1 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
7 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
8 instead of -1U in last predictors element's probability member.
10 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
13 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
14 arguments if they are comparable.
16 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
18 PR tree-optimization/84634
19 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
20 masks and masked_loop_p with a single loop_masks, making sure it's
21 null for bb vectorization.
23 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
25 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
26 (vect_analyze_data_ref_access): Use loop->safe_len rather than
27 loop->force_vectorize to check whether there is no alias.
29 2018-03-02 Jakub Jelinek <jakub@redhat.com>
32 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
34 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
36 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
37 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
38 instead of a loop around prev_real_insn.
39 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
43 * config/i386/i386.c (ix86_print_operand): Use conditional
44 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
47 2018-03-02 Richard Biener <rguenther@suse.de>
49 PR tree-optimization/84427
50 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
51 (bitmap_set_subtract_values): Rewrite to handle multiple
54 (prune_clobbered_mems): Likewise.
55 (phi_translate): Take edge instead of pred/phiblock.
56 (phi_translate_1): Likewise.
57 (phi_translate_set): Likewise. Insert all translated
58 exprs for a value into the set, keeping possibly multiple
59 expressions per value.
60 (compute_antic_aux): Adjust for phi_translate changes.
61 When intersecting union the expressions and prune those
62 not in the final value set, keeping possibly multiple
63 expressions per value. Do not use value-insertion
64 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
65 all expressions. Add verification that the value-sets
66 only shrink during iteration.
67 (compute_partial_antic_aux): Adjust for the phi_translate changes.
68 (do_pre_regular_insertion): Likewise.
69 (do_pre_partial_partial_insertion): Likewise.
71 2018-03-02 Richard Biener <rguenther@suse.de>
74 * config/darwin.c (saved_debug_info_level): New static global.
75 (darwin_asm_lto_start): Disable debug info generation for LTO out.
76 (darwin_asm_lto_end): Restore debug info generation settings.
78 2018-03-01 Martin Liska <mliska@suse.cz>
81 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
84 2018-03-01 Richard Biener <rguenther@suse.de>
87 * dwarf2out.c (gen_variable_die): Properly handle late VLA
88 type annotation with LTO when debug was disabled at compile-time.
90 2018-03-01 Matthew Fortune <mfortune@gmail.com>
92 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
94 (mips_final_postscan_insn): Likewise.
96 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
98 PR rtl-optimization/84528
99 * alias.c (init_alias_target): Add commentary.
100 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
101 a unique base value if the frame pointer is not eliminated
102 to the stack pointer.
104 2018-03-01 Tom de Vries <tom@codesourcery.com>
106 PR rtl-optimization/83327
107 * lra-int.h (hard_regs_spilled_into): Declare.
108 * lra.c (hard_regs_spilled_into): Define.
109 (init_reg_info): Init hard_regs_spilled_into.
110 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
111 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
112 (process_bb_lives): Handle hard_regs_spilled_into.
113 (lra_create_live_ranges_1): Before doing liveness propagation, clear
114 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
116 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
118 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
119 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
120 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
121 * config/rs6000/aix72.h: New file.
123 2018-02-28 Jakub Jelinek <jakub@redhat.com>
125 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
126 instead of warning_at with conditional singular and plural messages
130 * stor-layout.c (update_alignment_for_field): For
131 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
132 && !DECL_PACKED (field), do the alignment update, just use
133 only desired_align instead of MAX (type_align, desired_align)
135 (place_field): Don't do known_align < desired_align handling
136 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
137 is non-NULL, instead do it after rli->prev_field handling and
138 only if not within a bitfield word. For DECL_PACKED (field)
139 use type_align of BITS_PER_UNIT.
141 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
143 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
144 superfluous parentheses and trailing spaces.
146 2018-02-28 Richard Biener <rguenther@suse.de>
148 PR tree-optimization/84584
149 * graphite-scop-detection.c (scop_detection::add_scop): Discard
150 SCoPs with fake exit edge.
152 2018-02-28 Martin Liska <mliska@suse.cz>
155 * timevar.c (timer::print): Fix format to properly print 100%
158 2018-02-28 Richard Biener <rguenther@suse.de>
161 * genmatch.c (capture_info::walk_match): Do not mark
162 captured expressions without operands as expr_p given
163 they act more like predicates and should be subject to
164 "lost tail" side-effect preserving.
166 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
168 PR rtl-optimization/81611
169 * auto-inc-dec.c (attempt_change): Move dead note from
170 mem_insn if it's the next use of regno
171 (find_address): Take address use of reg holding
172 non-incremented value. Add parm to limit search to the named
174 (merge_in_block): Attempt to use a mem insn that is the next
175 use of the original regno.
177 2018-02-27 Martin Sebor <msebor@redhat.com>
180 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
181 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
183 2018-02-27 Martin Sebor <msebor@redhat.com>
186 * diagnostic-core.h (warning_n, error_n, inform_n): Change
187 n argument to unsigned HOST_WIDE_INT.
188 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
189 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
190 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
191 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
193 2018-02-27 Richard Biener <rguenther@suse.de>
195 PR tree-optimization/84512
196 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
197 Do not use the estimate returned from record_stmt_cost for
198 the scalar iteration cost but sum properly using add_stmt_cost.
200 2018-02-27 Richard Biener <rguenther@suse.de>
202 PR tree-optimization/84466
203 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
204 Adjust last change to less strictly validate use operands.
206 2018-02-27 Martin Liska <mliska@suse.cz>
208 PR gcov-profile/84548
209 * gcov.c (process_file): Allow partial overlap and consider it
210 also as group functions.
211 (output_lines): Properly calculate range of lines for a group.
213 2018-02-27 Martin Liska <mliska@suse.cz>
215 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
216 'ggc' suffixes. Change first column width.
217 (timer::print): Fix formatting of the column.
219 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
221 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
222 preserve inline entry blocks for the sake of debug inline
223 entry point markers alone.
224 (remove_unused_locals): Suggest in comments a better place to
225 force the preservation of inline entry blocks that are
226 otherwise unused, but do not preserve them.
228 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
230 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
232 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
235 * config/i386/constraints.md (Bs): Replace
236 ix86_indirect_branch_register with
237 TARGET_INDIRECT_BRANCH_REGISTER.
239 * config/i386/i386.md (indirect_jump): Likewise.
240 (tablejump): Likewise.
241 (*sibcall_memory): Likewise.
242 (*sibcall_value_memory): Likewise.
243 Peepholes of indirect call and jump via memory: Likewise.
244 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
245 (*sibcall_value_GOT_32): Likewise.
246 * config/i386/predicates.md (indirect_branch_operand): Likewise.
247 (GOT_memory_operand): Likewise.
248 (call_insn_operand): Likewise.
249 (sibcall_insn_operand): Likewise.
250 (GOT32_symbol_operand): Likewise.
251 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
253 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
255 PR rtl-optimization/83496
256 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
257 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
258 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
259 redundant insn, if any.
260 (relax_delay_slots): Likewise.
261 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
263 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
265 PR tree-optimization/83965
266 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
267 that grouped statements are part of a reduction chain. Return
268 true if the statement is not marked as a reduction itself but
270 (vect_recog_dot_prod_pattern): Don't check whether the statement
271 is part of a group here.
272 (vect_recog_sad_pattern): Likewise.
273 (vect_recog_widen_sum_pattern): Likewise.
275 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
278 * final.c (rest_of_clean_state): Also look for calls inside sequences.
280 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
283 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
285 (ix86_output_indirect_function_return): New prototype.
286 (ix86_split_simple_return_pop_internal): Likewise.
287 * config/i386/i386.c (indirect_return_via_cx): New.
288 (indirect_return_via_cx_bnd): Likewise.
289 (indirect_thunk_name): Handle return va CX_REG.
290 (output_indirect_thunk_function): Create alias for
291 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
292 (ix86_output_indirect_jmp): Remove the bool argument.
293 (ix86_output_indirect_function_return): New function.
294 (ix86_split_simple_return_pop_internal): Likewise.
295 * config/i386/i386.md (*indirect_jump): Don't pass false
296 to ix86_output_indirect_jmp.
297 (*tablejump_1): Likewise.
298 (simple_return_pop_internal): Change it to define_insn_and_split.
299 Call ix86_split_simple_return_pop_internal to split it for
301 (simple_return_indirect_internal): Call
302 ix86_output_indirect_function_return instead of
303 ix86_output_indirect_jmp.
305 2018-02-26 Jakub Jelinek <jakub@redhat.com>
308 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
309 memset and value initialization afterwards.
311 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
313 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
315 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
318 * common/config/aarch64/aarch64-common.c
319 (aarch_option_optimization_table[]): Switch
320 off fomit-frame-pointer
322 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
323 Chung-Ju Wu <jasonwucj@gmail.com>
325 * config/nds32/nds32-multiple.md (load_multiple): Disallow
327 (store_multiple): Ditto.
329 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
331 * config.gcc: Add --with-cpu support for nds32 target.
332 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
333 * config/nds32/nds32.opt: Add -mcpu= option.
335 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
337 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
338 isel=yes): Warn for these deprecated options.
340 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
342 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
343 ISA_2_5_MASKS_EMBEDDED.
345 2018-02-23 Jakub Jelinek <jakub@redhat.com>
347 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
348 p->max as pointers rather than using iterative_hash_expr.
350 2018-02-23 Carl Love <cel@us.ibm.com>
352 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
353 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
355 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
356 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
357 P8V_BUILTIN_VEC_VUNSIGNED2.
359 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
362 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
363 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
365 * lra-constraints.c (curr_insn_transform): Set up
366 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
368 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
369 * lra-eliminations.c (spill_pseudos): Ditto.
370 (process_insn_for_elimination): Ditto.
371 * lra-lives.c (reg_early_clobber_p): Use the new macros.
372 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
373 LRA_NON_CLOBBERED_ALT.
375 2018-02-22 Martin Sebor <msebor@redhat.com>
377 PR tree-optimization/84480
378 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
379 to maybe_diag_stxncpy_trunc. Call it.
380 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
381 from gimple_fold_builtin_strcpy. Print inlining stack.
382 (handle_builtin_stxncpy): Print inlining stack.
383 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
385 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
388 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
389 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
390 and -fcheck-pointer-bounds are used together.
391 (indirect_thunk_prefix): New enum.
392 (indirect_thunk_need_prefix): New function.
393 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
394 "_nt" instead of "_bnd" for NOTRACK prefix.
395 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
396 (output_indirect_thunk_function): Likewise.
398 (ix86_code_end): Update output_indirect_thunk_function calls.
399 (ix86_output_indirect_branch_via_reg): Replace
400 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
401 (ix86_output_indirect_branch_via_push): Likewise.
402 (ix86_output_function_return): Likewise.
403 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
404 incompatible with -fcf-protection=branch and
405 -fcheck-pointer-bounds.
407 2018-02-22 Steve Ellcey <sellcey@cavium.com>
410 * config/aarch64/aarch64.c (aarch64_print_address_internal):
411 Change gcc_assert call to output_operand_lossage.
413 2018-02-22 Steve Ellcey <sellcey@cavium.com>
415 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
417 2018-02-22 DJ Delorie <dj@redhat.com>
418 Sebastian Perta <sebastian.perta@renesas.com>
419 Oleg Endo <olegendo@gcc.gnu.org>
421 * config/rx/rx.c (rx_rtx_costs): New function.
422 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
424 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
426 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
428 2018-02-22 Martin Liska <mliska@suse.cz>
431 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
432 Add "native" as a possible value.
434 2018-02-22 Martin Liska <mliska@suse.cz>
437 * config/i386/i386.c (ix86_option_override_internal):
438 Add "native" as a possible value for -march and -mtune.
440 2018-02-22 Jakub Jelinek <jakub@redhat.com>
443 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
444 to all type variants.
446 PR tree-optimization/84503
447 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
448 width as info->bitpos + info->bitsize - start.
449 (merged_store_group::merge_overlapping): Simplify width computation.
450 (check_no_overlap): New function.
451 (imm_store_chain_info::try_coalesce_bswap): Compute expected
452 start + width and last_order of the group, fail if check_no_overlap
454 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
455 to group if check_no_overlap fails.
457 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
459 * config/rs6000/altivec.md: Delete contraint arguments to
460 define_expand, define_split, and define_peephole2, and in
461 define_insn_and_split if always unused.
462 * config/rs6000/darwin.md: Ditto.
463 * config/rs6000/dfp.md: Ditto.
464 * config/rs6000/rs6000.md: Ditto.
465 * config/rs6000/sync.md: Ditto.
466 * config/rs6000/vector.md: Ditto.
467 * config/rs6000/vsx.md: Ditto.
469 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
471 * config/rs6000/altivec.md: Write output control strings as braced
472 blocks instead of double-quoted strings.
473 * config/rs6000/darwin.md: Ditto.
474 * config/rs6000/rs6000.md: Ditto.
475 * config/rs6000/vector.md: Ditto.
476 * config/rs6000/vsx.md: Ditto.
478 2018-02-21 Jason Merrill <jason@redhat.com>
480 PR c++/84314 - ICE with templates and fastcall attribute.
481 * attribs.c (build_type_attribute_qual_variant): Remove assert.
483 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
485 * ipa-cp.c (determine_versionability): Fix comment typos.
487 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
490 * ipa-cp.c (determine_versionability): Do not version functions caling
493 2018-02-21 Martin Liska <mliska@suse.cz>
496 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
497 Add "native" as a possible value.
498 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
499 the macro when native cpu detection is available.
501 2018-02-21 Martin Liska <mliska@suse.cz>
504 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
505 Add "native" as a possible value.
506 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
507 when native cpu detection is available.
509 2018-02-21 Jakub Jelinek <jakub@redhat.com>
510 Martin Sebor <msebor@redhat.com>
512 PR tree-optimization/84478
513 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
515 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
516 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
517 support which is conservatively correct, for 2 only stay conservative
518 for maxlen. Formatting and comment capitalization fixes. Add STRICT
519 argument to the 2 argument get_range_strlen, adjust 6 arg
520 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
522 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
523 (gimple_fold_builtin_strlen): Pass true as last argument to
526 2018-02-20 Martin Sebor <msebor@redhat.com>
529 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
530 (builtin_memref::set_base_and_offset): Same. Handle inner references.
531 (builtin_memref::builtin_memref): Factor out parts into
532 set_base_and_offset and call it.
534 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
537 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
538 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
539 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
540 search at the associated MODE_INT.
542 2018-02-20 Jeff Law <law@redhat.com>
545 PR tree-optimization/81592
547 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
548 for range data rather than using global data.
549 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
550 range data rather than using global data.
551 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
552 pass it to children as needed.
553 (struct directive::fmtresult): Similarly.
554 (struct directive::set_width): Similarly.
555 (struct directive::set_precision): Similarly.
556 (format_integer, format_directive, parse_directive): Similarly.
557 (format_none): Accept unnamed vr_values parameter.
558 (format_percent, format_floating, format_character): Similarly.
559 (format_string, format_plain): Similarly.
560 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
561 the EVRP range analyzer for range data rather than using global data.
562 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
563 gimple-ssa-evrp-analyze.h
564 (class sprintf_dom_walker): Add after_dom_children member function.
565 Add evrp_range_analyzer member.
566 (sprintf_dom_walker::before_dom_children): Call into the EVRP
567 range analyzer as needed.
568 (sprintf_dom_walker::after_dom_children): New member function.
569 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
571 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
572 (evrp_range_analyzer::pop_to_marker): Likewise.
574 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
576 PR tree-optimization/84419
577 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
578 with the required type if its current type is compatible but
581 2018-02-20 Jakub Jelinek <jakub@redhat.com>
584 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
587 2018-02-20 Martin Liska <mliska@suse.cz>
590 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
591 possible values if we don't have a hint.
593 2018-02-20 Martin Liska <mliska@suse.cz>
597 * final.c (shorten_branches): Build align_tab array with one
599 * opts.c (finish_options): Add alignment option limit check.
600 (MAX_CODE_ALIGN): Likewise.
601 (MAX_CODE_ALIGN_VALUE): Likewise.
602 * doc/invoke.texi: Document maximum allowed option value for
603 all -falign-* options.
605 2018-02-19 Jakub Jelinek <jakub@redhat.com>
608 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
609 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
610 * var-tracking.c (emit_note_insn_var_location): Remove all references
611 to NOTE_INSN_CALL_ARG_LOCATION.
612 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
613 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
614 Use copy_rtx_if_shared.
615 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
616 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
617 (dwarf2out_var_location): Remove handling of
618 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
620 * final.c (final_scan_insn): Remove all references to
621 NOTE_INSN_CALL_ARG_LOCATION.
622 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
623 before dumping final insns.
624 * except.c (emit_note_eh_region_end): Remove all references to
625 NOTE_INSN_CALL_ARG_LOCATION.
626 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
627 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
628 * config/arc/arc.c (hwloop_optimize): Likewise.
629 * config/arm/arm.c (create_fix_barrier): Likewise.
630 * config/s390/s390.c (s390_chunkify_start): Likewise.
631 * config/sh/sh.c (find_barrier): Likewise.
632 * config/i386/i386.c (rest_of_insert_endbranch,
633 ix86_seh_fixup_eh_fallthru): Likewise.
634 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
635 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
636 * config/frv/frv.c (frv_function_prologue): Likewise.
637 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
639 (note_outside_basic_block_p): Remove all references to
640 NOTE_INSN_CALL_ARG_LOCATION.
641 * gengtype.c (adjust_field_rtx_def): Likewise.
642 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
644 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
645 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
648 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
651 PR tree-optimization/84452
652 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
653 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
656 2018-02-19 Martin Liska <mliska@suse.cz>
659 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
661 2018-02-19 Martin Liska <mliska@suse.cz>
662 Richard Sandiford <richard.sandiford@linaro.org>
664 PR tree-optimization/82491
665 * gimple-fold.c (get_base_constructor): Make earlier bail out
668 2018-02-19 Carl Love <cel@us.ibm.com>
670 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
671 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
673 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
676 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
678 * config/rl78/rl78.md (movdf): New define expand.
680 2018-02-19 Martin Liska <mliska@suse.cz>
683 * doc/invoke.texi: Fix typo.
684 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
686 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
688 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
689 handle rs6000_single_float and rs6000_double_float specially for
692 2018-02-16 Jeff Law <law@redhat.com>
694 * config/rx/rx.c (add_pop_cfi_notes): New function.;
697 2018-02-16 Jakub Jelinek <jakub@redhat.com>
700 * ipa-inline.c (inline_small_functions): Fix a typo.
702 2018-02-16 Nathan Sidwell <nathan@acm.org>
704 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
706 2018-02-16 Carl Love <cel@us.ibm.com>
708 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
709 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
710 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
711 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
712 expansion to P8V_BUILTIN_VEC_FLOAT2.
714 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
716 PR rtl-optimization/70023
717 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
718 src_regno into account.
720 2018-02-16 Carl Love <cel@us.ibm.com>
722 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
723 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
724 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
725 * config/rs6000/rs6000.c: Remove case statements for
726 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
727 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
728 and P9V_BUILTIN_VEC_VINSERT4B.
729 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
730 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
731 * config/rs6000/vsx.md:
732 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
735 2018-02-16 Carl Love <cel@us.ibm.com>
737 * config/rs6000/altivec.h: Add builtin names vec_extract4b
739 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
741 * config/rs6000/rs6000-c.c: Add the definitions for
742 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
743 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
744 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
745 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
746 definition for insert4b and define insn *insert3b_internal.
747 * doc/extend.texi: Add documentation for vec_extract4b.
749 2018-02-16 Nathan Sidwell <nathan@acm.org>
751 * doc/extend.texi (Backwards Compatibility): Mention friend
752 injection. Note for-scope is deprecated.
753 * doc/invoke.texi (-ffriend-injection): Deprecate.
755 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
757 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
758 that moved to I2, also allow destinations that are a paradoxical
759 subreg (instead of a normal reg).
761 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
764 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
767 2018-02-16 Richard Biener <rguenther@suse.de>
769 PR tree-optimization/84037
770 PR tree-optimization/84016
772 * config/i386/i386.c (ix86_builtin_vectorization_cost):
773 Adjust vec_construct for the fact we need additional higher latency
774 128bit inserts for AVX256 and AVX512 vector builds.
775 (ix86_add_stmt_cost): Scale vector construction cost for
778 2018-02-16 Richard Biener <rguenther@suse.de>
780 PR tree-optimization/84417
781 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
782 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
783 (non_rewritable_lvalue_p): Likewise, use poly-ints.
785 2018-02-16 Martin Liska <mliska@suse.cz>
788 * internal-fn.def (ASAN_CHECK): Set proper flags.
789 (ASAN_MARK): Likewise.
791 2018-02-16 Julia Koval <julia.koval@intel.com>
793 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
796 2018-02-16 Jakub Jelinek <jakub@redhat.com>
799 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
800 Use ++iter rather than iter++ for std::list iterators.
801 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
802 defer deleting them until all nodes in the forest are processed. Do
803 free even leaf nodes. Change to_process into auto_vec.
806 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
807 * vec.h (vec_default_construct): Use memset instead of placement new
808 if BROKEN_VALUE_INITIALIZATION is defined.
809 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
810 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
813 PR rtl-optimization/83723
814 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
815 * lra.c (lra_substitute_pseudo): Likewise. If true, use
816 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
818 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
820 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
822 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
824 PR rtl-optimization/81443
825 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
826 from inner REGs to paradoxical SUBREGs.
828 2018-02-16 Richard Biener <rguenther@suse.de>
830 PR tree-optimization/84399
831 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
832 For operands we can analyze at their definition make sure we can
833 analyze them at each use as well.
835 2018-02-16 Richard Biener <rguenther@suse.de>
837 PR tree-optimization/84190
838 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
839 volatile accesses if the decl isn't volatile.
841 2018-02-15 Jason Merrill <jason@redhat.com>
843 PR c++/84314 - ICE with templates and fastcall attribute.
844 * attribs.c (build_type_attribute_qual_variant): Don't clobber
845 TYPE_CANONICAL on an existing type.
847 2018-02-15 Jakub Jelinek <jakub@redhat.com>
849 PR tree-optimization/84383
850 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
851 dstoff nor call operand_equal_p if dstbase is NULL.
853 PR tree-optimization/84334
854 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
855 also a CONSTANT_CLASS_P, punt.
857 2018-02-14 Jim Wilson <jimw@sifive.com>
859 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
860 first SMALL_OPERAND check. New local min_second_step. Move assert
861 to where locals are set. Add TARGET_RVC support.
862 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
864 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
866 * doc/invoke.texi: Correct -Wformat-overflow code sample.
868 2018-02-14 Martin Sebor <msebor@redhat.com>
870 PR tree-optimization/83698
871 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
872 arrays constrain the offset range to their bounds.
873 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
874 (builtin_access::overlap): Avoid setting the size of overlap if it's
876 (maybe_diag_overlap): Also consider arrays when deciding what values
877 of offsets to include in diagnostics.
879 2018-02-14 Martin Sebor <msebor@redhat.com>
882 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
883 that correspond to the kind of a declaration.
885 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
888 * config/pa/pa.md: Load address of PIC label using the linkage table
889 if the label is nonlocal.
891 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
893 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
894 warning message if user requests -maltivec=be.
895 * doc/invoke.texi: Document deprecation of -maltivec=be.
897 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
900 * config/rs6000/rs6000-c.c: Update definitions for
901 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
902 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
904 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
907 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
908 add _get_ssp intrinsics. Remove argument from
909 __builtin_ia32_rdssp[d|q].
910 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
911 * config/i386/i386-builtin.def: Remove argument from
912 __builtin_ia32_rdssp[d|q].
913 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
914 ix86_expand_special_args_builtin for _rdssp[d|q].
915 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
916 Clear register before usage.
917 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
918 Add documentation for new _get_ssp and _inc_ssp intrinsics.
920 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
922 PR tree-optimization/84357
923 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
924 operand 1 of an ARRAY_REF too.
926 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
929 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
930 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
932 (set_of_reg): New struct.
933 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
934 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
935 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
937 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
938 Split into bitclr, bitset, bitinvert patterns if appropriate.
939 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
940 use rx_fuse_in_memory_bitop.
941 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
942 to named insn, correct maximum insn length.
944 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
947 * machmode.def: Define a complex mode for PARTIAL_INT.
948 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
950 * doc/rtl.texi: Document CSPImode.
951 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
953 (msp430_hard_regno_nregs_with_padding): Likewise.
955 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
958 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
960 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
962 PR rtl-optimization/84169
963 * combine.c (try_combine): New variable split_i2i3. Set it to true if
964 we generated a parallel as new i3 and we split that to new i2 and i3
965 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
966 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
967 those to i2, not i1. Partially rewrite this scan code.
969 2018-02-13 Jakub Jelinek <jakub@redhat.com>
972 * stor-layout.c (place_field): For variable length fields, adjust
973 offset_align afterwards not just based on the field's alignment,
974 but also on the size.
977 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
978 of exps and logs in the use_exp2 case.
980 2018-02-13 Jeff Law <law@redhat.com>
982 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
985 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
988 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
992 * common.opt (gas-loc-support, gas-locview-support): New.
993 (ginline-points, ginternal-reset-location-views): New.
994 * doc/invoke.texi: Document them. Use @itemx where intended.
995 (gvariable-location-views): Adjust.
996 * target.def (reset_location_view): New.
997 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
998 (TARGET_RESET_LOCATION_VIEW): New.
999 * doc/tm.texi: Rebuilt.
1000 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
1001 (dwarf2out_default_as_locview_support): New.
1002 (output_asm_line_debug_info): Use option variables.
1003 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
1004 (output_loc_list): Likewise.
1005 (add_high_low_attributes): Check option variables.
1006 Don't output entry view attribute in strict mode.
1007 (gen_inlined_subroutine_die): Check option variables.
1008 (dwarf2out_inline_entry): Likewise.
1009 (init_sections_and_labels): Likewise.
1010 (dwarf2out_early_finish): Likewise.
1011 (maybe_reset_location_view): New, from...
1012 (dwarf2out_var_location): ... here. Call it.
1013 * debug.h (dwarf2out_default_as_loc_support): Declare.
1014 (dwarf2out_default_as_locview_support): Declare.
1015 * hooks.c (hook_int_rtx_insn_0): New.
1016 * hooks.h (hook_int_rtx_insn_0): Declare.
1017 * toplev.c (process_options): Take -gas-loc-support and
1018 -gas-locview-support from dwarf2out. Enable
1019 -gvariable-location-views by default only with locview
1020 assembler support. Enable -ginternal-reset-location-views by
1021 default only if the target defines the corresponding hook.
1022 Enable -ginline-points by default if location views are
1023 enabled; force it disabled if statement frontiers are
1025 * tree-inline.c (expand_call_inline): Check option variables.
1026 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1028 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
1030 PR tree-optimization/84321
1031 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
1032 handling. Also check whether the anti-range contains any values
1033 that satisfy the mask; switch to a VR_RANGE if not.
1035 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
1038 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
1040 2018-02-13 Martin Jambor <mjambor@suse.cz>
1043 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
1044 of call statements, also set location of a load to a temporary.
1046 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
1048 * config/rl78/rl78.c (add_vector_labels): New function.
1049 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
1050 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
1051 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
1052 which checks that no arguments are passed.
1053 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
1054 * doc/extend.texi: Documentation for the new attribute.
1056 2018-02-13 Andreas Schwab <schwab@suse.de>
1058 * config/riscv/linux.h (CPP_SPEC): Define.
1060 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1063 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
1064 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
1065 OPTION_MASK_ISA_AES as first argument to def_builtin_const
1066 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
1067 instead of OPTION_MASK_ISA_PCLMUL as first argument to
1068 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
1069 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
1070 temporarily for AES and PCLMUL builtins.
1072 PR tree-optimization/84339
1073 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
1074 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
1078 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
1079 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
1080 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
1082 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
1084 * omp-simd-clone.h: New file.
1085 * omp-simd-clone.c: Include omp-simd-clone.h.
1086 (expand_simd_clones): No longer static.
1087 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
1088 cgraph.h and omp-simd-clone.h.
1089 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
1090 (vect_recog_widen_shift_pattern): Formatting fix.
1091 (vect_pattern_recog_1): Don't check optab for calls.
1094 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
1095 operands[2] into a REG before using gen_lowpart on it.
1097 2018-02-12 Jeff Law <law@redhat.com>
1100 * config/sh/sh.c (find_barrier): Consider a sibling call
1103 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
1104 successfully back substituting a reg.
1106 2018-02-12 Richard Biener <rguenther@suse.de>
1108 PR tree-optimization/84037
1109 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
1110 parameter, move visited init to caller.
1111 (vect_slp_analyze_operations): Separate cost from validity
1112 check, initialize visited once for all instances.
1113 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
1115 * tree-vect-stmts.c (vect_model_simple_cost): Make early
1117 (vect_model_promotion_demotion_cost): Likewise.
1118 (vectorizable_bswap): Guard cost modeling with !slp_node
1119 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
1121 (vectorizable_call): Likewise.
1122 (vectorizable_conversion): Likewise.
1123 (vectorizable_assignment): Likewise.
1124 (vectorizable_shift): Likewise.
1125 (vectorizable_operation): Likewise.
1126 (vectorizable_store): Likewise.
1127 (vectorizable_load): Likewise.
1128 (vectorizable_condition): Likewise.
1129 (vectorizable_comparison): Likewise.
1131 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
1134 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
1135 (ASAN_MARK): Fix fnspec to account for return value, change pointer
1136 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
1138 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1141 * params.def (inline-min-speedup): Increase from 8 to 15.
1142 (max-inline-insns-auto): Decrease from 40 to 30.
1143 * ipa-split.c (consider_split): Add some buffer for function to
1144 be considered inlining candidate.
1145 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
1148 2018-02-12 Richard Biener <rguenther@suse.de>
1150 PR tree-optimization/84037
1151 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
1152 matched stmts if we cannot swap the non-matched ones.
1154 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1156 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
1157 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
1158 _mm_maskz_scalef_round_ss): New intrinsics.
1159 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
1160 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
1161 __builtin_ia32_scalefss_round): Remove.
1162 (__builtin_ia32_scalefsd_mask_round,
1163 __builtin_ia32_scalefss_mask_round): New intrinsics.
1164 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
1165 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
1166 ((match_operand:VF_128 2 "<round_nimm_predicate>"
1167 "<round_constraint>")): Changed to ...
1168 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
1169 "<round_scalar_constraint>")): ... this.
1170 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
1171 %0, %1, %2<round_op3>}"): Changed to ...
1172 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
1173 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
1174 %2<round_scalar_mask_op3>}"): ... this.
1175 * config/i386/subst.md (round_scalar_nimm_predicate): New.
1177 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1179 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
1180 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
1181 (_mm_maskz_sqrt_round_ss): New intrinsics.
1182 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
1183 (__builtin_ia32_sqrtsd_mask_round)
1184 (__builtin_ia32_sqrtss_mask_round): New builtins.
1185 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
1186 (__builtin_ia32_sqrtss_round): Remove.
1187 (__builtin_ia32_sqrtsd_mask_round)
1188 (__builtin_ia32_sqrtss_mask_round): New builtins.
1189 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
1190 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
1191 ((match_operand:VF_128 1 "vector_operand"
1192 "xBm,<round_constraint>")): Changed to ...
1193 ((match_operand:VF_128 1 "vector_operand"
1194 "xBm,<round_scalar_constraint>")): ... this.
1195 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
1196 %0, %2, %<iptr>1<round_op3>}): Changed to ...
1197 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
1198 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
1199 %<iptr>1<round_scalar_mask_op3>}): ... this.
1200 ((set_attr "prefix" "<round_prefix>")): Changed to ...
1201 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
1203 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
1206 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
1207 Cast vec_cmpeq result to correct type.
1208 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
1209 Cast vec_cmpgt result to correct type.
1211 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
1213 * final.c (final_scan_insn_1): Renamed from...
1214 (final_scan_insn): ... this. New wrapper, to recover
1215 seen from the outermost call in recursive ones.
1216 * config/sparc/sparc.c (output_return): Drop seen from call.
1217 (output_sibcall): Likewise.
1218 * config/visium/visium.c (output_branch): Likewise.
1220 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
1222 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
1225 2018-02-10 Alan Modra <amodra@gmail.com>
1228 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
1229 Specify LR as an input.
1231 2018-02-10 Jakub Jelinek <jakub@redhat.com>
1234 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
1235 remove_member_access_dummy_vars): New functions.
1236 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
1237 lower_omp_1, execute_lower_omp): Use them.
1239 PR rtl-optimization/84308
1240 * shrink-wrap.c (spread_components): Release todo vector.
1242 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
1244 PR rtl-optimization/57193
1245 * ira-color.c (struct allocno_color_data): Add member
1246 conflict_allocno_hard_prefs.
1247 (update_conflict_allocno_hard_prefs): New.
1248 (bucket_allocno_compare_func): Add a preference based on
1249 conflict_allocno_hard_prefs.
1250 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
1251 (color_allocnos): Remove a dead code. Initiate
1252 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
1254 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1257 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
1258 constraint from =wa to wa. Avoid a subreg on the output operand,
1259 instead use a pseudo and subreg it in a move.
1260 (p9_xxbrd_<mode>): Changed to ...
1261 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
1262 (p9_xxbrd_v2df): New expander.
1263 (p9_xxbrw_<mode>): Changed to ...
1264 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
1265 (p9_xxbrw_v4sf): New expander.
1267 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1269 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
1271 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
1274 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
1275 multiply in 32-bit mode.
1276 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
1277 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
1280 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1282 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
1283 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
1284 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
1285 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
1287 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
1290 * dwarf2out.c (is_trivial_indirect_ref): New function.
1291 (dwarf2out_late_global_decl): Do not generate a location
1292 attribute for variables that have a non-trivial DECL_VALUE_EXPR
1293 and that are not defined in the current unit.
1295 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1297 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
1298 instead of a libcall for UNORDERED.
1300 2018-02-09 Tamar Christina <tamar.christina@arm.com>
1303 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
1304 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
1306 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1309 * config/s390/s390.c (s390_set_current_function): Invoke
1310 s390_indirect_branch_settings also if fndecl didn't change.
1312 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1314 * config/rs6000/rs6000.md (blockage): Set length to zero.
1316 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1318 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
1320 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1323 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
1324 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
1328 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
1329 PARALLEL incoming that failed vt_get_decl_and_offset check.
1332 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
1333 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
1335 (get_variable_section): For decls in named .bss* sections pass true as
1336 second argument to bss_initializer_p.
1338 2018-02-09 Marek Polacek <polacek@redhat.com>
1339 Jakub Jelinek <jakub@redhat.com>
1342 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
1343 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
1344 Sync some changes from cxx_fold_indirect_ref.
1346 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1348 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
1350 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
1351 (BLOCK_INLINE_ENTRY_LABEL): New.
1352 (dwarf2out_var_location): Disregard inline entry markers.
1353 (inline_entry_data): New struct.
1354 (inline_entry_data_hasher): New hashtable type.
1355 (inline_entry_data_hasher::hash): New.
1356 (inline_entry_data_hasher::equal): New.
1357 (inline_entry_data_table): New variable.
1358 (add_high_low_attributes): Add DW_AT_entry_pc and
1359 DW_AT_GNU_entry_view attributes if a pending entry is found
1360 in inline_entry_data_table. Add old entry_pc attribute only
1361 if debug nonbinding markers are disabled.
1362 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
1363 markers are enabled.
1364 (block_within_block_p, dwarf2out_inline_entry): New.
1365 (dwarf2out_finish): Check that no entries remained in
1366 inline_entry_data_table.
1367 * final.c (reemit_insn_block_notes): Handle inline entry notes.
1368 (final_scan_insn, notice_source_line): Likewise.
1369 (rest_of_clean_state): Skip inline entry markers.
1370 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
1372 * gimple.c (gimple_build_debug_inline_entry): New.
1373 * gimple.h (enum gimple_debug_subcode): Add
1374 GIMPLE_DEBUG_INLINE_ENTRY.
1375 (gimple_build_debug_inline_entry): Declare.
1376 (gimple_debug_inline_entry_p): New.
1377 (gimple_debug_nonbind_marker_p): Adjust.
1378 * insn-notes.def (INLINE_ENTRY): New.
1379 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
1380 inline entry marker notes.
1381 (print_insn): Likewise.
1382 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
1383 (INSN_DEBUG_MARKER_KIND): Likewise.
1384 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
1385 * tree-inline.c (expand_call_inline): Build and insert
1386 debug_inline_entry stmt.
1387 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
1388 inline entry blocks early, if nonbind markers are enabled.
1389 (dump_scope_block): Dump fragment info.
1390 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
1391 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
1392 (gimple_build_debug_inline_entry): New.
1393 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
1394 Enable/disable inline entry points too.
1395 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
1396 (DEBUG_INSN): Describe inline entry markers.
1398 * common.opt (gvariable-location-views): New.
1399 (gvariable-location-views=incompat5): New.
1400 * config.in: Rebuilt.
1401 * configure: Rebuilt.
1402 * configure.ac: Test assembler for view support.
1403 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
1404 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
1405 * dwarf2out.c (var_loc_view): New typedef.
1406 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
1407 (dwarf2out_locviews_in_attribute): New.
1408 (dwarf2out_locviews_in_loclist): New.
1409 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
1410 (enum dw_line_info_opcode): Add LI_adv_address.
1411 (struct dw_line_info_table): Add view.
1412 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
1413 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
1414 (zero_view_p): New variable.
1415 (ZERO_VIEW_P): New macro.
1416 (output_asm_line_debug_info): New.
1417 (struct var_loc_node): Add view.
1418 (add_AT_view_list, AT_loc_list): New.
1419 (add_var_loc_to_decl): Add view param. Test it against last.
1420 (new_loc_list): Add view params. Record them.
1421 (AT_loc_list_ptr): Handle loc and view lists.
1422 (view_list_to_loc_list_val_node): New.
1423 (print_dw_val): Handle dw_val_class_view_list.
1424 (size_of_die): Likewise.
1425 (value_format): Likewise.
1426 (loc_list_has_views): New.
1427 (gen_llsym): Set vl_symbol too.
1428 (maybe_gen_llsym, skip_loc_list_entry): New.
1429 (dwarf2out_maybe_output_loclist_view_pair): New.
1430 (output_loc_list): Output view list or entries too.
1431 (output_view_list_offset): New.
1432 (output_die): Handle dw_val_class_view_list.
1433 (output_dwarf_version): New.
1434 (output_compilation_unit_header): Use it.
1435 (output_skeleton_debug_sections): Likewise.
1436 (output_rnglists, output_line_info): Likewise.
1437 (output_pubnames, output_aranges): Update version comments.
1438 (output_one_line_info_table): Output view numbers in asm comments.
1439 (dw_loc_list): Determine current endview, pass it to new_loc_list.
1440 Call maybe_gen_llsym.
1441 (loc_list_from_tree_1): Adjust.
1442 (add_AT_location_description): Create view list attribute if
1443 needed, check it's absent otherwise.
1444 (convert_cfa_to_fb_loc_list): Adjust.
1445 (maybe_emit_file): Call output_asm_line_debug_info for test.
1446 (dwarf2out_var_location): Reset views as needed. Precompute
1447 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
1448 attribute. Set view.
1449 (new_line_info_table): Reset next view.
1450 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
1451 (dwarf2out_source_line): Likewise. Output view resets and labels to
1452 the assembler, or select appropriate line info opcodes.
1453 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
1454 (optimize_string_length): Catch it. Adjust.
1455 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
1456 dw_val_class_view_list, and remove it if no longer needed.
1457 (hash_loc_list): Hash view numbers.
1458 (loc_list_hasher::equal): Compare them.
1459 (optimize_location_lists): Check whether a view list symbol is
1460 needed, and whether the locview attribute is present, and
1461 whether they match. Remove the locview attribute if no longer
1463 (index_location_lists): Call skip_loc_list_entry for test.
1464 (dwarf2out_finish): Call output_asm_line_debug_info for test.
1465 Use output_dwarf_version.
1466 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
1467 (struct dw_val_node): Add val_view_list.
1468 * final.c (SEEN_NEXT_VIEW): New.
1469 (set_next_view_needed): New.
1470 (clear_next_view_needed): New.
1471 (maybe_output_next_view): New.
1472 (final_start_function): Rename to...
1473 (final_start_function_1): ... this. Take pointer to FIRST,
1474 add SEEN parameter. Emit param bindings in the initial view.
1475 (final_start_function): Reintroduce SEEN-less interface.
1476 (final): Rename to...
1477 (final_1): ... this. Take SEEN parameter. Output final pending
1478 next view at the end.
1479 (final): Reintroduce seen-less interface.
1480 (final_scan_insn): Output pending next view before switching
1481 sections or ending a block. Mark the next view as needed when
1482 outputting variable locations. Notify debug backend of section
1483 changes, and of location view changes.
1484 (rest_of_handle_final): Adjust.
1485 * toplev.c (process_options): Autodetect value for debug variable
1486 location views option. Warn on incompat5 without -gdwarf-5.
1487 * doc/invoke.texi (gvariable-location-views): New.
1488 (gvariable-location-views=incompat5): New.
1489 (gno-variable-location-views): New.
1491 2018-02-08 David Malcolm <dmalcolm@redhat.com>
1493 PR tree-optimization/84136
1494 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
1495 that the result of find_edge is non-NULL.
1497 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1500 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
1501 storing integer register in SImode. Fix cost of 256 and 512
1502 byte aligned SSE register store.
1504 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1506 * config/i386/i386.c (ix86_multiplication_cost): Fix
1507 multiplication cost for TARGET_AVX512DQ.
1509 2018-02-08 Marek Polacek <polacek@redhat.com>
1511 PR tree-optimization/84238
1512 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
1515 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1517 PR tree-optimization/84265
1518 * tree-vect-stmts.c (vectorizable_store): Don't treat
1519 VMAT_CONTIGUOUS accesses as grouped.
1520 (vectorizable_load): Likewise.
1522 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1524 PR tree-optimization/81635
1525 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
1526 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
1527 (test_round_for_mask): New functions.
1528 (wide_int_cc_tests): Call test_round_for_mask.
1529 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
1530 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
1531 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
1532 range returned by get_range_info.
1534 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1537 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
1538 * symtab.c: Include builtins.h
1539 (symtab_node::output_to_lto_symbol_table_p): Move here
1540 from lto-streamer-out.c:output_symbol_p.
1541 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
1542 (output_symbol_p): Move all logic to symtab.c
1543 (produce_symtab): Update.
1545 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1547 * config/s390/s390-opts.h (enum indirect_branch): Define.
1548 * config/s390/s390-protos.h (s390_return_addr_from_memory)
1549 (s390_indirect_branch_via_thunk)
1550 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
1551 (enum s390_indirect_branch_type): Define.
1552 * config/s390/s390.c (struct s390_frame_layout, struct
1553 machine_function): Remove.
1554 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
1555 (indirect_branch_table_label_no, indirect_branch_table_name):
1557 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
1558 (enum s390_indirect_branch_option): Define.
1559 (s390_return_addr_from_memory): New function.
1560 (s390_handle_string_attribute): New function.
1561 (s390_attribute_table): Add new attribute handler.
1562 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
1563 (s390_indirect_branch_via_thunk): New function.
1564 (s390_indirect_branch_via_inline_thunk): New function.
1565 (s390_function_ok_for_sibcall): When jumping via thunk disallow
1566 sibling call optimization for non z10 compiles.
1567 (s390_emit_call): Force indirect branch target to be a single
1568 register. Add r1 clobber for non-z10 compiles.
1569 (s390_emit_epilogue): Emit return jump via return_use expander.
1570 (s390_reorg): Handle JUMP_INSNs as execute targets.
1571 (s390_option_override_internal): Perform validity checks for the
1572 new command line options.
1573 (s390_indirect_branch_attrvalue): New function.
1574 (s390_indirect_branch_settings): New function.
1575 (s390_set_current_function): Invoke s390_indirect_branch_settings.
1576 (s390_output_indirect_thunk_function): New function.
1577 (s390_code_end): Implement target hook.
1578 (s390_case_values_threshold): Implement target hook.
1579 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
1581 * config/s390/s390.h (struct s390_frame_layout)
1582 (struct machine_function): Move here from s390.c.
1583 (TARGET_INDIRECT_BRANCH_NOBP_RET)
1584 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
1585 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
1586 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
1587 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
1588 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
1589 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
1590 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
1591 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
1592 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
1593 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
1594 (mnemonic attribute): Add values which aren't recognized
1596 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
1597 pattern for branch conversion. Fix mnemonic attribute.
1598 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
1599 indirect branch via thunk if requested.
1600 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
1601 ("*indirect_jump"): Disable for branch conversion using out of
1603 ("indirect_jump_via_thunk<mode>_z10")
1604 ("indirect_jump_via_thunk<mode>")
1605 ("indirect_jump_via_inlinethunk<mode>_z10")
1606 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
1607 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
1608 ("casesi_jump_via_inlinethunk<mode>_z10")
1609 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
1610 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
1611 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
1612 ("*indirect2_jump"): Disable for branch conversion.
1613 ("casesi_jump"): Turn into expander and expand patterns for branch
1615 ("return_use"): New expander.
1616 ("*return"): Emit return via thunk and rename it to ...
1617 ("*return<mode>"): ... this one.
1618 * config/s390/s390.opt: Add new options and and enum for the
1621 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1623 * lra-constraints.c (match_reload): Unconditionally use
1624 gen_lowpart_SUBREG, rather than selecting between that
1625 and equivalent gen_rtx_SUBREG code.
1627 2018-02-08 Richard Biener <rguenther@suse.de>
1629 PR tree-optimization/84233
1630 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
1631 changed flag instead of boguously re-using phi_inserted.
1633 2018-02-08 Martin Jambor <mjambor@suse.cz>
1635 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
1636 static local variables.
1638 2018-02-08 Richard Biener <rguenther@suse.de>
1640 PR tree-optimization/84278
1641 * tree-vect-stmts.c (vectorizable_store): When looking for
1642 smaller vector types to perform grouped strided loads/stores
1643 make sure the mode is supported by the target.
1644 (vectorizable_load): Likewise.
1646 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1648 * config/aarch64/aarch64.c (aarch64_components_for_bb):
1649 Increase LDP/STP opportunities by adding adjacent callee-saves.
1651 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1653 PR rtl-optimization/84068
1654 PR rtl-optimization/83459
1655 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
1657 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
1659 PR tree-optimization/84224
1660 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
1661 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
1664 2018-02-07 Iain Sandoe <iain@codesourcery.com>
1667 * config/rs6000/altivec.md (*restore_world): Remove LR use.
1668 * config/rs6000/predicates.md (restore_world_operation): Adjust op
1669 count, remove one USE.
1671 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
1673 * doc/install.texi (Configuration): Document the
1674 --with-long-double-format={ibm,ieee} PowerPC configuration
1678 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
1679 Convert from define_expand to be define_insn_and_split. Rework
1680 float/double/_Float128 conversions to QI/HI/SImode to work with
1681 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
1682 conversions to QI/HImode types did a store and then a load to
1683 truncate the value. For conversions to VSX registers, don't split
1684 the insn, instead emit the code directly. Use the code iterator
1685 any_fix to combine signed and unsigned conversions.
1686 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
1687 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
1688 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
1689 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
1690 (fix_<mode>di2_hw): Likewise.
1691 (fixuns_<mode>di2_hw): Likewise.
1692 (fix_<mode>si2_hw): Likewise.
1693 (fixuns_<mode>si2_hw): Likewise.
1694 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
1695 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
1696 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
1697 fix<uns>_trunc<SFDF:mode>si2_p8.
1698 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
1700 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
1701 (fix<uns>_<mode>_mem): Likewise.
1702 (fctiw<u>z_<mode>_mem): Likewise.
1703 (fix<uns>_<mode>_mem): Likewise.
1704 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
1705 the register allocator from doing a direct move to the GPRs to do
1706 a store, and instead use the ISA 3.0 store byte/half-word from
1707 vector register instruction. For IEEE 128-bit floating point,
1708 also optimize stores of 32-bit ints.
1709 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
1711 2018-02-07 Alan Hayward <alan.hayward@arm.com>
1713 * genextract.c (push_pathstr_operand): New function to support
1715 (walk_rtx): Call push_pathstr_operand.
1716 (print_path): Support [a-zA-Z].
1718 2018-02-07 Richard Biener <rguenther@suse.de>
1720 PR tree-optimization/84037
1721 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
1722 (cse_and_gimplify_to_preheader): Declare.
1723 (vect_get_place_in_interleaving_chain): Likewise.
1724 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
1726 (_loop_vec_info::~_loop_vec_info): Delete it.
1727 (cse_and_gimplify_to_preheader): New function.
1728 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
1729 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
1730 (vectorizable_load): Likewise. For grouped stores always base
1731 the IV on the first element.
1732 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
1733 condition before gimplifying.
1735 2018-02-07 Jakub Jelinek <jakub@redhat.com>
1737 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
1738 *DIV_EXPR and *MOD_EXPR.
1740 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
1743 * config/i386/i386.c (ix86_option_override_internal): Mask out
1744 the CF_SET bit when checking -fcf-protection.
1746 2018-02-07 Tom de Vries <tom@codesourcery.com>
1749 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
1752 2018-02-07 Richard Biener <rguenther@suse.de>
1754 PR tree-optimization/84204
1755 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
1758 PR tree-optimization/84205
1759 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
1760 special-case isl_ast_op_zdiv_r.
1762 PR tree-optimization/84223
1763 * graphite-scop-detection.c (gather_bbs::before_dom_children):
1764 Only add conditions from within the region.
1765 (gather_bbs::after_dom_children): Adjust.
1767 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
1770 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
1771 * config/avr/avr.md: Only post-reload split REG-REG moves if
1772 either register is REGERAL_REG_P.
1774 2018-02-07 Jakub Jelinek <jakub@redhat.com>
1776 PR tree-optimization/84235
1777 * tree-ssa-scopedtables.c
1778 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
1779 if the subtraction is performed in floating point type where NaNs are
1780 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
1781 build 1. Formatting fix.
1783 2018-02-06 Jakub Jelinek <jakub@redhat.com>
1786 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
1787 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
1788 and skip it regardless of bb boundaries. Use CALL_P macro,
1789 don't test INSN_P (insn) together with CALL_P or JUMP_P check
1790 unnecessarily, formatting fix.
1792 2018-02-06 Michael Collison <michael.collison@arm.com>
1794 * config/arm/thumb2.md:
1795 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
1796 (*thumb_mov_notscc): Ditto.
1798 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
1801 * config/rs6000/rs6000.md (su code attribute): Use "u" for
1802 unsigned_fix, not "s".
1804 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1806 * configure.ac (gcc_fn_eh_frame_ro): New function.
1807 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
1808 correct .eh_frame permissions.
1809 * configure: Regenerate.
1811 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
1813 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
1816 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1818 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1819 Display warning message for -mno-speculate-indirect-jumps.
1821 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
1823 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
1825 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
1827 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
1829 PR tree-optimization/84225
1830 * tree-eh.c (find_trapping_overflow): Only call
1831 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
1833 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1836 * config/i386/i386.c: Reimplement the check of possible options
1837 -mibt/-mshstk conbination. Change error messages.
1838 * doc/invoke.texi: Fix a typo: remove extra '='.
1840 2018-02-06 Marek Polacek <polacek@redhat.com>
1842 PR tree-optimization/84228
1843 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
1845 2018-02-06 Tamar Christina <tamar.christina@arm.com>
1848 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
1849 emitted arch directives.
1850 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
1851 __ARM_FEATURE_COPROC before changing architectures.
1853 2018-02-06 Richard Biener <rguenther@suse.de>
1855 * config/i386/i386.c (print_reg): Fix typo.
1856 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
1858 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
1860 * configure: Regenerate.
1862 2018-02-05 Martin Sebor <msebor@redhat.com>
1864 PR tree-optimization/83369
1865 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
1868 2018-02-05 Martin Liska <mliska@suse.cz>
1870 * doc/invoke.texi: Cherry-pick upstream r323995.
1872 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
1874 * ira.c (ira_init_register_move_cost): Adjust comment.
1876 2018-02-05 Martin Liska <mliska@suse.cz>
1878 PR gcov-profile/84137
1879 * doc/gcov.texi: Fix typo in documentation.
1881 2018-02-05 Martin Liska <mliska@suse.cz>
1883 PR gcov-profile/83879
1884 * doc/gcov.texi: Document necessity of --dynamic-list-data when
1885 using dlopen functionality.
1887 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
1889 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
1890 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
1891 _mm_maskz_range_ss, _mm_mask_range_round_ss,
1892 _mm_maskz_range_round_ss): New intrinsics.
1893 (__builtin_ia32_rangesd128_round)
1894 (__builtin_ia32_rangess128_round): Remove.
1895 (__builtin_ia32_rangesd128_mask_round,
1896 __builtin_ia32_rangess128_mask_round): New builtins.
1897 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
1898 __builtin_ia32_rangess128_round): Remove.
1899 (__builtin_ia32_rangesd128_mask_round,
1900 __builtin_ia32_rangess128_mask_round): New builtins.
1901 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
1902 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
1903 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
1904 "<round_saeonly_constraint>")): Changed to ...
1905 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
1906 "<round_saeonly_scalar_constraint>")): ... this.
1907 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
1908 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
1909 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
1910 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
1911 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
1913 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
1915 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
1917 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
1918 Remove all values except native, 8540 and 8548.
1920 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
1922 * config/i386/i386.c (ix86_output_function_return): Pass
1923 INVALID_REGNUM, instead of -1, as invalid register number to
1924 indirect_thunk_name and output_indirect_thunk.
1926 2018-02-02 Julia Koval <julia.koval@intel.com>
1928 * config.gcc: Add -march=icelake.
1929 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
1930 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
1931 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
1932 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
1933 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
1934 (processor_target_table): Add icelake.
1935 (ix86_option_override_internal): Handle new PTAs.
1936 (get_builtin_code_for_version): Handle icelake.
1937 (M_INTEL_COREI7_ICELAKE): New.
1938 (fold_builtin_cpu): Handle icelake.
1939 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
1940 * doc/invoke.texi: Add -march=icelake.
1942 2018-02-02 Julia Koval <julia.koval@intel.com>
1944 * config/i386/i386.c (ix86_option_override_internal): Change flags type
1945 to wide_int_bitmask.
1946 * wide-int-bitmask.h: New.
1948 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1951 * config/i386/i386.md: Replace Pmode with word_mode in
1952 builtin_setjmp_setup and builtin_longjmp to support x32.
1954 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
1958 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
1960 (rs6000_supported_cpu_names): New static variable.
1961 (linux_cpu_translation_table): Likewise.
1962 (elf_platform) <cpu>: Define new static variable and use it.
1963 Translate kernel AT_PLATFORM name to canonical name if needed.
1964 Error if platform name is unknown.
1966 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
1969 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
1971 2018-02-01 Jeff Law <law@redhat.com>
1974 * config/i386/i386.c (release_scratch_register_on_entry): Add new
1975 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
1976 the scratch if RELEASE_VIA_POP is false.
1977 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
1978 If we have to save a temporary register, decrement SIZE appropriately.
1979 Pass new arguments to release_scratch_register_on_entry.
1980 (ix86_adjust_stack_and_probe): Likewise.
1981 (ix86_emit_probe_stack_range): Pass new arguments to
1982 release_scratch_register_on_entry.
1984 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
1986 PR rtl-optimization/84157
1987 * combine.c (change_zero_ext): Use REG_P predicate in
1988 front of HARD_REGISTER_P predicate.
1990 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
1992 * config/avr/avr.c (avr_option_override): Move disabling of
1993 -fdelete-null-pointer-checks to...
1994 * common/config/avr/avr-common.c (avr_option_optimization_table):
1997 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1999 PR tree-optimization/81635
2000 * tree-data-ref.c (split_constant_offset_1): For types that
2001 wrap on overflow, try to use range info to prove that wrapping
2004 2018-02-01 Renlin Li <renlin.li@arm.com>
2007 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
2009 (aarch64_register_move_cost): Likewise.
2010 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
2012 (REG_CLASS_NAMES): Likewise.
2013 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
2014 TAILCALL_ADDR_REGS. Remove IP registers.
2015 * config/aarch64/aarch64.md (Ucs): Update register constraint.
2017 2018-02-01 Richard Biener <rguenther@suse.de>
2019 * domwalk.h (dom_walker::dom_walker): Add additional constructor
2020 for specifying RPO order and allow NULL for that.
2021 * domwalk.c (dom_walker::dom_walker): Likewise.
2022 (dom_walker::walk): Handle NULL RPO order.
2023 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
2025 (rewrite_update_dom_walker): Likewise.
2026 (mark_def_dom_walker): Likewise.
2028 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2030 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
2031 (aarch64_maybe_expand_sve_subreg_move): Declare.
2032 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
2033 * config/aarch64/predicates.md (aarch64_any_register_operand): New
2035 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
2036 that are semantically a reverse operation.
2037 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
2038 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
2039 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
2041 (aarch64_can_change_mode_class): For big-endian, forbid changes
2042 between two SVE modes if they have different element sizes.
2044 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2046 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
2047 the TImode handling for big-endian targets.
2049 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2051 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
2052 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
2054 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
2055 Remove BSWAP handing for big-endian targets and use the form of
2056 LD1RQ appropariate for the mode.
2058 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2060 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
2061 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
2064 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2067 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
2068 check for operands that need to go through aarch64_sve_reload_be.
2070 2018-02-01 Jakub Jelinek <jakub@redhat.com>
2072 PR tree-optimization/81661
2073 PR tree-optimization/84117
2074 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
2075 * tree-eh.c: Include gimplify.h.
2076 (find_trapping_overflow, replace_trapping_overflow,
2077 rewrite_to_non_trapping_overflow): New functions.
2078 * tree-vect-loop.c: Include tree-eh.h.
2079 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
2080 * tree-data-ref.c: Include tree-eh.h.
2081 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
2083 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
2085 PR rtl-optimization/84123
2086 * combine.c (change_zero_ext): Check if hard register satisfies
2087 can_change_dest_mode before calling gen_lowpart_SUBREG.
2089 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
2092 * ira.c (ira_init_register_move_cost): Remove assert.
2094 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2096 PR rtl-optimization/84071
2097 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
2098 * doc/tm.texi: Regenerate.
2100 2018-01-31 Richard Biener <rguenther@suse.de>
2102 PR tree-optimization/84132
2103 * tree-data-ref.c (analyze_miv_subscript): Properly
2104 check whether evolution_function_is_affine_multivariate_p
2105 before calling gcd_of_steps_may_divide_p.
2107 2018-01-31 Julia Koval <julia.koval@intel.com>
2110 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
2111 * config/i386/i386.md (rdpid_rex64) New.
2112 (rdpid): Make 32bit only.
2114 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
2117 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
2118 an IDENTIFIER_NODE for FUNCTION_TYPE's.
2120 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2123 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
2125 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
2127 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2129 PR rtl-optimization/84071
2130 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
2131 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
2133 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2135 * config/arc/arc.c (arc_handle_aux_attribute): New function.
2136 (arc_attribute_table): Add 'aux' attribute.
2137 (arc_in_small_data_p): Consider aux like variables.
2138 (arc_is_aux_reg_p): New function.
2139 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
2140 (arc_get_aux_arg): New function.
2141 (prepare_move_operands): Handle aux-register access.
2142 (arc_handle_aux_attribute): New function.
2143 * doc/extend.texi (ARC Variable attributes): Add subsection.
2145 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2147 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
2148 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
2149 (arc_attribute_table): Add 'uncached' attribute.
2150 (arc_print_operand): Print '.di' flag for uncached memory
2152 (arc_in_small_data_p): Do not consider for small data the uncached
2154 (arc_is_uncached_mem_p): New function.
2155 * config/arc/predicates.md (compact_store_memory_operand): Check
2156 for uncached memory accesses.
2157 (nonvol_nonimm_operand): Likewise.
2158 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
2160 2018-01-31 Jakub Jelinek <jakub@redhat.com>
2163 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
2164 falign-loops=): Add Optimization flag.
2166 2018-01-30 Jeff Law <law@redhat.com>
2169 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
2170 INT_REGISTERS_SAVED. Check it prior to calling
2171 get_scratch_register_on_entry.
2172 (ix86_adjust_stack_and_probe): Similarly.
2173 (ix86_emit_probe_stack_range): Similarly.
2174 (ix86_expand_prologue): Corresponding changes.
2176 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2179 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
2180 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
2182 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
2185 * lra-constraints.c (curr_insn_transform): Process AND in the
2188 2018-01-30 Jakub Jelinek <jakub@redhat.com>
2190 PR rtl-optimization/83986
2191 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
2192 dependence against last_pending_memory_flush in addition to
2195 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
2197 PR tree-optimization/81611
2198 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
2201 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2204 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
2207 2018-01-30 Richard Biener <rguenther@suse.de>
2208 Jakub Jelinek <jakub@redhat.com>
2210 PR tree-optimization/84111
2211 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
2212 inner loops added during recursion, as they don't have up-to-date
2215 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2218 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
2219 (can_inline_edge_by_limits_p): ... here.
2220 (can_early_inline_edge_p, check_callers,
2221 update_caller_keys, update_callee_keys, recursive_inlining,
2222 add_new_edges_to_heap, speculation_useful_p,
2223 inline_small_functions,
2224 inline_small_functions, flatten_function,
2225 inline_to_all_callers_1): Update.
2227 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2229 * profile-count.c (profile_count::combine_with_ipa_count): Handle
2232 2018-01-30 Richard Biener <rguenther@suse.de>
2234 PR tree-optimization/83008
2235 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
2236 invariant and constant vector uses in stmts when they need
2239 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2242 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
2243 * configure: Regenerate.
2245 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2247 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
2249 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
2250 Use gen_rtx_REG rather than gen_lowpart.
2252 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2254 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
2255 rather than 0 when creating partial subregs.
2257 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2259 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
2262 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
2265 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
2266 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
2267 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
2268 flags. This restores the settings used before the 2017-07-24.
2269 Turning off pre increment/decrement/modify allows IVOPTS to
2270 optimize DF/SF loops where the index is an int.
2272 2018-01-29 Richard Biener <rguenther@suse.de>
2273 Kelvin Nilsen <kelvin@gcc.gnu.org>
2276 * tree-vect-stmts.c (vectorizable_call): Don't call
2277 targetm.vectorize_builtin_md_vectorized_function if callee is
2280 2018-01-22 Carl Love <cel@us.ibm.com>
2282 * doc/extend.tex: Fix typo in second arg in
2283 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
2285 2018-01-29 Richard Biener <rguenther@suse.de>
2287 PR tree-optimization/84086
2288 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
2289 (flush_ssaname_freelist): When SSA names were released reset
2290 the SCEV hash table.
2292 2018-01-29 Richard Biener <rguenther@suse.de>
2294 PR tree-optimization/84057
2295 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
2296 removed paths when removing edges.
2298 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
2300 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
2301 -mfunction-return=@var{choice}.
2303 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
2306 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
2308 (layout::print_source_line): Likewise.
2309 (test_get_line_width_without_trailing_whitespace): Add test cases.
2311 2018-01-27 Jakub Jelinek <jakub@redhat.com>
2314 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
2317 2018-01-26 Jim Wilson <jimw@sifive.com>
2319 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
2321 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
2324 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2326 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
2327 and CMP + SUB-immediate -> SUBS.
2329 2018-01-26 Martin Sebor <msebor@redhat.com>
2331 PR tree-optimization/83896
2332 * tree-ssa-strlen.c (get_string_len): Rename...
2333 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
2334 Avoid assuming length is constant.
2335 (handle_char_store): Use HOST_WIDE_INT for string length.
2337 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
2340 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
2341 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
2343 2018-01-26 Richard Biener <rguenther@suse.de>
2345 PR rtl-optimization/84003
2346 * dse.c (record_store): Only record redundant stores when
2347 the earlier store aliases at least all accesses the later one does.
2349 2018-01-26 Jakub Jelinek <jakub@redhat.com>
2351 PR rtl-optimization/83985
2352 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
2353 REG_CFA_RESTORE insns.
2354 (delete_unmarked_insns): Don't ignore separate shrink wrapping
2355 REG_CFA_RESTORE insns here.
2358 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
2359 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
2361 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2363 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
2364 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
2365 (arc_init): Likewise.
2366 (arc_override_options): Likewise.
2367 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
2369 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
2371 * config/arc/arc.h (TARGET_DBNZ): Define.
2372 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
2373 properly set the tune attribute.
2374 (dbnz): Use TARGET_DBNZ guard.
2375 * config/arc/arc.opt (mtune): Add core3 option.
2377 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2379 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
2380 recognize new pic like addresses.
2381 (arc_delegitimize_address): Clean up.
2383 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2385 * config/arc/arc-arches.def: Option mrf16 valid for all
2387 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
2388 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
2389 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
2390 * config/arc/arc-tables.opt: Regenerate.
2391 * config/arc/arc.c (arc_conditional_register_usage): Handle
2392 reduced register file case.
2393 (arc_file_start): Set must have build attributes.
2394 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
2396 * config/arc/arc.opt (mrf16): Add new option.
2397 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
2398 * config/arc/genmultilib.awk: Handle new mrf16 option.
2399 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
2400 * config/arc/t-multilib: Regenerate.
2401 * doc/invoke.texi (ARC Options): Document mrf16 option.
2403 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2405 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
2406 * config/arc/arc.c (arc_handle_secure_attribute): New function.
2407 (arc_attribute_table): Add 'secure_call' attribute.
2408 (arc_print_operand): Print secure call operand.
2409 (arc_function_ok_for_sibcall): Don't optimize tail calls when
2411 (arc_is_secure_call_p): New function. * config/arc/arc.md
2412 (call_i): Add support for sjli instruction.
2413 (call_value_i): Likewise.
2414 * config/arc/constraints.md (Csc): New constraint.
2416 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2417 John Eric Martin <John.Martin@emmicro-us.com>
2419 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
2420 * config/arc/arc.c (_arc_jli_section): New struct.
2421 (arc_jli_section): New type.
2422 (rc_jli_sections): New static variable.
2423 (arc_handle_jli_attribute): New function.
2424 (arc_attribute_table): Add jli_always and jli_fixed attribute.
2425 (arc_file_end): New function.
2426 (TARGET_ASM_FILE_END): Define.
2427 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
2428 (arc_add_jli_section): New function.
2429 (jli_call_scan): Likewise.
2430 (arc_reorg): Call jli_call_scan.
2431 (arc_output_addsi): Remove 'S' from printing asm operand.
2432 (arc_is_jli_call_p): New function.
2433 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
2435 (movhi_insn): Likewise.
2436 (movsi_insn): Likewise.
2437 (movsi_set_cc_insn): Likewise.
2438 (loadqi_update): Likewise.
2439 (load_zeroextendqisi_update): Likewise.
2440 (load_signextendqisi_update): Likewise.
2441 (loadhi_update): Likewise.
2442 (load_zeroextendhisi_update): Likewise.
2443 (load_signextendhisi_update): Likewise.
2444 (loadsi_update): Likewise.
2445 (loadsf_update): Likewise.
2446 (movsicc_insn): Likewise.
2447 (bset_insn): Likewise.
2448 (bxor_insn): Likewise.
2449 (bclr_insn): Likewise.
2450 (bmsk_insn): Likewise.
2451 (bicsi3_insn): Likewise.
2452 (cmpsi_cc_c_insn): Likewise.
2453 (movsi_ne): Likewise.
2454 (movsi_cond_exec): Likewise.
2455 (clrsbsi2): Likewise.
2466 (call_i): Remove 'S' asm letter, add jli instruction.
2467 (call_value_i): Likewise.
2468 * config/arc/arc.op (mjli-always): New option.
2469 * config/arc/constraints.md (Cji): New constraint.
2470 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
2472 (subsf3_fpx): Likewise.
2473 (mulsf3_fpx): Likewise.
2474 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
2476 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
2478 * doc/invoke.texi (ARC): Document mjli-always option.
2480 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
2482 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
2483 avoid addition with 0 and use incw and decw where possible.
2485 2018-01-26 Richard Biener <rguenther@suse.de>
2487 PR tree-optimization/81082
2488 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
2489 association if it requires casting to unsigned.
2490 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
2491 from fold_plusminus_mult_expr to catch important cases late when
2492 range info is available.
2494 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2496 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
2497 * configure.ac (hidden_linkonce): New test.
2498 * configure: Regenerate.
2499 * config.in: Regenerate.
2501 2018-01-26 Julia Koval <julia.koval@intel.com>
2503 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
2504 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
2505 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
2506 _mm_mask_bitshuffle_epi64_mask): Fix type.
2507 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
2508 USI_FTYPE_V4DI_V4DI_USI): Remove.
2509 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
2510 __builtin_ia32_vpshufbitqmb256_mask,
2511 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
2512 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
2513 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
2515 2018-01-26 Alan Modra <amodra@gmail.com>
2518 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
2519 UNSPEC_VBPERMQ. Sort other unspecs.
2521 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
2523 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
2525 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
2528 * predict.c (drop_profile): Do not push/pop cfun; update also
2530 (handle_missing_profiles): Fix logic looking for zero profiles.
2532 2018-01-25 Jakub Jelinek <jakub@redhat.com>
2535 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
2536 on functions with #pragma omp declare simd or functions with simd
2538 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
2539 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
2540 Remove trailing \n from warning_at calls.
2542 2018-01-25 Tom de Vries <tom@codesourcery.com>
2545 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2546 for neutered workers.
2548 2018-01-24 Joseph Myers <joseph@codesourcery.com>
2551 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
2552 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
2554 2018-01-24 Jeff Law <law@redhat.com>
2557 * i386.c (get_probe_interval): Move to earlier point.
2558 (ix86_compute_frame_layout): If -fstack-clash-protection and
2559 the frame is larger than the probe interval, then use pushes
2560 to save registers rather than reg->mem moves.
2561 (ix86_expand_prologue): Remove conditional for int_registers_saved
2564 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
2567 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
2568 min/max for never referenced object.
2570 2018-01-24 Jakub Jelinek <jakub@redhat.com>
2573 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
2575 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
2576 attributes from DECL_ATTRIBUTES (decl) without affecting
2577 DECL_ATTRIBUTES (current_function_decl).
2578 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
2579 functions with non-NULL DECL_ABSTRACT_ORIGIN.
2581 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
2583 PR tree-optimization/83979
2584 * fold-const.c (fold_comparison): Use constant_boolean_node
2585 instead of boolean_{true,false}_node.
2587 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
2589 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
2592 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2594 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
2595 Simplify the clause that sets the length attribute.
2596 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
2597 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
2598 clause that sets the length attribute.
2599 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
2601 2018-01-24 Tom de Vries <tom@codesourcery.com>
2604 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
2605 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
2606 Add strict parameter.
2607 (prevent_branch_around_nothing): Insert dummy insn between branch to
2608 label and label with no ptx insn inbetween.
2609 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
2611 2018-01-24 Tom de Vries <tom@codesourcery.com>
2614 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2615 for neutered threads in warp.
2616 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
2618 2018-01-24 Richard Biener <rguenther@suse.de>
2620 PR tree-optimization/83176
2621 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
2624 2018-01-24 Richard Biener <rguenther@suse.de>
2626 PR tree-optimization/82819
2627 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
2628 code generating pluses that are no-ops in the target precision.
2630 2018-01-24 Richard Biener <rguenther@suse.de>
2633 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
2635 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2637 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
2638 to merge probabilities.
2639 * predict.c (probably_never_executed): Also mark as cold functions
2640 with global 0 profile and guessed local profile.
2641 * profile-count.c (profile_probability::combine_with_count): New
2643 * profile-count.h (profile_probability::operator*,
2644 profile_probability::operator*=, profile_probability::operator/,
2645 profile_probability::operator/=): Reduce precision to adjusted
2646 and set value to guessed on contradictory divisions.
2647 (profile_probability::combine_with_freq): Remove.
2648 (profile_probability::combine_wiht_count): Declare.
2649 (profile_count::force_nonzero):: Set to adjusted.
2650 (profile_count::probability_in):: Set quality to adjusted.
2651 * tree-ssa-tail-merge.c (replace_block_by): Use
2654 2018-01-23 Andrew Waterman <andrew@sifive.com>
2655 Jim Wilson <jimw@sifive.com>
2657 * config/riscv/riscv.c (riscv_stack_boundary): New.
2658 (riscv_option_override): Set riscv_stack_boundary. Handle
2659 riscv_preferred_stack_boundary_arg.
2660 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
2661 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
2662 (STACK_BOUNDARY): Set to riscv_stack_boundary.
2663 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
2664 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
2665 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
2667 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
2670 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
2671 of struct ix86_frame.
2672 (ix86_expand_epilogue): Likewise. Add a local variable for
2673 the reg_save_offset field in struct ix86_frame.
2675 2018-01-23 Bin Cheng <bin.cheng@arm.com>
2677 PR tree-optimization/82604
2678 * tree-loop-distribution.c (enum partition_kind): New enum item
2679 PKIND_PARTIAL_MEMSET.
2680 (partition_builtin_p): Support above new enum item.
2681 (generate_code_for_partition): Ditto.
2682 (compute_access_range): Differentiate cases that equality can be
2683 proven at all loops, the innermost loops or no loops.
2684 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
2685 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
2686 (finalize_partitions, distribute_loop): Don't fuse partition of
2687 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
2688 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
2691 2018-01-23 Martin Liska <mliska@suse.cz>
2693 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
2694 order to ignore the predictor.
2695 (PRED_POLYMORPHIC_CALL): Likewise.
2696 (PRED_RECURSIVE_CALL): Likewise.
2698 2018-01-23 Martin Liska <mliska@suse.cz>
2700 * tree-profile.c (tree_profiling): Print function header to
2701 aware reader which function we are working on.
2702 * value-prof.c (gimple_find_values_to_profile): Do not print
2703 not interesting value histograms.
2705 2018-01-23 Martin Liska <mliska@suse.cz>
2707 * profile-count.h (enum profile_quality): Add
2708 profile_uninitialized as the first value. Do not number values
2709 as they are zero based.
2710 (profile_count::verify): Update sanity check.
2711 (profile_probability::verify): Likewise.
2713 2018-01-23 Nathan Sidwell <nathan@acm.org>
2715 * doc/invoke.texi (ffor-scope): Deprecate.
2717 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2719 PR tree-optimization/83510
2720 * domwalk.c (set_all_edges_as_executable): New function.
2721 (dom_walker::dom_walker): Convert bool param
2722 "skip_unreachable_blocks" to enum reachability. Move setup of
2723 edge flags to set_all_edges_as_executable and only do it when
2724 reachability is REACHABLE_BLOCKS.
2725 * domwalk.h (enum dom_walker::reachability): New enum.
2726 (dom_walker::dom_walker): Convert bool param
2727 "skip_unreachable_blocks" to enum reachability.
2728 (set_all_edges_as_executable): New decl.
2729 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
2730 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
2732 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
2733 but converting true to REACHABLE_BLOCKS.
2734 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
2736 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
2737 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
2738 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
2740 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
2741 if check_all_array_refs will be called.
2743 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2745 * tree.c (selftest::test_location_wrappers): Add more test
2748 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2750 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
2751 (selftest::test_bit_in_range): Likewise.
2753 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
2756 * doc/sourcebuild.texi (vect_float): Say that the selector
2757 only describes the situation when -funsafe-math-optimizations is on.
2758 (vect_float_strict): Document.
2760 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
2762 PR tree-optimization/83965
2763 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
2764 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
2765 instead of checking only for a reduction.
2766 (vect_recog_widen_sum_pattern): Likewise.
2768 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2770 * predict.c (probably_never_executed): Only use precise profile info.
2771 (compute_function_frequency): Skip after inlining hack since we now
2772 have quality checking.
2774 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2776 * profile-count.h (profile_probability::very_unlikely,
2777 profile_probability::unlikely, profile_probability::even): Set
2778 precision to guessed.
2780 2018-01-23 Richard Biener <rguenther@suse.de>
2782 PR tree-optimization/83963
2783 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
2784 Properly terminate dominator walk when crossing the exit edge not
2785 when visiting its source block.
2787 2018-01-23 Jakub Jelinek <jakub@redhat.com>
2790 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
2791 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
2793 2018-01-22 Jakub Jelinek <jakub@redhat.com>
2795 PR tree-optimization/83957
2796 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
2797 semicolon after for body surrounded by braces.
2799 PR tree-optimization/83081
2800 * profile-count.h (profile_probability::split): New method.
2801 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
2802 Use profile_probability::split.
2803 (do_compare_rtx_and_jump): Fix adjustment of probabilities
2804 when splitting a single conditional jump into 2.
2806 2018-01-22 David Malcolm <dmalcolm@redhat.com>
2808 PR tree-optimization/69452
2809 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
2812 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2814 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
2815 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
2816 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
2818 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2820 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
2821 * config/rl78/rl78.md (movdi): New define_expand.
2822 * config/rl78/rl78.c (rl78_split_movdi): New function.
2824 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
2827 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
2829 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
2830 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
2831 128-bit to produce an UNSPEC move to get the double word with the
2832 signbit and then a shift directly to do signbit.
2833 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
2834 implementation with a new version that just does either a direct
2835 move or a regular move. Move memory interface to separate insns.
2836 Move insns so they are next to the expander.
2837 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
2838 with signbit move. Split big and little endian case.
2839 (signbit<mode>2_dm_mem_le): Likewise.
2840 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
2841 (signbit<mode>2_dm2): Likewise.
2843 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2845 * config/rl78/rl78.md (anddi3): New define_expand.
2847 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2849 * config/rl78/rl78.md (umindi3): New define_expand.
2851 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2853 * config/rl78/rl78.md (smindi3): New define_expand.
2855 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2857 * config/rl78/rl78.md (smaxdi3): New define_expand.
2859 2018-01-22 Carl Love <cel@us.ibm.com>
2861 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
2862 LVX_V1TI): Add macro expansion.
2863 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
2864 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
2865 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
2866 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
2867 Change check to determine if the instruction is a byte reversing
2868 entry. Fix typo in comment.
2869 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
2870 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
2871 Add def_builtin calls for new builtins.
2872 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
2873 Add define_insn expansion.
2875 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2877 * config/rl78/rl78.md (umaxdi3): New define_expand.
2879 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2881 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
2882 for non-QImode registers.
2884 2018-01-22 Richard Biener <rguenther@suse.de>
2886 PR tree-optimization/83963
2887 * graphite-scop-detection.c (scop_detection::get_sese): Delay
2888 including the loop exit block.
2889 (scop_detection::merge_sese): Likewise.
2890 (scop_detection::add_scop): Do it here instead.
2892 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2894 * doc/sourcebuild.texi (arm_softfloat): Document.
2896 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
2899 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
2900 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
2901 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
2903 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2904 David Edelsohn <dje.gcc@gmail.com>
2907 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
2908 Change "crset eq" to "crset 2".
2909 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
2910 (*call_indirect_aix<mode>_nospec): Likewise.
2911 (*call_value_indirect_aix<mode>_nospec): Likewise.
2912 (*call_indirect_elfv2<mode>_nospec): Likewise.
2913 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
2914 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
2915 change assembly output from . to $.
2916 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
2917 (indirect_jump<mode>_nospec): Change assembly output from . to $.
2918 (*tablejump<mode>_internal1_nospec): Likewise.
2920 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
2923 * config/sh/sh_optimize_sett_clrt.cc:
2924 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
2926 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
2928 PR tree-optimization/83940
2929 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
2930 offset_dt to vect_constant_def rather than vect_unknown_def_type.
2931 (vect_check_load_store_mask): Add a mask_dt_out parameter and
2932 use it to pass back the definition type.
2933 (vect_check_store_rhs): Likewise rhs_dt_out.
2934 (vect_build_gather_load_calls): Add a mask_dt argument and use
2935 it instead of a call to vect_is_simple_use.
2936 (vectorizable_store): Update calls to vect_check_load_store_mask
2937 and vect_check_store_rhs. Use the dt returned by the latter instead
2938 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
2939 instead of calls to vect_is_simple_use. Pass the scalar rather
2940 than the vector operand to vect_is_simple_use when handling
2941 second and subsequent copies of an rhs value.
2942 (vectorizable_load): Update calls to vect_check_load_store_mask
2943 and vect_build_gather_load_calls. Use the cached mask_dt and
2944 gs_info.offset_dt instead of calls to vect_is_simple_use.
2946 2018-01-20 Jakub Jelinek <jakub@redhat.com>
2949 * tree-emutls.c: Include gimplify.h.
2950 (lower_emutls_2): New function.
2951 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
2952 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
2953 it before further processing.
2956 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
2957 UINTVAL (trueop1) instead of INTVAL (op1).
2959 2018-01-19 Jakub Jelinek <jakub@redhat.com>
2963 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
2964 INCOMING_FRAME_SP_OFFSET if not defined.
2965 (scan_trace): Add ENTRY argument. If true and
2966 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
2967 emit a note to adjust the CFA offset.
2968 (create_cfi_notes): Adjust scan_trace callers.
2969 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
2970 INCOMING_FRAME_SP_OFFSET in the CIE.
2971 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
2972 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
2974 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
2975 * doc/tm.texi: Regenerated.
2977 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2979 PR rtl-optimization/83147
2980 * lra-constraints.c (remove_inheritance_pseudos): Use
2981 lra_substitute_pseudo_within_insn.
2983 2018-01-19 Tom de Vries <tom@codesourcery.com>
2984 Cesar Philippidis <cesar@codesourcery.com>
2987 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
2989 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
2992 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
2993 spaces for function labels.
2995 2018-01-19 Martin Liska <mliska@suse.cz>
2997 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
2998 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
2999 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
3000 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
3001 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
3002 (PRED_CONST_RETURN): Change from 69 to 65.
3003 (PRED_NULL_RETURN): Change from 91 to 71.
3004 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
3005 (PRED_LOOP_GUARD): Change from 66 to 73.
3007 2018-01-19 Martin Liska <mliska@suse.cz>
3009 * predict.c (predict_insn_def): Add new assert.
3010 (struct branch_predictor): Change type to signed integer.
3011 (test_prediction_value_range): Amend test to cover
3013 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
3014 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
3015 (PRED_LOOP_ITERATIONS_MAX): Likewise.
3016 (PRED_LOOP_IV_COMPARE): Likewise.
3017 * predict.h (PROB_UNINITIALIZED): Define new constant.
3019 2018-01-19 Martin Liska <mliska@suse.cz>
3021 * predict.c (dump_prediction): Add new format for
3022 analyze_brprob.py script which is enabled with -details
3024 * profile-count.h (precise_p): New function.
3026 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3028 PR tree-optimization/83922
3029 * tree-vect-loop.c (vect_verify_full_masking): Return false if
3030 there are no statements that need masking.
3031 (vect_active_double_reduction_p): New function.
3032 (vect_analyze_loop_operations): Use it when handling phis that
3033 are not in the loop header.
3035 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3037 PR tree-optimization/83914
3038 * tree-vect-loop.c (vectorizable_induction): Don't convert
3039 init_expr or apply the peeling adjustment for inductions
3040 that are nested within the vectorized loop.
3042 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3044 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
3047 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3051 * function.h (gimplify_parameters): Add gimple_seq * argument.
3052 * function.c: Include gimple.h and options.h.
3053 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
3054 for the added local temporaries if needed.
3055 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
3056 if there are any parameter cleanups, wrap whole body into a
3057 try/finally with the cleanups.
3059 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
3062 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
3063 Use GET_MODE_CLASS for scalar floating point.
3065 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3069 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3070 Fix call of call_cgraph_insertion_hooks.
3072 2018-01-18 Martin Sebor <msebor@redhat.com>
3074 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
3076 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3079 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
3082 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
3085 * common.opt: (-ffile-prefix-map): New option.
3086 * opts.c (common_handle_option): Defer it.
3087 * opts-global.c (handle_common_deferred_options): Handle it.
3088 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
3089 * file-prefix-map.h: New file.
3090 (remap_debug_filename, add_debug_prefix_map): ...here.
3091 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
3092 * final.c (debug_prefix_map, add_debug_prefix_map
3093 remap_debug_filename): Move to...
3094 * file-prefix-map.c: New file.
3095 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
3096 generalize, get rid of alloca(), use strrchr() instead of strchr().
3097 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
3098 Implement in terms of add_prefix_map().
3099 (remap_macro_filename, remap_debug_filename): Implement in term of
3101 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
3102 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
3103 * dbxout.c: Include file-prefix-map.h.
3104 * varasm.c: Likewise.
3105 * vmsdbgout.c: Likewise.
3106 * xcoffout.c: Likewise.
3107 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
3108 * doc/cppopts.texi (-fmacro-prefix-map): Document.
3109 * doc/invoke.texi (-ffile-prefix-map): Document.
3110 (-fdebug-prefix-map): Update description.
3112 2018-01-18 Martin Liska <mliska@suse.cz>
3114 * config/i386/i386.c (indirect_thunk_name): Document that also
3116 (output_indirect_thunk): Document why both instructions
3117 (pause and lfence) are generated.
3119 2018-01-18 Richard Biener <rguenther@suse.de>
3121 PR tree-optimization/83887
3122 * graphite-scop-detection.c
3123 (scop_detection::get_nearest_dom_with_single_entry): Remove.
3124 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
3125 (scop_detection::merge_sese): Re-implement with a flood-fill
3126 algorithm that properly finds a SESE region if it exists.
3128 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3131 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
3132 pointer_diff optimizations use view_convert instead of convert.
3134 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3136 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3137 Generate different code for -mno-speculate-indirect-jumps.
3138 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3139 (*call_indirect_aix<mode>): Disable for
3140 -mno-speculate-indirect-jumps.
3141 (*call_indirect_aix<mode>_nospec): New define_insn.
3142 (*call_value_indirect_aix<mode>): Disable for
3143 -mno-speculate-indirect-jumps.
3144 (*call_value_indirect_aix<mode>_nospec): New define_insn.
3145 (*sibcall_nonlocal_sysv<mode>): Generate different code for
3146 -mno-speculate-indirect-jumps.
3147 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3149 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3151 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
3152 long double type, set the flags for noting the default long double
3153 type, even if we don't pass or return a long double type.
3155 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
3158 * ipa-inline.c (flatten_function): Do not overwrite final inlining
3161 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
3163 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
3164 support for merge[hl].
3165 (fold_mergehl_helper): New helper function.
3166 (tree-vector-builder.h): New #include for tree_vector_builder usage.
3167 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
3168 (altivec_vmrglw_direct): Add xxmrglw insn.
3170 2018-01-17 Andrew Waterman <andrew@sifive.com>
3172 * config/riscv/riscv.c (riscv_conditional_register_usage): If
3173 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
3175 2018-01-17 David Malcolm <dmalcolm@redhat.com>
3178 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
3179 call the lto_location_cache before reading the
3180 DECL_SOURCE_LOCATION of the types.
3182 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3183 Richard Sandiford <richard.sandiford@linaro.org>
3185 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
3186 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
3187 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
3188 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
3189 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
3191 * config/aarch64/constraints.md (aarch64_movti_operand):
3193 * config/aarch64/predicates.md (Uti): Add new constraint.
3195 2018-01-17 Carl Love <cel@us.ibm.com>
3197 * config/rs6000/vsx.md (define_expand xl_len_r,
3198 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
3199 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
3201 (define_expand, define_insn): Move the shift left from the
3202 define_insn to the define_expand for lxvl and stxvl instructions.
3203 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
3204 and XL_LEN_R definitions to PURE.
3206 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
3208 * config/i386/i386.c (indirect_thunk_name): Declare regno
3209 as unsigned int. Compare regno with INVALID_REGNUM.
3210 (output_indirect_thunk): Ditto.
3211 (output_indirect_thunk_function): Ditto.
3212 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
3213 in the call to output_indirect_thunk_function.
3215 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
3218 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
3219 rather than the size of inner_type to determine the stack slot size
3220 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
3222 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
3225 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
3228 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
3230 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
3231 endian Linux systems to optionally enable multilibs for selecting
3232 the long double type if the user configured an explicit type.
3233 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
3234 have no long double multilibs if not defined.
3235 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
3236 warn if the user used -mabi={ieee,ibm}longdouble and we built
3237 multilibs for long double.
3238 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
3239 appropriate multilib option.
3240 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
3242 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
3243 for building long double multilibs.
3244 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
3246 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
3248 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
3251 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
3253 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
3256 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
3259 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
3262 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3264 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
3265 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
3267 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3269 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
3270 different rtl trees depending on TARGET_64BIT.
3271 (rs6000_gen_lvx): Likewise.
3273 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3275 * config/visium/visium.md (nop): Tweak comment.
3276 (hazard_nop): Likewise.
3278 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3280 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
3281 -mspeculate-indirect-jumps.
3282 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
3283 for -mno-speculate-indirect-jumps.
3284 (*call_indirect_elfv2<mode>_nospec): New define_insn.
3285 (*call_value_indirect_elfv2<mode>): Disable for
3286 -mno-speculate-indirect-jumps.
3287 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
3288 (indirect_jump): Emit different RTL for
3289 -mno-speculate-indirect-jumps.
3290 (*indirect_jump<mode>): Disable for
3291 -mno-speculate-indirect-jumps.
3292 (*indirect_jump<mode>_nospec): New define_insn.
3293 (tablejump): Emit different RTL for
3294 -mno-speculate-indirect-jumps.
3295 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
3296 (tablejumpsi_nospec): New define_expand.
3297 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
3298 (tablejumpdi_nospec): New define_expand.
3299 (*tablejump<mode>_internal1): Disable for
3300 -mno-speculate-indirect-jumps.
3301 (*tablejump<mode>_internal1_nospec): New define_insn.
3302 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
3305 2018-01-16 Artyom Skrobov tyomitch@gmail.com
3307 * caller-save.c (insert_save): Drop unnecessary parameter. All
3310 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3311 Richard Biener <rguenth@suse.de>
3314 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
3315 return early, inline manually is_gimple_sizepos. Make sure if we
3316 call gimplify_expr we don't end up with a gimple constant.
3317 * tree.c (variably_modified_type_p): Don't return true for
3318 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
3319 * gimplify.h (is_gimple_sizepos): Remove.
3321 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3323 PR tree-optimization/83857
3324 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
3325 vectorizable_live_operation for pure SLP statements.
3326 (vectorizable_live_operation): Handle PHIs.
3328 2018-01-16 Richard Biener <rguenther@suse.de>
3330 PR tree-optimization/83867
3331 * tree-vect-stmts.c (vect_transform_stmt): Precompute
3332 nested_in_vect_loop_p since the scalar stmt may get invalidated.
3334 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3337 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
3338 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
3339 If off is not INTEGER_CST, issue a may not be aligned warning
3340 rather than isn't aligned. Use isn%'t rather than isn't.
3341 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
3343 <case MULT_EXPR>: Improve the case when bottom and one of the
3344 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
3345 operand, in that case check if the other operand is multiple of
3346 bottom divided by the INTEGER_CST operand.
3348 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3351 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
3352 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
3353 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
3354 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
3355 * config/pa/pa.c (pa_function_arg_advance): Likewise.
3356 (pa_function_arg, pa_arg_partial_bytes): Likewise.
3357 (pa_function_arg_size): New function.
3359 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3361 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
3362 in a separate statement.
3364 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3366 PR tree-optimization/83847
3367 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
3368 group gathers and scatters.
3370 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3372 PR rtl-optimization/86620
3373 * params.def (max-sched-ready-insns): Bump minimum value to 1.
3375 PR rtl-optimization/83213
3376 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
3377 to last if both are JUMP_INSNs.
3379 PR tree-optimization/83843
3380 * gimple-ssa-store-merging.c
3381 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
3382 store_immediate_info for bswap/nop orig_stores.
3384 2018-01-15 Andrew Waterman <andrew@sifive.com>
3386 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
3388 <UDIV>: Increase cost if !TARGET_DIV.
3390 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
3392 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
3393 (define_attr "cr_logical_3op"): New.
3394 (cceq_ior_compare): Adjust.
3395 (cceq_ior_compare_complement): Adjust.
3396 (*cceq_rev_compare): Adjust.
3397 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
3398 (is_cracked_insn): Adjust.
3399 (insn_must_be_first_in_group): Adjust.
3400 * config/rs6000/40x.md: Adjust.
3401 * config/rs6000/440.md: Adjust.
3402 * config/rs6000/476.md: Adjust.
3403 * config/rs6000/601.md: Adjust.
3404 * config/rs6000/603.md: Adjust.
3405 * config/rs6000/6xx.md: Adjust.
3406 * config/rs6000/7450.md: Adjust.
3407 * config/rs6000/7xx.md: Adjust.
3408 * config/rs6000/8540.md: Adjust.
3409 * config/rs6000/cell.md: Adjust.
3410 * config/rs6000/e300c2c3.md: Adjust.
3411 * config/rs6000/e500mc.md: Adjust.
3412 * config/rs6000/e500mc64.md: Adjust.
3413 * config/rs6000/e5500.md: Adjust.
3414 * config/rs6000/e6500.md: Adjust.
3415 * config/rs6000/mpc.md: Adjust.
3416 * config/rs6000/power4.md: Adjust.
3417 * config/rs6000/power5.md: Adjust.
3418 * config/rs6000/power6.md: Adjust.
3419 * config/rs6000/power7.md: Adjust.
3420 * config/rs6000/power8.md: Adjust.
3421 * config/rs6000/power9.md: Adjust.
3422 * config/rs6000/rs64.md: Adjust.
3423 * config/rs6000/titan.md: Adjust.
3425 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3427 * config/i386/predicates.md (indirect_branch_operand): Rewrite
3428 ix86_indirect_branch_register logic.
3430 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3432 * config/i386/constraints.md (Bs): Update
3433 ix86_indirect_branch_register check. Don't check
3434 ix86_indirect_branch_register with GOT_memory_operand.
3436 * config/i386/predicates.md (GOT_memory_operand): Don't check
3437 ix86_indirect_branch_register here.
3438 (GOT32_symbol_operand): Likewise.
3440 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3442 * config/i386/predicates.md (constant_call_address_operand):
3443 Rewrite ix86_indirect_branch_register logic.
3444 (sibcall_insn_operand): Likewise.
3446 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3448 * config/i386/constraints.md (Bs): Replace
3449 ix86_indirect_branch_thunk_register with
3450 ix86_indirect_branch_register.
3452 * config/i386/i386.md (indirect_jump): Likewise.
3453 (tablejump): Likewise.
3454 (*sibcall_memory): Likewise.
3455 (*sibcall_value_memory): Likewise.
3456 Peepholes of indirect call and jump via memory: Likewise.
3457 * config/i386/i386.opt: Likewise.
3458 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3459 (GOT_memory_operand): Likewise.
3460 (call_insn_operand): Likewise.
3461 (sibcall_insn_operand): Likewise.
3462 (GOT32_symbol_operand): Likewise.
3464 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3467 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
3468 type rather than type addr's type points to.
3469 (expand_omp_atomic_mutex): Likewise.
3470 (expand_omp_atomic): Likewise.
3472 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3475 * config/i386/i386.c (output_indirect_thunk_function): Use
3476 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
3477 for __x86_return_thunk.
3479 2018-01-15 Richard Biener <rguenther@suse.de>
3482 * expmed.c (extract_bit_field_1): Fix typo.
3484 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3487 * config/arm/iterators.md (VF): New mode iterator.
3488 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
3489 Remove integer-related logic from pattern.
3490 (neon_vabd<mode>_3): Likewise.
3492 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3495 * common.opt (fstrict-overflow): No longer an alias.
3496 (fwrapv-pointer): New option.
3497 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
3498 also for pointer types based on flag_wrapv_pointer.
3499 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
3500 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
3501 opts->x_flag_wrapv got set.
3502 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
3503 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
3504 POINTER_TYPE_OVERFLOW_UNDEFINED.
3505 * match.pd: Likewise in address comparison pattern.
3506 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
3508 2018-01-15 Richard Biener <rguenther@suse.de>
3511 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
3512 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
3513 Reset type names to their identifier if their TYPE_DECL doesn't
3514 have linkage (and thus is used for ODR and devirt).
3515 (save_debug_info_for_decl): Remove.
3516 (save_debug_info_for_type): Likewise.
3517 (add_tree_to_fld_list): Adjust.
3518 * tree-pretty-print.c (dump_generic_node): Make dumping of
3519 type names more robust.
3521 2018-01-15 Richard Biener <rguenther@suse.de>
3523 * BASE-VER: Bump to 8.0.1.
3525 2018-01-14 Martin Sebor <msebor@redhat.com>
3528 * builtins.c (check_access): Avoid warning when the no-warning bit
3531 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
3533 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
3534 * ira-color (allocno_hard_regs_compare): Likewise.
3536 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
3539 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
3540 Use .pushsection/.popsection.
3542 2018-01-14 Martin Sebor <msebor@redhat.com>
3545 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
3547 2018-01-14 Jakub Jelinek <jakub@redhat.com>
3549 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
3550 entry from extra_headers.
3551 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
3552 extra_headers, make the list bitwise identical to the i?86-*-* one.
3554 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3556 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
3557 -mcmodel=large with -mindirect-branch=thunk,
3558 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
3559 -mfunction-return=thunk-extern.
3560 * doc/invoke.texi: Document -mcmodel=large is incompatible with
3561 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
3562 -mfunction-return=thunk and -mfunction-return=thunk-extern.
3564 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3566 * config/i386/i386.c (print_reg): Print the name of the full
3567 integer register without '%'.
3568 (ix86_print_operand): Handle 'V'.
3569 * doc/extend.texi: Document 'V' modifier.
3571 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3573 * config/i386/constraints.md (Bs): Disallow memory operand for
3574 -mindirect-branch-register.
3576 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3577 (GOT_memory_operand): Likewise.
3578 (call_insn_operand): Likewise.
3579 (sibcall_insn_operand): Likewise.
3580 (GOT32_symbol_operand): Likewise.
3581 * config/i386/i386.md (indirect_jump): Call convert_memory_address
3582 for -mindirect-branch-register.
3583 (tablejump): Likewise.
3584 (*sibcall_memory): Likewise.
3585 (*sibcall_value_memory): Likewise.
3586 Disallow peepholes of indirect call and jump via memory for
3587 -mindirect-branch-register.
3588 (*call_pop): Replace m with Bw.
3589 (*call_value_pop): Likewise.
3590 (*sibcall_pop_memory): Replace m with Bs.
3591 * config/i386/i386.opt (mindirect-branch-register): New option.
3592 * doc/invoke.texi: Document -mindirect-branch-register option.
3594 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3596 * config/i386/i386-protos.h (ix86_output_function_return): New.
3597 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
3598 set function_return_type.
3599 (indirect_thunk_name): Add ret_p to indicate thunk for function
3601 (output_indirect_thunk_function): Pass false to
3602 indirect_thunk_name.
3603 (ix86_output_indirect_branch_via_reg): Likewise.
3604 (ix86_output_indirect_branch_via_push): Likewise.
3605 (output_indirect_thunk_function): Create alias for function
3606 return thunk if regno < 0.
3607 (ix86_output_function_return): New function.
3608 (ix86_handle_fndecl_attribute): Handle function_return.
3609 (ix86_attribute_table): Add function_return.
3610 * config/i386/i386.h (machine_function): Add
3611 function_return_type.
3612 * config/i386/i386.md (simple_return_internal): Use
3613 ix86_output_function_return.
3614 (simple_return_internal_long): Likewise.
3615 * config/i386/i386.opt (mfunction-return=): New option.
3616 (indirect_branch): Mention -mfunction-return=.
3617 * doc/extend.texi: Document function_return function attribute.
3618 * doc/invoke.texi: Document -mfunction-return= option.
3620 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3622 * config/i386/i386-opts.h (indirect_branch): New.
3623 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
3624 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
3625 with local indirect jump when converting indirect call and jump.
3626 (ix86_set_indirect_branch_type): New.
3627 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
3628 (indirectlabelno): New.
3629 (indirect_thunk_needed): Likewise.
3630 (indirect_thunk_bnd_needed): Likewise.
3631 (indirect_thunks_used): Likewise.
3632 (indirect_thunks_bnd_used): Likewise.
3633 (INDIRECT_LABEL): Likewise.
3634 (indirect_thunk_name): Likewise.
3635 (output_indirect_thunk): Likewise.
3636 (output_indirect_thunk_function): Likewise.
3637 (ix86_output_indirect_branch_via_reg): Likewise.
3638 (ix86_output_indirect_branch_via_push): Likewise.
3639 (ix86_output_indirect_branch): Likewise.
3640 (ix86_output_indirect_jmp): Likewise.
3641 (ix86_code_end): Call output_indirect_thunk_function if needed.
3642 (ix86_output_call_insn): Call ix86_output_indirect_branch if
3644 (ix86_handle_fndecl_attribute): Handle indirect_branch.
3645 (ix86_attribute_table): Add indirect_branch.
3646 * config/i386/i386.h (machine_function): Add indirect_branch_type
3647 and has_local_indirect_jump.
3648 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
3650 (tablejump): Likewise.
3651 (*indirect_jump): Use ix86_output_indirect_jmp.
3652 (*tablejump_1): Likewise.
3653 (simple_return_indirect_internal): Likewise.
3654 * config/i386/i386.opt (mindirect-branch=): New option.
3655 (indirect_branch): New.
3658 (thunk-inline): Likewise.
3659 (thunk-extern): Likewise.
3660 * doc/extend.texi: Document indirect_branch function attribute.
3661 * doc/invoke.texi: Document -mindirect-branch= option.
3663 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
3666 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
3668 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
3670 * ipa-inline.c (want_inline_small_function_p): Return false if
3671 inlining has already failed with CIF_FINAL_ERROR.
3672 (update_caller_keys): Call want_inline_small_function_p before
3674 (update_callee_keys): Likewise.
3676 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3678 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
3680 (rs6000_quadword_masked_address_p): Likewise.
3681 (quad_aligned_load_p): Likewise.
3682 (quad_aligned_store_p): Likewise.
3683 (const_load_sequence_p): Add comment to describe the outer-most loop.
3684 (mimic_memory_attributes_and_flags): New function.
3685 (rs6000_gen_stvx): Likewise.
3686 (replace_swapped_aligned_store): Likewise.
3687 (rs6000_gen_lvx): Likewise.
3688 (replace_swapped_aligned_load): Likewise.
3689 (replace_swapped_load_constant): Capitalize argument name in
3690 comment describing this function.
3691 (rs6000_analyze_swaps): Add a third pass to search for vector loads
3692 and stores that access quad-word aligned addresses and replace
3693 with stvx or lvx instructions when appropriate.
3694 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
3695 New function prototype.
3696 (rs6000_quadword_masked_address_p): Likewise.
3697 (rs6000_gen_lvx): Likewise.
3698 (rs6000_gen_stvx): Likewise.
3699 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
3700 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
3701 when memory address is aligned.
3702 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
3703 this split to select lvx instruction when memory address is aligned.
3704 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
3705 instruction when memory address is aligned.
3706 (*vsx_le_perm_load_v16qi): Likewise.
3707 (four unnamed splitters): Modify to select the stvx instruction
3708 when memory is aligned.
3710 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
3712 * predict.c (determine_unlikely_bbs): Handle correctly BBs
3713 which appears in the queue multiple times.
3715 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3716 Alan Hayward <alan.hayward@arm.com>
3717 David Sherwood <david.sherwood@arm.com>
3719 * tree-vectorizer.h (vec_lower_bound): New structure.
3720 (_loop_vec_info): Add check_nonzero and lower_bounds.
3721 (LOOP_VINFO_CHECK_NONZERO): New macro.
3722 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
3723 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
3724 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
3725 fields. Make seg_len the distance travelled, not including the
3727 (dr_direction_indicator): Declare.
3728 (dr_zero_step_indicator): Likewise.
3729 (dr_known_forward_stride_p): Likewise.
3730 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
3732 (runtime_alias_check_p): Allow runtime alias checks with
3734 (operator ==): Compare access_size and align.
3735 (prune_runtime_alias_test_list): Rework for new distinction between
3736 the access_size and seg_len.
3737 (create_intersect_range_checks_index): Likewise. Cope with polynomial
3739 (get_segment_min_max): New function.
3740 (create_intersect_range_checks): Use it.
3741 (dr_step_indicator): New function.
3742 (dr_direction_indicator): Likewise.
3743 (dr_zero_step_indicator): Likewise.
3744 (dr_known_forward_stride_p): Likewise.
3745 * tree-loop-distribution.c (data_ref_segment_size): Return
3746 DR_STEP * (niters - 1).
3747 (compute_alias_check_pairs): Update call to the dr_with_seg_len
3749 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
3750 (vect_preserves_scalar_order_p): New function, split out from...
3751 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
3752 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
3753 (vect_vfa_access_size): New function.
3754 (vect_vfa_align): Likewise.
3755 (vect_compile_time_alias): Take access_size_a and access_b arguments.
3756 (dump_lower_bound): New function.
3757 (vect_check_lower_bound): Likewise.
3758 (vect_small_gap_p): Likewise.
3759 (vectorizable_with_step_bound_p): Likewise.
3760 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
3761 depencies if the vectorization factor is 1. Convert the checks
3762 for nonzero steps into checks on the bounds of DR_STEP. Try using
3763 a bunds check for variable steps if the minimum required step is
3764 relatively small. Update calls to the dr_with_seg_len
3765 constructor and to vect_compile_time_alias.
3766 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
3768 (vect_loop_versioning): Call it.
3769 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
3771 (vect_estimate_min_profitable_iters): Account for any bounds checks.
3773 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3774 Alan Hayward <alan.hayward@arm.com>
3775 David Sherwood <david.sherwood@arm.com>
3777 * doc/sourcebuild.texi (vect_scatter_store): Document.
3778 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
3780 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
3782 * genopinit.c (main): Add supports_vec_scatter_store and
3783 supports_vec_scatter_store_cached to target_optabs.
3784 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
3785 IFN_MASK_SCATTER_STORE.
3786 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
3788 * internal-fn.h (internal_store_fn_p): Declare.
3789 (internal_fn_stored_value_index): Likewise.
3790 * internal-fn.c (scatter_store_direct): New macro.
3791 (expand_scatter_store_optab_fn): New function.
3792 (direct_scatter_store_optab_supported_p): New macro.
3793 (internal_store_fn_p): New function.
3794 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
3795 IFN_MASK_SCATTER_STORE.
3796 (internal_fn_mask_index): Likewise.
3797 (internal_fn_stored_value_index): New function.
3798 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
3800 * optabs-query.h (supports_vec_scatter_store_p): Declare.
3801 * optabs-query.c (supports_vec_scatter_store_p): New function.
3802 * tree-vectorizer.h (vect_get_store_rhs): Declare.
3803 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
3804 true for scatter stores.
3805 (vect_gather_scatter_fn_p): Handle scatter stores too.
3806 (vect_check_gather_scatter): Consider using scatter stores if
3807 supports_vec_scatter_store_p.
3808 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
3810 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
3811 internal_fn_stored_value_index.
3812 (check_load_store_masking): Handle scatter stores too.
3813 (vect_get_store_rhs): Make public.
3814 (vectorizable_call): Use internal_store_fn_p.
3815 (vectorizable_store): Handle scatter store internal functions.
3816 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
3817 when deciding whether the end of the group has been reached.
3818 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
3819 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
3820 (mask_scatter_store<mode>): New insns.
3822 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3823 Alan Hayward <alan.hayward@arm.com>
3824 David Sherwood <david.sherwood@arm.com>
3826 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
3827 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
3828 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
3830 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
3831 Use vect_truncate_gather_scatter_offset if we can't treat the
3832 operation as a normal gather load or scatter store.
3833 (get_group_load_store_type): Take the gather_scatter_info
3834 as argument. Try using a gather load or scatter store for
3835 single-element groups.
3836 (get_load_store_type): Update calls to get_group_load_store_type
3837 and vect_use_strided_gather_scatters_p.
3839 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3840 Alan Hayward <alan.hayward@arm.com>
3841 David Sherwood <david.sherwood@arm.com>
3843 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
3844 optional tree argument.
3845 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
3847 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
3848 but continue to use the current value as a fallback.
3849 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
3850 to compare the updates.
3851 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
3852 (get_load_store_type): Use it when handling a strided access.
3853 (vect_get_strided_load_store_ops): New function.
3854 (vect_get_data_ptr_increment): Likewise.
3855 (vectorizable_load): Handle strided gather loads. Always pass
3856 a step to vect_create_data_ref_ptr and bump_vector_ptr.
3858 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3859 Alan Hayward <alan.hayward@arm.com>
3860 David Sherwood <david.sherwood@arm.com>
3862 * doc/md.texi (gather_load@var{m}): Document.
3863 (mask_gather_load@var{m}): Likewise.
3864 * genopinit.c (main): Add supports_vec_gather_load and
3865 supports_vec_gather_load_cached to target_optabs.
3866 * optabs-tree.c (init_tree_optimization_optabs): Use
3867 ggc_cleared_alloc to allocate target_optabs.
3868 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
3869 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
3871 * internal-fn.h (internal_load_fn_p): Declare.
3872 (internal_gather_scatter_fn_p): Likewise.
3873 (internal_fn_mask_index): Likewise.
3874 (internal_gather_scatter_fn_supported_p): Likewise.
3875 * internal-fn.c (gather_load_direct): New macro.
3876 (expand_gather_load_optab_fn): New function.
3877 (direct_gather_load_optab_supported_p): New macro.
3878 (direct_internal_fn_optab): New function.
3879 (internal_load_fn_p): Likewise.
3880 (internal_gather_scatter_fn_p): Likewise.
3881 (internal_fn_mask_index): Likewise.
3882 (internal_gather_scatter_fn_supported_p): Likewise.
3883 * optabs-query.c (supports_at_least_one_mode_p): New function.
3884 (supports_vec_gather_load_p): Likewise.
3885 * optabs-query.h (supports_vec_gather_load_p): Declare.
3886 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
3887 and memory_type field.
3888 (NUM_PATTERNS): Bump to 15.
3889 * tree-vect-data-refs.c: Include internal-fn.h.
3890 (vect_gather_scatter_fn_p): New function.
3891 (vect_describe_gather_scatter_call): Likewise.
3892 (vect_check_gather_scatter): Try using internal functions for
3893 gather loads. Recognize existing calls to a gather load function.
3894 (vect_analyze_data_refs): Consider using gather loads if
3895 supports_vec_gather_load_p.
3896 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
3897 (vect_get_gather_scatter_offset_type): Likewise.
3898 (vect_convert_mask_for_vectype): Likewise.
3899 (vect_add_conversion_to_patterm): Likewise.
3900 (vect_try_gather_scatter_pattern): Likewise.
3901 (vect_recog_gather_scatter_pattern): New pattern recognizer.
3902 (vect_vect_recog_func_ptrs): Add it.
3903 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
3904 internal_fn_mask_index and internal_gather_scatter_fn_p.
3905 (check_load_store_masking): Take the gather_scatter_info as an
3906 argument and handle gather loads.
3907 (vect_get_gather_scatter_ops): New function.
3908 (vectorizable_call): Check internal_load_fn_p.
3909 (vectorizable_load): Likewise. Handle gather load internal
3911 (vectorizable_store): Update call to check_load_store_masking.
3912 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
3913 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
3914 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
3915 (aarch64_gather_scale_operand_d): New predicates.
3916 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
3917 (mask_gather_load<mode>): New insns.
3919 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3920 Alan Hayward <alan.hayward@arm.com>
3921 David Sherwood <david.sherwood@arm.com>
3923 * optabs.def (fold_left_plus_optab): New optab.
3924 * doc/md.texi (fold_left_plus_@var{m}): Document.
3925 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
3926 * internal-fn.c (fold_left_direct): Define.
3927 (expand_fold_left_optab_fn): Likewise.
3928 (direct_fold_left_optab_supported_p): Likewise.
3929 * fold-const-call.c (fold_const_fold_left): New function.
3930 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
3931 * tree-parloops.c (valid_reduction_p): New function.
3932 (gather_scalar_reductions): Use it.
3933 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
3934 (vect_finish_replace_stmt): Declare.
3935 * tree-vect-loop.c (fold_left_reduction_fn): New function.
3936 (needs_fold_left_reduction_p): New function, split out from...
3937 (vect_is_simple_reduction): ...here. Accept reductions that
3938 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
3939 (vect_force_simple_reduction): Also store the reduction type in
3940 the assignment's STMT_VINFO_REDUC_TYPE.
3941 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
3942 (merge_with_identity): New function.
3943 (vect_expand_fold_left): Likewise.
3944 (vectorize_fold_left_reduction): Likewise.
3945 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
3946 scalar phi in place for it. Check for target support and reject
3947 cases that would reassociate the operation. Defer the transform
3948 phase to vectorize_fold_left_reduction.
3949 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
3950 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
3951 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
3953 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3955 * tree-if-conv.c (predicate_mem_writes): Remove redundant
3956 call to ifc_temp_var.
3958 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3959 Alan Hayward <alan.hayward@arm.com>
3960 David Sherwood <david.sherwood@arm.com>
3962 * target.def (legitimize_address_displacement): Take the original
3963 offset as a poly_int.
3964 * targhooks.h (default_legitimize_address_displacement): Update
3966 * targhooks.c (default_legitimize_address_displacement): Likewise.
3967 * doc/tm.texi: Regenerate.
3968 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
3969 as an argument, moving assert of ad->disp == ad->disp_term to...
3970 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
3971 Try calling targetm.legitimize_address_displacement before expanding
3972 the address rather than afterwards, and adjust for the new interface.
3973 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
3974 Match the new hook interface. Handle SVE addresses.
3975 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
3978 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3980 * Makefile.in (OBJS): Add early-remat.o.
3981 * target.def (select_early_remat_modes): New hook.
3982 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
3983 * doc/tm.texi: Regenerate.
3984 * targhooks.h (default_select_early_remat_modes): Declare.
3985 * targhooks.c (default_select_early_remat_modes): New function.
3986 * timevar.def (TV_EARLY_REMAT): New timevar.
3987 * passes.def (pass_early_remat): New pass.
3988 * tree-pass.h (make_pass_early_remat): Declare.
3989 * early-remat.c: New file.
3990 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
3992 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
3994 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3995 Alan Hayward <alan.hayward@arm.com>
3996 David Sherwood <david.sherwood@arm.com>
3998 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
3999 vfm1 with a bound_epilog parameter.
4000 (vect_do_peeling): Update calls accordingly, and move the prologue
4001 call earlier in the function. Treat the base bound_epilog as 0 for
4002 fully-masked loops and retain vf - 1 for other loops. Add 1 to
4003 this base when peeling for gaps.
4004 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
4005 with fully-masked loops.
4006 (vect_estimate_min_profitable_iters): Handle the single peeled
4007 iteration in that case.
4009 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4010 Alan Hayward <alan.hayward@arm.com>
4011 David Sherwood <david.sherwood@arm.com>
4013 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
4014 single-element interleaving even if the size is not a power of 2.
4015 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
4016 accesses for single-element interleaving if the group size is
4019 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4020 Alan Hayward <alan.hayward@arm.com>
4021 David Sherwood <david.sherwood@arm.com>
4023 * doc/md.texi (fold_extract_last_@var{m}): Document.
4024 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
4025 * optabs.def (fold_extract_last_optab): New optab.
4026 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
4027 * internal-fn.c (fold_extract_direct): New macro.
4028 (expand_fold_extract_optab_fn): Likewise.
4029 (direct_fold_extract_optab_supported_p): Likewise.
4030 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
4031 * tree-vect-loop.c (vect_model_reduction_cost): Handle
4032 EXTRACT_LAST_REDUCTION.
4033 (get_initial_def_for_reduction): Do not create an initial vector
4034 for EXTRACT_LAST_REDUCTION reductions.
4035 (vectorizable_reduction): Leave the scalar phi in place for
4036 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
4037 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
4038 epilogue code for EXTRACT_LAST_REDUCTION and defer the
4039 transform phase to vectorizable_condition.
4040 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
4042 (vect_finish_stmt_generation): ...here.
4043 (vect_finish_replace_stmt): New function.
4044 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
4045 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
4047 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
4049 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4050 Alan Hayward <alan.hayward@arm.com>
4051 David Sherwood <david.sherwood@arm.com>
4053 * doc/md.texi (extract_last_@var{m}): Document.
4054 * optabs.def (extract_last_optab): New optab.
4055 * internal-fn.def (EXTRACT_LAST): New internal function.
4056 * internal-fn.c (cond_unary_direct): New macro.
4057 (expand_cond_unary_optab_fn): Likewise.
4058 (direct_cond_unary_optab_supported_p): Likewise.
4059 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
4060 loops using EXTRACT_LAST.
4061 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
4062 (extract_last_<mode>): ...this optab.
4063 (vec_extract<mode><Vel>): Update accordingly.
4065 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4066 Alan Hayward <alan.hayward@arm.com>
4067 David Sherwood <david.sherwood@arm.com>
4069 * target.def (empty_mask_is_expensive): New hook.
4070 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
4071 * doc/tm.texi: Regenerate.
4072 * targhooks.h (default_empty_mask_is_expensive): Declare.
4073 * targhooks.c (default_empty_mask_is_expensive): New function.
4074 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
4075 if the target says that empty masks are expensive.
4076 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
4078 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
4080 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4081 Alan Hayward <alan.hayward@arm.com>
4082 David Sherwood <david.sherwood@arm.com>
4084 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
4085 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
4086 (vect_use_loop_mask_for_alignment_p): New function.
4087 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
4088 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
4089 niters_skip argument. Make sure that the first niters_skip elements
4090 of the first iteration are inactive.
4091 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
4092 Update call to vect_set_loop_masks_directly.
4093 (get_misalign_in_elems): New function, split out from...
4094 (vect_gen_prolog_loop_niters): ...here.
4095 (vect_update_init_of_dr): Take a code argument that specifies whether
4096 the adjustment should be added or subtracted.
4097 (vect_update_init_of_drs): Likewise.
4098 (vect_prepare_for_masked_peels): New function.
4099 (vect_do_peeling): Skip prologue peeling if we're using a mask
4100 instead. Update call to vect_update_inits_of_drs.
4101 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4103 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
4104 alignment. Do not include the number of peeled iterations in
4105 the minimum threshold in that case.
4106 (vectorizable_induction): Adjust the start value down by
4107 LOOP_VINFO_MASK_SKIP_NITERS iterations.
4108 (vect_transform_loop): Call vect_prepare_for_masked_peels.
4109 Take the number of skipped iterations into account when calculating
4111 * tree-vect-stmts.c (vect_gen_while_not): New function.
4113 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4114 Alan Hayward <alan.hayward@arm.com>
4115 David Sherwood <david.sherwood@arm.com>
4117 * doc/sourcebuild.texi (vect_fully_masked): Document.
4118 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
4120 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
4122 (vect_analyze_loop_2): ...here. Don't check the vectorization
4123 factor against the number of loop iterations if the loop is
4126 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4127 Alan Hayward <alan.hayward@arm.com>
4128 David Sherwood <david.sherwood@arm.com>
4130 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
4131 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
4132 (dump_groups): Update accordingly.
4133 (iv_use::mem_type): New member variable.
4134 (address_p): New function.
4135 (record_use): Add a mem_type argument and initialize the new
4137 (record_group_use): Add a mem_type argument. Use address_p.
4138 Remove obsolete null checks of base_object. Update call to record_use.
4139 (find_interesting_uses_op): Update call to record_group_use.
4140 (find_interesting_uses_cond): Likewise.
4141 (find_interesting_uses_address): Likewise.
4142 (get_mem_type_for_internal_fn): New function.
4143 (find_address_like_use): Likewise.
4144 (find_interesting_uses_stmt): Try find_address_like_use before
4145 calling find_interesting_uses_op.
4146 (addr_offset_valid_p): Use the iv mem_type field as the type
4147 of the addressed memory.
4148 (add_autoinc_candidates): Likewise.
4149 (get_address_cost): Likewise.
4150 (split_small_address_groups_p): Use address_p.
4151 (split_address_groups): Likewise.
4152 (add_iv_candidate_for_use): Likewise.
4153 (autoinc_possible_for_pair): Likewise.
4154 (rewrite_groups): Likewise.
4155 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
4156 (determine_group_iv_cost): Update after split of USE_ADDRESS.
4157 (get_alias_ptr_type_for_ptr_address): New function.
4158 (rewrite_use_address): Rewrite address uses in calls that were
4159 identified by find_address_like_use.
4161 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4162 Alan Hayward <alan.hayward@arm.com>
4163 David Sherwood <david.sherwood@arm.com>
4165 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
4167 * gimple-expr.h (is_gimple_addressable: Likewise.
4168 * gimple-expr.c (is_gimple_address): Likewise.
4169 * internal-fn.c (expand_call_mem_ref): New function.
4170 (expand_mask_load_optab_fn): Use it.
4171 (expand_mask_store_optab_fn): Likewise.
4173 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4174 Alan Hayward <alan.hayward@arm.com>
4175 David Sherwood <david.sherwood@arm.com>
4177 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
4178 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
4179 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
4180 (cond_umax@var{mode}): Document.
4181 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
4182 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
4183 (cond_umin_optab, cond_umax_optab): New optabs.
4184 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
4185 (COND_IOR, COND_XOR): New internal functions.
4186 * internal-fn.h (get_conditional_internal_fn): Declare.
4187 * internal-fn.c (cond_binary_direct): New macro.
4188 (expand_cond_binary_optab_fn): Likewise.
4189 (direct_cond_binary_optab_supported_p): Likewise.
4190 (get_conditional_internal_fn): New function.
4191 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
4192 Cope with reduction statements that are vectorized as calls rather
4194 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
4195 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
4196 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
4197 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4198 (UNSPEC_COND_EOR): New unspecs.
4199 (optab): Add mappings for them.
4200 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
4201 (sve_int_op, sve_fp_op): New int attributes.
4203 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4204 Alan Hayward <alan.hayward@arm.com>
4205 David Sherwood <david.sherwood@arm.com>
4207 * optabs.def (while_ult_optab): New optab.
4208 * doc/md.texi (while_ult@var{m}@var{n}): Document.
4209 * internal-fn.def (WHILE_ULT): New internal function.
4210 * internal-fn.h (direct_internal_fn_supported_p): New override
4211 that takes two types as argument.
4212 * internal-fn.c (while_direct): New macro.
4213 (expand_while_optab_fn): New function.
4214 (convert_optab_supported_p): Likewise.
4215 (direct_while_optab_supported_p): New macro.
4216 * wide-int.h (wi::udiv_ceil): New function.
4217 * tree-vectorizer.h (rgroup_masks): New structure.
4218 (vec_loop_masks): New typedef.
4219 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
4221 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
4222 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
4223 (vect_max_vf): New function.
4224 (slpeel_make_loop_iterate_ntimes): Delete.
4225 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
4226 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
4227 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
4228 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
4229 internal-fn.h, stor-layout.h and optabs-query.h.
4230 (vect_set_loop_mask): New function.
4231 (add_preheader_seq): Likewise.
4232 (add_header_seq): Likewise.
4233 (interleave_supported_p): Likewise.
4234 (vect_maybe_permute_loop_masks): Likewise.
4235 (vect_set_loop_masks_directly): Likewise.
4236 (vect_set_loop_condition_masked): Likewise.
4237 (vect_set_loop_condition_unmasked): New function, split out from
4238 slpeel_make_loop_iterate_ntimes.
4239 (slpeel_make_loop_iterate_ntimes): Rename to..
4240 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
4241 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
4242 (vect_do_peeling): Update call accordingly.
4243 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
4245 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4246 mask_compare_type, can_fully_mask_p and fully_masked_p.
4247 (release_vec_loop_masks): New function.
4248 (_loop_vec_info): Use it to free the loop masks.
4249 (can_produce_all_loop_masks_p): New function.
4250 (vect_get_max_nscalars_per_iter): Likewise.
4251 (vect_verify_full_masking): Likewise.
4252 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
4253 retries, and free the mask rgroups before retrying. Check loop-wide
4254 reasons for disallowing fully-masked loops. Make the final decision
4255 about whether use a fully-masked loop or not.
4256 (vect_estimate_min_profitable_iters): Do not assume that peeling
4257 for the number of iterations will be needed for fully-masked loops.
4258 (vectorizable_reduction): Disable fully-masked loops.
4259 (vectorizable_live_operation): Likewise.
4260 (vect_halve_mask_nunits): New function.
4261 (vect_double_mask_nunits): Likewise.
4262 (vect_record_loop_mask): Likewise.
4263 (vect_get_loop_mask): Likewise.
4264 (vect_transform_loop): Handle the case in which the final loop
4265 iteration might handle a partial vector. Call vect_set_loop_condition
4266 instead of slpeel_make_loop_iterate_ntimes.
4267 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
4268 (check_load_store_masking): New function.
4269 (prepare_load_store_mask): Likewise.
4270 (vectorizable_store): Handle fully-masked loops.
4271 (vectorizable_load): Likewise.
4272 (supportable_widening_operation): Use vect_halve_mask_nunits for
4274 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
4275 (vect_gen_while): New function.
4276 * config/aarch64/aarch64.md (umax<mode>3): New expander.
4277 (aarch64_uqdec<mode>): New insn.
4279 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4280 Alan Hayward <alan.hayward@arm.com>
4281 David Sherwood <david.sherwood@arm.com>
4283 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
4284 (reduc_xor_scal_optab): New optabs.
4285 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
4286 (reduc_xor_scal_@var{m}): Document.
4287 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
4288 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
4290 * fold-const-call.c (fold_const_call): Handle them.
4291 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
4292 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
4293 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
4294 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
4295 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
4296 (UNSPEC_XORV): New unspecs.
4297 (optab): Add entries for them.
4298 (BITWISEV): New int iterator.
4299 (bit_reduc_op): New int attributes.
4301 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4302 Alan Hayward <alan.hayward@arm.com>
4303 David Sherwood <david.sherwood@arm.com>
4305 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
4306 * internal-fn.def (VEC_SHL_INSERT): New internal function.
4307 * optabs.def (vec_shl_insert_optab): New optab.
4308 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
4309 (duplicate_and_interleave): Likewise.
4310 * tree-vect-loop.c: Include internal-fn.h.
4311 (neutral_op_for_slp_reduction): New function, split out from
4312 get_initial_defs_for_reduction.
4313 (get_initial_def_for_reduction): Handle option 2 for variable-length
4314 vectors by loading the neutral value into a vector and then shifting
4315 the initial value into element 0.
4316 (get_initial_defs_for_reduction): Replace the code argument with
4317 the neutral value calculated by neutral_op_for_slp_reduction.
4318 Use gimple_build_vector for constant-length vectors.
4319 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
4320 but the first group_size elements have a neutral value.
4321 Use duplicate_and_interleave otherwise.
4322 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
4323 Update call to get_initial_defs_for_reduction. Handle SLP
4324 reductions for variable-length vectors by creating one vector
4325 result for each scalar result, with the elements associated
4326 with other scalar results stubbed out with the neutral value.
4327 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
4328 Require IFN_VEC_SHL_INSERT for double reductions on
4329 variable-length vectors, or SLP reductions that have
4330 a neutral value. Require can_duplicate_and_interleave_p
4331 support for variable-length unchained SLP reductions if there
4332 is no neutral value, such as for MIN/MAX reductions. Also require
4333 the number of vector elements to be a multiple of the number of
4334 SLP statements when doing variable-length unchained SLP reductions.
4335 Update call to vect_create_epilog_for_reduction.
4336 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
4337 and remove initial values.
4338 (duplicate_and_interleave): Make public.
4339 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
4340 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
4342 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4343 Alan Hayward <alan.hayward@arm.com>
4344 David Sherwood <david.sherwood@arm.com>
4346 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
4347 (can_duplicate_and_interleave_p): New function.
4348 (vect_get_and_check_slp_defs): Take the vector of statements
4349 rather than just the current one. Remove excess parentheses.
4350 Restriction rejectinon of vect_constant_def and vect_external_def
4351 for variable-length vectors to boolean types, or types for which
4352 can_duplicate_and_interleave_p is false.
4353 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
4354 (duplicate_and_interleave): New function.
4355 (vect_get_constant_vectors): Use gimple_build_vector for
4356 constant-length vectors and suitable variable-length constant
4357 vectors. Use duplicate_and_interleave for other variable-length
4358 vectors. Don't defer the update when inserting new statements.
4360 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4361 Alan Hayward <alan.hayward@arm.com>
4362 David Sherwood <david.sherwood@arm.com>
4364 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
4365 min_profitable_iters doesn't go negative.
4367 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4368 Alan Hayward <alan.hayward@arm.com>
4369 David Sherwood <david.sherwood@arm.com>
4371 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
4372 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
4373 * optabs.def (vec_mask_load_lanes_optab): New optab.
4374 (vec_mask_store_lanes_optab): Likewise.
4375 * internal-fn.def (MASK_LOAD_LANES): New internal function.
4376 (MASK_STORE_LANES): Likewise.
4377 * internal-fn.c (mask_load_lanes_direct): New macro.
4378 (mask_store_lanes_direct): Likewise.
4379 (expand_mask_load_optab_fn): Handle masked operations.
4380 (expand_mask_load_lanes_optab_fn): New macro.
4381 (expand_mask_store_optab_fn): Handle masked operations.
4382 (expand_mask_store_lanes_optab_fn): New macro.
4383 (direct_mask_load_lanes_optab_supported_p): Likewise.
4384 (direct_mask_store_lanes_optab_supported_p): Likewise.
4385 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
4387 (vect_load_lanes_supported): Likewise.
4388 * tree-vect-data-refs.c (strip_conversion): New function.
4389 (can_group_stmts_p): Likewise.
4390 (vect_analyze_data_ref_accesses): Use it instead of checking
4391 for a pair of assignments.
4392 (vect_store_lanes_supported): Take a masked_p parameter.
4393 (vect_load_lanes_supported): Likewise.
4394 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
4395 vect_store_lanes_supported and vect_load_lanes_supported.
4396 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4397 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
4398 parameter. Don't allow gaps for masked accesses.
4399 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
4400 and vect_load_lanes_supported.
4401 (get_load_store_type): Take a masked_p parameter and update
4402 call to get_group_load_store_type.
4403 (vectorizable_store): Update call to get_load_store_type.
4404 Handle IFN_MASK_STORE_LANES.
4405 (vectorizable_load): Update call to get_load_store_type.
4406 Handle IFN_MASK_LOAD_LANES.
4408 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4409 Alan Hayward <alan.hayward@arm.com>
4410 David Sherwood <david.sherwood@arm.com>
4412 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
4414 * config/aarch64/aarch64-protos.h
4415 (aarch64_sve_struct_memory_operand_p): Declare.
4416 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
4417 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
4418 (VPRED, vpred): Handle SVE structure modes.
4419 * config/aarch64/constraints.md (Utx): New constraint.
4420 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
4421 (aarch64_sve_struct_nonimmediate_operand): New predicates.
4422 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
4423 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
4424 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
4425 structure modes. Split into pieces after RA.
4426 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
4427 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
4429 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
4430 SVE structure modes.
4431 (aarch64_classify_address): Likewise.
4432 (sizetochar): Move earlier in file.
4433 (aarch64_print_operand): Handle SVE register lists.
4434 (aarch64_array_mode): New function.
4435 (aarch64_sve_struct_memory_operand_p): Likewise.
4436 (TARGET_ARRAY_MODE): Redefine.
4438 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4439 Alan Hayward <alan.hayward@arm.com>
4440 David Sherwood <david.sherwood@arm.com>
4442 * target.def (array_mode): New target hook.
4443 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
4444 * doc/tm.texi: Regenerate.
4445 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
4446 * hooks.c (hook_optmode_mode_uhwi_none): New function.
4447 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
4449 * stor-layout.c (mode_for_array): Likewise. Support polynomial
4452 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4453 Alan Hayward <alan.hayward@arm.com>
4454 David Sherwood <david.sherwood@arm.com>
4456 * fold-const.c (fold_binary_loc): Check the argument types
4457 rather than the result type when testing for a vector operation.
4459 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4461 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
4462 * doc/tm.texi: Regenerate.
4464 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4465 Alan Hayward <alan.hayward@arm.com>
4466 David Sherwood <david.sherwood@arm.com>
4468 * doc/invoke.texi (-msve-vector-bits=): Document new option.
4469 (sve): Document new AArch64 extension.
4470 * doc/md.texi (w): Extend the description of the AArch64
4471 constraint to include SVE vectors.
4472 (Upl, Upa): Document new AArch64 predicate constraints.
4473 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
4475 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
4476 (msve-vector-bits=): New option.
4477 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
4478 SVE when these are disabled.
4479 (sve): New extension.
4480 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
4481 modes. Adjust their number of units based on aarch64_sve_vg.
4482 (MAX_BITSIZE_MODE_ANY_MODE): Define.
4483 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
4484 aarch64_addr_query_type.
4485 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
4486 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
4487 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
4488 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
4489 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
4490 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
4491 (aarch64_simd_imm_zero_p): Delete.
4492 (aarch64_check_zero_based_sve_index_immediate): Declare.
4493 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4494 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4495 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4496 (aarch64_sve_float_mul_immediate_p): Likewise.
4497 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4499 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
4500 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
4501 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
4502 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
4503 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
4504 (aarch64_regmode_natural_size): Likewise.
4505 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
4506 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
4508 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
4509 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
4510 for VG and the SVE predicate registers.
4511 (V_ALIASES): Add a "z"-prefixed alias.
4512 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
4513 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
4514 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
4515 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
4516 (REG_CLASS_NAMES): Add entries for them.
4517 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
4518 and the predicate registers.
4519 (aarch64_sve_vg): Declare.
4520 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
4521 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
4522 (REGMODE_NATURAL_SIZE): Define.
4523 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
4525 * config/aarch64/aarch64.c: Include cfgrtl.h.
4526 (simd_immediate_info): Add a constructor for series vectors,
4527 and an associated step field.
4528 (aarch64_sve_vg): New variable.
4529 (aarch64_dbx_register_number): Handle VG and the predicate registers.
4530 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
4531 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
4532 (VEC_ANY_DATA, VEC_STRUCT): New constants.
4533 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
4534 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
4535 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
4536 (aarch64_get_mask_mode): New functions.
4537 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
4538 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
4539 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
4540 predicate modes and predicate registers. Explicitly restrict
4541 GPRs to modes of 16 bytes or smaller. Only allow FP registers
4542 to store a vector mode if it is recognized by
4543 aarch64_classify_vector_mode.
4544 (aarch64_regmode_natural_size): New function.
4545 (aarch64_hard_regno_caller_save_mode): Return the original mode
4547 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
4548 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
4549 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
4550 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
4552 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
4553 does not overlap dest if the function is frame-related. Handle
4555 (aarch64_split_add_offset): New function.
4556 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
4557 them aarch64_add_offset.
4558 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
4559 and update call to aarch64_sub_sp.
4560 (aarch64_add_cfa_expression): New function.
4561 (aarch64_expand_prologue): Pass extra temporary registers to the
4562 functions above. Handle the case in which we need to emit new
4563 DW_CFA_expressions for registers that were originally saved
4564 relative to the stack pointer, but now have to be expressed
4565 relative to the frame pointer.
4566 (aarch64_output_mi_thunk): Pass extra temporary registers to the
4568 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
4569 IP0 and IP1 values for SVE frames.
4570 (aarch64_expand_vec_series): New function.
4571 (aarch64_expand_sve_widened_duplicate): Likewise.
4572 (aarch64_expand_sve_const_vector): Likewise.
4573 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
4574 Handle SVE constants. Use emit_move_insn to move a force_const_mem
4575 into the register, rather than emitting a SET directly.
4576 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
4577 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
4578 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
4579 (offset_9bit_signed_scaled_p): New functions.
4580 (aarch64_replicate_bitmask_imm): New function.
4581 (aarch64_bitmask_imm): Use it.
4582 (aarch64_cannot_force_const_mem): Reject expressions involving
4583 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
4584 (aarch64_classify_index): Handle SVE indices, by requiring
4585 a plain register index with a scale that matches the element size.
4586 (aarch64_classify_address): Handle SVE addresses. Assert that
4587 the mode of the address is VOIDmode or an integer mode.
4588 Update call to aarch64_classify_symbol.
4589 (aarch64_classify_symbolic_expression): Update call to
4590 aarch64_classify_symbol.
4591 (aarch64_const_vec_all_in_range_p): New function.
4592 (aarch64_print_vector_float_operand): Likewise.
4593 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
4594 "vN" for FP registers with SVE modes. Handle (const ...) vectors
4595 and the FP immediates 1.0 and 0.5.
4596 (aarch64_print_address_internal): Handle SVE addresses.
4597 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
4598 (aarch64_regno_regclass): Handle predicate registers.
4599 (aarch64_secondary_reload): Handle big-endian reloads of SVE
4601 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
4602 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
4603 (aarch64_convert_sve_vector_bits): New function.
4604 (aarch64_override_options): Use it to handle -msve-vector-bits=.
4605 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4607 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
4608 Handle SVE vector and predicate modes. Accept VL-based constants
4609 that need only one temporary register, and VL offsets that require
4610 no temporary registers.
4611 (aarch64_conditional_register_usage): Mark the predicate registers
4612 as fixed if SVE isn't available.
4613 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
4614 Return true for SVE vector and predicate modes.
4615 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
4616 rather than an unsigned int. Handle SVE modes.
4617 (aarch64_preferred_simd_mode): Update call accordingly. Handle
4619 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
4621 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4622 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4623 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4624 (aarch64_sve_float_mul_immediate_p): New functions.
4625 (aarch64_sve_valid_immediate): New function.
4626 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
4627 Explicitly reject structure modes. Check for INDEX constants.
4628 Handle PTRUE and PFALSE constants.
4629 (aarch64_check_zero_based_sve_index_immediate): New function.
4630 (aarch64_simd_imm_zero_p): Delete.
4631 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
4632 vector modes. Accept constants in the range of CNT[BHWD].
4633 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
4634 ask for an Advanced SIMD mode.
4635 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
4636 (aarch64_simd_vector_alignment): Handle SVE predicates.
4637 (aarch64_vectorize_preferred_vector_alignment): New function.
4638 (aarch64_simd_vector_alignment_reachable): Use it instead of
4640 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
4641 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
4643 (MAX_VECT_LEN): Delete.
4644 (expand_vec_perm_d): Add a vec_flags field.
4645 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
4646 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
4647 (aarch64_evpc_ext): Don't apply a big-endian lane correction
4649 (aarch64_evpc_rev): Rename to...
4650 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
4651 (aarch64_evpc_rev_global): New function.
4652 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
4653 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
4655 (aarch64_evpc_sve_tbl): New function.
4656 (aarch64_expand_vec_perm_const_1): Update after rename of
4657 aarch64_evpc_rev. Handle SVE permutes too, trying
4658 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
4659 than aarch64_evpc_tbl.
4660 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
4661 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
4662 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
4663 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
4664 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
4665 (aarch64_expand_sve_vcond): New functions.
4666 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
4667 of aarch64_vector_mode_p.
4668 (aarch64_dwarf_poly_indeterminate_value): New function.
4669 (aarch64_compute_pressure_classes): Likewise.
4670 (aarch64_can_change_mode_class): Likewise.
4671 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
4672 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
4673 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
4674 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
4675 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
4676 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
4677 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
4678 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
4680 (Dn, Dl, Dr): Accept const as well as const_vector.
4681 (Dz): Likewise. Compare against CONST0_RTX.
4682 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
4683 of "vector" where appropriate.
4684 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
4685 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
4686 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
4687 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
4688 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
4689 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
4690 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
4691 (v_int_equiv): Extend to SVE modes.
4692 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
4694 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
4695 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
4696 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
4697 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
4698 (SVE_COND_FP_CMP): New int iterators.
4699 (perm_hilo): Handle the new unpack unspecs.
4700 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
4702 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
4703 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
4704 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
4705 (aarch64_equality_operator, aarch64_constant_vector_operand)
4706 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
4707 (aarch64_sve_nonimmediate_operand): Likewise.
4708 (aarch64_sve_general_operand): Likewise.
4709 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
4710 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
4711 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
4712 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
4713 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
4714 (aarch64_sve_float_arith_immediate): Likewise.
4715 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
4716 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
4717 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
4718 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
4719 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
4720 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
4721 (aarch64_sve_float_arith_operand): Likewise.
4722 (aarch64_sve_float_arith_with_sub_operand): Likewise.
4723 (aarch64_sve_float_mul_operand): Likewise.
4724 (aarch64_sve_vec_perm_operand): Likewise.
4725 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
4726 (aarch64_mov_operand): Accept const_poly_int and const_vector.
4727 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
4728 as well as const_vector.
4729 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
4730 in file. Use CONST0_RTX and CONSTM1_RTX.
4731 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
4732 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
4733 Use aarch64_simd_imm_zero.
4734 * config/aarch64/aarch64-sve.md: New file.
4735 * config/aarch64/aarch64.md: Include it.
4736 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
4737 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
4738 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
4739 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
4740 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
4741 (sve): New attribute.
4742 (enabled): Disable instructions with the sve attribute unless
4744 (movqi, movhi): Pass CONST_POLY_INT operaneds through
4745 aarch64_expand_mov_immediate.
4746 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
4747 CNT[BHSD] immediates.
4748 (movti): Split CONST_POLY_INT moves into two halves.
4749 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
4750 Split additions that need a temporary here if the destination
4751 is the stack pointer.
4752 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
4753 (*add<mode>3_poly_1): New instruction.
4754 (set_clobber_cc): New expander.
4756 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4758 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
4759 parameter and use it instead of GET_MODE_SIZE (innermode). Use
4760 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
4761 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
4762 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
4763 Change innermode from fixed_mode_size to machine_mode.
4764 (simplify_subreg): Update call accordingly. Handle a constant-sized
4765 subreg of a variable-length CONST_VECTOR.
4767 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4768 Alan Hayward <alan.hayward@arm.com>
4769 David Sherwood <david.sherwood@arm.com>
4771 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
4772 (add_offset_to_base): New function, split out from...
4773 (create_mem_ref): ...here. When handling a scale other than 1,
4774 check first whether the address is valid without the offset.
4775 Add it into the base if so, leaving the index and scale as-is.
4777 2018-01-12 Jakub Jelinek <jakub@redhat.com>
4780 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
4781 fold_for_warn before checking if arg2 is INTEGER_CST.
4783 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
4785 * config/rs6000/predicates.md (load_multiple_operation): Delete.
4786 (store_multiple_operation): Delete.
4787 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
4788 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
4789 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
4790 guarded by TARGET_STRING.
4791 (rs6000_output_load_multiple): Delete.
4792 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
4793 OPTION_MASK_STRING / TARGET_STRING handling.
4794 (print_operand) <'N', 'O'>: Add comment that these are unused now.
4795 (const rs6000_opt_masks) <"string">: Change mask to 0.
4796 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
4797 (MASK_STRING): Delete.
4798 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
4800 (load_multiple): Delete.
4807 (store_multiple): Delete.
4814 (movmemsi_8reg): Delete.
4815 (corresponding unnamed define_insn): Delete.
4816 (movmemsi_6reg): Delete.
4817 (corresponding unnamed define_insn): Delete.
4818 (movmemsi_4reg): Delete.
4819 (corresponding unnamed define_insn): Delete.
4820 (movmemsi_2reg): Delete.
4821 (corresponding unnamed define_insn): Delete.
4822 (movmemsi_1reg): Delete.
4823 (corresponding unnamed define_insn): Delete.
4824 * config/rs6000/rs6000.opt (mno-string): New.
4825 (mstring): Replace by deprecation warning stub.
4826 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
4828 2018-01-12 Jakub Jelinek <jakub@redhat.com>
4830 * regrename.c (regrename_do_replace): If replacing the same
4831 reg multiple times, try to reuse last created gen_raw_REG.
4834 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
4835 main to workaround a bug in GDB.
4837 2018-01-12 Tom de Vries <tom@codesourcery.com>
4840 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
4842 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
4844 PR rtl-optimization/80481
4845 * ira-color.c (get_cap_member): New function.
4846 (allocnos_conflict_by_live_ranges_p): Use it.
4847 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
4848 (setup_slot_coalesced_allocno_live_ranges): Ditto.
4850 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
4853 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
4854 (*saddl_se_1): Ditto.
4856 (*ssubl_se_1): Ditto.
4858 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
4860 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
4861 rather than wi::to_widest for DR_INITs.
4862 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
4863 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
4864 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
4866 (vect_analyze_group_access_1): Note that here.
4868 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
4870 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
4871 polynomial type sizes.
4873 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
4875 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
4876 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
4877 (gimple_add_tmp_var): Likewise.
4879 2018-01-12 Martin Liska <mliska@suse.cz>
4881 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
4882 (gimple_alloc_sizes): Likewise.
4883 (dump_gimple_statistics): Use PRIu64 in printf format.
4884 * gimple.h: Change uint64_t to int.
4886 2018-01-12 Martin Liska <mliska@suse.cz>
4888 * tree-core.h: Use uint64_t instead of int.
4889 * tree.c (tree_node_counts): Likewise.
4890 (tree_node_sizes): Likewise.
4891 (dump_tree_statistics): Use PRIu64 in printf format.
4893 2018-01-12 Martin Liska <mliska@suse.cz>
4895 * Makefile.in: As qsort_chk is implemented in vec.c, add
4896 vec.o to linkage of gencfn-macros.
4897 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
4898 passing the info to record_node_allocation_statistics.
4899 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
4901 * ggc-common.c (struct ggc_usage): Add operator== and use
4902 it in operator< and compare function.
4903 * mem-stats.h (struct mem_usage): Likewise.
4904 * vec.c (struct vec_usage): Remove operator< and compare
4905 function. Can be simply inherited.
4907 2018-01-12 Martin Jambor <mjambor@suse.cz>
4910 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
4911 * tree-ssa-math-opts.c: Include domwalk.h.
4912 (convert_mult_to_fma_1): New function.
4913 (fma_transformation_info): New type.
4914 (fma_deferring_state): Likewise.
4915 (cancel_fma_deferring): New function.
4916 (result_of_phi): Likewise.
4917 (last_fma_candidate_feeds_initial_phi): Likewise.
4918 (convert_mult_to_fma): Added deferring logic, split actual
4919 transformation to convert_mult_to_fma_1.
4920 (math_opts_dom_walker): New type.
4921 (math_opts_dom_walker::after_dom_children): New method, body moved
4922 here from pass_optimize_widening_mul::execute, added deferring logic
4924 (pass_optimize_widening_mul::execute): Moved most of code to
4925 math_opts_dom_walker::after_dom_children.
4926 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
4927 * config/i386/i386.c (ix86_option_override_internal): Added
4928 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
4930 2018-01-12 Richard Biener <rguenther@suse.de>
4933 * dwarf2out.c (gen_variable_die): Do not reset old_die for
4934 inline instance vars.
4936 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
4939 * config/rx/rx.c (rx_is_restricted_memory_address):
4942 2018-01-12 Richard Biener <rguenther@suse.de>
4944 PR tree-optimization/80846
4945 * target.def (split_reduction): New target hook.
4946 * targhooks.c (default_split_reduction): New function.
4947 * targhooks.h (default_split_reduction): Declare.
4948 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
4949 target requests first reduce vectors by combining low and high
4951 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
4952 (get_vectype_for_scalar_type_and_size): Export.
4953 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
4954 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
4955 * doc/tm.texi: Regenerate.
4956 * config/i386/i386.c (ix86_split_reduction): Implement
4957 TARGET_VECTORIZE_SPLIT_REDUCTION.
4959 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4962 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
4963 in PIC mode except for TARGET_VXWORKS_RTP.
4964 * config/sparc/sparc.c: Include cfgrtl.h.
4965 (TARGET_INIT_PIC_REG): Define.
4966 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
4967 (sparc_pic_register_p): New predicate.
4968 (sparc_legitimate_address_p): Use it.
4969 (sparc_legitimize_pic_address): Likewise.
4970 (sparc_delegitimize_address): Likewise.
4971 (sparc_mode_dependent_address_p): Likewise.
4972 (gen_load_pcrel_sym): Remove 4th parameter.
4973 (load_got_register): Adjust call to above. Remove obsolete stuff.
4974 (sparc_expand_prologue): Do not call load_got_register here.
4975 (sparc_flat_expand_prologue): Likewise.
4976 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
4977 (sparc_use_pseudo_pic_reg): New function.
4978 (sparc_init_pic_reg): Likewise.
4979 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
4980 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
4982 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
4984 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
4985 Add item for branch_cost.
4987 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4989 PR rtl-optimization/83565
4990 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
4991 not extend the result to a larger mode for rotate operations.
4992 (num_sign_bit_copies1): Likewise.
4994 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4997 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
4999 Use values-Xc.o for -pedantic.
5000 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
5002 2018-01-12 Martin Liska <mliska@suse.cz>
5005 * ipa-devirt.c (final_warning_record::grow_type_warnings):
5007 (possible_polymorphic_call_targets): Use it.
5008 (ipa_devirt): Likewise.
5010 2018-01-12 Martin Liska <mliska@suse.cz>
5012 * profile-count.h (enum profile_quality): Use 0 as invalid
5013 enum value of profile_quality.
5015 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
5017 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
5018 -mext-string options.
5020 2018-01-12 Richard Biener <rguenther@suse.de>
5022 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
5023 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
5024 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
5026 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
5028 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5030 * configure.ac (--with-long-double-format): Add support for the
5031 configuration option to change the default long double format on
5033 * config.gcc (powerpc*-linux*-*): Likewise.
5034 * configure: Regenerate.
5035 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
5036 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
5037 used without modification.
5039 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5041 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
5042 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
5043 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
5044 MISC_BUILTIN_SPEC_BARRIER.
5045 (rs6000_init_builtins): Likewise.
5046 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
5048 (speculation_barrier): New define_insn.
5049 * doc/extend.texi: Document __builtin_speculation_barrier.
5051 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5054 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
5055 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
5056 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
5058 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
5059 integral modes instead of "ss" and "sd".
5060 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
5061 vectors with 32-bit and 64-bit elements.
5062 (vecdupssescalarmodesuffix): New mode attribute.
5063 (vec_dup<mode>): Use it.
5065 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
5068 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
5069 frame if argument is passed on stack.
5071 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5074 * ree.c (combine_reaching_defs): Optimize also
5075 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
5076 reg2=any_extend(exp); reg1=reg2;, formatting fix.
5078 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5081 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
5083 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5086 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
5087 after they are computed.
5089 2018-01-11 Bin Cheng <bin.cheng@arm.com>
5091 PR tree-optimization/83695
5092 * gimple-loop-linterchange.cc
5093 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
5094 reset cached scev information after interchange.
5095 (pass_linterchange::execute): Remove call to scev_reset_htab.
5097 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5099 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
5100 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
5101 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
5102 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
5103 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
5104 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
5105 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
5106 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
5107 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
5108 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
5109 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
5110 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
5111 (V_lane_reg): Likewise.
5112 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
5114 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
5115 (vfmal_lane_low<mode>_intrinsic,
5116 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
5117 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
5118 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
5119 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
5120 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
5121 vfmsl_lane_high<mode>_intrinsic): New define_insns.
5123 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5125 * config/arm/arm-cpus.in (fp16fml): New feature.
5126 (ALL_SIMD): Add fp16fml.
5127 (armv8.2-a): Add fp16fml as an option.
5128 (armv8.3-a): Likewise.
5129 (armv8.4-a): Add fp16fml as part of fp16.
5130 * config/arm/arm.h (TARGET_FP16FML): Define.
5131 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
5133 * config/arm/arm-modes.def (V2HF): Define.
5134 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
5135 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
5136 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
5137 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
5138 vfmsl_low, vfmsl_high): New set of builtins.
5139 * config/arm/iterators.md (PLUSMINUS): New code iterator.
5140 (vfml_op): New code attribute.
5141 (VFMLHALVES): New int iterator.
5142 (VFML, VFMLSEL): New mode attributes.
5143 (V_reg): Define mapping for V2HF.
5144 (V_hi, V_lo): New mode attributes.
5145 (VF_constraint): Likewise.
5146 (vfml_half, vfml_half_selector): New int attributes.
5147 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
5149 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
5150 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
5152 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
5153 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
5154 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
5155 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
5157 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
5158 Document new effective target and option set.
5160 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5162 * config/arm/arm-cpus.in (armv8_4): New feature.
5163 (ARMv8_4a): New fgroup.
5164 (armv8.4-a): New arch.
5165 * config/arm/arm-tables.opt: Regenerate.
5166 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
5167 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
5168 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
5169 Add matching rules for -march=armv8.4-a and extensions.
5170 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
5172 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
5175 * config/rx/rx.md (BW): New mode attribute.
5176 (sync_lock_test_and_setsi): Add mode suffix to insn output.
5178 2018-01-11 Richard Biener <rguenther@suse.de>
5180 PR tree-optimization/83435
5181 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
5182 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
5183 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
5185 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5186 Alan Hayward <alan.hayward@arm.com>
5187 David Sherwood <david.sherwood@arm.com>
5189 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
5191 (aarch64_classify_address): Initialize it. Track polynomial offsets.
5192 (aarch64_print_address_internal): Use it to check for a zero offset.
5194 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5195 Alan Hayward <alan.hayward@arm.com>
5196 David Sherwood <david.sherwood@arm.com>
5198 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
5199 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
5200 Return a poly_int64 rather than a HOST_WIDE_INT.
5201 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
5202 rather than a HOST_WIDE_INT.
5203 * config/aarch64/aarch64.h (aarch64_frame): Protect with
5204 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
5205 hard_fp_offset, frame_size, initial_adjust, callee_offset and
5206 final_offset from HOST_WIDE_INT to poly_int64.
5207 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
5208 to_constant when getting the number of units in an Advanced SIMD
5210 (aarch64_builtin_vectorized_function): Check for a constant number
5212 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
5214 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
5215 attribute instead of GET_MODE_NUNITS.
5216 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
5217 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
5218 GET_MODE_SIZE for fixed-size registers.
5219 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
5220 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
5221 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
5222 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
5223 (aarch64_print_operand, aarch64_print_address_internal)
5224 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
5225 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
5226 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
5227 Handle polynomial GET_MODE_SIZE.
5228 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
5229 wider than SImode without modification.
5230 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
5231 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
5232 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
5233 passing and returning SVE modes.
5234 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
5235 rather than GEN_INT.
5236 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
5237 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
5238 (aarch64_allocate_and_probe_stack_space): Likewise.
5239 (aarch64_layout_frame): Cope with polynomial offsets.
5240 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
5241 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
5243 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
5244 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
5245 poly_int64 rather than a HOST_WIDE_INT.
5246 (aarch64_get_separate_components, aarch64_process_components)
5247 (aarch64_expand_prologue, aarch64_expand_epilogue)
5248 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
5249 (aarch64_anchor_offset): New function, split out from...
5250 (aarch64_legitimize_address): ...here.
5251 (aarch64_builtin_vectorization_cost): Handle polynomial
5252 TYPE_VECTOR_SUBPARTS.
5253 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
5255 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
5256 number of elements from the PARALLEL rather than the mode.
5257 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
5258 rather than GET_MODE_BITSIZE.
5259 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
5260 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
5261 (aarch64_expand_vec_perm_const_1): Handle polynomial
5262 d->perm.length () and d->perm elements.
5263 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
5264 Apply to_constant to d->perm elements.
5265 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
5266 polynomial CONST_VECTOR_NUNITS.
5267 (aarch64_move_pointer): Take amount as a poly_int64 rather
5269 (aarch64_progress_pointer): Avoid temporary variable.
5270 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
5271 the mode attribute instead of GET_MODE.
5273 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5274 Alan Hayward <alan.hayward@arm.com>
5275 David Sherwood <david.sherwood@arm.com>
5277 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
5278 x exists before using it.
5279 (aarch64_add_constant_internal): Rename to...
5280 (aarch64_add_offset_1): ...this. Replace regnum with separate
5281 src and dest rtxes. Handle the case in which they're different,
5282 including when the offset is zero. Replace scratchreg with an rtx.
5283 Use 2 additions if there is no spare register into which we can
5284 move a 16-bit constant.
5285 (aarch64_add_constant): Delete.
5286 (aarch64_add_offset): Replace reg with separate src and dest
5287 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
5288 Use aarch64_add_offset_1.
5289 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
5290 an rtx rather than an int. Take the delta as a poly_int64
5291 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
5292 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
5293 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
5294 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
5295 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
5297 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
5298 aarch64_add_constant.
5300 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5302 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
5303 Use scalar_float_mode.
5305 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5307 * config/aarch64/aarch64-simd.md
5308 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
5309 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
5310 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
5311 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
5312 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
5313 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
5314 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
5315 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
5316 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
5317 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
5319 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5322 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
5323 targ_options->x_arm_arch_string is non NULL.
5325 2018-01-11 Tamar Christina <tamar.christina@arm.com>
5327 * config/aarch64/aarch64.h
5328 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
5330 2018-01-11 Sudakshina Das <sudi.das@arm.com>
5333 * expmed.c (emit_store_flag_force): Swap if const op0
5334 and change VOIDmode to mode of op0.
5336 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5338 PR rtl-optimization/83761
5339 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
5340 than bytes to mode_for_size.
5342 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5345 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
5346 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
5349 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5352 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
5353 when in layout mode.
5354 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
5355 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
5358 2018-01-10 Michael Collison <michael.collison@arm.com>
5360 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
5361 * config/aarch64/aarch64-option-extension.def: Add
5362 AARCH64_OPT_EXTENSION of 'fp16fml'.
5363 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5364 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
5365 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
5366 * config/aarch64/constraints.md (Ui7): New constraint.
5367 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
5368 (VFMLA_SEL_W): Ditto.
5371 (VFMLA16_LOW): New int iterator.
5372 (VFMLA16_HIGH): Ditto.
5373 (UNSPEC_FMLAL): New unspec.
5374 (UNSPEC_FMLSL): Ditto.
5375 (UNSPEC_FMLAL2): Ditto.
5376 (UNSPEC_FMLSL2): Ditto.
5377 (f16mac): New code attribute.
5378 * config/aarch64/aarch64-simd-builtins.def
5379 (aarch64_fmlal_lowv2sf): Ditto.
5380 (aarch64_fmlsl_lowv2sf): Ditto.
5381 (aarch64_fmlalq_lowv4sf): Ditto.
5382 (aarch64_fmlslq_lowv4sf): Ditto.
5383 (aarch64_fmlal_highv2sf): Ditto.
5384 (aarch64_fmlsl_highv2sf): Ditto.
5385 (aarch64_fmlalq_highv4sf): Ditto.
5386 (aarch64_fmlslq_highv4sf): Ditto.
5387 (aarch64_fmlal_lane_lowv2sf): Ditto.
5388 (aarch64_fmlsl_lane_lowv2sf): Ditto.
5389 (aarch64_fmlal_laneq_lowv2sf): Ditto.
5390 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
5391 (aarch64_fmlalq_lane_lowv4sf): Ditto.
5392 (aarch64_fmlsl_lane_lowv4sf): Ditto.
5393 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
5394 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
5395 (aarch64_fmlal_lane_highv2sf): Ditto.
5396 (aarch64_fmlsl_lane_highv2sf): Ditto.
5397 (aarch64_fmlal_laneq_highv2sf): Ditto.
5398 (aarch64_fmlsl_laneq_highv2sf): Ditto.
5399 (aarch64_fmlalq_lane_highv4sf): Ditto.
5400 (aarch64_fmlsl_lane_highv4sf): Ditto.
5401 (aarch64_fmlalq_laneq_highv4sf): Ditto.
5402 (aarch64_fmlsl_laneq_highv4sf): Ditto.
5403 * config/aarch64/aarch64-simd.md:
5404 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
5405 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5406 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
5407 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5408 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
5409 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
5410 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
5411 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
5412 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
5413 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
5414 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
5415 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
5416 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
5417 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
5418 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
5419 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
5420 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
5421 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
5422 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
5423 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
5424 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
5425 (vfmlsl_low_u32): Ditto.
5426 (vfmlalq_low_u32): Ditto.
5427 (vfmlslq_low_u32): Ditto.
5428 (vfmlal_high_u32): Ditto.
5429 (vfmlsl_high_u32): Ditto.
5430 (vfmlalq_high_u32): Ditto.
5431 (vfmlslq_high_u32): Ditto.
5432 (vfmlal_lane_low_u32): Ditto.
5433 (vfmlsl_lane_low_u32): Ditto.
5434 (vfmlal_laneq_low_u32): Ditto.
5435 (vfmlsl_laneq_low_u32): Ditto.
5436 (vfmlalq_lane_low_u32): Ditto.
5437 (vfmlslq_lane_low_u32): Ditto.
5438 (vfmlalq_laneq_low_u32): Ditto.
5439 (vfmlslq_laneq_low_u32): Ditto.
5440 (vfmlal_lane_high_u32): Ditto.
5441 (vfmlsl_lane_high_u32): Ditto.
5442 (vfmlal_laneq_high_u32): Ditto.
5443 (vfmlsl_laneq_high_u32): Ditto.
5444 (vfmlalq_lane_high_u32): Ditto.
5445 (vfmlslq_lane_high_u32): Ditto.
5446 (vfmlalq_laneq_high_u32): Ditto.
5447 (vfmlslq_laneq_high_u32): Ditto.
5448 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
5449 (AARCH64_FL_FOR_ARCH8_4): New.
5450 (AARCH64_ISA_F16FML): New ISA flag.
5451 (TARGET_F16FML): New feature flag for fp16fml.
5452 (doc/invoke.texi): Document new fp16fml option.
5454 2018-01-10 Michael Collison <michael.collison@arm.com>
5456 * config/aarch64/aarch64-builtins.c:
5457 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
5458 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5459 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
5460 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
5461 (AARCH64_ISA_SHA3): New ISA flag.
5462 (TARGET_SHA3): New feature flag for sha3.
5463 * config/aarch64/iterators.md (sha512_op): New int attribute.
5464 (CRYPTO_SHA512): New int iterator.
5465 (UNSPEC_SHA512H): New unspec.
5466 (UNSPEC_SHA512H2): Ditto.
5467 (UNSPEC_SHA512SU0): Ditto.
5468 (UNSPEC_SHA512SU1): Ditto.
5469 * config/aarch64/aarch64-simd-builtins.def
5470 (aarch64_crypto_sha512hqv2di): New builtin.
5471 (aarch64_crypto_sha512h2qv2di): Ditto.
5472 (aarch64_crypto_sha512su0qv2di): Ditto.
5473 (aarch64_crypto_sha512su1qv2di): Ditto.
5474 (aarch64_eor3qv8hi): Ditto.
5475 (aarch64_rax1qv2di): Ditto.
5476 (aarch64_xarqv2di): Ditto.
5477 (aarch64_bcaxqv8hi): Ditto.
5478 * config/aarch64/aarch64-simd.md:
5479 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
5480 (aarch64_crypto_sha512su0qv2di): Ditto.
5481 (aarch64_crypto_sha512su1qv2di): Ditto.
5482 (aarch64_eor3qv8hi): Ditto.
5483 (aarch64_rax1qv2di): Ditto.
5484 (aarch64_xarqv2di): Ditto.
5485 (aarch64_bcaxqv8hi): Ditto.
5486 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
5487 (vsha512h2q_u64): Ditto.
5488 (vsha512su0q_u64): Ditto.
5489 (vsha512su1q_u64): Ditto.
5490 (veor3q_u16): Ditto.
5491 (vrax1q_u64): Ditto.
5493 (vbcaxq_u16): Ditto.
5494 * config/arm/types.md (crypto_sha512): New type attribute.
5495 (crypto_sha3): Ditto.
5496 (doc/invoke.texi): Document new sha3 option.
5498 2018-01-10 Michael Collison <michael.collison@arm.com>
5500 * config/aarch64/aarch64-builtins.c:
5501 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
5502 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5503 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
5504 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
5505 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
5506 (AARCH64_ISA_SM4): New ISA flag.
5507 (TARGET_SM4): New feature flag for sm4.
5508 * config/aarch64/aarch64-simd-builtins.def
5509 (aarch64_sm3ss1qv4si): Ditto.
5510 (aarch64_sm3tt1aq4si): Ditto.
5511 (aarch64_sm3tt1bq4si): Ditto.
5512 (aarch64_sm3tt2aq4si): Ditto.
5513 (aarch64_sm3tt2bq4si): Ditto.
5514 (aarch64_sm3partw1qv4si): Ditto.
5515 (aarch64_sm3partw2qv4si): Ditto.
5516 (aarch64_sm4eqv4si): Ditto.
5517 (aarch64_sm4ekeyqv4si): Ditto.
5518 * config/aarch64/aarch64-simd.md:
5519 (aarch64_sm3ss1qv4si): Ditto.
5520 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
5521 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
5522 (aarch64_sm4eqv4si): Ditto.
5523 (aarch64_sm4ekeyqv4si): Ditto.
5524 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
5525 (sm3part_op): Ditto.
5526 (CRYPTO_SM3TT): Ditto.
5527 (CRYPTO_SM3PART): Ditto.
5528 (UNSPEC_SM3SS1): New unspec.
5529 (UNSPEC_SM3TT1A): Ditto.
5530 (UNSPEC_SM3TT1B): Ditto.
5531 (UNSPEC_SM3TT2A): Ditto.
5532 (UNSPEC_SM3TT2B): Ditto.
5533 (UNSPEC_SM3PARTW1): Ditto.
5534 (UNSPEC_SM3PARTW2): Ditto.
5535 (UNSPEC_SM4E): Ditto.
5536 (UNSPEC_SM4EKEY): Ditto.
5537 * config/aarch64/constraints.md (Ui2): New constraint.
5538 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
5539 * config/arm/types.md (crypto_sm3): New type attribute.
5540 (crypto_sm4): Ditto.
5541 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
5542 (vsm3tt1aq_u32): Ditto.
5543 (vsm3tt1bq_u32): Ditto.
5544 (vsm3tt2aq_u32): Ditto.
5545 (vsm3tt2bq_u32): Ditto.
5546 (vsm3partw1q_u32): Ditto.
5547 (vsm3partw2q_u32): Ditto.
5548 (vsm4eq_u32): Ditto.
5549 (vsm4ekeyq_u32): Ditto.
5550 (doc/invoke.texi): Document new sm4 option.
5552 2018-01-10 Michael Collison <michael.collison@arm.com>
5554 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
5555 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
5556 (AARCH64_FL_FOR_ARCH8_4): New.
5557 (AARCH64_FL_V8_4): New flag.
5558 (doc/invoke.texi): Document new armv8.4-a option.
5560 2018-01-10 Michael Collison <michael.collison@arm.com>
5562 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5563 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
5564 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
5565 * config/aarch64/aarch64-option-extension.def: Add
5566 AARCH64_OPT_EXTENSION of 'sha2'.
5567 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
5568 (crypto): Disable sha2 and aes if crypto disabled.
5569 (crypto): Enable aes and sha2 if enabled.
5570 (simd): Disable sha2 and aes if simd disabled.
5571 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
5573 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
5574 (TARGET_SHA2): New feature flag for sha2.
5575 (TARGET_AES): New feature flag for aes.
5576 * config/aarch64/aarch64-simd.md:
5577 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
5578 conditional on TARGET_AES.
5579 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
5580 (aarch64_crypto_sha1hsi): Make pattern conditional
5582 (aarch64_crypto_sha1hv4si): Ditto.
5583 (aarch64_be_crypto_sha1hv4si): Ditto.
5584 (aarch64_crypto_sha1su1v4si): Ditto.
5585 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
5586 (aarch64_crypto_sha1su0v4si): Ditto.
5587 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
5588 (aarch64_crypto_sha256su0v4si): Ditto.
5589 (aarch64_crypto_sha256su1v4si): Ditto.
5590 (doc/invoke.texi): Document new aes and sha2 options.
5592 2018-01-10 Martin Sebor <msebor@redhat.com>
5594 PR tree-optimization/83781
5595 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
5598 2018-01-11 Martin Sebor <msebor@gmail.com>
5599 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5601 PR tree-optimization/83501
5602 PR tree-optimization/81703
5604 * tree-ssa-strlen.c (get_string_cst): Rename...
5605 (get_string_len): ...to this. Handle global constants.
5606 (handle_char_store): Adjust.
5608 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
5609 Jim Wilson <jimw@sifive.com>
5611 * config/riscv/riscv-protos.h (riscv_output_return): New.
5612 * config/riscv/riscv.c (struct machine_function): New naked_p field.
5613 (riscv_attribute_table, riscv_output_return),
5614 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
5615 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
5616 (riscv_compute_frame_info): Only compute frame->mask if not a naked
5618 (riscv_expand_prologue): Add early return for naked function.
5619 (riscv_expand_epilogue): Likewise.
5620 (riscv_function_ok_for_sibcall): Return false for naked function.
5621 (riscv_set_current_function): New.
5622 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
5623 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
5624 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
5625 * doc/extend.texi (RISC-V Function Attributes): New.
5627 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
5629 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
5630 check for 128-bit long double before checking TCmode.
5631 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
5632 128-bit long doubles before checking TFmode or TCmode.
5633 (FLOAT128_IBM_P): Likewise.
5635 2018-01-10 Martin Sebor <msebor@redhat.com>
5637 PR tree-optimization/83671
5638 * builtins.c (c_strlen): Unconditionally return zero for the empty
5640 Use -Warray-bounds for warnings.
5641 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
5642 for non-constant array indices with COMPONENT_REF, arrays of
5643 arrays, and pointers to arrays.
5644 (gimple_fold_builtin_strlen): Determine and set length range for
5645 non-constant character arrays.
5647 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
5650 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
5653 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
5655 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
5657 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5660 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
5661 VECTOR_MEM_ALTIVEC_OR_VSX_P.
5662 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
5663 indexed_or_indirect_operand predicate.
5664 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
5665 (*vsx_le_perm_load_v8hi): Likewise.
5666 (*vsx_le_perm_load_v16qi): Likewise.
5667 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
5668 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
5669 (*vsx_le_perm_store_v8hi): Likewise.
5670 (*vsx_le_perm_store_v16qi): Likewise.
5671 (eight unnamed splitters): Likewise.
5673 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5675 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
5676 * config/rs6000/emmintrin.h: Likewise.
5677 * config/rs6000/mmintrin.h: Likewise.
5678 * config/rs6000/xmmintrin.h: Likewise.
5680 2018-01-10 David Malcolm <dmalcolm@redhat.com>
5683 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
5685 * tree.c (tree_nop_conversion): Return true for location wrapper
5687 (maybe_wrap_with_location): New function.
5688 (selftest::check_strip_nops): New function.
5689 (selftest::test_location_wrappers): New function.
5690 (selftest::tree_c_tests): Call it.
5691 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
5692 (maybe_wrap_with_location): New decl.
5693 (EXPR_LOCATION_WRAPPER_P): New macro.
5694 (location_wrapper_p): New inline function.
5695 (tree_strip_any_location_wrapper): New inline function.
5697 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
5700 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
5701 stack_realign_offset for the largest alignment of stack slot
5703 (ix86_find_max_used_stack_alignment): New function.
5704 (ix86_finalize_stack_frame_flags): Use it. Set
5705 max_used_stack_alignment if we don't realign stack.
5706 * config/i386/i386.h (machine_function): Add
5707 max_used_stack_alignment.
5709 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
5711 * config/arm/arm.opt (-mbranch-cost): New option.
5712 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
5715 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
5718 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
5719 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
5721 2018-01-10 Richard Biener <rguenther@suse.de>
5724 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
5725 early out so it also covers the case where we have a non-NULL
5728 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
5730 PR tree-optimization/83753
5731 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
5732 for non-strided grouped accesses if the number of elements is 1.
5734 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5737 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
5738 * i386.h (TARGET_USE_GATHER): Define.
5739 * x86-tune.def (X86_TUNE_USE_GATHER): New.
5741 2018-01-10 Martin Liska <mliska@suse.cz>
5744 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
5745 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
5747 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
5748 CLEANUP_NO_PARTITIONING is not set.
5750 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
5752 * doc/rtl.texi: Remove documentation of (const ...) wrappers
5753 for vectors, as a partial revert of r254296.
5754 * rtl.h (const_vec_p): Delete.
5755 (const_vec_duplicate_p): Don't test for vector CONSTs.
5756 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
5757 * expmed.c (make_tree): Likewise.
5760 * common.md (E, F): Use CONSTANT_P instead of checking for
5762 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
5763 checking for CONST_VECTOR.
5765 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5768 * predict.c (force_edge_cold): Handle in more sane way edges
5771 2018-01-09 Carl Love <cel@us.ibm.com>
5773 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
5775 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
5776 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
5777 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
5778 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
5779 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
5780 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
5781 * config/rs6000/rs6000-protos.h: Add extern defition for
5782 rs6000_generate_float2_double_code.
5783 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
5785 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
5786 (float2_v2df): Add define_expand.
5788 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
5791 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
5792 op_mode in the force_to_mode call.
5794 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5796 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
5797 instead of checking each element individually.
5798 (aarch64_evpc_uzp): Likewise.
5799 (aarch64_evpc_zip): Likewise.
5800 (aarch64_evpc_ext): Likewise.
5801 (aarch64_evpc_rev): Likewise.
5802 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
5803 instead of checking each element individually. Return true without
5805 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
5806 whether all selected elements come from the same input, instead of
5807 checking each element individually. Remove calls to gen_rtx_REG,
5808 start_sequence and end_sequence and instead assert that no rtl is
5811 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5813 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
5814 order of HIGH and CONST checks.
5816 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5818 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
5819 if the destination isn't an SSA_NAME.
5821 2018-01-09 Richard Biener <rguenther@suse.de>
5823 PR tree-optimization/83668
5824 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
5826 (canonicalize_loop_form): ... here, renamed from ...
5827 (canonicalize_loop_closed_ssa_form): ... this and amended to
5828 swap successor edges for loop exit blocks to make us use
5829 the RPO order we need for initial schedule generation.
5831 2018-01-09 Joseph Myers <joseph@codesourcery.com>
5833 PR tree-optimization/64811
5834 * match.pd: When optimizing comparisons with Inf, avoid
5835 introducing or losing exceptions from comparisons with NaN.
5837 2018-01-09 Martin Liska <mliska@suse.cz>
5840 * asan.c (shadow_mem_size): Add gcc_assert.
5842 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
5844 Don't save registers in main().
5847 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
5848 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
5849 * config/avr/avr.c (avr_set_current_function): Don't error if
5850 naked, OS_task or OS_main are specified at the same time.
5851 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
5853 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
5855 * common/config/avr/avr-common.c (avr_option_optimization_table):
5856 Switch on -mmain-is-OS_task for optimizing compilations.
5858 2018-01-09 Richard Biener <rguenther@suse.de>
5860 PR tree-optimization/83572
5861 * graphite.c: Include cfganal.h.
5862 (graphite_transform_loops): Connect infinite loops to exit
5863 and remove fake edges at the end.
5865 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5867 * ipa-inline.c (edge_badness): Revert accidental checkin.
5869 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5872 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
5873 symbols; not inline clones.
5875 2018-01-09 Jakub Jelinek <jakub@redhat.com>
5878 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
5879 hard registers. Formatting fixes.
5881 PR preprocessor/83722
5882 * gcc.c (try_generate_repro): Pass
5883 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
5884 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
5887 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
5888 Kito Cheng <kito.cheng@gmail.com>
5890 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
5891 (riscv_leaf_function_p): Delete.
5892 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
5894 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5896 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
5898 (do_ifelse): New function.
5899 (do_isel): New function.
5900 (do_sub3): New function.
5901 (do_add3): New function.
5902 (do_load_mask_compare): New function.
5903 (do_overlap_load_compare): New function.
5904 (expand_compare_loop): New function.
5905 (expand_block_compare): Call expand_compare_loop() when appropriate.
5906 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
5908 (-mblock-compare-inline-loop-limit): New option.
5910 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5913 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
5914 Reverse order of second and third operands in first alternative.
5915 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
5916 of first and second elements in UNSPEC_VPERMR vector.
5917 (altivec_expand_vec_perm_le): Likewise.
5919 2018-01-08 Jeff Law <law@redhat.com>
5921 PR rtl-optimizatin/81308
5922 * tree-switch-conversion.c (cfg_altered): New file scoped static.
5923 (process_switch): If group_case_labels makes a change, then set
5925 (pass_convert_switch::execute): If a switch is converted, then
5926 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
5928 PR rtl-optimization/81308
5929 * recog.c (split_all_insns): Conditionally cleanup the CFG after
5932 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
5934 PR target/83663 - Revert r255946
5935 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
5936 generation for cases where splatting a value is not useful.
5937 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
5938 across a vec_duplicate and a paradoxical subreg forming a vector
5939 mode to a vec_concat.
5941 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5943 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
5944 -march=armv8.3-a variants.
5945 * config/arm/t-multilib: Likewise.
5946 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
5948 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5950 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
5952 (cceq_ior_compare_complement): Give it a name so I can use it, and
5953 change boolean_or_operator predicate to boolean_operator so it can
5954 be used to generate a crand.
5955 (eqne): New code iterator.
5956 (bd/bd_neg): New code_attrs.
5957 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
5958 a single define_insn.
5959 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
5960 decrement (bdnzt/bdnzf/bdzt/bdzf).
5961 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
5962 with the new names of the branch decrement patterns, and added the
5963 names of the branch decrement conditional patterns.
5965 2018-01-08 Richard Biener <rguenther@suse.de>
5967 PR tree-optimization/83563
5968 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
5971 2018-01-08 Richard Biener <rguenther@suse.de>
5974 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
5976 2018-01-08 Richard Biener <rguenther@suse.de>
5978 PR tree-optimization/83685
5979 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
5980 references to abnormals.
5982 2018-01-08 Richard Biener <rguenther@suse.de>
5985 * dwarf2out.c (output_indirect_strings): Handle empty
5986 skeleton_debug_str_hash.
5987 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
5989 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
5991 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
5992 (emit_store_direct): Likewise.
5993 (arc_trampoline_adjust_address): Likewise.
5994 (arc_asm_trampoline_template): New function.
5995 (arc_initialize_trampoline): Use asm_trampoline_template.
5996 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
5997 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
5998 * config/arc/arc.md (flush_icache): Delete pattern.
6000 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6002 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
6003 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
6006 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6009 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
6010 by not USED_FOR_TARGET.
6011 (make_pass_resolve_sw_modes): Likewise.
6013 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6015 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
6018 2018-01-08 Richard Biener <rguenther@suse.de>
6021 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
6023 2018-01-08 Richard Biener <rguenther@suse.de>
6026 * match.pd ((t * 2) / 2) -> t): Add missing :c.
6028 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
6031 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
6032 basic blocks with a small number of successors.
6033 (convert_control_dep_chain_into_preds): Improve handling of
6035 (dump_predicates): Split apart into...
6036 (dump_pred_chain): ...here...
6037 (dump_pred_info): ...and here.
6038 (can_one_predicate_be_invalidated_p): Add debugging printfs.
6039 (can_chain_union_be_invalidated_p): Improve check for invalidation
6041 (uninit_uses_cannot_happen): Avoid unnecessary if
6042 convert_control_dep_chain_into_preds yielded nothing.
6044 2018-01-06 Martin Sebor <msebor@redhat.com>
6046 PR tree-optimization/83640
6047 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
6048 subtracting negative offset from size.
6049 (builtin_access::overlap): Adjust offset bounds of the access to fall
6050 within the size of the object if possible.
6052 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
6054 PR rtl-optimization/83699
6055 * expmed.c (extract_bit_field_1): Restrict the vector usage of
6056 extract_bit_field_as_subreg to cases in which the extracted
6057 value is also a vector.
6059 * lra-constraints.c (process_alt_operands): Test for the equivalence
6060 substitutions when detecting a possible reload cycle.
6062 2018-01-06 Jakub Jelinek <jakub@redhat.com>
6065 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
6066 by default if flag_selective_schedling{,2}. Formatting fixes.
6068 PR rtl-optimization/83682
6069 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
6070 if it has non-VECTOR_MODE element mode.
6071 (vec_duplicate_p): Likewise.
6074 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
6075 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
6077 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6080 * config/i386/i386-builtin.def
6081 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
6082 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
6083 Require also OPTION_MASK_ISA_AVX512F in addition to
6084 OPTION_MASK_ISA_GFNI.
6085 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
6086 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
6087 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
6088 to OPTION_MASK_ISA_GFNI.
6089 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
6090 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
6091 OPTION_MASK_ISA_AVX512BW.
6092 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
6093 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
6094 addition to OPTION_MASK_ISA_GFNI.
6095 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
6096 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
6097 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
6098 to OPTION_MASK_ISA_GFNI.
6099 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
6100 a requirement for all ISAs rather than any of them with a few
6102 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
6104 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
6105 bitmasks to be enabled with 3 exceptions, instead of requiring any
6106 enabled ISA with lots of exceptions.
6107 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
6108 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
6109 Change avx512bw in isa attribute to avx512f.
6110 * config/i386/sgxintrin.h: Add license boilerplate.
6111 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
6112 to __AVX512F__ and __AVX512VL to __AVX512VL__.
6113 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
6114 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
6116 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
6117 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
6118 temporarily sse2 rather than sse if not enabled already.
6121 * config/i386/sse.md (VI248_VLBW): Rename to ...
6122 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
6123 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
6124 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
6125 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
6126 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
6127 mode iterator instead of VI248_VLBW.
6129 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
6131 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
6132 (record_modified): Skip clobbers; add debug output.
6133 (param_change_prob): Use sreal frequencies.
6135 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6137 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
6138 punt for user-aligned variables.
6140 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6142 * tree-chrec.c (chrec_contains_symbols): Return true for
6145 2018-01-05 Sudakshina Das <sudi.das@arm.com>
6148 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
6149 of (x|y) == x for BICS pattern.
6151 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6153 PR tree-optimization/83605
6154 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
6155 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
6158 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
6160 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
6161 * config/epiphany/rtems.h: New file.
6163 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6164 Uros Bizjak <ubizjak@gmail.com>
6167 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
6168 QIreg_operand instead of register_operand predicate.
6169 * config/i386/i386.c (ix86_rop_should_change_byte_p,
6170 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
6171 comments instead of -fmitigate[-_]rop.
6173 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6176 * cgraphunit.c (symbol_table::compile): Switch to text_section
6177 before calling assembly_start debug hook.
6178 * run-rtl-passes.c (run_rtl_passes): Likewise.
6181 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6183 * tree-vrp.c (extract_range_from_binary_expr_1): Check
6184 range_int_cst_p rather than !symbolic_range_p before calling
6185 extract_range_from_multiplicative_op_1.
6187 2018-01-04 Jeff Law <law@redhat.com>
6189 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
6190 redundant test in assertion.
6192 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6194 * doc/rtl.texi: Document machine_mode wrapper classes.
6196 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6198 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
6201 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6203 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
6204 the VEC_PERM_EXPR fold to fail.
6206 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6209 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
6210 to switched_sections.
6212 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6215 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
6218 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
6221 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
6222 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
6224 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6227 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
6228 is BLKmode and bitpos not zero or mode change is needed.
6230 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6233 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
6236 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
6239 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
6240 instead of MULT rtx. Update all corresponding splitters.
6242 (*ssub<modesuffix>): Ditto.
6244 (*cmp_sadd_di): Update split patterns.
6245 (*cmp_sadd_si): Ditto.
6246 (*cmp_sadd_sidi): Ditto.
6247 (*cmp_ssub_di): Ditto.
6248 (*cmp_ssub_si): Ditto.
6249 (*cmp_ssub_sidi): Ditto.
6250 * config/alpha/predicates.md (const23_operand): New predicate.
6251 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
6252 Look for ASHIFT, not MULT inner operand.
6253 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
6255 2018-01-04 Martin Liska <mliska@suse.cz>
6257 PR gcov-profile/83669
6258 * gcov.c (output_intermediate_file): Add version to intermediate
6260 * doc/gcov.texi: Document new field 'version' in intermediate
6261 file format. Fix location of '-k' option of gcov command.
6263 2018-01-04 Martin Liska <mliska@suse.cz>
6266 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
6268 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6270 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
6272 2018-01-03 Martin Sebor <msebor@redhat.com>
6274 PR tree-optimization/83655
6275 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
6276 checking calls with invalid arguments.
6278 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6280 * tree-vect-stmts.c (vect_get_store_rhs): New function.
6281 (vectorizable_mask_load_store): Delete.
6282 (vectorizable_call): Return false for masked loads and stores.
6283 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
6284 instead of gimple_assign_rhs1.
6285 (vectorizable_load): Handle IFN_MASK_LOAD.
6286 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
6288 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6290 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
6292 (vectorizable_mask_load_store): ...here.
6293 (vectorizable_load): ...and here.
6295 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6297 * tree-vect-stmts.c (vect_build_all_ones_mask)
6298 (vect_build_zero_merge_argument): New functions, split out from...
6299 (vectorizable_load): ...here.
6301 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6303 * tree-vect-stmts.c (vect_check_store_rhs): New function,
6305 (vectorizable_mask_load_store): ...here.
6306 (vectorizable_store): ...and here.
6308 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6310 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
6312 (vectorizable_mask_load_store): ...here.
6314 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6316 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
6317 (vect_model_store_cost): Take a vec_load_store_type instead of a
6319 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
6320 (vect_model_store_cost): Take a vec_load_store_type instead of a
6322 (vectorizable_mask_load_store): Update accordingly.
6323 (vectorizable_store): Likewise.
6324 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
6326 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6328 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
6329 IFN_MASK_LOAD calls here rather than...
6330 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
6332 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6333 Alan Hayward <alan.hayward@arm.com>
6334 David Sherwood <david.sherwood@arm.com>
6336 * expmed.c (extract_bit_field_1): For vector extracts,
6337 fall back to extract_bit_field_as_subreg if vec_extract
6340 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6341 Alan Hayward <alan.hayward@arm.com>
6342 David Sherwood <david.sherwood@arm.com>
6344 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
6345 they are variable or constant sized.
6346 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
6347 slots for constant-sized data.
6349 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6350 Alan Hayward <alan.hayward@arm.com>
6351 David Sherwood <david.sherwood@arm.com>
6353 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
6354 handling COND_EXPRs with boolean comparisons, try to find a better
6355 basis for the mask type than the boolean itself.
6357 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6359 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
6360 is calculated and how it can be overridden.
6361 * genmodes.c (max_bitsize_mode_any_mode): New variable.
6362 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
6364 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
6367 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6368 Alan Hayward <alan.hayward@arm.com>
6369 David Sherwood <david.sherwood@arm.com>
6371 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
6372 Remove the mode argument.
6373 (aarch64_simd_valid_immediate): Remove the mode and inverse
6375 * config/aarch64/iterators.md (bitsize): New iterator.
6376 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
6377 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
6378 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
6379 aarch64_simd_valid_immediate.
6380 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
6381 (aarch64_reg_or_bic_imm): Likewise.
6382 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
6383 with an insn_type enum and msl with a modifier_type enum.
6384 Replace element_width with a scalar_mode. Change the shift
6385 to unsigned int. Add constructors for scalar_float_mode and
6386 scalar_int_mode elements.
6387 (aarch64_vect_float_const_representable_p): Delete.
6388 (aarch64_can_const_movi_rtx_p)
6389 (aarch64_simd_scalar_immediate_valid_for_move)
6390 (aarch64_simd_make_constant): Update call to
6391 aarch64_simd_valid_immediate.
6392 (aarch64_advsimd_valid_immediate_hs): New function.
6393 (aarch64_advsimd_valid_immediate): Likewise.
6394 (aarch64_simd_valid_immediate): Remove mode and inverse
6395 arguments. Rewrite to use the above. Use const_vec_duplicate_p
6396 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
6397 and aarch64_float_const_representable_p on the result.
6398 (aarch64_output_simd_mov_immediate): Remove mode argument.
6399 Update call to aarch64_simd_valid_immediate and use of
6400 simd_immediate_info.
6401 (aarch64_output_scalar_simd_mov_immediate): Update call
6404 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6405 Alan Hayward <alan.hayward@arm.com>
6406 David Sherwood <david.sherwood@arm.com>
6408 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
6409 (mode_nunits): Likewise CONST_MODE_NUNITS.
6410 * machmode.def (ADJUST_NUNITS): Document.
6411 * genmodes.c (mode_data::need_nunits_adj): New field.
6412 (blank_mode): Update accordingly.
6413 (adj_nunits): New variable.
6414 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
6416 (emit_mode_size_inline): Set need_bytesize_adj for all modes
6417 listed in adj_nunits.
6418 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
6419 listed in adj_nunits. Don't emit case statements for such modes.
6420 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
6421 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
6422 nothing if adj_nunits is nonnull.
6423 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
6424 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
6425 (emit_mode_fbit): Update use of print_maybe_const_decl.
6426 (emit_move_size): Likewise. Treat the array as non-const
6428 (emit_mode_adjustments): Handle adj_nunits.
6430 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6432 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
6433 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
6434 (VECTOR_MODES): Use it.
6435 (make_vector_modes): Take the prefix as an argument.
6437 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6438 Alan Hayward <alan.hayward@arm.com>
6439 David Sherwood <david.sherwood@arm.com>
6441 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
6442 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
6443 for MODE_VECTOR_BOOL.
6444 * machmode.def (VECTOR_BOOL_MODE): Document.
6445 * genmodes.c (VECTOR_BOOL_MODE): New macro.
6446 (make_vector_bool_mode): New function.
6447 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
6449 * lto-streamer-in.c (lto_input_mode_table): Likewise.
6450 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
6452 * stor-layout.c (int_mode_for_mode): Likewise.
6453 * tree.c (build_vector_type_for_mode): Likewise.
6454 * varasm.c (output_constant_pool_2): Likewise.
6455 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
6456 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
6457 for MODE_VECTOR_BOOL.
6458 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
6459 of mode class checks.
6460 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
6461 instead of a list of mode class checks.
6462 (expand_vector_scalar_condition): Likewise.
6463 (type_for_widest_vector_mode): Handle BImode as an inner mode.
6465 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6466 Alan Hayward <alan.hayward@arm.com>
6467 David Sherwood <david.sherwood@arm.com>
6469 * machmode.h (mode_size): Change from unsigned short to
6471 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
6472 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6473 or if measurement_type is not polynomial.
6474 (fixed_size_mode::includes_p): Check for constant-sized modes.
6475 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
6476 return a poly_uint16 rather than an unsigned short.
6477 (emit_mode_size): Change the type of mode_size from unsigned short
6478 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
6479 (emit_mode_adjustments): Cope with polynomial vector sizes.
6480 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6482 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6484 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
6485 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
6486 * caller-save.c (setup_save_areas): Likewise.
6487 (replace_reg_with_saved_mem): Likewise.
6488 * calls.c (emit_library_call_value_1): Likewise.
6489 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
6490 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
6491 (gen_lowpart_for_combine): Likewise.
6492 * convert.c (convert_to_integer_1): Likewise.
6493 * cse.c (equiv_constant, cse_insn): Likewise.
6494 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
6495 (cselib_subst_to_values): Likewise.
6496 * dce.c (word_dce_process_block): Likewise.
6497 * df-problems.c (df_word_lr_mark_ref): Likewise.
6498 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
6499 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
6500 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
6501 (rtl_for_decl_location): Likewise.
6502 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
6503 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
6504 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
6505 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
6506 (expand_expr_real_1): Likewise.
6507 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
6508 (pad_below): Likewise.
6509 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6510 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
6511 * ira.c (get_subreg_tracking_sizes): Likewise.
6512 * ira-build.c (ira_create_allocno_objects): Likewise.
6513 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
6514 (ira_sort_regnos_for_alter_reg): Likewise.
6515 * ira-costs.c (record_operand_costs): Likewise.
6516 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
6517 (resolve_simple_move): Likewise.
6518 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
6519 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
6520 (lra_constraints): Likewise.
6521 (CONST_POOL_OK_P): Reject variable-sized modes.
6522 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
6523 (add_pseudo_to_slot, lra_spill): Likewise.
6524 * omp-low.c (omp_clause_aligned_alignment): Likewise.
6525 * optabs-query.c (get_best_extraction_insn): Likewise.
6526 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6527 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
6528 (expand_mult_highpart, valid_multiword_target_p): Likewise.
6529 * recog.c (offsettable_address_addr_space_p): Likewise.
6530 * regcprop.c (maybe_mode_change): Likewise.
6531 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
6532 * regrename.c (build_def_use): Likewise.
6533 * regstat.c (dump_reg_info): Likewise.
6534 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
6535 (find_reloads, find_reloads_subreg_address): Likewise.
6536 * reload1.c (eliminate_regs_1): Likewise.
6537 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
6538 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
6539 (simplify_binary_operation_1, simplify_subreg): Likewise.
6540 * targhooks.c (default_function_arg_padding): Likewise.
6541 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
6542 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
6543 (verify_gimple_assign_ternary): Likewise.
6544 * tree-inline.c (estimate_move_cost): Likewise.
6545 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6546 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
6547 (get_address_cost_ainc): Likewise.
6548 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
6549 (vect_supportable_dr_alignment): Likewise.
6550 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6551 (vectorizable_reduction): Likewise.
6552 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
6553 (vectorizable_operation, vectorizable_load): Likewise.
6554 * tree.c (build_same_sized_truth_vector_type): Likewise.
6555 * valtrack.c (cleanup_auto_inc_dec): Likewise.
6556 * var-tracking.c (emit_note_insn_var_location): Likewise.
6557 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
6558 (ADDR_VEC_ALIGN): Likewise.
6560 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6561 Alan Hayward <alan.hayward@arm.com>
6562 David Sherwood <david.sherwood@arm.com>
6564 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
6566 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6567 or if measurement_type is polynomial.
6568 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
6569 * combine.c (make_extraction): Likewise.
6570 * dse.c (find_shift_sequence): Likewise.
6571 * dwarf2out.c (mem_loc_descriptor): Likewise.
6572 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
6573 (extract_bit_field, extract_low_bits): Likewise.
6574 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
6575 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
6576 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
6577 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
6578 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6579 * reload.c (find_reloads): Likewise.
6580 * reload1.c (alter_reg): Likewise.
6581 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
6582 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
6583 * tree-if-conv.c (predicate_mem_writes): Likewise.
6584 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
6585 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
6586 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
6587 * valtrack.c (dead_debug_insert_temp): Likewise.
6588 * varasm.c (mergeable_constant_section): Likewise.
6589 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
6591 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6592 Alan Hayward <alan.hayward@arm.com>
6593 David Sherwood <david.sherwood@arm.com>
6595 * expr.c (expand_assignment): Cope with polynomial mode sizes
6596 when assigning to a CONCAT.
6598 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6599 Alan Hayward <alan.hayward@arm.com>
6600 David Sherwood <david.sherwood@arm.com>
6602 * machmode.h (mode_precision): Change from unsigned short to
6604 (mode_to_precision): Return a poly_uint16 rather than an unsigned
6606 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
6607 or if measurement_type is not polynomial.
6608 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
6609 in which the mode is already known to be a scalar_int_mode.
6610 * genmodes.c (emit_mode_precision): Change the type of mode_precision
6611 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
6613 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6614 for GET_MODE_PRECISION.
6615 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6616 for GET_MODE_PRECISION.
6617 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
6619 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
6620 (expand_field_assignment, make_extraction): Likewise.
6621 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
6622 (get_last_value): Likewise.
6623 * convert.c (convert_to_integer_1): Likewise.
6624 * cse.c (cse_insn): Likewise.
6625 * expr.c (expand_expr_real_1): Likewise.
6626 * lra-constraints.c (simplify_operand_subreg): Likewise.
6627 * optabs-query.c (can_atomic_load_p): Likewise.
6628 * optabs.c (expand_atomic_load): Likewise.
6629 (expand_atomic_store): Likewise.
6630 * ree.c (combine_reaching_defs): Likewise.
6631 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
6632 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
6633 * tree.h (type_has_mode_precision_p): Likewise.
6634 * ubsan.c (instrument_si_overflow): Likewise.
6636 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6637 Alan Hayward <alan.hayward@arm.com>
6638 David Sherwood <david.sherwood@arm.com>
6640 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
6641 polynomial numbers of units.
6642 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
6643 (valid_vector_subparts_p): New function.
6644 (build_vector_type): Remove temporary shim and take the number
6645 of units as a poly_uint64 rather than an int.
6646 (build_opaque_vector_type): Take the number of units as a
6647 poly_uint64 rather than an int.
6648 * tree.c (build_vector_from_ctor): Handle polynomial
6649 TYPE_VECTOR_SUBPARTS.
6650 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
6651 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
6652 (build_vector_from_val): If the number of units is variable,
6653 use build_vec_duplicate_cst for constant operands and
6654 VEC_DUPLICATE_EXPR otherwise.
6655 (make_vector_type): Remove temporary is_constant ().
6656 (build_vector_type, build_opaque_vector_type): Take the number of
6657 units as a poly_uint64 rather than an int.
6658 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
6660 * cfgexpand.c (expand_debug_expr): Likewise.
6661 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
6662 (store_constructor, expand_expr_real_1): Likewise.
6663 (const_scalar_mask_from_tree): Likewise.
6664 * fold-const-call.c (fold_const_reduction): Likewise.
6665 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
6666 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
6667 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
6668 (fold_relational_const): Likewise.
6669 (native_interpret_vector): Likewise. Change the size from an
6670 int to an unsigned int.
6671 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
6672 TYPE_VECTOR_SUBPARTS.
6673 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
6674 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
6675 duplicating a non-constant operand into a variable-length vector.
6676 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
6677 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
6678 * ipa-icf.c (sem_variable::equals): Likewise.
6679 * match.pd: Likewise.
6680 * omp-simd-clone.c (simd_clone_subparts): Likewise.
6681 * print-tree.c (print_node): Likewise.
6682 * stor-layout.c (layout_type): Likewise.
6683 * targhooks.c (default_builtin_vectorization_cost): Likewise.
6684 * tree-cfg.c (verify_gimple_comparison): Likewise.
6685 (verify_gimple_assign_binary): Likewise.
6686 (verify_gimple_assign_ternary): Likewise.
6687 (verify_gimple_assign_single): Likewise.
6688 * tree-pretty-print.c (dump_generic_node): Likewise.
6689 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6690 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
6691 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
6692 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
6693 (vect_shift_permute_load_chain): Likewise.
6694 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
6695 (expand_vector_condition, optimize_vector_constructor): Likewise.
6696 (lower_vec_perm, get_compute_type): Likewise.
6697 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6698 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
6699 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
6700 (vect_recog_mask_conversion_pattern): Likewise.
6701 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
6702 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
6703 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
6704 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
6705 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
6706 (vectorizable_shift, vectorizable_operation, vectorizable_store)
6707 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
6708 (supportable_widening_operation): Likewise.
6709 (supportable_narrowing_operation): Likewise.
6710 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
6712 * varasm.c (output_constant): Likewise.
6714 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6715 Alan Hayward <alan.hayward@arm.com>
6716 David Sherwood <david.sherwood@arm.com>
6718 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
6719 so that both the length == 3 and length != 3 cases set up their
6720 own permute vectors. Add comments explaining why we know the
6721 number of elements is constant.
6722 (vect_permute_load_chain): Likewise.
6724 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6725 Alan Hayward <alan.hayward@arm.com>
6726 David Sherwood <david.sherwood@arm.com>
6728 * machmode.h (mode_nunits): Change from unsigned char to
6730 (ONLY_FIXED_SIZE_MODES): New macro.
6731 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
6732 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
6733 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
6735 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
6736 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
6737 or if measurement_type is not polynomial.
6738 * genmodes.c (ZERO_COEFFS): New macro.
6739 (emit_mode_nunits_inline): Make mode_nunits_inline return a
6741 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
6742 Use ZERO_COEFFS when emitting initializers.
6743 * data-streamer.h (bp_pack_poly_value): New function.
6744 (bp_unpack_poly_value): Likewise.
6745 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6746 for GET_MODE_NUNITS.
6747 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6748 for GET_MODE_NUNITS.
6749 * tree.c (make_vector_type): Remove temporary shim and make
6750 the real function take the number of units as a poly_uint64
6752 (build_vector_type_for_mode): Handle polynomial nunits.
6753 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
6754 * emit-rtl.c (const_vec_series_p_1): Likewise.
6755 (gen_rtx_CONST_VECTOR): Likewise.
6756 * fold-const.c (test_vec_duplicate_folding): Likewise.
6757 * genrecog.c (validate_pattern): Likewise.
6758 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
6759 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6760 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
6761 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
6762 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
6763 * rtlanal.c (subreg_get_info): Likewise.
6764 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
6765 (vect_grouped_load_supported): Likewise.
6766 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
6767 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
6768 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
6769 (simplify_const_unary_operation, simplify_binary_operation_1)
6770 (simplify_const_binary_operation, simplify_ternary_operation)
6771 (test_vector_ops_duplicate, test_vector_ops): Likewise.
6772 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
6773 instead of CONST_VECTOR_NUNITS.
6774 * varasm.c (output_constant_pool_2): Likewise.
6775 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
6776 explicit-encoded elements in the XVEC for variable-length vectors.
6778 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6780 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
6782 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6783 Alan Hayward <alan.hayward@arm.com>
6784 David Sherwood <david.sherwood@arm.com>
6786 * coretypes.h (fixed_size_mode): Declare.
6787 (fixed_size_mode_pod): New typedef.
6788 * builtins.h (target_builtins::x_apply_args_mode)
6789 (target_builtins::x_apply_result_mode): Change type to
6790 fixed_size_mode_pod.
6791 * builtins.c (apply_args_size, apply_result_size, result_vector)
6792 (expand_builtin_apply_args_1, expand_builtin_apply)
6793 (expand_builtin_return): Update accordingly.
6795 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6797 * cse.c (hash_rtx_cb): Hash only the encoded elements.
6798 * cselib.c (cselib_hash_rtx): Likewise.
6799 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
6800 CONST_VECTOR encoding.
6802 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6803 Jeff Law <law@redhat.com>
6806 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
6807 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
6808 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
6809 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
6812 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
6813 explicitly probe *sp in a noreturn function if there were any callee
6814 register saves or frame pointer is needed.
6816 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6819 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
6820 BLKmode for ternary, binary or unary expressions.
6823 * var-tracking.c (delete_vta_debug_insn): New inline function.
6824 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
6825 insns from get_insns () to NULL instead of each bb separately.
6826 Use delete_vta_debug_insn. No longer static.
6827 (vt_debug_insns_local, variable_tracking_main_1): Adjust
6828 delete_vta_debug_insns callers.
6829 * rtl.h (delete_vta_debug_insns): Declare.
6830 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
6831 instead of variable_tracking_main.
6833 2018-01-03 Martin Sebor <msebor@redhat.com>
6835 PR tree-optimization/83603
6836 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
6837 arguments past the endof the argument list in functions declared
6838 without a prototype.
6839 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
6840 Avoid checking when arguments are null.
6842 2018-01-03 Martin Sebor <msebor@redhat.com>
6845 * doc/extend.texi (attribute const): Fix a typo.
6846 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
6847 issuing -Wsuggest-attribute for void functions.
6849 2018-01-03 Martin Sebor <msebor@redhat.com>
6851 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
6852 offset_int::from instead of wide_int::to_shwi.
6853 (maybe_diag_overlap): Remove assertion.
6854 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
6855 * gimple-ssa-sprintf.c (format_directive): Same.
6856 (parse_directive): Same.
6857 (sprintf_dom_walker::compute_format_length): Same.
6858 (try_substitute_return_value): Same.
6860 2018-01-03 Jeff Law <law@redhat.com>
6863 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
6864 non-constant residual for zero at runtime and avoid probing in
6865 that case. Reorganize code for trailing problem to mirror handling
6868 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6870 PR tree-optimization/83501
6871 * tree-ssa-strlen.c (get_string_cst): New.
6872 (handle_char_store): Call get_string_cst.
6874 2018-01-03 Martin Liska <mliska@suse.cz>
6876 PR tree-optimization/83593
6877 * tree-ssa-strlen.c: Include tree-cfg.h.
6878 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
6879 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
6880 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
6882 (strlen_dom_walker::before_dom_children): Call
6883 gimple_purge_dead_eh_edges. Dump tranformation with details
6885 (strlen_dom_walker::before_dom_children): Update call by adding
6886 new argument cleanup_eh.
6887 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
6889 2018-01-03 Martin Liska <mliska@suse.cz>
6892 * cif-code.def (VARIADIC_THUNK): New enum value.
6893 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
6896 2018-01-03 Jan Beulich <jbeulich@suse.com>
6898 * sse.md (mov<mode>_internal): Tighten condition for when to use
6899 vmovdqu<ssescalarsize> for TI and OI modes.
6901 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6903 Update copyright years.
6905 2018-01-03 Martin Liska <mliska@suse.cz>
6908 * ipa-visibility.c (function_and_variable_visibility): Skip
6909 functions with noipa attribure.
6911 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6913 * gcc.c (process_command): Update copyright notice dates.
6914 * gcov-dump.c (print_version): Ditto.
6915 * gcov.c (print_version): Ditto.
6916 * gcov-tool.c (print_version): Ditto.
6917 * gengtype.c (create_file): Ditto.
6918 * doc/cpp.texi: Bump @copying's copyright year.
6919 * doc/cppinternals.texi: Ditto.
6920 * doc/gcc.texi: Ditto.
6921 * doc/gccint.texi: Ditto.
6922 * doc/gcov.texi: Ditto.
6923 * doc/install.texi: Ditto.
6924 * doc/invoke.texi: Ditto.
6926 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6928 * vector-builder.h (vector_builder::m_full_nelts): Change from
6929 unsigned int to poly_uint64.
6930 (vector_builder::full_nelts): Update prototype accordingly.
6931 (vector_builder::new_vector): Likewise.
6932 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
6933 (vector_builder::operator ==): Likewise.
6934 (vector_builder::finalize): Likewise.
6935 * int-vector-builder.h (int_vector_builder::int_vector_builder):
6936 Take the number of elements as a poly_uint64 rather than an
6938 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
6939 from unsigned int to poly_uint64.
6940 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
6941 (vec_perm_indices::new_vector): Likewise.
6942 (vec_perm_indices::length): Likewise.
6943 (vec_perm_indices::nelts_per_input): Likewise.
6944 (vec_perm_indices::input_nelts): Likewise.
6945 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
6946 number of elements per input as a poly_uint64 rather than an
6947 unsigned int. Use the original encoding for variable-length
6948 vectors, rather than clamping each individual element.
6949 For the second and subsequent elements in each pattern,
6950 clamp the step and base before clamping their sum.
6951 (vec_perm_indices::series_p): Handle polynomial element counts.
6952 (vec_perm_indices::all_in_range_p): Likewise.
6953 (vec_perm_indices_to_tree): Likewise.
6954 (vec_perm_indices_to_rtx): Likewise.
6955 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
6956 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
6957 (tree_vector_builder::new_binary_operation): Handle polynomial
6958 element counts. Return false if we need to know the number
6959 of elements at compile time.
6960 * fold-const.c (fold_vec_perm): Punt if the number of elements
6961 isn't known at compile time.
6963 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6965 * vec-perm-indices.h (vec_perm_builder): Change element type
6966 from HOST_WIDE_INT to poly_int64.
6967 (vec_perm_indices::element_type): Update accordingly.
6968 (vec_perm_indices::clamp): Handle polynomial element_types.
6969 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
6970 (vec_perm_indices::all_in_range_p): Likewise.
6971 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
6973 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
6974 polynomial vec_perm_indices element types.
6975 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
6976 * fold-const.c (fold_vec_perm): Likewise.
6977 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
6978 * tree-vect-generic.c (lower_vec_perm): Likewise.
6979 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
6980 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
6981 element type to HOST_WIDE_INT.
6983 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6984 Alan Hayward <alan.hayward@arm.com>
6985 David Sherwood <david.sherwood@arm.com>
6987 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
6988 rather than an int. Use plus_constant.
6989 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
6990 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
6992 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6993 Alan Hayward <alan.hayward@arm.com>
6994 David Sherwood <david.sherwood@arm.com>
6996 * calls.c (emit_call_1, expand_call): Change struct_value_size from
6997 a HOST_WIDE_INT to a poly_int64.
6999 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7000 Alan Hayward <alan.hayward@arm.com>
7001 David Sherwood <david.sherwood@arm.com>
7003 * calls.c (load_register_parameters): Cope with polynomial
7004 mode sizes. Require a constant size for BLKmode parameters
7005 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
7006 forces a parameter to be padded at the lsb end in order to
7007 fill a complete number of words, require the parameter size
7008 to be ordered wrt UNITS_PER_WORD.
7010 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7011 Alan Hayward <alan.hayward@arm.com>
7012 David Sherwood <david.sherwood@arm.com>
7014 * reload1.c (spill_stack_slot_width): Change element type
7015 from unsigned int to poly_uint64_pod.
7016 (alter_reg): Treat mode sizes as polynomial.
7018 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7019 Alan Hayward <alan.hayward@arm.com>
7020 David Sherwood <david.sherwood@arm.com>
7022 * reload.c (complex_word_subreg_p): New function.
7023 (reload_inner_reg_of_subreg, push_reload): Use it.
7025 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7026 Alan Hayward <alan.hayward@arm.com>
7027 David Sherwood <david.sherwood@arm.com>
7029 * lra-constraints.c (process_alt_operands): Reject matched
7030 operands whose sizes aren't ordered.
7031 (match_reload): Refer to this check here.
7033 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7034 Alan Hayward <alan.hayward@arm.com>
7035 David Sherwood <david.sherwood@arm.com>
7037 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
7038 that the mode size is in the set {1, 2, 4, 8, 16}.
7040 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7041 Alan Hayward <alan.hayward@arm.com>
7042 David Sherwood <david.sherwood@arm.com>
7044 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
7045 Use plus_constant instead of gen_rtx_PLUS.
7047 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7048 Alan Hayward <alan.hayward@arm.com>
7049 David Sherwood <david.sherwood@arm.com>
7051 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
7052 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
7053 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
7054 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
7055 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
7056 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
7057 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
7058 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
7059 * config/i386/i386.c (ix86_push_rounding): ...this new function.
7060 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
7062 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
7063 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
7064 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
7065 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
7066 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
7067 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
7068 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
7069 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
7070 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
7071 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
7073 * expr.c (emit_move_resolve_push): Treat the input and result
7074 of PUSH_ROUNDING as a poly_int64.
7075 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
7076 (emit_push_insn): Likewise.
7077 * lra-eliminations.c (mark_not_eliminable): Likewise.
7078 * recog.c (push_operand): Likewise.
7079 * reload1.c (elimination_effects): Likewise.
7080 * rtlanal.c (nonzero_bits1): Likewise.
7081 * calls.c (store_one_arg): Likewise. Require the padding to be
7082 known at compile time.
7084 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7085 Alan Hayward <alan.hayward@arm.com>
7086 David Sherwood <david.sherwood@arm.com>
7088 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
7089 Use plus_constant instead of gen_rtx_PLUS.
7091 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7092 Alan Hayward <alan.hayward@arm.com>
7093 David Sherwood <david.sherwood@arm.com>
7095 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
7098 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7099 Alan Hayward <alan.hayward@arm.com>
7100 David Sherwood <david.sherwood@arm.com>
7102 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
7103 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
7104 via stack temporaries. Treat the mode size as polynomial too.
7106 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7107 Alan Hayward <alan.hayward@arm.com>
7108 David Sherwood <david.sherwood@arm.com>
7110 * expr.c (expand_expr_real_2): When handling conversions involving
7111 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
7112 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
7113 as a poly_uint64 too.
7115 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7116 Alan Hayward <alan.hayward@arm.com>
7117 David Sherwood <david.sherwood@arm.com>
7119 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
7121 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7122 Alan Hayward <alan.hayward@arm.com>
7123 David Sherwood <david.sherwood@arm.com>
7125 * combine.c (can_change_dest_mode): Handle polynomial
7126 REGMODE_NATURAL_SIZE.
7127 * expmed.c (store_bit_field_1): Likewise.
7128 * expr.c (store_constructor): Likewise.
7129 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
7130 and polynomial REGMODE_NATURAL_SIZE.
7131 (gen_lowpart_common): Likewise.
7132 * reginfo.c (record_subregs_of_mode): Likewise.
7133 * rtlanal.c (read_modify_subreg_p): Likewise.
7135 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7136 Alan Hayward <alan.hayward@arm.com>
7137 David Sherwood <david.sherwood@arm.com>
7139 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
7140 numbers of elements.
7142 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7143 Alan Hayward <alan.hayward@arm.com>
7144 David Sherwood <david.sherwood@arm.com>
7146 * match.pd: Cope with polynomial numbers of vector elements.
7148 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7149 Alan Hayward <alan.hayward@arm.com>
7150 David Sherwood <david.sherwood@arm.com>
7152 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
7153 in a POINTER_PLUS_EXPR.
7155 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7156 Alan Hayward <alan.hayward@arm.com>
7157 David Sherwood <david.sherwood@arm.com>
7159 * omp-simd-clone.c (simd_clone_subparts): New function.
7160 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
7161 (ipa_simd_modify_function_body): Likewise.
7163 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7164 Alan Hayward <alan.hayward@arm.com>
7165 David Sherwood <david.sherwood@arm.com>
7167 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
7168 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
7169 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
7170 (expand_vector_condition, vector_element): Likewise.
7171 (subparts_gt): New function.
7172 (get_compute_type): Use subparts_gt.
7173 (count_type_subparts): Delete.
7174 (expand_vector_operations_1): Use subparts_gt instead of
7175 count_type_subparts.
7177 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7178 Alan Hayward <alan.hayward@arm.com>
7179 David Sherwood <david.sherwood@arm.com>
7181 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
7182 (vect_compile_time_alias): ...this new function. Do the calculation
7183 on poly_ints rather than trees.
7184 (vect_prune_runtime_alias_test_list): Update call accordingly.
7186 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7187 Alan Hayward <alan.hayward@arm.com>
7188 David Sherwood <david.sherwood@arm.com>
7190 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
7192 (vect_schedule_slp_instance): Likewise.
7194 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7195 Alan Hayward <alan.hayward@arm.com>
7196 David Sherwood <david.sherwood@arm.com>
7198 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
7199 constant and extern definitions for variable-length vectors.
7200 (vect_get_constant_vectors): Note that the number of units
7201 is known to be constant.
7203 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7204 Alan Hayward <alan.hayward@arm.com>
7205 David Sherwood <david.sherwood@arm.com>
7207 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
7208 of units as polynomial. Choose between WIDE and NARROW based
7211 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7212 Alan Hayward <alan.hayward@arm.com>
7213 David Sherwood <david.sherwood@arm.com>
7215 * tree-vect-stmts.c (simd_clone_subparts): New function.
7216 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
7218 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7219 Alan Hayward <alan.hayward@arm.com>
7220 David Sherwood <david.sherwood@arm.com>
7222 * tree-vect-stmts.c (vectorizable_call): Treat the number of
7223 vectors as polynomial. Use build_index_vector for
7226 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7227 Alan Hayward <alan.hayward@arm.com>
7228 David Sherwood <david.sherwood@arm.com>
7230 * tree-vect-stmts.c (get_load_store_type): Treat the number of
7231 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
7232 for variable-length vectors.
7233 (vectorizable_mask_load_store): Treat the number of units as
7234 polynomial, asserting that it is constant if the condition has
7235 already been enforced.
7236 (vectorizable_store, vectorizable_load): Likewise.
7238 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7239 Alan Hayward <alan.hayward@arm.com>
7240 David Sherwood <david.sherwood@arm.com>
7242 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
7243 of units as polynomial. Punt if we can't tell at compile time
7244 which vector contains the final result.
7246 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7247 Alan Hayward <alan.hayward@arm.com>
7248 David Sherwood <david.sherwood@arm.com>
7250 * tree-vect-loop.c (vectorizable_induction): Treat the number
7251 of units as polynomial. Punt on SLP inductions. Use an integer
7252 VEC_SERIES_EXPR for variable-length integer reductions. Use a
7253 cast of such a series for variable-length floating-point
7256 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7257 Alan Hayward <alan.hayward@arm.com>
7258 David Sherwood <david.sherwood@arm.com>
7260 * tree.h (build_index_vector): Declare.
7261 * tree.c (build_index_vector): New function.
7262 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
7263 of units as polynomial, forcibly converting it to a constant if
7264 vectorizable_reduction has already enforced the condition.
7265 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
7266 to create a {1,2,3,...} vector.
7267 (vectorizable_reduction): Treat the number of units as polynomial.
7268 Choose vectype_in based on the largest scalar element size rather
7269 than the smallest number of units. Enforce the restrictions
7272 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7273 Alan Hayward <alan.hayward@arm.com>
7274 David Sherwood <david.sherwood@arm.com>
7276 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
7277 number of units as polynomial.
7279 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7280 Alan Hayward <alan.hayward@arm.com>
7281 David Sherwood <david.sherwood@arm.com>
7283 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
7284 * target.def (autovectorize_vector_sizes): Return the vector sizes
7285 by pointer, using vector_sizes rather than a bitmask.
7286 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
7287 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
7288 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
7290 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
7291 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
7292 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
7293 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
7294 * omp-general.c (omp_max_vf): Likewise.
7295 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7296 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
7297 * tree-vect-loop.c (vect_analyze_loop): Likewise.
7298 * tree-vect-slp.c (vect_slp_bb): Likewise.
7299 * doc/tm.texi: Regenerate.
7300 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
7302 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
7303 the vector size as a poly_uint64 rather than an unsigned int.
7304 (current_vector_size): Change from an unsigned int to a poly_uint64.
7305 (get_vectype_for_scalar_type): Update accordingly.
7306 * tree.h (build_truth_vector_type): Take the size and number of
7307 units as a poly_uint64 rather than an unsigned int.
7308 (build_vector_type): Add a temporary overload that takes
7309 the number of units as a poly_uint64 rather than an unsigned int.
7310 * tree.c (make_vector_type): Likewise.
7311 (build_truth_vector_type): Take the number of units as a poly_uint64
7312 rather than an unsigned int.
7314 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7315 Alan Hayward <alan.hayward@arm.com>
7316 David Sherwood <david.sherwood@arm.com>
7318 * target.def (get_mask_mode): Take the number of units and length
7319 as poly_uint64s rather than unsigned ints.
7320 * targhooks.h (default_get_mask_mode): Update accordingly.
7321 * targhooks.c (default_get_mask_mode): Likewise.
7322 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
7323 * doc/tm.texi: Regenerate.
7325 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7326 Alan Hayward <alan.hayward@arm.com>
7327 David Sherwood <david.sherwood@arm.com>
7329 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
7330 * omp-general.c (omp_max_vf): Likewise.
7331 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
7332 (expand_omp_simd): Handle polynomial safelen.
7333 * omp-low.c (omplow_simd_context): Add a default constructor.
7334 (omplow_simd_context::max_vf): Change from int to poly_uint64.
7335 (lower_rec_simd_input_clauses): Update accordingly.
7336 (lower_rec_input_clauses): Likewise.
7338 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7339 Alan Hayward <alan.hayward@arm.com>
7340 David Sherwood <david.sherwood@arm.com>
7342 * tree-vectorizer.h (vect_nunits_for_cost): New function.
7343 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
7344 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
7345 (vect_analyze_slp_cost): Likewise.
7346 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
7347 (vect_model_load_cost): Likewise.
7349 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7350 Alan Hayward <alan.hayward@arm.com>
7351 David Sherwood <david.sherwood@arm.com>
7353 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
7354 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
7355 from an unsigned int * to a poly_uint64_pod *.
7356 (calculate_unrolling_factor): New function.
7357 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
7359 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7360 Alan Hayward <alan.hayward@arm.com>
7361 David Sherwood <david.sherwood@arm.com>
7363 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
7364 from an unsigned int to a poly_uint64.
7365 (_loop_vec_info::slp_unrolling_factor): Likewise.
7366 (_loop_vec_info::vectorization_factor): Change from an int
7368 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
7369 (vect_get_num_vectors): New function.
7370 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
7371 (vect_get_num_copies): Use vect_get_num_vectors.
7372 (vect_analyze_data_ref_dependences): Change max_vf from an int *
7373 to an unsigned int *.
7374 (vect_analyze_data_refs): Change min_vf from an int * to a
7376 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7377 than an unsigned HOST_WIDE_INT.
7378 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7379 (vect_analyze_data_ref_dependence): Change max_vf from an int *
7380 to an unsigned int *.
7381 (vect_analyze_data_ref_dependences): Likewise.
7382 (vect_compute_data_ref_alignment): Handle polynomial vf.
7383 (vect_enhance_data_refs_alignment): Likewise.
7384 (vect_prune_runtime_alias_test_list): Likewise.
7385 (vect_shift_permute_load_chain): Likewise.
7386 (vect_supportable_dr_alignment): Likewise.
7387 (dependence_distance_ge_vf): Take the vectorization factor as a
7388 poly_uint64 rather than an unsigned HOST_WIDE_INT.
7389 (vect_analyze_data_refs): Change min_vf from an int * to a
7391 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
7392 vfm1 as a poly_uint64 rather than an int. Make the same change
7393 for the returned bound_scalar.
7394 (vect_gen_vector_loop_niters): Handle polynomial vf.
7395 (vect_do_peeling): Likewise. Update call to
7396 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
7397 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
7399 * tree-vect-loop.c (vect_determine_vectorization_factor)
7400 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
7401 (vect_get_known_peeling_cost): Likewise.
7402 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
7403 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
7404 (vect_transform_loop): Likewise. Use the lowest possible VF when
7405 updating the upper bounds of the loop.
7406 (vect_min_worthwhile_factor): Make static. Return an unsigned int
7408 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
7409 polynomial unroll factors.
7410 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
7411 (vect_make_slp_decision): Likewise.
7412 (vect_supported_load_permutation_p): Likewise, and polynomial
7414 (vect_analyze_slp_cost): Handle polynomial vf.
7415 (vect_slp_analyze_node_operations): Likewise.
7416 (vect_slp_analyze_bb_1): Likewise.
7417 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7418 than an unsigned HOST_WIDE_INT.
7419 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
7420 (vectorizable_load): Handle polynomial vf.
7421 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
7423 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
7425 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7426 Alan Hayward <alan.hayward@arm.com>
7427 David Sherwood <david.sherwood@arm.com>
7429 * match.pd: Handle bit operations involving three constants
7430 and try to fold one pair.
7432 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7434 * tree-vect-loop-manip.c: Include gimple-fold.h.
7435 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
7436 niters_maybe_zero parameters. Handle other cases besides a step of 1.
7437 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
7438 Add a path that uses a step of VF instead of 1, but disable it
7440 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
7441 and niters_no_overflow parameters. Update calls to
7442 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
7443 Create a new SSA name if the latter choses to use a ste other
7444 than zero, and return it via niters_vector_mult_vf_var.
7445 * tree-vect-loop.c (vect_transform_loop): Update calls to
7446 vect_do_peeling, vect_gen_vector_loop_niters and
7447 slpeel_make_loop_iterate_ntimes.
7448 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
7449 (vect_gen_vector_loop_niters): Update declarations after above changes.
7451 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
7453 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
7454 128-bit round to integer instructions.
7455 (ceil<mode>2): Likewise.
7456 (btrunc<mode>2): Likewise.
7457 (round<mode>2): Likewise.
7459 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7461 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
7462 unaligned VSX load/store on P8/P9.
7463 (expand_block_clear): Allow the use of unaligned VSX
7464 load/store on P8/P9.
7466 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7468 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
7470 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
7471 swap associated with both a load and a store.
7473 2018-01-02 Andrew Waterman <andrew@sifive.com>
7475 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
7476 * config/riscv/riscv.md (clear_cache): Use it.
7478 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
7480 * web.c: Remove out-of-date comment.
7482 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7484 * expr.c (fixup_args_size_notes): Check that any existing
7485 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
7486 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
7487 (emit_single_push_insn): ...here.
7489 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7491 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
7492 (const_vector_encoded_nelts): New function.
7493 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
7494 (const_vector_int_elt, const_vector_elt): Declare.
7495 * emit-rtl.c (const_vector_int_elt_1): New function.
7496 (const_vector_elt): Likewise.
7497 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
7498 of CONST_VECTOR_ELT.
7500 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7502 * expr.c: Include rtx-vector-builder.h.
7503 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
7504 directly on the tree encoding.
7505 (const_vector_from_tree): Likewise.
7506 * optabs.c: Include rtx-vector-builder.h.
7507 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
7508 sequence of "u" values.
7509 * vec-perm-indices.c: Include rtx-vector-builder.h.
7510 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
7511 directly on the vec_perm_indices encoding.
7513 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7515 * doc/rtl.texi (const_vector): Describe new encoding scheme.
7516 * Makefile.in (OBJS): Add rtx-vector-builder.o.
7517 * rtx-vector-builder.h: New file.
7518 * rtx-vector-builder.c: Likewise.
7519 * rtl.h (rtx_def::u2): Add a const_vector field.
7520 (CONST_VECTOR_NPATTERNS): New macro.
7521 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
7522 (CONST_VECTOR_DUPLICATE_P): Likewise.
7523 (CONST_VECTOR_STEPPED_P): Likewise.
7524 (CONST_VECTOR_ENCODED_ELT): Likewise.
7525 (const_vec_duplicate_p): Check for a duplicated vector encoding.
7526 (unwrap_const_vec_duplicate): Likewise.
7527 (const_vec_series_p): Check for a non-duplicated vector encoding.
7528 Say that the function only returns true for integer vectors.
7529 * emit-rtl.c: Include rtx-vector-builder.h.
7530 (gen_const_vec_duplicate_1): Delete.
7531 (gen_const_vector): Call gen_const_vec_duplicate instead of
7532 gen_const_vec_duplicate_1.
7533 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
7534 (gen_const_vec_duplicate): Use rtx_vector_builder.
7535 (gen_const_vec_series): Likewise.
7536 (gen_rtx_CONST_VECTOR): Likewise.
7537 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
7538 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7539 Build a new vector rather than modifying a CONST_VECTOR in-place.
7540 (handle_special_swappables): Update call accordingly.
7541 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
7542 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7543 Build a new vector rather than modifying a CONST_VECTOR in-place.
7544 (handle_special_swappables): Update call accordingly.
7546 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7548 * simplify-rtx.c (simplify_const_binary_operation): Use
7549 CONST_VECTOR_ELT instead of XVECEXP.
7551 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7553 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
7554 the selector elements to be different from the data elements
7555 if the selector is a VECTOR_CST.
7556 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
7557 ssizetype for the selector.
7559 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7561 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
7562 before testing each element individually.
7563 * tree-vect-generic.c (lower_vec_perm): Likewise.
7565 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7567 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
7568 * selftest-run-tests.c (selftest::run_tests): Call it.
7569 * vector-builder.h (vector_builder::operator ==): New function.
7570 (vector_builder::operator !=): Likewise.
7571 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
7572 (vec_perm_indices::all_from_input_p): New function.
7573 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7574 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
7575 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
7576 instead of reading the VECTOR_CST directly. Detect whether both
7577 vector inputs are the same before constructing the vec_perm_indices,
7578 and update the number of inputs argument accordingly. Use the
7579 utility functions added above. Only construct sel2 if we need to.
7581 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7583 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
7584 the broadcast of the low byte.
7585 (expand_mult_highpart): Use an explicit encoding for the permutes.
7586 * optabs-query.c (can_mult_highpart_p): Likewise.
7587 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
7588 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7589 (vectorizable_bswap): Likewise.
7590 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
7591 explicit encoding for the power-of-2 permutes.
7592 (vect_permute_store_chain): Likewise.
7593 (vect_grouped_load_supported): Likewise.
7594 (vect_permute_load_chain): Likewise.
7596 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7598 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
7599 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
7600 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
7601 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7602 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
7603 (vect_gen_perm_mask_any): Likewise.
7605 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7607 * int-vector-builder.h: New file.
7608 * vec-perm-indices.h: Include int-vector-builder.h.
7609 (vec_perm_indices): Redefine as an int_vector_builder.
7610 (auto_vec_perm_indices): Delete.
7611 (vec_perm_builder): Redefine as a stand-alone class.
7612 (vec_perm_indices::vec_perm_indices): New function.
7613 (vec_perm_indices::clamp): Likewise.
7614 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
7615 (vec_perm_indices::new_vector): New function.
7616 (vec_perm_indices::new_expanded_vector): Update for new
7617 vec_perm_indices class.
7618 (vec_perm_indices::rotate_inputs): New function.
7619 (vec_perm_indices::all_in_range_p): Operate directly on the
7620 encoded form, without computing elided elements.
7621 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
7622 encoding. Update for new vec_perm_indices class.
7623 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
7624 the given vec_perm_builder.
7625 (expand_vec_perm_var): Update vec_perm_builder constructor.
7626 (expand_mult_highpart): Use vec_perm_builder instead of
7627 auto_vec_perm_indices.
7628 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
7629 vec_perm_indices instead of auto_vec_perm_indices. Use a single
7630 or double series encoding as appropriate.
7631 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
7632 vec_perm_indices instead of auto_vec_perm_indices.
7633 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7634 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7635 (vect_permute_store_chain): Likewise.
7636 (vect_grouped_load_supported): Likewise.
7637 (vect_permute_load_chain): Likewise.
7638 (vect_shift_permute_load_chain): Likewise.
7639 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
7640 (vect_transform_slp_perm_load): Likewise.
7641 (vect_schedule_slp_instance): Likewise.
7642 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7643 (vectorizable_mask_load_store): Likewise.
7644 (vectorizable_bswap): Likewise.
7645 (vectorizable_store): Likewise.
7646 (vectorizable_load): Likewise.
7647 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
7648 vec_perm_indices instead of auto_vec_perm_indices. Use
7649 tree_to_vec_perm_builder to read the vector from a tree.
7650 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
7651 vec_perm_builder instead of a vec_perm_indices.
7652 (have_whole_vector_shift): Use vec_perm_builder and
7653 vec_perm_indices instead of auto_vec_perm_indices. Leave the
7654 truncation to calc_vec_perm_mask_for_shift.
7655 (vect_create_epilog_for_reduction): Likewise.
7656 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
7657 from auto_vec_perm_indices to vec_perm_indices.
7658 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7659 instead of changing individual elements.
7660 (aarch64_vectorize_vec_perm_const): Use new_vector to install
7661 the vector in d.perm.
7662 * config/arm/arm.c (expand_vec_perm_d::perm): Change
7663 from auto_vec_perm_indices to vec_perm_indices.
7664 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7665 instead of changing individual elements.
7666 (arm_vectorize_vec_perm_const): Use new_vector to install
7667 the vector in d.perm.
7668 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
7669 Update vec_perm_builder constructor.
7670 (rs6000_expand_interleave): Likewise.
7671 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
7672 (rs6000_expand_interleave): Likewise.
7674 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7676 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
7677 to qimode could truncate the indices.
7678 * optabs.c (expand_vec_perm_var): Likewise.
7680 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7682 * Makefile.in (OBJS): Add vec-perm-indices.o.
7683 * vec-perm-indices.h: New file.
7684 * vec-perm-indices.c: Likewise.
7685 * target.h (vec_perm_indices): Replace with a forward class
7687 (auto_vec_perm_indices): Move to vec-perm-indices.h.
7688 * optabs.h: Include vec-perm-indices.h.
7689 (expand_vec_perm): Delete.
7690 (selector_fits_mode_p, expand_vec_perm_var): Declare.
7691 (expand_vec_perm_const): Declare.
7692 * target.def (vec_perm_const_ok): Replace with...
7693 (vec_perm_const): ...this new hook.
7694 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
7695 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
7696 * doc/tm.texi: Regenerate.
7697 * optabs.def (vec_perm_const): Delete.
7698 * doc/md.texi (vec_perm_const): Likewise.
7699 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
7700 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
7701 expand_vec_perm for constant permutation vectors. Assert that
7702 the mode of variable permutation vectors is the integer equivalent
7703 of the mode that is being permuted.
7704 * optabs-query.h (selector_fits_mode_p): Declare.
7705 * optabs-query.c: Include vec-perm-indices.h.
7706 (selector_fits_mode_p): New function.
7707 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
7708 is defined, instead of checking whether the vec_perm_const_optab
7709 exists. Use targetm.vectorize.vec_perm_const instead of
7710 targetm.vectorize.vec_perm_const_ok. Check whether the indices
7711 fit in the vector mode before using a variable permute.
7712 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
7713 vec_perm_indices instead of an rtx.
7714 (expand_vec_perm): Replace with...
7715 (expand_vec_perm_const): ...this new function. Take the selector
7716 as a vec_perm_indices rather than an rtx. Also take the mode of
7717 the selector. Update call to shift_amt_for_vec_perm_mask.
7718 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
7719 Use vec_perm_indices::new_expanded_vector to expand the original
7720 selector into bytes. Check whether the indices fit in the vector
7721 mode before using a variable permute.
7722 (expand_vec_perm_var): Make global.
7723 (expand_mult_highpart): Use expand_vec_perm_const.
7724 * fold-const.c: Includes vec-perm-indices.h.
7725 * tree-ssa-forwprop.c: Likewise.
7726 * tree-vect-data-refs.c: Likewise.
7727 * tree-vect-generic.c: Likewise.
7728 * tree-vect-loop.c: Likewise.
7729 * tree-vect-slp.c: Likewise.
7730 * tree-vect-stmts.c: Likewise.
7731 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
7733 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
7734 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
7735 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
7736 (aarch64_vectorize_vec_perm_const): ...this new function.
7737 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7738 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7739 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
7740 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
7741 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7742 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7743 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
7745 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
7746 check for NEON modes.
7747 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
7748 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
7749 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
7750 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
7752 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
7753 the old VEC_PERM_CONST conditions.
7754 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
7755 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
7756 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
7757 (ia64_vectorize_vec_perm_const_ok): Merge into...
7758 (ia64_vectorize_vec_perm_const): ...this new function.
7759 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
7760 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
7761 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
7762 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
7763 * config/mips/mips.c (mips_expand_vec_perm_const)
7764 (mips_vectorize_vec_perm_const_ok): Merge into...
7765 (mips_vectorize_vec_perm_const): ...this new function.
7766 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
7767 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
7768 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
7769 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
7770 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
7771 (rs6000_expand_vec_perm_const): Delete.
7772 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
7774 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7775 (altivec_expand_vec_perm_const_le): Take each operand individually.
7776 Operate on constant selectors rather than rtxes.
7777 (altivec_expand_vec_perm_const): Likewise. Update call to
7778 altivec_expand_vec_perm_const_le.
7779 (rs6000_expand_vec_perm_const): Delete.
7780 (rs6000_vectorize_vec_perm_const_ok): Delete.
7781 (rs6000_vectorize_vec_perm_const): New function.
7782 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
7783 an element count and rtx array.
7784 (rs6000_expand_extract_even): Update call accordingly.
7785 (rs6000_expand_interleave): Likewise.
7786 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
7787 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
7788 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
7789 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
7790 (rs6000_expand_vec_perm_const): Delete.
7791 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7792 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7793 (altivec_expand_vec_perm_const_le): Take each operand individually.
7794 Operate on constant selectors rather than rtxes.
7795 (altivec_expand_vec_perm_const): Likewise. Update call to
7796 altivec_expand_vec_perm_const_le.
7797 (rs6000_expand_vec_perm_const): Delete.
7798 (rs6000_vectorize_vec_perm_const_ok): Delete.
7799 (rs6000_vectorize_vec_perm_const): New function. Remove stray
7800 reference to the SPE evmerge intructions.
7801 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
7802 an element count and rtx array.
7803 (rs6000_expand_extract_even): Update call accordingly.
7804 (rs6000_expand_interleave): Likewise.
7805 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
7806 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
7808 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7810 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7812 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
7813 vector mode and that that mode matches the mode of the data
7815 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
7816 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
7817 directly using expand_vec_perm_1 when forcing selectors into
7819 (expand_vec_perm_var): New function, split out from expand_vec_perm.
7821 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7823 * optabs-query.h (can_vec_perm_p): Delete.
7824 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
7825 * optabs-query.c (can_vec_perm_p): Split into...
7826 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
7827 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
7828 particular selector is valid.
7829 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7830 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7831 (vect_grouped_load_supported): Likewise.
7832 (vect_shift_permute_load_chain): Likewise.
7833 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
7834 (vect_transform_slp_perm_load): Likewise.
7835 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7836 (vectorizable_bswap): Likewise.
7837 (vect_gen_perm_mask_checked): Likewise.
7838 * fold-const.c (fold_ternary_loc): Likewise. Don't take
7839 implementations of variable permutation vectors into account
7840 when deciding which selector to use.
7841 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
7842 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
7843 with a false third argument.
7844 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
7845 to test whether the constant selector is valid and can_vec_perm_var_p
7846 to test whether a variable selector is valid.
7848 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7850 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
7851 * optabs-query.c (can_vec_perm_p): Likewise.
7852 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
7853 instead of vec_perm_indices.
7854 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
7855 (vect_gen_perm_mask_checked): Likewise,
7856 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
7857 (vect_gen_perm_mask_checked): Likewise,
7859 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7861 * optabs-query.h (qimode_for_vec_perm): Declare.
7862 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
7863 (qimode_for_vec_perm): ...this new function.
7864 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
7866 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7868 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
7869 does not have a conditional at the top.
7871 2018-01-02 Richard Biener <rguenther@suse.de>
7873 * ipa-inline.c (big_speedup_p): Fix expression.
7875 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
7878 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
7881 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
7885 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
7886 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
7887 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
7888 cond_taken_branch_cost 3->4.
7890 2018-01-01 Jakub Jelinek <jakub@redhat.com>
7892 PR tree-optimization/83581
7893 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
7894 TODO_cleanup_cfg if any changes have been made.
7897 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
7898 convert_modes if target mode has the right side, but different mode
7902 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
7903 last argument when extracting from CONCAT. If either from_real or
7904 from_imag is NULL, use expansion through memory. If result is not
7905 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
7906 the parts directly to inner mode, if even that fails, use expansion
7910 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
7911 check for bswap in mode rather than HImode and use that in expand_unop
7914 Copyright (C) 2018 Free Software Foundation, Inc.
7916 Copying and distribution of this file, with or without modification,
7917 are permitted in any medium without royalty provided the copyright
7918 notice and this notice are preserved.