load_got_register): Adjust call to above. Remove obsolete stuff.
4885 (sparc_expand_prologue): Do not call load_got_register here.
4886 (sparc_flat_expand_prologue): Likewise.
4887 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
4888 (sparc_use_pseudo_pic_reg): New function.
4889 (sparc_init_pic_reg): Likewise.
4890 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
4891 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
4893 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
4895 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
4896 Add item for branch_cost.
4898 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4900 PR rtl-optimization/83565
4901 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
4902 not extend the result to a larger mode for rotate operations.
4903 (num_sign_bit_copies1): Likewise.
4905 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4908 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
4910 Use values-Xc.o for -pedantic.
4911 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
4913 2018-01-12 Martin Liska <mliska@suse.cz>
4916 * ipa-devirt.c (final_warning_record::grow_type_warnings):
4918 (possible_polymorphic_call_targets): Use it.
4919 (ipa_devirt): Likewise.
4921 2018-01-12 Martin Liska <mliska@suse.cz>
4923 * profile-count.h (enum profile_quality): Use 0 as invalid
4924 enum value of profile_quality.
4926 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
4928 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
4929 -mext-string options.
4931 2018-01-12 Richard Biener <rguenther@suse.de>
4933 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
4934 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
4935 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
4937 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
4939 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
4941 * configure.ac (--with-long-double-format): Add support for the
4942 configuration option to change the default long double format on
4944 * config.gcc (powerpc*-linux*-*): Likewise.
4945 * configure: Regenerate.
4946 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
4947 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
4948 used without modification.
4950 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4952 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
4953 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
4954 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
4955 MISC_BUILTIN_SPEC_BARRIER.
4956 (rs6000_init_builtins): Likewise.
4957 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
4959 (speculation_barrier): New define_insn.
4960 * doc/extend.texi: Document __builtin_speculation_barrier.
4962 2018-01-11 Jakub Jelinek <jakub@redhat.com>
4965 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
4966 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
4967 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
4969 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
4970 integral modes instead of "ss" and "sd".
4971 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
4972 vectors with 32-bit and 64-bit elements.
4973 (vecdupssescalarmodesuffix): New mode attribute.
4974 (vec_dup<mode>): Use it.
4976 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
4979 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
4980 frame if argument is passed on stack.
4982 2018-01-11 Jakub Jelinek <jakub@redhat.com>
4985 * ree.c (combine_reaching_defs): Optimize also
4986 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
4987 reg2=any_extend(exp); reg1=reg2;, formatting fix.
4989 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
4992 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
4994 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
4997 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
4998 after they are computed.
5000 2018-01-11 Bin Cheng <bin.cheng@arm.com>
5002 PR tree-optimization/83695
5003 * gimple-loop-linterchange.cc
5004 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
5005 reset cached scev information after interchange.
5006 (pass_linterchange::execute): Remove call to scev_reset_htab.
5008 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5010 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
5011 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
5012 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
5013 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
5014 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
5015 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
5016 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
5017 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
5018 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
5019 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
5020 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
5021 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
5022 (V_lane_reg): Likewise.
5023 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
5025 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
5026 (vfmal_lane_low<mode>_intrinsic,
5027 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
5028 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
5029 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
5030 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
5031 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
5032 vfmsl_lane_high<mode>_intrinsic): New define_insns.
5034 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5036 * config/arm/arm-cpus.in (fp16fml): New feature.
5037 (ALL_SIMD): Add fp16fml.
5038 (armv8.2-a): Add fp16fml as an option.
5039 (armv8.3-a): Likewise.
5040 (armv8.4-a): Add fp16fml as part of fp16.
5041 * config/arm/arm.h (TARGET_FP16FML): Define.
5042 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
5044 * config/arm/arm-modes.def (V2HF): Define.
5045 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
5046 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
5047 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
5048 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
5049 vfmsl_low, vfmsl_high): New set of builtins.
5050 * config/arm/iterators.md (PLUSMINUS): New code iterator.
5051 (vfml_op): New code attribute.
5052 (VFMLHALVES): New int iterator.
5053 (VFML, VFMLSEL): New mode attributes.
5054 (V_reg): Define mapping for V2HF.
5055 (V_hi, V_lo): New mode attributes.
5056 (VF_constraint): Likewise.
5057 (vfml_half, vfml_half_selector): New int attributes.
5058 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
5060 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
5061 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
5063 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
5064 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
5065 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
5066 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
5068 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
5069 Document new effective target and option set.
5071 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5073 * config/arm/arm-cpus.in (armv8_4): New feature.
5074 (ARMv8_4a): New fgroup.
5075 (armv8.4-a): New arch.
5076 * config/arm/arm-tables.opt: Regenerate.
5077 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
5078 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
5079 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
5080 Add matching rules for -march=armv8.4-a and extensions.
5081 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
5083 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
5086 * config/rx/rx.md (BW): New mode attribute.
5087 (sync_lock_test_and_setsi): Add mode suffix to insn output.
5089 2018-01-11 Richard Biener <rguenther@suse.de>
5091 PR tree-optimization/83435
5092 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
5093 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
5094 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
5096 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5097 Alan Hayward <alan.hayward@arm.com>
5098 David Sherwood <david.sherwood@arm.com>
5100 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
5102 (aarch64_classify_address): Initialize it. Track polynomial offsets.
5103 (aarch64_print_address_internal): Use it to check for a zero offset.
5105 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5106 Alan Hayward <alan.hayward@arm.com>
5107 David Sherwood <david.sherwood@arm.com>
5109 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
5110 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
5111 Return a poly_int64 rather than a HOST_WIDE_INT.
5112 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
5113 rather than a HOST_WIDE_INT.
5114 * config/aarch64/aarch64.h (aarch64_frame): Protect with
5115 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
5116 hard_fp_offset, frame_size, initial_adjust, callee_offset and
5117 final_offset from HOST_WIDE_INT to poly_int64.
5118 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
5119 to_constant when getting the number of units in an Advanced SIMD
5121 (aarch64_builtin_vectorized_function): Check for a constant number
5123 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
5125 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
5126 attribute instead of GET_MODE_NUNITS.
5127 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
5128 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
5129 GET_MODE_SIZE for fixed-size registers.
5130 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
5131 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
5132 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
5133 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
5134 (aarch64_print_operand, aarch64_print_address_internal)
5135 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
5136 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
5137 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
5138 Handle polynomial GET_MODE_SIZE.
5139 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
5140 wider than SImode without modification.
5141 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
5142 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
5143 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
5144 passing and returning SVE modes.
5145 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
5146 rather than GEN_INT.
5147 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
5148 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
5149 (aarch64_allocate_and_probe_stack_space): Likewise.
5150 (aarch64_layout_frame): Cope with polynomial offsets.
5151 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
5152 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
5154 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
5155 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
5156 poly_int64 rather than a HOST_WIDE_INT.
5157 (aarch64_get_separate_components, aarch64_process_components)
5158 (aarch64_expand_prologue, aarch64_expand_epilogue)
5159 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
5160 (aarch64_anchor_offset): New function, split out from...
5161 (aarch64_legitimize_address): ...here.
5162 (aarch64_builtin_vectorization_cost): Handle polynomial
5163 TYPE_VECTOR_SUBPARTS.
5164 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
5166 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
5167 number of elements from the PARALLEL rather than the mode.
5168 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
5169 rather than GET_MODE_BITSIZE.
5170 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
5171 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
5172 (aarch64_expand_vec_perm_const_1): Handle polynomial
5173 d->perm.length () and d->perm elements.
5174 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
5175 Apply to_constant to d->perm elements.
5176 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
5177 polynomial CONST_VECTOR_NUNITS.
5178 (aarch64_move_pointer): Take amount as a poly_int64 rather
5180 (aarch64_progress_pointer): Avoid temporary variable.
5181 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
5182 the mode attribute instead of GET_MODE.
5184 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5185 Alan Hayward <alan.hayward@arm.com>
5186 David Sherwood <david.sherwood@arm.com>
5188 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
5189 x exists before using it.
5190 (aarch64_add_constant_internal): Rename to...
5191 (aarch64_add_offset_1): ...this. Replace regnum with separate
5192 src and dest rtxes. Handle the case in which they're different,
5193 including when the offset is zero. Replace scratchreg with an rtx.
5194 Use 2 additions if there is no spare register into which we can
5195 move a 16-bit constant.
5196 (aarch64_add_constant): Delete.
5197 (aarch64_add_offset): Replace reg with separate src and dest
5198 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
5199 Use aarch64_add_offset_1.
5200 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
5201 an rtx rather than an int. Take the delta as a poly_int64
5202 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
5203 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
5204 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
5205 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
5206 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
5208 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
5209 aarch64_add_constant.
5211 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5213 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
5214 Use scalar_float_mode.
5216 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5218 * config/aarch64/aarch64-simd.md
5219 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
5220 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
5221 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
5222 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
5223 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
5224 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
5225 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
5226 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
5227 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
5228 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
5230 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5233 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
5234 targ_options->x_arm_arch_string is non NULL.
5236 2018-01-11 Tamar Christina <tamar.christina@arm.com>
5238 * config/aarch64/aarch64.h
5239 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
5241 2018-01-11 Sudakshina Das <sudi.das@arm.com>
5244 * expmed.c (emit_store_flag_force): Swap if const op0
5245 and change VOIDmode to mode of op0.
5247 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5249 PR rtl-optimization/83761
5250 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
5251 than bytes to mode_for_size.
5253 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5256 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
5257 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
5260 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5263 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
5264 when in layout mode.
5265 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
5266 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
5269 2018-01-10 Michael Collison <michael.collison@arm.com>
5271 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
5272 * config/aarch64/aarch64-option-extension.def: Add
5273 AARCH64_OPT_EXTENSION of 'fp16fml'.
5274 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5275 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
5276 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
5277 * config/aarch64/constraints.md (Ui7): New constraint.
5278 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
5279 (VFMLA_SEL_W): Ditto.
5282 (VFMLA16_LOW): New int iterator.
5283 (VFMLA16_HIGH): Ditto.
5284 (UNSPEC_FMLAL): New unspec.
5285 (UNSPEC_FMLSL): Ditto.
5286 (UNSPEC_FMLAL2): Ditto.
5287 (UNSPEC_FMLSL2): Ditto.
5288 (f16mac): New code attribute.
5289 * config/aarch64/aarch64-simd-builtins.def
5290 (aarch64_fmlal_lowv2sf): Ditto.
5291 (aarch64_fmlsl_lowv2sf): Ditto.
5292 (aarch64_fmlalq_lowv4sf): Ditto.
5293 (aarch64_fmlslq_lowv4sf): Ditto.
5294 (aarch64_fmlal_highv2sf): Ditto.
5295 (aarch64_fmlsl_highv2sf): Ditto.
5296 (aarch64_fmlalq_highv4sf): Ditto.
5297 (aarch64_fmlslq_highv4sf): Ditto.
5298 (aarch64_fmlal_lane_lowv2sf): Ditto.
5299 (aarch64_fmlsl_lane_lowv2sf): Ditto.
5300 (aarch64_fmlal_laneq_lowv2sf): Ditto.
5301 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
5302 (aarch64_fmlalq_lane_lowv4sf): Ditto.
5303 (aarch64_fmlsl_lane_lowv4sf): Ditto.
5304 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
5305 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
5306 (aarch64_fmlal_lane_highv2sf): Ditto.
5307 (aarch64_fmlsl_lane_highv2sf): Ditto.
5308 (aarch64_fmlal_laneq_highv2sf): Ditto.
5309 (aarch64_fmlsl_laneq_highv2sf): Ditto.
5310 (aarch64_fmlalq_lane_highv4sf): Ditto.
5311 (aarch64_fmlsl_lane_highv4sf): Ditto.
5312 (aarch64_fmlalq_laneq_highv4sf): Ditto.
5313 (aarch64_fmlsl_laneq_highv4sf): Ditto.
5314 * config/aarch64/aarch64-simd.md:
5315 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
5316 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5317 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
5318 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5319 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
5320 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
5321 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
5322 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
5323 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
5324 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
5325 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
5326 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
5327 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
5328 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
5329 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
5330 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
5331 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
5332 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
5333 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
5334 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
5335 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
5336 (vfmlsl_low_u32): Ditto.
5337 (vfmlalq_low_u32): Ditto.
5338 (vfmlslq_low_u32): Ditto.
5339 (vfmlal_high_u32): Ditto.
5340 (vfmlsl_high_u32): Ditto.
5341 (vfmlalq_high_u32): Ditto.
5342 (vfmlslq_high_u32): Ditto.
5343 (vfmlal_lane_low_u32): Ditto.
5344 (vfmlsl_lane_low_u32): Ditto.
5345 (vfmlal_laneq_low_u32): Ditto.
5346 (vfmlsl_laneq_low_u32): Ditto.
5347 (vfmlalq_lane_low_u32): Ditto.
5348 (vfmlslq_lane_low_u32): Ditto.
5349 (vfmlalq_laneq_low_u32): Ditto.
5350 (vfmlslq_laneq_low_u32): Ditto.
5351 (vfmlal_lane_high_u32): Ditto.
5352 (vfmlsl_lane_high_u32): Ditto.
5353 (vfmlal_laneq_high_u32): Ditto.
5354 (vfmlsl_laneq_high_u32): Ditto.
5355 (vfmlalq_lane_high_u32): Ditto.
5356 (vfmlslq_lane_high_u32): Ditto.
5357 (vfmlalq_laneq_high_u32): Ditto.
5358 (vfmlslq_laneq_high_u32): Ditto.
5359 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
5360 (AARCH64_FL_FOR_ARCH8_4): New.
5361 (AARCH64_ISA_F16FML): New ISA flag.
5362 (TARGET_F16FML): New feature flag for fp16fml.
5363 (doc/invoke.texi): Document new fp16fml option.
5365 2018-01-10 Michael Collison <michael.collison@arm.com>
5367 * config/aarch64/aarch64-builtins.c:
5368 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
5369 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5370 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
5371 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
5372 (AARCH64_ISA_SHA3): New ISA flag.
5373 (TARGET_SHA3): New feature flag for sha3.
5374 * config/aarch64/iterators.md (sha512_op): New int attribute.
5375 (CRYPTO_SHA512): New int iterator.
5376 (UNSPEC_SHA512H): New unspec.
5377 (UNSPEC_SHA512H2): Ditto.
5378 (UNSPEC_SHA512SU0): Ditto.
5379 (UNSPEC_SHA512SU1): Ditto.
5380 * config/aarch64/aarch64-simd-builtins.def
5381 (aarch64_crypto_sha512hqv2di): New builtin.
5382 (aarch64_crypto_sha512h2qv2di): Ditto.
5383 (aarch64_crypto_sha512su0qv2di): Ditto.
5384 (aarch64_crypto_sha512su1qv2di): Ditto.
5385 (aarch64_eor3qv8hi): Ditto.
5386 (aarch64_rax1qv2di): Ditto.
5387 (aarch64_xarqv2di): Ditto.
5388 (aarch64_bcaxqv8hi): Ditto.
5389 * config/aarch64/aarch64-simd.md:
5390 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
5391 (aarch64_crypto_sha512su0qv2di): Ditto.
5392 (aarch64_crypto_sha512su1qv2di): Ditto.
5393 (aarch64_eor3qv8hi): Ditto.
5394 (aarch64_rax1qv2di): Ditto.
5395 (aarch64_xarqv2di): Ditto.
5396 (aarch64_bcaxqv8hi): Ditto.
5397 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
5398 (vsha512h2q_u64): Ditto.
5399 (vsha512su0q_u64): Ditto.
5400 (vsha512su1q_u64): Ditto.
5401 (veor3q_u16): Ditto.
5402 (vrax1q_u64): Ditto.
5404 (vbcaxq_u16): Ditto.
5405 * config/arm/types.md (crypto_sha512): New type attribute.
5406 (crypto_sha3): Ditto.
5407 (doc/invoke.texi): Document new sha3 option.
5409 2018-01-10 Michael Collison <michael.collison@arm.com>
5411 * config/aarch64/aarch64-builtins.c:
5412 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
5413 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5414 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
5415 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
5416 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
5417 (AARCH64_ISA_SM4): New ISA flag.
5418 (TARGET_SM4): New feature flag for sm4.
5419 * config/aarch64/aarch64-simd-builtins.def
5420 (aarch64_sm3ss1qv4si): Ditto.
5421 (aarch64_sm3tt1aq4si): Ditto.
5422 (aarch64_sm3tt1bq4si): Ditto.
5423 (aarch64_sm3tt2aq4si): Ditto.
5424 (aarch64_sm3tt2bq4si): Ditto.
5425 (aarch64_sm3partw1qv4si): Ditto.
5426 (aarch64_sm3partw2qv4si): Ditto.
5427 (aarch64_sm4eqv4si): Ditto.
5428 (aarch64_sm4ekeyqv4si): Ditto.
5429 * config/aarch64/aarch64-simd.md:
5430 (aarch64_sm3ss1qv4si): Ditto.
5431 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
5432 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
5433 (aarch64_sm4eqv4si): Ditto.
5434 (aarch64_sm4ekeyqv4si): Ditto.
5435 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
5436 (sm3part_op): Ditto.
5437 (CRYPTO_SM3TT): Ditto.
5438 (CRYPTO_SM3PART): Ditto.
5439 (UNSPEC_SM3SS1): New unspec.
5440 (UNSPEC_SM3TT1A): Ditto.
5441 (UNSPEC_SM3TT1B): Ditto.
5442 (UNSPEC_SM3TT2A): Ditto.
5443 (UNSPEC_SM3TT2B): Ditto.
5444 (UNSPEC_SM3PARTW1): Ditto.
5445 (UNSPEC_SM3PARTW2): Ditto.
5446 (UNSPEC_SM4E): Ditto.
5447 (UNSPEC_SM4EKEY): Ditto.
5448 * config/aarch64/constraints.md (Ui2): New constraint.
5449 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
5450 * config/arm/types.md (crypto_sm3): New type attribute.
5451 (crypto_sm4): Ditto.
5452 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
5453 (vsm3tt1aq_u32): Ditto.
5454 (vsm3tt1bq_u32): Ditto.
5455 (vsm3tt2aq_u32): Ditto.
5456 (vsm3tt2bq_u32): Ditto.
5457 (vsm3partw1q_u32): Ditto.
5458 (vsm3partw2q_u32): Ditto.
5459 (vsm4eq_u32): Ditto.
5460 (vsm4ekeyq_u32): Ditto.
5461 (doc/invoke.texi): Document new sm4 option.
5463 2018-01-10 Michael Collison <michael.collison@arm.com>
5465 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
5466 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
5467 (AARCH64_FL_FOR_ARCH8_4): New.
5468 (AARCH64_FL_V8_4): New flag.
5469 (doc/invoke.texi): Document new armv8.4-a option.
5471 2018-01-10 Michael Collison <michael.collison@arm.com>
5473 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5474 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
5475 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
5476 * config/aarch64/aarch64-option-extension.def: Add
5477 AARCH64_OPT_EXTENSION of 'sha2'.
5478 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
5479 (crypto): Disable sha2 and aes if crypto disabled.
5480 (crypto): Enable aes and sha2 if enabled.
5481 (simd): Disable sha2 and aes if simd disabled.
5482 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
5484 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
5485 (TARGET_SHA2): New feature flag for sha2.
5486 (TARGET_AES): New feature flag for aes.
5487 * config/aarch64/aarch64-simd.md:
5488 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
5489 conditional on TARGET_AES.
5490 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
5491 (aarch64_crypto_sha1hsi): Make pattern conditional
5493 (aarch64_crypto_sha1hv4si): Ditto.
5494 (aarch64_be_crypto_sha1hv4si): Ditto.
5495 (aarch64_crypto_sha1su1v4si): Ditto.
5496 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
5497 (aarch64_crypto_sha1su0v4si): Ditto.
5498 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
5499 (aarch64_crypto_sha256su0v4si): Ditto.
5500 (aarch64_crypto_sha256su1v4si): Ditto.
5501 (doc/invoke.texi): Document new aes and sha2 options.
5503 2018-01-10 Martin Sebor <msebor@redhat.com>
5505 PR tree-optimization/83781
5506 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
5509 2018-01-11 Martin Sebor <msebor@gmail.com>
5510 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5512 PR tree-optimization/83501
5513 PR tree-optimization/81703
5515 * tree-ssa-strlen.c (get_string_cst): Rename...
5516 (get_string_len): ...to this. Handle global constants.
5517 (handle_char_store): Adjust.
5519 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
5520 Jim Wilson <jimw@sifive.com>
5522 * config/riscv/riscv-protos.h (riscv_output_return): New.
5523 * config/riscv/riscv.c (struct machine_function): New naked_p field.
5524 (riscv_attribute_table, riscv_output_return),
5525 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
5526 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
5527 (riscv_compute_frame_info): Only compute frame->mask if not a naked
5529 (riscv_expand_prologue): Add early return for naked function.
5530 (riscv_expand_epilogue): Likewise.
5531 (riscv_function_ok_for_sibcall): Return false for naked function.
5532 (riscv_set_current_function): New.
5533 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
5534 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
5535 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
5536 * doc/extend.texi (RISC-V Function Attributes): New.
5538 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
5540 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
5541 check for 128-bit long double before checking TCmode.
5542 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
5543 128-bit long doubles before checking TFmode or TCmode.
5544 (FLOAT128_IBM_P): Likewise.
5546 2018-01-10 Martin Sebor <msebor@redhat.com>
5548 PR tree-optimization/83671
5549 * builtins.c (c_strlen): Unconditionally return zero for the empty
5551 Use -Warray-bounds for warnings.
5552 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
5553 for non-constant array indices with COMPONENT_REF, arrays of
5554 arrays, and pointers to arrays.
5555 (gimple_fold_builtin_strlen): Determine and set length range for
5556 non-constant character arrays.
5558 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
5561 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
5564 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
5566 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
5568 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5571 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
5572 VECTOR_MEM_ALTIVEC_OR_VSX_P.
5573 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
5574 indexed_or_indirect_operand predicate.
5575 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
5576 (*vsx_le_perm_load_v8hi): Likewise.
5577 (*vsx_le_perm_load_v16qi): Likewise.
5578 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
5579 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
5580 (*vsx_le_perm_store_v8hi): Likewise.
5581 (*vsx_le_perm_store_v16qi): Likewise.
5582 (eight unnamed splitters): Likewise.
5584 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5586 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
5587 * config/rs6000/emmintrin.h: Likewise.
5588 * config/rs6000/mmintrin.h: Likewise.
5589 * config/rs6000/xmmintrin.h: Likewise.
5591 2018-01-10 David Malcolm <dmalcolm@redhat.com>
5594 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
5596 * tree.c (tree_nop_conversion): Return true for location wrapper
5598 (maybe_wrap_with_location): New function.
5599 (selftest::check_strip_nops): New function.
5600 (selftest::test_location_wrappers): New function.
5601 (selftest::tree_c_tests): Call it.
5602 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
5603 (maybe_wrap_with_location): New decl.
5604 (EXPR_LOCATION_WRAPPER_P): New macro.
5605 (location_wrapper_p): New inline function.
5606 (tree_strip_any_location_wrapper): New inline function.
5608 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
5611 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
5612 stack_realign_offset for the largest alignment of stack slot
5614 (ix86_find_max_used_stack_alignment): New function.
5615 (ix86_finalize_stack_frame_flags): Use it. Set
5616 max_used_stack_alignment if we don't realign stack.
5617 * config/i386/i386.h (machine_function): Add
5618 max_used_stack_alignment.
5620 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
5622 * config/arm/arm.opt (-mbranch-cost): New option.
5623 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
5626 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
5629 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
5630 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
5632 2018-01-10 Richard Biener <rguenther@suse.de>
5635 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
5636 early out so it also covers the case where we have a non-NULL
5639 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
5641 PR tree-optimization/83753
5642 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
5643 for non-strided grouped accesses if the number of elements is 1.
5645 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5648 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
5649 * i386.h (TARGET_USE_GATHER): Define.
5650 * x86-tune.def (X86_TUNE_USE_GATHER): New.
5652 2018-01-10 Martin Liska <mliska@suse.cz>
5655 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
5656 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
5658 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
5659 CLEANUP_NO_PARTITIONING is not set.
5661 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
5663 * doc/rtl.texi: Remove documentation of (const ...) wrappers
5664 for vectors, as a partial revert of r254296.
5665 * rtl.h (const_vec_p): Delete.
5666 (const_vec_duplicate_p): Don't test for vector CONSTs.
5667 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
5668 * expmed.c (make_tree): Likewise.
5671 * common.md (E, F): Use CONSTANT_P instead of checking for
5673 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
5674 checking for CONST_VECTOR.
5676 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5679 * predict.c (force_edge_cold): Handle in more sane way edges
5682 2018-01-09 Carl Love <cel@us.ibm.com>
5684 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
5686 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
5687 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
5688 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
5689 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
5690 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
5691 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
5692 * config/rs6000/rs6000-protos.h: Add extern defition for
5693 rs6000_generate_float2_double_code.
5694 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
5696 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
5697 (float2_v2df): Add define_expand.
5699 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
5702 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
5703 op_mode in the force_to_mode call.
5705 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5707 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
5708 instead of checking each element individually.
5709 (aarch64_evpc_uzp): Likewise.
5710 (aarch64_evpc_zip): Likewise.
5711 (aarch64_evpc_ext): Likewise.
5712 (aarch64_evpc_rev): Likewise.
5713 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
5714 instead of checking each element individually. Return true without
5716 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
5717 whether all selected elements come from the same input, instead of
5718 checking each element individually. Remove calls to gen_rtx_REG,
5719 start_sequence and end_sequence and instead assert that no rtl is
5722 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5724 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
5725 order of HIGH and CONST checks.
5727 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5729 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
5730 if the destination isn't an SSA_NAME.
5732 2018-01-09 Richard Biener <rguenther@suse.de>
5734 PR tree-optimization/83668
5735 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
5737 (canonicalize_loop_form): ... here, renamed from ...
5738 (canonicalize_loop_closed_ssa_form): ... this and amended to
5739 swap successor edges for loop exit blocks to make us use
5740 the RPO order we need for initial schedule generation.
5742 2018-01-09 Joseph Myers <joseph@codesourcery.com>
5744 PR tree-optimization/64811
5745 * match.pd: When optimizing comparisons with Inf, avoid
5746 introducing or losing exceptions from comparisons with NaN.
5748 2018-01-09 Martin Liska <mliska@suse.cz>
5751 * asan.c (shadow_mem_size): Add gcc_assert.
5753 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
5755 Don't save registers in main().
5758 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
5759 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
5760 * config/avr/avr.c (avr_set_current_function): Don't error if
5761 naked, OS_task or OS_main are specified at the same time.
5762 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
5764 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
5766 * common/config/avr/avr-common.c (avr_option_optimization_table):
5767 Switch on -mmain-is-OS_task for optimizing compilations.
5769 2018-01-09 Richard Biener <rguenther@suse.de>
5771 PR tree-optimization/83572
5772 * graphite.c: Include cfganal.h.
5773 (graphite_transform_loops): Connect infinite loops to exit
5774 and remove fake edges at the end.
5776 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5778 * ipa-inline.c (edge_badness): Revert accidental checkin.
5780 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5783 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
5784 symbols; not inline clones.
5786 2018-01-09 Jakub Jelinek <jakub@redhat.com>
5789 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
5790 hard registers. Formatting fixes.
5792 PR preprocessor/83722
5793 * gcc.c (try_generate_repro): Pass
5794 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
5795 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
5798 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
5799 Kito Cheng <kito.cheng@gmail.com>
5801 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
5802 (riscv_leaf_function_p): Delete.
5803 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
5805 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5807 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
5809 (do_ifelse): New function.
5810 (do_isel): New function.
5811 (do_sub3): New function.
5812 (do_add3): New function.
5813 (do_load_mask_compare): New function.
5814 (do_overlap_load_compare): New function.
5815 (expand_compare_loop): New function.
5816 (expand_block_compare): Call expand_compare_loop() when appropriate.
5817 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
5819 (-mblock-compare-inline-loop-limit): New option.
5821 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5824 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
5825 Reverse order of second and third operands in first alternative.
5826 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
5827 of first and second elements in UNSPEC_VPERMR vector.
5828 (altivec_expand_vec_perm_le): Likewise.
5830 2018-01-08 Jeff Law <law@redhat.com>
5832 PR rtl-optimizatin/81308
5833 * tree-switch-conversion.c (cfg_altered): New file scoped static.
5834 (process_switch): If group_case_labels makes a change, then set
5836 (pass_convert_switch::execute): If a switch is converted, then
5837 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
5839 PR rtl-optimization/81308
5840 * recog.c (split_all_insns): Conditionally cleanup the CFG after
5843 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
5845 PR target/83663 - Revert r255946
5846 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
5847 generation for cases where splatting a value is not useful.
5848 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
5849 across a vec_duplicate and a paradoxical subreg forming a vector
5850 mode to a vec_concat.
5852 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5854 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
5855 -march=armv8.3-a variants.
5856 * config/arm/t-multilib: Likewise.
5857 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
5859 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5861 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
5863 (cceq_ior_compare_complement): Give it a name so I can use it, and
5864 change boolean_or_operator predicate to boolean_operator so it can
5865 be used to generate a crand.
5866 (eqne): New code iterator.
5867 (bd/bd_neg): New code_attrs.
5868 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
5869 a single define_insn.
5870 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
5871 decrement (bdnzt/bdnzf/bdzt/bdzf).
5872 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
5873 with the new names of the branch decrement patterns, and added the
5874 names of the branch decrement conditional patterns.
5876 2018-01-08 Richard Biener <rguenther@suse.de>
5878 PR tree-optimization/83563
5879 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
5882 2018-01-08 Richard Biener <rguenther@suse.de>
5885 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
5887 2018-01-08 Richard Biener <rguenther@suse.de>
5889 PR tree-optimization/83685
5890 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
5891 references to abnormals.
5893 2018-01-08 Richard Biener <rguenther@suse.de>
5896 * dwarf2out.c (output_indirect_strings): Handle empty
5897 skeleton_debug_str_hash.
5898 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
5900 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
5902 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
5903 (emit_store_direct): Likewise.
5904 (arc_trampoline_adjust_address): Likewise.
5905 (arc_asm_trampoline_template): New function.
5906 (arc_initialize_trampoline): Use asm_trampoline_template.
5907 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
5908 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
5909 * config/arc/arc.md (flush_icache): Delete pattern.
5911 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
5913 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
5914 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
5917 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
5920 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
5921 by not USED_FOR_TARGET.
5922 (make_pass_resolve_sw_modes): Likewise.
5924 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
5926 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
5929 2018-01-08 Richard Biener <rguenther@suse.de>
5932 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
5934 2018-01-08 Richard Biener <rguenther@suse.de>
5937 * match.pd ((t * 2) / 2) -> t): Add missing :c.
5939 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
5942 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
5943 basic blocks with a small number of successors.
5944 (convert_control_dep_chain_into_preds): Improve handling of
5946 (dump_predicates): Split apart into...
5947 (dump_pred_chain): ...here...
5948 (dump_pred_info): ...and here.
5949 (can_one_predicate_be_invalidated_p): Add debugging printfs.
5950 (can_chain_union_be_invalidated_p): Improve check for invalidation
5952 (uninit_uses_cannot_happen): Avoid unnecessary if
5953 convert_control_dep_chain_into_preds yielded nothing.
5955 2018-01-06 Martin Sebor <msebor@redhat.com>
5957 PR tree-optimization/83640
5958 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
5959 subtracting negative offset from size.
5960 (builtin_access::overlap): Adjust offset bounds of the access to fall
5961 within the size of the object if possible.
5963 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
5965 PR rtl-optimization/83699
5966 * expmed.c (extract_bit_field_1): Restrict the vector usage of
5967 extract_bit_field_as_subreg to cases in which the extracted
5968 value is also a vector.
5970 * lra-constraints.c (process_alt_operands): Test for the equivalence
5971 substitutions when detecting a possible reload cycle.
5973 2018-01-06 Jakub Jelinek <jakub@redhat.com>
5976 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
5977 by default if flag_selective_schedling{,2}. Formatting fixes.
5979 PR rtl-optimization/83682
5980 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
5981 if it has non-VECTOR_MODE element mode.
5982 (vec_duplicate_p): Likewise.
5985 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
5986 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
5988 2018-01-05 Jakub Jelinek <jakub@redhat.com>
5991 * config/i386/i386-builtin.def
5992 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
5993 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
5994 Require also OPTION_MASK_ISA_AVX512F in addition to
5995 OPTION_MASK_ISA_GFNI.
5996 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
5997 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
5998 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
5999 to OPTION_MASK_ISA_GFNI.
6000 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
6001 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
6002 OPTION_MASK_ISA_AVX512BW.
6003 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
6004 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
6005 addition to OPTION_MASK_ISA_GFNI.
6006 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
6007 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
6008 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
6009 to OPTION_MASK_ISA_GFNI.
6010 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
6011 a requirement for all ISAs rather than any of them with a few
6013 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
6015 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
6016 bitmasks to be enabled with 3 exceptions, instead of requiring any
6017 enabled ISA with lots of exceptions.
6018 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
6019 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
6020 Change avx512bw in isa attribute to avx512f.
6021 * config/i386/sgxintrin.h: Add license boilerplate.
6022 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
6023 to __AVX512F__ and __AVX512VL to __AVX512VL__.
6024 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
6025 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
6027 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
6028 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
6029 temporarily sse2 rather than sse if not enabled already.
6032 * config/i386/sse.md (VI248_VLBW): Rename to ...
6033 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
6034 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
6035 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
6036 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
6037 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
6038 mode iterator instead of VI248_VLBW.
6040 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
6042 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
6043 (record_modified): Skip clobbers; add debug output.
6044 (param_change_prob): Use sreal frequencies.
6046 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6048 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
6049 punt for user-aligned variables.
6051 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6053 * tree-chrec.c (chrec_contains_symbols): Return true for
6056 2018-01-05 Sudakshina Das <sudi.das@arm.com>
6059 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
6060 of (x|y) == x for BICS pattern.
6062 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6064 PR tree-optimization/83605
6065 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
6066 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
6069 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
6071 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
6072 * config/epiphany/rtems.h: New file.
6074 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6075 Uros Bizjak <ubizjak@gmail.com>
6078 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
6079 QIreg_operand instead of register_operand predicate.
6080 * config/i386/i386.c (ix86_rop_should_change_byte_p,
6081 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
6082 comments instead of -fmitigate[-_]rop.
6084 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6087 * cgraphunit.c (symbol_table::compile): Switch to text_section
6088 before calling assembly_start debug hook.
6089 * run-rtl-passes.c (run_rtl_passes): Likewise.
6092 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6094 * tree-vrp.c (extract_range_from_binary_expr_1): Check
6095 range_int_cst_p rather than !symbolic_range_p before calling
6096 extract_range_from_multiplicative_op_1.
6098 2018-01-04 Jeff Law <law@redhat.com>
6100 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
6101 redundant test in assertion.
6103 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6105 * doc/rtl.texi: Document machine_mode wrapper classes.
6107 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6109 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
6112 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6114 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
6115 the VEC_PERM_EXPR fold to fail.
6117 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6120 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
6121 to switched_sections.
6123 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6126 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
6129 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
6132 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
6133 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
6135 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6138 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
6139 is BLKmode and bitpos not zero or mode change is needed.
6141 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6144 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
6147 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
6150 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
6151 instead of MULT rtx. Update all corresponding splitters.
6153 (*ssub<modesuffix>): Ditto.
6155 (*cmp_sadd_di): Update split patterns.
6156 (*cmp_sadd_si): Ditto.
6157 (*cmp_sadd_sidi): Ditto.
6158 (*cmp_ssub_di): Ditto.
6159 (*cmp_ssub_si): Ditto.
6160 (*cmp_ssub_sidi): Ditto.
6161 * config/alpha/predicates.md (const23_operand): New predicate.
6162 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
6163 Look for ASHIFT, not MULT inner operand.
6164 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
6166 2018-01-04 Martin Liska <mliska@suse.cz>
6168 PR gcov-profile/83669
6169 * gcov.c (output_intermediate_file): Add version to intermediate
6171 * doc/gcov.texi: Document new field 'version' in intermediate
6172 file format. Fix location of '-k' option of gcov command.
6174 2018-01-04 Martin Liska <mliska@suse.cz>
6177 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
6179 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6181 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
6183 2018-01-03 Martin Sebor <msebor@redhat.com>
6185 PR tree-optimization/83655
6186 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
6187 checking calls with invalid arguments.
6189 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6191 * tree-vect-stmts.c (vect_get_store_rhs): New function.
6192 (vectorizable_mask_load_store): Delete.
6193 (vectorizable_call): Return false for masked loads and stores.
6194 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
6195 instead of gimple_assign_rhs1.
6196 (vectorizable_load): Handle IFN_MASK_LOAD.
6197 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
6199 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6201 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
6203 (vectorizable_mask_load_store): ...here.
6204 (vectorizable_load): ...and here.
6206 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6208 * tree-vect-stmts.c (vect_build_all_ones_mask)
6209 (vect_build_zero_merge_argument): New functions, split out from...
6210 (vectorizable_load): ...here.
6212 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6214 * tree-vect-stmts.c (vect_check_store_rhs): New function,
6216 (vectorizable_mask_load_store): ...here.
6217 (vectorizable_store): ...and here.
6219 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6221 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
6223 (vectorizable_mask_load_store): ...here.
6225 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6227 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
6228 (vect_model_store_cost): Take a vec_load_store_type instead of a
6230 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
6231 (vect_model_store_cost): Take a vec_load_store_type instead of a
6233 (vectorizable_mask_load_store): Update accordingly.
6234 (vectorizable_store): Likewise.
6235 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
6237 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6239 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
6240 IFN_MASK_LOAD calls here rather than...
6241 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
6243 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6244 Alan Hayward <alan.hayward@arm.com>
6245 David Sherwood <david.sherwood@arm.com>
6247 * expmed.c (extract_bit_field_1): For vector extracts,
6248 fall back to extract_bit_field_as_subreg if vec_extract
6251 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6252 Alan Hayward <alan.hayward@arm.com>
6253 David Sherwood <david.sherwood@arm.com>
6255 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
6256 they are variable or constant sized.
6257 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
6258 slots for constant-sized data.
6260 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6261 Alan Hayward <alan.hayward@arm.com>
6262 David Sherwood <david.sherwood@arm.com>
6264 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
6265 handling COND_EXPRs with boolean comparisons, try to find a better
6266 basis for the mask type than the boolean itself.
6268 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6270 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
6271 is calculated and how it can be overridden.
6272 * genmodes.c (max_bitsize_mode_any_mode): New variable.
6273 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
6275 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
6278 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6279 Alan Hayward <alan.hayward@arm.com>
6280 David Sherwood <david.sherwood@arm.com>
6282 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
6283 Remove the mode argument.
6284 (aarch64_simd_valid_immediate): Remove the mode and inverse
6286 * config/aarch64/iterators.md (bitsize): New iterator.
6287 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
6288 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
6289 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
6290 aarch64_simd_valid_immediate.
6291 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
6292 (aarch64_reg_or_bic_imm): Likewise.
6293 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
6294 with an insn_type enum and msl with a modifier_type enum.
6295 Replace element_width with a scalar_mode. Change the shift
6296 to unsigned int. Add constructors for scalar_float_mode and
6297 scalar_int_mode elements.
6298 (aarch64_vect_float_const_representable_p): Delete.
6299 (aarch64_can_const_movi_rtx_p)
6300 (aarch64_simd_scalar_immediate_valid_for_move)
6301 (aarch64_simd_make_constant): Update call to
6302 aarch64_simd_valid_immediate.
6303 (aarch64_advsimd_valid_immediate_hs): New function.
6304 (aarch64_advsimd_valid_immediate): Likewise.
6305 (aarch64_simd_valid_immediate): Remove mode and inverse
6306 arguments. Rewrite to use the above. Use const_vec_duplicate_p
6307 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
6308 and aarch64_float_const_representable_p on the result.
6309 (aarch64_output_simd_mov_immediate): Remove mode argument.
6310 Update call to aarch64_simd_valid_immediate and use of
6311 simd_immediate_info.
6312 (aarch64_output_scalar_simd_mov_immediate): Update call
6315 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6316 Alan Hayward <alan.hayward@arm.com>
6317 David Sherwood <david.sherwood@arm.com>
6319 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
6320 (mode_nunits): Likewise CONST_MODE_NUNITS.
6321 * machmode.def (ADJUST_NUNITS): Document.
6322 * genmodes.c (mode_data::need_nunits_adj): New field.
6323 (blank_mode): Update accordingly.
6324 (adj_nunits): New variable.
6325 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
6327 (emit_mode_size_inline): Set need_bytesize_adj for all modes
6328 listed in adj_nunits.
6329 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
6330 listed in adj_nunits. Don't emit case statements for such modes.
6331 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
6332 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
6333 nothing if adj_nunits is nonnull.
6334 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
6335 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
6336 (emit_mode_fbit): Update use of print_maybe_const_decl.
6337 (emit_move_size): Likewise. Treat the array as non-const
6339 (emit_mode_adjustments): Handle adj_nunits.
6341 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6343 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
6344 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
6345 (VECTOR_MODES): Use it.
6346 (make_vector_modes): Take the prefix as an argument.
6348 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6349 Alan Hayward <alan.hayward@arm.com>
6350 David Sherwood <david.sherwood@arm.com>
6352 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
6353 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
6354 for MODE_VECTOR_BOOL.
6355 * machmode.def (VECTOR_BOOL_MODE): Document.
6356 * genmodes.c (VECTOR_BOOL_MODE): New macro.
6357 (make_vector_bool_mode): New function.
6358 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
6360 * lto-streamer-in.c (lto_input_mode_table): Likewise.
6361 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
6363 * stor-layout.c (int_mode_for_mode): Likewise.
6364 * tree.c (build_vector_type_for_mode): Likewise.
6365 * varasm.c (output_constant_pool_2): Likewise.
6366 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
6367 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
6368 for MODE_VECTOR_BOOL.
6369 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
6370 of mode class checks.
6371 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
6372 instead of a list of mode class checks.
6373 (expand_vector_scalar_condition): Likewise.
6374 (type_for_widest_vector_mode): Handle BImode as an inner mode.
6376 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6377 Alan Hayward <alan.hayward@arm.com>
6378 David Sherwood <david.sherwood@arm.com>
6380 * machmode.h (mode_size): Change from unsigned short to
6382 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
6383 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6384 or if measurement_type is not polynomial.
6385 (fixed_size_mode::includes_p): Check for constant-sized modes.
6386 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
6387 return a poly_uint16 rather than an unsigned short.
6388 (emit_mode_size): Change the type of mode_size from unsigned short
6389 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
6390 (emit_mode_adjustments): Cope with polynomial vector sizes.
6391 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6393 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6395 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
6396 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
6397 * caller-save.c (setup_save_areas): Likewise.
6398 (replace_reg_with_saved_mem): Likewise.
6399 * calls.c (emit_library_call_value_1): Likewise.
6400 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
6401 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
6402 (gen_lowpart_for_combine): Likewise.
6403 * convert.c (convert_to_integer_1): Likewise.
6404 * cse.c (equiv_constant, cse_insn): Likewise.
6405 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
6406 (cselib_subst_to_values): Likewise.
6407 * dce.c (word_dce_process_block): Likewise.
6408 * df-problems.c (df_word_lr_mark_ref): Likewise.
6409 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
6410 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
6411 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
6412 (rtl_for_decl_location): Likewise.
6413 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
6414 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
6415 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
6416 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
6417 (expand_expr_real_1): Likewise.
6418 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
6419 (pad_below): Likewise.
6420 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6421 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
6422 * ira.c (get_subreg_tracking_sizes): Likewise.
6423 * ira-build.c (ira_create_allocno_objects): Likewise.
6424 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
6425 (ira_sort_regnos_for_alter_reg): Likewise.
6426 * ira-costs.c (record_operand_costs): Likewise.
6427 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
6428 (resolve_simple_move): Likewise.
6429 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
6430 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
6431 (lra_constraints): Likewise.
6432 (CONST_POOL_OK_P): Reject variable-sized modes.
6433 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
6434 (add_pseudo_to_slot, lra_spill): Likewise.
6435 * omp-low.c (omp_clause_aligned_alignment): Likewise.
6436 * optabs-query.c (get_best_extraction_insn): Likewise.
6437 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6438 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
6439 (expand_mult_highpart, valid_multiword_target_p): Likewise.
6440 * recog.c (offsettable_address_addr_space_p): Likewise.
6441 * regcprop.c (maybe_mode_change): Likewise.
6442 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
6443 * regrename.c (build_def_use): Likewise.
6444 * regstat.c (dump_reg_info): Likewise.
6445 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
6446 (find_reloads, find_reloads_subreg_address): Likewise.
6447 * reload1.c (eliminate_regs_1): Likewise.
6448 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
6449 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
6450 (simplify_binary_operation_1, simplify_subreg): Likewise.
6451 * targhooks.c (default_function_arg_padding): Likewise.
6452 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
6453 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
6454 (verify_gimple_assign_ternary): Likewise.
6455 * tree-inline.c (estimate_move_cost): Likewise.
6456 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6457 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
6458 (get_address_cost_ainc): Likewise.
6459 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
6460 (vect_supportable_dr_alignment): Likewise.
6461 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6462 (vectorizable_reduction): Likewise.
6463 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
6464 (vectorizable_operation, vectorizable_load): Likewise.
6465 * tree.c (build_same_sized_truth_vector_type): Likewise.
6466 * valtrack.c (cleanup_auto_inc_dec): Likewise.
6467 * var-tracking.c (emit_note_insn_var_location): Likewise.
6468 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
6469 (ADDR_VEC_ALIGN): Likewise.
6471 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6472 Alan Hayward <alan.hayward@arm.com>
6473 David Sherwood <david.sherwood@arm.com>
6475 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
6477 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6478 or if measurement_type is polynomial.
6479 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
6480 * combine.c (make_extraction): Likewise.
6481 * dse.c (find_shift_sequence): Likewise.
6482 * dwarf2out.c (mem_loc_descriptor): Likewise.
6483 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
6484 (extract_bit_field, extract_low_bits): Likewise.
6485 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
6486 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
6487 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
6488 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
6489 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6490 * reload.c (find_reloads): Likewise.
6491 * reload1.c (alter_reg): Likewise.
6492 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
6493 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
6494 * tree-if-conv.c (predicate_mem_writes): Likewise.
6495 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
6496 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
6497 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
6498 * valtrack.c (dead_debug_insert_temp): Likewise.
6499 * varasm.c (mergeable_constant_section): Likewise.
6500 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
6502 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6503 Alan Hayward <alan.hayward@arm.com>
6504 David Sherwood <david.sherwood@arm.com>
6506 * expr.c (expand_assignment): Cope with polynomial mode sizes
6507 when assigning to a CONCAT.
6509 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6510 Alan Hayward <alan.hayward@arm.com>
6511 David Sherwood <david.sherwood@arm.com>
6513 * machmode.h (mode_precision): Change from unsigned short to
6515 (mode_to_precision): Return a poly_uint16 rather than an unsigned
6517 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
6518 or if measurement_type is not polynomial.
6519 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
6520 in which the mode is already known to be a scalar_int_mode.
6521 * genmodes.c (emit_mode_precision): Change the type of mode_precision
6522 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
6524 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6525 for GET_MODE_PRECISION.
6526 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6527 for GET_MODE_PRECISION.
6528 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
6530 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
6531 (expand_field_assignment, make_extraction): Likewise.
6532 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
6533 (get_last_value): Likewise.
6534 * convert.c (convert_to_integer_1): Likewise.
6535 * cse.c (cse_insn): Likewise.
6536 * expr.c (expand_expr_real_1): Likewise.
6537 * lra-constraints.c (simplify_operand_subreg): Likewise.
6538 * optabs-query.c (can_atomic_load_p): Likewise.
6539 * optabs.c (expand_atomic_load): Likewise.
6540 (expand_atomic_store): Likewise.
6541 * ree.c (combine_reaching_defs): Likewise.
6542 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
6543 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
6544 * tree.h (type_has_mode_precision_p): Likewise.
6545 * ubsan.c (instrument_si_overflow): Likewise.
6547 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6548 Alan Hayward <alan.hayward@arm.com>
6549 David Sherwood <david.sherwood@arm.com>
6551 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
6552 polynomial numbers of units.
6553 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
6554 (valid_vector_subparts_p): New function.
6555 (build_vector_type): Remove temporary shim and take the number
6556 of units as a poly_uint64 rather than an int.
6557 (build_opaque_vector_type): Take the number of units as a
6558 poly_uint64 rather than an int.
6559 * tree.c (build_vector_from_ctor): Handle polynomial
6560 TYPE_VECTOR_SUBPARTS.
6561 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
6562 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
6563 (build_vector_from_val): If the number of units is variable,
6564 use build_vec_duplicate_cst for constant operands and
6565 VEC_DUPLICATE_EXPR otherwise.
6566 (make_vector_type): Remove temporary is_constant ().
6567 (build_vector_type, build_opaque_vector_type): Take the number of
6568 units as a poly_uint64 rather than an int.
6569 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
6571 * cfgexpand.c (expand_debug_expr): Likewise.
6572 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
6573 (store_constructor, expand_expr_real_1): Likewise.
6574 (const_scalar_mask_from_tree): Likewise.
6575 * fold-const-call.c (fold_const_reduction): Likewise.
6576 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
6577 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
6578 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
6579 (fold_relational_const): Likewise.
6580 (native_interpret_vector): Likewise. Change the size from an
6581 int to an unsigned int.
6582 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
6583 TYPE_VECTOR_SUBPARTS.
6584 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
6585 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
6586 duplicating a non-constant operand into a variable-length vector.
6587 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
6588 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
6589 * ipa-icf.c (sem_variable::equals): Likewise.
6590 * match.pd: Likewise.
6591 * omp-simd-clone.c (simd_clone_subparts): Likewise.
6592 * print-tree.c (print_node): Likewise.
6593 * stor-layout.c (layout_type): Likewise.
6594 * targhooks.c (default_builtin_vectorization_cost): Likewise.
6595 * tree-cfg.c (verify_gimple_comparison): Likewise.
6596 (verify_gimple_assign_binary): Likewise.
6597 (verify_gimple_assign_ternary): Likewise.
6598 (verify_gimple_assign_single): Likewise.
6599 * tree-pretty-print.c (dump_generic_node): Likewise.
6600 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6601 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
6602 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
6603 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
6604 (vect_shift_permute_load_chain): Likewise.
6605 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
6606 (expand_vector_condition, optimize_vector_constructor): Likewise.
6607 (lower_vec_perm, get_compute_type): Likewise.
6608 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6609 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
6610 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
6611 (vect_recog_mask_conversion_pattern): Likewise.
6612 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
6613 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
6614 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
6615 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
6616 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
6617 (vectorizable_shift, vectorizable_operation, vectorizable_store)
6618 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
6619 (supportable_widening_operation): Likewise.
6620 (supportable_narrowing_operation): Likewise.
6621 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
6623 * varasm.c (output_constant): Likewise.
6625 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6626 Alan Hayward <alan.hayward@arm.com>
6627 David Sherwood <david.sherwood@arm.com>
6629 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
6630 so that both the length == 3 and length != 3 cases set up their
6631 own permute vectors. Add comments explaining why we know the
6632 number of elements is constant.
6633 (vect_permute_load_chain): Likewise.
6635 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6636 Alan Hayward <alan.hayward@arm.com>
6637 David Sherwood <david.sherwood@arm.com>
6639 * machmode.h (mode_nunits): Change from unsigned char to
6641 (ONLY_FIXED_SIZE_MODES): New macro.
6642 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
6643 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
6644 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
6646 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
6647 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
6648 or if measurement_type is not polynomial.
6649 * genmodes.c (ZERO_COEFFS): New macro.
6650 (emit_mode_nunits_inline): Make mode_nunits_inline return a
6652 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
6653 Use ZERO_COEFFS when emitting initializers.
6654 * data-streamer.h (bp_pack_poly_value): New function.
6655 (bp_unpack_poly_value): Likewise.
6656 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6657 for GET_MODE_NUNITS.
6658 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6659 for GET_MODE_NUNITS.
6660 * tree.c (make_vector_type): Remove temporary shim and make
6661 the real function take the number of units as a poly_uint64
6663 (build_vector_type_for_mode): Handle polynomial nunits.
6664 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
6665 * emit-rtl.c (const_vec_series_p_1): Likewise.
6666 (gen_rtx_CONST_VECTOR): Likewise.
6667 * fold-const.c (test_vec_duplicate_folding): Likewise.
6668 * genrecog.c (validate_pattern): Likewise.
6669 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
6670 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6671 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
6672 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
6673 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
6674 * rtlanal.c (subreg_get_info): Likewise.
6675 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
6676 (vect_grouped_load_supported): Likewise.
6677 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
6678 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
6679 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
6680 (simplify_const_unary_operation, simplify_binary_operation_1)
6681 (simplify_const_binary_operation, simplify_ternary_operation)
6682 (test_vector_ops_duplicate, test_vector_ops): Likewise.
6683 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
6684 instead of CONST_VECTOR_NUNITS.
6685 * varasm.c (output_constant_pool_2): Likewise.
6686 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
6687 explicit-encoded elements in the XVEC for variable-length vectors.
6689 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6691 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
6693 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6694 Alan Hayward <alan.hayward@arm.com>
6695 David Sherwood <david.sherwood@arm.com>
6697 * coretypes.h (fixed_size_mode): Declare.
6698 (fixed_size_mode_pod): New typedef.
6699 * builtins.h (target_builtins::x_apply_args_mode)
6700 (target_builtins::x_apply_result_mode): Change type to
6701 fixed_size_mode_pod.
6702 * builtins.c (apply_args_size, apply_result_size, result_vector)
6703 (expand_builtin_apply_args_1, expand_builtin_apply)
6704 (expand_builtin_return): Update accordingly.
6706 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6708 * cse.c (hash_rtx_cb): Hash only the encoded elements.
6709 * cselib.c (cselib_hash_rtx): Likewise.
6710 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
6711 CONST_VECTOR encoding.
6713 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6714 Jeff Law <law@redhat.com>
6717 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
6718 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
6719 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
6720 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
6723 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
6724 explicitly probe *sp in a noreturn function if there were any callee
6725 register saves or frame pointer is needed.
6727 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6730 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
6731 BLKmode for ternary, binary or unary expressions.
6734 * var-tracking.c (delete_vta_debug_insn): New inline function.
6735 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
6736 insns from get_insns () to NULL instead of each bb separately.
6737 Use delete_vta_debug_insn. No longer static.
6738 (vt_debug_insns_local, variable_tracking_main_1): Adjust
6739 delete_vta_debug_insns callers.
6740 * rtl.h (delete_vta_debug_insns): Declare.
6741 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
6742 instead of variable_tracking_main.
6744 2018-01-03 Martin Sebor <msebor@redhat.com>
6746 PR tree-optimization/83603
6747 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
6748 arguments past the endof the argument list in functions declared
6749 without a prototype.
6750 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
6751 Avoid checking when arguments are null.
6753 2018-01-03 Martin Sebor <msebor@redhat.com>
6756 * doc/extend.texi (attribute const): Fix a typo.
6757 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
6758 issuing -Wsuggest-attribute for void functions.
6760 2018-01-03 Martin Sebor <msebor@redhat.com>
6762 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
6763 offset_int::from instead of wide_int::to_shwi.
6764 (maybe_diag_overlap): Remove assertion.
6765 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
6766 * gimple-ssa-sprintf.c (format_directive): Same.
6767 (parse_directive): Same.
6768 (sprintf_dom_walker::compute_format_length): Same.
6769 (try_substitute_return_value): Same.
6771 2018-01-03 Jeff Law <law@redhat.com>
6774 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
6775 non-constant residual for zero at runtime and avoid probing in
6776 that case. Reorganize code for trailing problem to mirror handling
6779 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6781 PR tree-optimization/83501
6782 * tree-ssa-strlen.c (get_string_cst): New.
6783 (handle_char_store): Call get_string_cst.
6785 2018-01-03 Martin Liska <mliska@suse.cz>
6787 PR tree-optimization/83593
6788 * tree-ssa-strlen.c: Include tree-cfg.h.
6789 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
6790 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
6791 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
6793 (strlen_dom_walker::before_dom_children): Call
6794 gimple_purge_dead_eh_edges. Dump tranformation with details
6796 (strlen_dom_walker::before_dom_children): Update call by adding
6797 new argument cleanup_eh.
6798 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
6800 2018-01-03 Martin Liska <mliska@suse.cz>
6803 * cif-code.def (VARIADIC_THUNK): New enum value.
6804 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
6807 2018-01-03 Jan Beulich <jbeulich@suse.com>
6809 * sse.md (mov<mode>_internal): Tighten condition for when to use
6810 vmovdqu<ssescalarsize> for TI and OI modes.
6812 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6814 Update copyright years.
6816 2018-01-03 Martin Liska <mliska@suse.cz>
6819 * ipa-visibility.c (function_and_variable_visibility): Skip
6820 functions with noipa attribure.
6822 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6824 * gcc.c (process_command): Update copyright notice dates.
6825 * gcov-dump.c (print_version): Ditto.
6826 * gcov.c (print_version): Ditto.
6827 * gcov-tool.c (print_version): Ditto.
6828 * gengtype.c (create_file): Ditto.
6829 * doc/cpp.texi: Bump @copying's copyright year.
6830 * doc/cppinternals.texi: Ditto.
6831 * doc/gcc.texi: Ditto.
6832 * doc/gccint.texi: Ditto.
6833 * doc/gcov.texi: Ditto.
6834 * doc/install.texi: Ditto.
6835 * doc/invoke.texi: Ditto.
6837 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6839 * vector-builder.h (vector_builder::m_full_nelts): Change from
6840 unsigned int to poly_uint64.
6841 (vector_builder::full_nelts): Update prototype accordingly.
6842 (vector_builder::new_vector): Likewise.
6843 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
6844 (vector_builder::operator ==): Likewise.
6845 (vector_builder::finalize): Likewise.
6846 * int-vector-builder.h (int_vector_builder::int_vector_builder):
6847 Take the number of elements as a poly_uint64 rather than an
6849 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
6850 from unsigned int to poly_uint64.
6851 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
6852 (vec_perm_indices::new_vector): Likewise.
6853 (vec_perm_indices::length): Likewise.
6854 (vec_perm_indices::nelts_per_input): Likewise.
6855 (vec_perm_indices::input_nelts): Likewise.
6856 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
6857 number of elements per input as a poly_uint64 rather than an
6858 unsigned int. Use the original encoding for variable-length
6859 vectors, rather than clamping each individual element.
6860 For the second and subsequent elements in each pattern,
6861 clamp the step and base before clamping their sum.
6862 (vec_perm_indices::series_p): Handle polynomial element counts.
6863 (vec_perm_indices::all_in_range_p): Likewise.
6864 (vec_perm_indices_to_tree): Likewise.
6865 (vec_perm_indices_to_rtx): Likewise.
6866 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
6867 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
6868 (tree_vector_builder::new_binary_operation): Handle polynomial
6869 element counts. Return false if we need to know the number
6870 of elements at compile time.
6871 * fold-const.c (fold_vec_perm): Punt if the number of elements
6872 isn't known at compile time.
6874 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6876 * vec-perm-indices.h (vec_perm_builder): Change element type
6877 from HOST_WIDE_INT to poly_int64.
6878 (vec_perm_indices::element_type): Update accordingly.
6879 (vec_perm_indices::clamp): Handle polynomial element_types.
6880 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
6881 (vec_perm_indices::all_in_range_p): Likewise.
6882 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
6884 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
6885 polynomial vec_perm_indices element types.
6886 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
6887 * fold-const.c (fold_vec_perm): Likewise.
6888 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
6889 * tree-vect-generic.c (lower_vec_perm): Likewise.
6890 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
6891 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
6892 element type to HOST_WIDE_INT.
6894 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6895 Alan Hayward <alan.hayward@arm.com>
6896 David Sherwood <david.sherwood@arm.com>
6898 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
6899 rather than an int. Use plus_constant.
6900 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
6901 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
6903 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6904 Alan Hayward <alan.hayward@arm.com>
6905 David Sherwood <david.sherwood@arm.com>
6907 * calls.c (emit_call_1, expand_call): Change struct_value_size from
6908 a HOST_WIDE_INT to a poly_int64.
6910 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6911 Alan Hayward <alan.hayward@arm.com>
6912 David Sherwood <david.sherwood@arm.com>
6914 * calls.c (load_register_parameters): Cope with polynomial
6915 mode sizes. Require a constant size for BLKmode parameters
6916 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
6917 forces a parameter to be padded at the lsb end in order to
6918 fill a complete number of words, require the parameter size
6919 to be ordered wrt UNITS_PER_WORD.
6921 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6922 Alan Hayward <alan.hayward@arm.com>
6923 David Sherwood <david.sherwood@arm.com>
6925 * reload1.c (spill_stack_slot_width): Change element type
6926 from unsigned int to poly_uint64_pod.
6927 (alter_reg): Treat mode sizes as polynomial.
6929 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6930 Alan Hayward <alan.hayward@arm.com>
6931 David Sherwood <david.sherwood@arm.com>
6933 * reload.c (complex_word_subreg_p): New function.
6934 (reload_inner_reg_of_subreg, push_reload): Use it.
6936 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6937 Alan Hayward <alan.hayward@arm.com>
6938 David Sherwood <david.sherwood@arm.com>
6940 * lra-constraints.c (process_alt_operands): Reject matched
6941 operands whose sizes aren't ordered.
6942 (match_reload): Refer to this check here.
6944 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6945 Alan Hayward <alan.hayward@arm.com>
6946 David Sherwood <david.sherwood@arm.com>
6948 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
6949 that the mode size is in the set {1, 2, 4, 8, 16}.
6951 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6952 Alan Hayward <alan.hayward@arm.com>
6953 David Sherwood <david.sherwood@arm.com>
6955 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
6956 Use plus_constant instead of gen_rtx_PLUS.
6958 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6959 Alan Hayward <alan.hayward@arm.com>
6960 David Sherwood <david.sherwood@arm.com>
6962 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
6963 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
6964 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
6965 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
6966 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
6967 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
6968 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
6969 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
6970 * config/i386/i386.c (ix86_push_rounding): ...this new function.
6971 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
6973 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
6974 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
6975 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
6976 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
6977 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
6978 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
6979 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
6980 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
6981 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
6982 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
6984 * expr.c (emit_move_resolve_push): Treat the input and result
6985 of PUSH_ROUNDING as a poly_int64.
6986 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
6987 (emit_push_insn): Likewise.
6988 * lra-eliminations.c (mark_not_eliminable): Likewise.
6989 * recog.c (push_operand): Likewise.
6990 * reload1.c (elimination_effects): Likewise.
6991 * rtlanal.c (nonzero_bits1): Likewise.
6992 * calls.c (store_one_arg): Likewise. Require the padding to be
6993 known at compile time.
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 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
7000 Use plus_constant instead of gen_rtx_PLUS.
7002 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7003 Alan Hayward <alan.hayward@arm.com>
7004 David Sherwood <david.sherwood@arm.com>
7006 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
7009 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7010 Alan Hayward <alan.hayward@arm.com>
7011 David Sherwood <david.sherwood@arm.com>
7013 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
7014 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
7015 via stack temporaries. Treat the mode size as polynomial too.
7017 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7018 Alan Hayward <alan.hayward@arm.com>
7019 David Sherwood <david.sherwood@arm.com>
7021 * expr.c (expand_expr_real_2): When handling conversions involving
7022 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
7023 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
7024 as a poly_uint64 too.
7026 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7027 Alan Hayward <alan.hayward@arm.com>
7028 David Sherwood <david.sherwood@arm.com>
7030 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
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 * combine.c (can_change_dest_mode): Handle polynomial
7037 REGMODE_NATURAL_SIZE.
7038 * expmed.c (store_bit_field_1): Likewise.
7039 * expr.c (store_constructor): Likewise.
7040 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
7041 and polynomial REGMODE_NATURAL_SIZE.
7042 (gen_lowpart_common): Likewise.
7043 * reginfo.c (record_subregs_of_mode): Likewise.
7044 * rtlanal.c (read_modify_subreg_p): Likewise.
7046 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7047 Alan Hayward <alan.hayward@arm.com>
7048 David Sherwood <david.sherwood@arm.com>
7050 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
7051 numbers of elements.
7053 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7054 Alan Hayward <alan.hayward@arm.com>
7055 David Sherwood <david.sherwood@arm.com>
7057 * match.pd: Cope with polynomial numbers of vector elements.
7059 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7060 Alan Hayward <alan.hayward@arm.com>
7061 David Sherwood <david.sherwood@arm.com>
7063 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
7064 in a POINTER_PLUS_EXPR.
7066 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7067 Alan Hayward <alan.hayward@arm.com>
7068 David Sherwood <david.sherwood@arm.com>
7070 * omp-simd-clone.c (simd_clone_subparts): New function.
7071 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
7072 (ipa_simd_modify_function_body): Likewise.
7074 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7075 Alan Hayward <alan.hayward@arm.com>
7076 David Sherwood <david.sherwood@arm.com>
7078 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
7079 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
7080 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
7081 (expand_vector_condition, vector_element): Likewise.
7082 (subparts_gt): New function.
7083 (get_compute_type): Use subparts_gt.
7084 (count_type_subparts): Delete.
7085 (expand_vector_operations_1): Use subparts_gt instead of
7086 count_type_subparts.
7088 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7089 Alan Hayward <alan.hayward@arm.com>
7090 David Sherwood <david.sherwood@arm.com>
7092 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
7093 (vect_compile_time_alias): ...this new function. Do the calculation
7094 on poly_ints rather than trees.
7095 (vect_prune_runtime_alias_test_list): Update call accordingly.
7097 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7098 Alan Hayward <alan.hayward@arm.com>
7099 David Sherwood <david.sherwood@arm.com>
7101 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
7103 (vect_schedule_slp_instance): Likewise.
7105 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7106 Alan Hayward <alan.hayward@arm.com>
7107 David Sherwood <david.sherwood@arm.com>
7109 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
7110 constant and extern definitions for variable-length vectors.
7111 (vect_get_constant_vectors): Note that the number of units
7112 is known to be constant.
7114 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7115 Alan Hayward <alan.hayward@arm.com>
7116 David Sherwood <david.sherwood@arm.com>
7118 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
7119 of units as polynomial. Choose between WIDE and NARROW based
7122 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7123 Alan Hayward <alan.hayward@arm.com>
7124 David Sherwood <david.sherwood@arm.com>
7126 * tree-vect-stmts.c (simd_clone_subparts): New function.
7127 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
7129 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7130 Alan Hayward <alan.hayward@arm.com>
7131 David Sherwood <david.sherwood@arm.com>
7133 * tree-vect-stmts.c (vectorizable_call): Treat the number of
7134 vectors as polynomial. Use build_index_vector for
7137 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7138 Alan Hayward <alan.hayward@arm.com>
7139 David Sherwood <david.sherwood@arm.com>
7141 * tree-vect-stmts.c (get_load_store_type): Treat the number of
7142 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
7143 for variable-length vectors.
7144 (vectorizable_mask_load_store): Treat the number of units as
7145 polynomial, asserting that it is constant if the condition has
7146 already been enforced.
7147 (vectorizable_store, vectorizable_load): Likewise.
7149 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7150 Alan Hayward <alan.hayward@arm.com>
7151 David Sherwood <david.sherwood@arm.com>
7153 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
7154 of units as polynomial. Punt if we can't tell at compile time
7155 which vector contains the final result.
7157 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7158 Alan Hayward <alan.hayward@arm.com>
7159 David Sherwood <david.sherwood@arm.com>
7161 * tree-vect-loop.c (vectorizable_induction): Treat the number
7162 of units as polynomial. Punt on SLP inductions. Use an integer
7163 VEC_SERIES_EXPR for variable-length integer reductions. Use a
7164 cast of such a series for variable-length floating-point
7167 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7168 Alan Hayward <alan.hayward@arm.com>
7169 David Sherwood <david.sherwood@arm.com>
7171 * tree.h (build_index_vector): Declare.
7172 * tree.c (build_index_vector): New function.
7173 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
7174 of units as polynomial, forcibly converting it to a constant if
7175 vectorizable_reduction has already enforced the condition.
7176 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
7177 to create a {1,2,3,...} vector.
7178 (vectorizable_reduction): Treat the number of units as polynomial.
7179 Choose vectype_in based on the largest scalar element size rather
7180 than the smallest number of units. Enforce the restrictions
7183 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7184 Alan Hayward <alan.hayward@arm.com>
7185 David Sherwood <david.sherwood@arm.com>
7187 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
7188 number of units as polynomial.
7190 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7191 Alan Hayward <alan.hayward@arm.com>
7192 David Sherwood <david.sherwood@arm.com>
7194 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
7195 * target.def (autovectorize_vector_sizes): Return the vector sizes
7196 by pointer, using vector_sizes rather than a bitmask.
7197 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
7198 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
7199 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
7201 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
7202 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
7203 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
7204 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
7205 * omp-general.c (omp_max_vf): Likewise.
7206 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7207 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
7208 * tree-vect-loop.c (vect_analyze_loop): Likewise.
7209 * tree-vect-slp.c (vect_slp_bb): Likewise.
7210 * doc/tm.texi: Regenerate.
7211 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
7213 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
7214 the vector size as a poly_uint64 rather than an unsigned int.
7215 (current_vector_size): Change from an unsigned int to a poly_uint64.
7216 (get_vectype_for_scalar_type): Update accordingly.
7217 * tree.h (build_truth_vector_type): Take the size and number of
7218 units as a poly_uint64 rather than an unsigned int.
7219 (build_vector_type): Add a temporary overload that takes
7220 the number of units as a poly_uint64 rather than an unsigned int.
7221 * tree.c (make_vector_type): Likewise.
7222 (build_truth_vector_type): Take the number of units as a poly_uint64
7223 rather than an unsigned int.
7225 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7226 Alan Hayward <alan.hayward@arm.com>
7227 David Sherwood <david.sherwood@arm.com>
7229 * target.def (get_mask_mode): Take the number of units and length
7230 as poly_uint64s rather than unsigned ints.
7231 * targhooks.h (default_get_mask_mode): Update accordingly.
7232 * targhooks.c (default_get_mask_mode): Likewise.
7233 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
7234 * doc/tm.texi: Regenerate.
7236 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7237 Alan Hayward <alan.hayward@arm.com>
7238 David Sherwood <david.sherwood@arm.com>
7240 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
7241 * omp-general.c (omp_max_vf): Likewise.
7242 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
7243 (expand_omp_simd): Handle polynomial safelen.
7244 * omp-low.c (omplow_simd_context): Add a default constructor.
7245 (omplow_simd_context::max_vf): Change from int to poly_uint64.
7246 (lower_rec_simd_input_clauses): Update accordingly.
7247 (lower_rec_input_clauses): Likewise.
7249 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7250 Alan Hayward <alan.hayward@arm.com>
7251 David Sherwood <david.sherwood@arm.com>
7253 * tree-vectorizer.h (vect_nunits_for_cost): New function.
7254 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
7255 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
7256 (vect_analyze_slp_cost): Likewise.
7257 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
7258 (vect_model_load_cost): Likewise.
7260 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7261 Alan Hayward <alan.hayward@arm.com>
7262 David Sherwood <david.sherwood@arm.com>
7264 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
7265 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
7266 from an unsigned int * to a poly_uint64_pod *.
7267 (calculate_unrolling_factor): New function.
7268 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
7270 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7271 Alan Hayward <alan.hayward@arm.com>
7272 David Sherwood <david.sherwood@arm.com>
7274 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
7275 from an unsigned int to a poly_uint64.
7276 (_loop_vec_info::slp_unrolling_factor): Likewise.
7277 (_loop_vec_info::vectorization_factor): Change from an int
7279 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
7280 (vect_get_num_vectors): New function.
7281 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
7282 (vect_get_num_copies): Use vect_get_num_vectors.
7283 (vect_analyze_data_ref_dependences): Change max_vf from an int *
7284 to an unsigned int *.
7285 (vect_analyze_data_refs): Change min_vf from an int * to a
7287 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7288 than an unsigned HOST_WIDE_INT.
7289 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7290 (vect_analyze_data_ref_dependence): Change max_vf from an int *
7291 to an unsigned int *.
7292 (vect_analyze_data_ref_dependences): Likewise.
7293 (vect_compute_data_ref_alignment): Handle polynomial vf.
7294 (vect_enhance_data_refs_alignment): Likewise.
7295 (vect_prune_runtime_alias_test_list): Likewise.
7296 (vect_shift_permute_load_chain): Likewise.
7297 (vect_supportable_dr_alignment): Likewise.
7298 (dependence_distance_ge_vf): Take the vectorization factor as a
7299 poly_uint64 rather than an unsigned HOST_WIDE_INT.
7300 (vect_analyze_data_refs): Change min_vf from an int * to a
7302 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
7303 vfm1 as a poly_uint64 rather than an int. Make the same change
7304 for the returned bound_scalar.
7305 (vect_gen_vector_loop_niters): Handle polynomial vf.
7306 (vect_do_peeling): Likewise. Update call to
7307 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
7308 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
7310 * tree-vect-loop.c (vect_determine_vectorization_factor)
7311 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
7312 (vect_get_known_peeling_cost): Likewise.
7313 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
7314 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
7315 (vect_transform_loop): Likewise. Use the lowest possible VF when
7316 updating the upper bounds of the loop.
7317 (vect_min_worthwhile_factor): Make static. Return an unsigned int
7319 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
7320 polynomial unroll factors.
7321 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
7322 (vect_make_slp_decision): Likewise.
7323 (vect_supported_load_permutation_p): Likewise, and polynomial
7325 (vect_analyze_slp_cost): Handle polynomial vf.
7326 (vect_slp_analyze_node_operations): Likewise.
7327 (vect_slp_analyze_bb_1): Likewise.
7328 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7329 than an unsigned HOST_WIDE_INT.
7330 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
7331 (vectorizable_load): Handle polynomial vf.
7332 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
7334 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
7336 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7337 Alan Hayward <alan.hayward@arm.com>
7338 David Sherwood <david.sherwood@arm.com>
7340 * match.pd: Handle bit operations involving three constants
7341 and try to fold one pair.
7343 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7345 * tree-vect-loop-manip.c: Include gimple-fold.h.
7346 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
7347 niters_maybe_zero parameters. Handle other cases besides a step of 1.
7348 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
7349 Add a path that uses a step of VF instead of 1, but disable it
7351 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
7352 and niters_no_overflow parameters. Update calls to
7353 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
7354 Create a new SSA name if the latter choses to use a ste other
7355 than zero, and return it via niters_vector_mult_vf_var.
7356 * tree-vect-loop.c (vect_transform_loop): Update calls to
7357 vect_do_peeling, vect_gen_vector_loop_niters and
7358 slpeel_make_loop_iterate_ntimes.
7359 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
7360 (vect_gen_vector_loop_niters): Update declarations after above changes.
7362 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
7364 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
7365 128-bit round to integer instructions.
7366 (ceil<mode>2): Likewise.
7367 (btrunc<mode>2): Likewise.
7368 (round<mode>2): Likewise.
7370 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7372 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
7373 unaligned VSX load/store on P8/P9.
7374 (expand_block_clear): Allow the use of unaligned VSX
7375 load/store on P8/P9.
7377 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7379 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
7381 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
7382 swap associated with both a load and a store.
7384 2018-01-02 Andrew Waterman <andrew@sifive.com>
7386 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
7387 * config/riscv/riscv.md (clear_cache): Use it.
7389 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
7391 * web.c: Remove out-of-date comment.
7393 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7395 * expr.c (fixup_args_size_notes): Check that any existing
7396 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
7397 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
7398 (emit_single_push_insn): ...here.
7400 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7402 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
7403 (const_vector_encoded_nelts): New function.
7404 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
7405 (const_vector_int_elt, const_vector_elt): Declare.
7406 * emit-rtl.c (const_vector_int_elt_1): New function.
7407 (const_vector_elt): Likewise.
7408 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
7409 of CONST_VECTOR_ELT.
7411 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7413 * expr.c: Include rtx-vector-builder.h.
7414 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
7415 directly on the tree encoding.
7416 (const_vector_from_tree): Likewise.
7417 * optabs.c: Include rtx-vector-builder.h.
7418 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
7419 sequence of "u" values.
7420 * vec-perm-indices.c: Include rtx-vector-builder.h.
7421 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
7422 directly on the vec_perm_indices encoding.
7424 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7426 * doc/rtl.texi (const_vector): Describe new encoding scheme.
7427 * Makefile.in (OBJS): Add rtx-vector-builder.o.
7428 * rtx-vector-builder.h: New file.
7429 * rtx-vector-builder.c: Likewise.
7430 * rtl.h (rtx_def::u2): Add a const_vector field.
7431 (CONST_VECTOR_NPATTERNS): New macro.
7432 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
7433 (CONST_VECTOR_DUPLICATE_P): Likewise.
7434 (CONST_VECTOR_STEPPED_P): Likewise.
7435 (CONST_VECTOR_ENCODED_ELT): Likewise.
7436 (const_vec_duplicate_p): Check for a duplicated vector encoding.
7437 (unwrap_const_vec_duplicate): Likewise.
7438 (const_vec_series_p): Check for a non-duplicated vector encoding.
7439 Say that the function only returns true for integer vectors.
7440 * emit-rtl.c: Include rtx-vector-builder.h.
7441 (gen_const_vec_duplicate_1): Delete.
7442 (gen_const_vector): Call gen_const_vec_duplicate instead of
7443 gen_const_vec_duplicate_1.
7444 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
7445 (gen_const_vec_duplicate): Use rtx_vector_builder.
7446 (gen_const_vec_series): Likewise.
7447 (gen_rtx_CONST_VECTOR): Likewise.
7448 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
7449 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7450 Build a new vector rather than modifying a CONST_VECTOR in-place.
7451 (handle_special_swappables): Update call accordingly.
7452 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
7453 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7454 Build a new vector rather than modifying a CONST_VECTOR in-place.
7455 (handle_special_swappables): Update call accordingly.
7457 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7459 * simplify-rtx.c (simplify_const_binary_operation): Use
7460 CONST_VECTOR_ELT instead of XVECEXP.
7462 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7464 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
7465 the selector elements to be different from the data elements
7466 if the selector is a VECTOR_CST.
7467 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
7468 ssizetype for the selector.
7470 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7472 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
7473 before testing each element individually.
7474 * tree-vect-generic.c (lower_vec_perm): Likewise.
7476 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7478 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
7479 * selftest-run-tests.c (selftest::run_tests): Call it.
7480 * vector-builder.h (vector_builder::operator ==): New function.
7481 (vector_builder::operator !=): Likewise.
7482 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
7483 (vec_perm_indices::all_from_input_p): New function.
7484 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7485 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
7486 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
7487 instead of reading the VECTOR_CST directly. Detect whether both
7488 vector inputs are the same before constructing the vec_perm_indices,
7489 and update the number of inputs argument accordingly. Use the
7490 utility functions added above. Only construct sel2 if we need to.
7492 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7494 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
7495 the broadcast of the low byte.
7496 (expand_mult_highpart): Use an explicit encoding for the permutes.
7497 * optabs-query.c (can_mult_highpart_p): Likewise.
7498 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
7499 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7500 (vectorizable_bswap): Likewise.
7501 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
7502 explicit encoding for the power-of-2 permutes.
7503 (vect_permute_store_chain): Likewise.
7504 (vect_grouped_load_supported): Likewise.
7505 (vect_permute_load_chain): Likewise.
7507 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7509 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
7510 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
7511 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
7512 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7513 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
7514 (vect_gen_perm_mask_any): Likewise.
7516 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7518 * int-vector-builder.h: New file.
7519 * vec-perm-indices.h: Include int-vector-builder.h.
7520 (vec_perm_indices): Redefine as an int_vector_builder.
7521 (auto_vec_perm_indices): Delete.
7522 (vec_perm_builder): Redefine as a stand-alone class.
7523 (vec_perm_indices::vec_perm_indices): New function.
7524 (vec_perm_indices::clamp): Likewise.
7525 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
7526 (vec_perm_indices::new_vector): New function.
7527 (vec_perm_indices::new_expanded_vector): Update for new
7528 vec_perm_indices class.
7529 (vec_perm_indices::rotate_inputs): New function.
7530 (vec_perm_indices::all_in_range_p): Operate directly on the
7531 encoded form, without computing elided elements.
7532 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
7533 encoding. Update for new vec_perm_indices class.
7534 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
7535 the given vec_perm_builder.
7536 (expand_vec_perm_var): Update vec_perm_builder constructor.
7537 (expand_mult_highpart): Use vec_perm_builder instead of
7538 auto_vec_perm_indices.
7539 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
7540 vec_perm_indices instead of auto_vec_perm_indices. Use a single
7541 or double series encoding as appropriate.
7542 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
7543 vec_perm_indices instead of auto_vec_perm_indices.
7544 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7545 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7546 (vect_permute_store_chain): Likewise.
7547 (vect_grouped_load_supported): Likewise.
7548 (vect_permute_load_chain): Likewise.
7549 (vect_shift_permute_load_chain): Likewise.
7550 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
7551 (vect_transform_slp_perm_load): Likewise.
7552 (vect_schedule_slp_instance): Likewise.
7553 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7554 (vectorizable_mask_load_store): Likewise.
7555 (vectorizable_bswap): Likewise.
7556 (vectorizable_store): Likewise.
7557 (vectorizable_load): Likewise.
7558 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
7559 vec_perm_indices instead of auto_vec_perm_indices. Use
7560 tree_to_vec_perm_builder to read the vector from a tree.
7561 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
7562 vec_perm_builder instead of a vec_perm_indices.
7563 (have_whole_vector_shift): Use vec_perm_builder and
7564 vec_perm_indices instead of auto_vec_perm_indices. Leave the
7565 truncation to calc_vec_perm_mask_for_shift.
7566 (vect_create_epilog_for_reduction): Likewise.
7567 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
7568 from auto_vec_perm_indices to vec_perm_indices.
7569 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7570 instead of changing individual elements.
7571 (aarch64_vectorize_vec_perm_const): Use new_vector to install
7572 the vector in d.perm.
7573 * config/arm/arm.c (expand_vec_perm_d::perm): Change
7574 from auto_vec_perm_indices to vec_perm_indices.
7575 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7576 instead of changing individual elements.
7577 (arm_vectorize_vec_perm_const): Use new_vector to install
7578 the vector in d.perm.
7579 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
7580 Update vec_perm_builder constructor.
7581 (rs6000_expand_interleave): Likewise.
7582 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
7583 (rs6000_expand_interleave): Likewise.
7585 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7587 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
7588 to qimode could truncate the indices.
7589 * optabs.c (expand_vec_perm_var): Likewise.
7591 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7593 * Makefile.in (OBJS): Add vec-perm-indices.o.
7594 * vec-perm-indices.h: New file.
7595 * vec-perm-indices.c: Likewise.
7596 * target.h (vec_perm_indices): Replace with a forward class
7598 (auto_vec_perm_indices): Move to vec-perm-indices.h.
7599 * optabs.h: Include vec-perm-indices.h.
7600 (expand_vec_perm): Delete.
7601 (selector_fits_mode_p, expand_vec_perm_var): Declare.
7602 (expand_vec_perm_const): Declare.
7603 * target.def (vec_perm_const_ok): Replace with...
7604 (vec_perm_const): ...this new hook.
7605 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
7606 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
7607 * doc/tm.texi: Regenerate.
7608 * optabs.def (vec_perm_const): Delete.
7609 * doc/md.texi (vec_perm_const): Likewise.
7610 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
7611 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
7612 expand_vec_perm for constant permutation vectors. Assert that
7613 the mode of variable permutation vectors is the integer equivalent
7614 of the mode that is being permuted.
7615 * optabs-query.h (selector_fits_mode_p): Declare.
7616 * optabs-query.c: Include vec-perm-indices.h.
7617 (selector_fits_mode_p): New function.
7618 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
7619 is defined, instead of checking whether the vec_perm_const_optab
7620 exists. Use targetm.vectorize.vec_perm_const instead of
7621 targetm.vectorize.vec_perm_const_ok. Check whether the indices
7622 fit in the vector mode before using a variable permute.
7623 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
7624 vec_perm_indices instead of an rtx.
7625 (expand_vec_perm): Replace with...
7626 (expand_vec_perm_const): ...this new function. Take the selector
7627 as a vec_perm_indices rather than an rtx. Also take the mode of
7628 the selector. Update call to shift_amt_for_vec_perm_mask.
7629 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
7630 Use vec_perm_indices::new_expanded_vector to expand the original
7631 selector into bytes. Check whether the indices fit in the vector
7632 mode before using a variable permute.
7633 (expand_vec_perm_var): Make global.
7634 (expand_mult_highpart): Use expand_vec_perm_const.
7635 * fold-const.c: Includes vec-perm-indices.h.
7636 * tree-ssa-forwprop.c: Likewise.
7637 * tree-vect-data-refs.c: Likewise.
7638 * tree-vect-generic.c: Likewise.
7639 * tree-vect-loop.c: Likewise.
7640 * tree-vect-slp.c: Likewise.
7641 * tree-vect-stmts.c: Likewise.
7642 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
7644 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
7645 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
7646 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
7647 (aarch64_vectorize_vec_perm_const): ...this new function.
7648 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7649 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7650 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
7651 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
7652 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7653 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7654 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
7656 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
7657 check for NEON modes.
7658 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
7659 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
7660 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
7661 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
7663 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
7664 the old VEC_PERM_CONST conditions.
7665 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
7666 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
7667 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
7668 (ia64_vectorize_vec_perm_const_ok): Merge into...
7669 (ia64_vectorize_vec_perm_const): ...this new function.
7670 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
7671 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
7672 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
7673 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
7674 * config/mips/mips.c (mips_expand_vec_perm_const)
7675 (mips_vectorize_vec_perm_const_ok): Merge into...
7676 (mips_vectorize_vec_perm_const): ...this new function.
7677 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
7678 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
7679 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
7680 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
7681 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
7682 (rs6000_expand_vec_perm_const): Delete.
7683 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
7685 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7686 (altivec_expand_vec_perm_const_le): Take each operand individually.
7687 Operate on constant selectors rather than rtxes.
7688 (altivec_expand_vec_perm_const): Likewise. Update call to
7689 altivec_expand_vec_perm_const_le.
7690 (rs6000_expand_vec_perm_const): Delete.
7691 (rs6000_vectorize_vec_perm_const_ok): Delete.
7692 (rs6000_vectorize_vec_perm_const): New function.
7693 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
7694 an element count and rtx array.
7695 (rs6000_expand_extract_even): Update call accordingly.
7696 (rs6000_expand_interleave): Likewise.
7697 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
7698 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
7699 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
7700 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
7701 (rs6000_expand_vec_perm_const): Delete.
7702 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7703 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7704 (altivec_expand_vec_perm_const_le): Take each operand individually.
7705 Operate on constant selectors rather than rtxes.
7706 (altivec_expand_vec_perm_const): Likewise. Update call to
7707 altivec_expand_vec_perm_const_le.
7708 (rs6000_expand_vec_perm_const): Delete.
7709 (rs6000_vectorize_vec_perm_const_ok): Delete.
7710 (rs6000_vectorize_vec_perm_const): New function. Remove stray
7711 reference to the SPE evmerge intructions.
7712 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
7713 an element count and rtx array.
7714 (rs6000_expand_extract_even): Update call accordingly.
7715 (rs6000_expand_interleave): Likewise.
7716 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
7717 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
7719 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7721 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7723 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
7724 vector mode and that that mode matches the mode of the data
7726 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
7727 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
7728 directly using expand_vec_perm_1 when forcing selectors into
7730 (expand_vec_perm_var): New function, split out from expand_vec_perm.
7732 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7734 * optabs-query.h (can_vec_perm_p): Delete.
7735 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
7736 * optabs-query.c (can_vec_perm_p): Split into...
7737 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
7738 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
7739 particular selector is valid.
7740 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7741 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7742 (vect_grouped_load_supported): Likewise.
7743 (vect_shift_permute_load_chain): Likewise.
7744 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
7745 (vect_transform_slp_perm_load): Likewise.
7746 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7747 (vectorizable_bswap): Likewise.
7748 (vect_gen_perm_mask_checked): Likewise.
7749 * fold-const.c (fold_ternary_loc): Likewise. Don't take
7750 implementations of variable permutation vectors into account
7751 when deciding which selector to use.
7752 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
7753 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
7754 with a false third argument.
7755 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
7756 to test whether the constant selector is valid and can_vec_perm_var_p
7757 to test whether a variable selector is valid.
7759 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7761 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
7762 * optabs-query.c (can_vec_perm_p): Likewise.
7763 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
7764 instead of vec_perm_indices.
7765 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
7766 (vect_gen_perm_mask_checked): Likewise,
7767 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
7768 (vect_gen_perm_mask_checked): Likewise,
7770 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7772 * optabs-query.h (qimode_for_vec_perm): Declare.
7773 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
7774 (qimode_for_vec_perm): ...this new function.
7775 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
7777 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7779 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
7780 does not have a conditional at the top.
7782 2018-01-02 Richard Biener <rguenther@suse.de>
7784 * ipa-inline.c (big_speedup_p): Fix expression.
7786 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
7789 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
7792 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
7796 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
7797 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
7798 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
7799 cond_taken_branch_cost 3->4.
7801 2018-01-01 Jakub Jelinek <jakub@redhat.com>
7803 PR tree-optimization/83581
7804 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
7805 TODO_cleanup_cfg if any changes have been made.
7808 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
7809 convert_modes if target mode has the right side, but different mode
7813 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
7814 last argument when extracting from CONCAT. If either from_real or
7815 from_imag is NULL, use expansion through memory. If result is not
7816 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
7817 the parts directly to inner mode, if even that fails, use expansion
7821 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
7822 check for bswap in mode rather than HImode and use that in expand_unop
7825 Copyright (C) 2018 Free Software Foundation, Inc.
7827 Copying and distribution of this file, with or without modification,
7828 are permitted in any medium without royalty provided the copyright
7829 notice and this notice are preserved.