1 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
3 * config/nds32/nds32.opt: Change -mcmodel= default value.
5 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
6 Monk Chiang <sh.chiang04@gmail.com>
7 Chung-Ju Wu <jasonwucj@gmail.com>
9 * config/nds32/constants.md (unspec_element): New enum.
10 * config/nds32/constraints.md (Umw): New constraint.
11 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
12 * config/nds32/nds32-intrinsic.md: Likewise.
13 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
14 (nds32_valid_smw_lwm_base_p): New.
15 (nds32_output_smw_single_word): New.
16 (nds32_output_lmw_single_word): New.
17 (nds32_expand_unaligned_load): New.
18 (nds32_expand_unaligned_store): New.
19 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
20 (nds32_output_smw_single_word): Declare.
21 (nds32_output_lmw_single_word): Declare.
22 (nds32_expand_unaligned_load): Declare.
23 (nds32_expand_unaligned_store): Declare.
24 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
25 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
26 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
27 NDS32_BUILTIN_UASTORE_DW.
28 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
31 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
32 Kito Cheng <kito.cheng@gmail.com>
33 Chung-Ju Wu <jasonwucj@gmail.com>
35 * config/nds32/nds32-intrinsic.c
36 (nds32_expand_builtin_null_ftype_reg): Delete.
37 (nds32_expand_builtin_reg_ftype_imm): Ditto.
38 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
39 (nds32_read_argument): New.
40 (nds32_legitimize_target): Ditto.
41 (nds32_legitimize_argument): Ditto.
42 (nds32_check_constant_argument): Ditto.
43 (nds32_expand_unop_builtin): Ditto.
44 (nds32_expand_unopimm_builtin): Ditto.
45 (nds32_expand_binop_builtin): Ditto.
46 (nds32_builtin_decl_impl): Ditto.
47 (builtin_description): Ditto.
48 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
49 (nds32_init_builtins_impl): Ditto.
50 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
51 (nds32_builtin_decl): New.
52 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
53 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
55 2018-03-02 Jeff Law <law@redhat.com>
57 * reorg.c (stop_search_p): Handle DEBUG_INSN.
58 (redundant_insn, fill_simple_delay_slots): Likewise.
59 (fill_slots_from_thread): Likewise.
60 * resource.c (mark_referenced_resources): Likewise.
61 (mark_set_resources, find_dead_or_set_registers): Likewise.
63 2018-03-02 Jakub Jelinek <jakub@redhat.com>
65 * substring-locations.h (format_warning_va): Formatting fix for
67 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
69 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
70 * substring-locations.c: Include intl.h.
71 (format_warning_va): Turned into small wrapper around
72 format_warning_n_va, renamed to ...
73 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
74 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
76 (format_warning_at_substring_n): New function.
77 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
78 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
79 format_warning_at_substring with just a shorter name instead of
80 const function pointer.
81 (fmtwarn_n): New function.
82 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
83 appropriate, get rid of all the fmtstr temporaries, move conditionals
84 with G_() wrapped string literals directly into fmtwarn arguments,
85 cast dir.len to (int), formatting fixes.
87 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
89 * doc/invoke.texi: Remove "Cilk Plus" references.
91 2018-03-02 Jakub Jelinek <jakub@redhat.com>
92 Richard Biener <rguenther@suse.de>
95 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
96 for error or warning attributes if CALL_FROM_THUNK_P is set.
99 2018-03-02 Jakub Jelinek <jakub@redhat.com>
102 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
103 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
106 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
107 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
109 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
110 instead of -1U in last predictors element's probability member.
112 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
115 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
116 arguments if they are comparable.
118 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
120 PR tree-optimization/84634
121 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
122 masks and masked_loop_p with a single loop_masks, making sure it's
123 null for bb vectorization.
125 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
127 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
128 (vect_analyze_data_ref_access): Use loop->safe_len rather than
129 loop->force_vectorize to check whether there is no alias.
131 2018-03-02 Jakub Jelinek <jakub@redhat.com>
134 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
136 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
138 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
139 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
140 instead of a loop around prev_real_insn.
141 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
145 * config/i386/i386.c (ix86_print_operand): Use conditional
146 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
149 2018-03-02 Richard Biener <rguenther@suse.de>
151 PR tree-optimization/84427
152 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
153 (bitmap_set_subtract_values): Rewrite to handle multiple
156 (prune_clobbered_mems): Likewise.
157 (phi_translate): Take edge instead of pred/phiblock.
158 (phi_translate_1): Likewise.
159 (phi_translate_set): Likewise. Insert all translated
160 exprs for a value into the set, keeping possibly multiple
161 expressions per value.
162 (compute_antic_aux): Adjust for phi_translate changes.
163 When intersecting union the expressions and prune those
164 not in the final value set, keeping possibly multiple
165 expressions per value. Do not use value-insertion
166 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
167 all expressions. Add verification that the value-sets
168 only shrink during iteration.
169 (compute_partial_antic_aux): Adjust for the phi_translate changes.
170 (do_pre_regular_insertion): Likewise.
171 (do_pre_partial_partial_insertion): Likewise.
173 2018-03-02 Richard Biener <rguenther@suse.de>
176 * config/darwin.c (saved_debug_info_level): New static global.
177 (darwin_asm_lto_start): Disable debug info generation for LTO out.
178 (darwin_asm_lto_end): Restore debug info generation settings.
180 2018-03-01 Martin Liska <mliska@suse.cz>
183 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
186 2018-03-01 Richard Biener <rguenther@suse.de>
189 * dwarf2out.c (gen_variable_die): Properly handle late VLA
190 type annotation with LTO when debug was disabled at compile-time.
192 2018-03-01 Matthew Fortune <mfortune@gmail.com>
194 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
196 (mips_final_postscan_insn): Likewise.
198 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
200 PR rtl-optimization/84528
201 * alias.c (init_alias_target): Add commentary.
202 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
203 a unique base value if the frame pointer is not eliminated
204 to the stack pointer.
206 2018-03-01 Tom de Vries <tom@codesourcery.com>
208 PR rtl-optimization/83327
209 * lra-int.h (hard_regs_spilled_into): Declare.
210 * lra.c (hard_regs_spilled_into): Define.
211 (init_reg_info): Init hard_regs_spilled_into.
212 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
213 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
214 (process_bb_lives): Handle hard_regs_spilled_into.
215 (lra_create_live_ranges_1): Before doing liveness propagation, clear
216 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
218 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
220 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
221 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
222 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
223 * config/rs6000/aix72.h: New file.
225 2018-02-28 Jakub Jelinek <jakub@redhat.com>
227 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
228 instead of warning_at with conditional singular and plural messages
232 * stor-layout.c (update_alignment_for_field): For
233 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
234 && !DECL_PACKED (field), do the alignment update, just use
235 only desired_align instead of MAX (type_align, desired_align)
237 (place_field): Don't do known_align < desired_align handling
238 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
239 is non-NULL, instead do it after rli->prev_field handling and
240 only if not within a bitfield word. For DECL_PACKED (field)
241 use type_align of BITS_PER_UNIT.
243 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
245 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
246 superfluous parentheses and trailing spaces.
248 2018-02-28 Richard Biener <rguenther@suse.de>
250 PR tree-optimization/84584
251 * graphite-scop-detection.c (scop_detection::add_scop): Discard
252 SCoPs with fake exit edge.
254 2018-02-28 Martin Liska <mliska@suse.cz>
257 * timevar.c (timer::print): Fix format to properly print 100%
260 2018-02-28 Richard Biener <rguenther@suse.de>
263 * genmatch.c (capture_info::walk_match): Do not mark
264 captured expressions without operands as expr_p given
265 they act more like predicates and should be subject to
266 "lost tail" side-effect preserving.
268 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
270 PR rtl-optimization/81611
271 * auto-inc-dec.c (attempt_change): Move dead note from
272 mem_insn if it's the next use of regno
273 (find_address): Take address use of reg holding
274 non-incremented value. Add parm to limit search to the named
276 (merge_in_block): Attempt to use a mem insn that is the next
277 use of the original regno.
279 2018-02-27 Martin Sebor <msebor@redhat.com>
282 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
283 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
285 2018-02-27 Martin Sebor <msebor@redhat.com>
288 * diagnostic-core.h (warning_n, error_n, inform_n): Change
289 n argument to unsigned HOST_WIDE_INT.
290 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
291 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
292 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
293 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
295 2018-02-27 Richard Biener <rguenther@suse.de>
297 PR tree-optimization/84512
298 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
299 Do not use the estimate returned from record_stmt_cost for
300 the scalar iteration cost but sum properly using add_stmt_cost.
302 2018-02-27 Richard Biener <rguenther@suse.de>
304 PR tree-optimization/84466
305 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
306 Adjust last change to less strictly validate use operands.
308 2018-02-27 Martin Liska <mliska@suse.cz>
310 PR gcov-profile/84548
311 * gcov.c (process_file): Allow partial overlap and consider it
312 also as group functions.
313 (output_lines): Properly calculate range of lines for a group.
315 2018-02-27 Martin Liska <mliska@suse.cz>
317 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
318 'ggc' suffixes. Change first column width.
319 (timer::print): Fix formatting of the column.
321 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
323 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
324 preserve inline entry blocks for the sake of debug inline
325 entry point markers alone.
326 (remove_unused_locals): Suggest in comments a better place to
327 force the preservation of inline entry blocks that are
328 otherwise unused, but do not preserve them.
330 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
332 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
334 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
337 * config/i386/constraints.md (Bs): Replace
338 ix86_indirect_branch_register with
339 TARGET_INDIRECT_BRANCH_REGISTER.
341 * config/i386/i386.md (indirect_jump): Likewise.
342 (tablejump): Likewise.
343 (*sibcall_memory): Likewise.
344 (*sibcall_value_memory): Likewise.
345 Peepholes of indirect call and jump via memory: Likewise.
346 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
347 (*sibcall_value_GOT_32): Likewise.
348 * config/i386/predicates.md (indirect_branch_operand): Likewise.
349 (GOT_memory_operand): Likewise.
350 (call_insn_operand): Likewise.
351 (sibcall_insn_operand): Likewise.
352 (GOT32_symbol_operand): Likewise.
353 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
355 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
357 PR rtl-optimization/83496
358 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
359 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
360 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
361 redundant insn, if any.
362 (relax_delay_slots): Likewise.
363 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
365 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
367 PR tree-optimization/83965
368 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
369 that grouped statements are part of a reduction chain. Return
370 true if the statement is not marked as a reduction itself but
372 (vect_recog_dot_prod_pattern): Don't check whether the statement
373 is part of a group here.
374 (vect_recog_sad_pattern): Likewise.
375 (vect_recog_widen_sum_pattern): Likewise.
377 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
380 * final.c (rest_of_clean_state): Also look for calls inside sequences.
382 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
385 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
387 (ix86_output_indirect_function_return): New prototype.
388 (ix86_split_simple_return_pop_internal): Likewise.
389 * config/i386/i386.c (indirect_return_via_cx): New.
390 (indirect_return_via_cx_bnd): Likewise.
391 (indirect_thunk_name): Handle return va CX_REG.
392 (output_indirect_thunk_function): Create alias for
393 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
394 (ix86_output_indirect_jmp): Remove the bool argument.
395 (ix86_output_indirect_function_return): New function.
396 (ix86_split_simple_return_pop_internal): Likewise.
397 * config/i386/i386.md (*indirect_jump): Don't pass false
398 to ix86_output_indirect_jmp.
399 (*tablejump_1): Likewise.
400 (simple_return_pop_internal): Change it to define_insn_and_split.
401 Call ix86_split_simple_return_pop_internal to split it for
403 (simple_return_indirect_internal): Call
404 ix86_output_indirect_function_return instead of
405 ix86_output_indirect_jmp.
407 2018-02-26 Jakub Jelinek <jakub@redhat.com>
410 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
411 memset and value initialization afterwards.
413 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
415 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
417 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
420 * common/config/aarch64/aarch64-common.c
421 (aarch_option_optimization_table[]): Switch
422 off fomit-frame-pointer
424 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
425 Chung-Ju Wu <jasonwucj@gmail.com>
427 * config/nds32/nds32-multiple.md (load_multiple): Disallow
429 (store_multiple): Ditto.
431 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
433 * config.gcc: Add --with-cpu support for nds32 target.
434 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
435 * config/nds32/nds32.opt: Add -mcpu= option.
437 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
439 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
440 isel=yes): Warn for these deprecated options.
442 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
444 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
445 ISA_2_5_MASKS_EMBEDDED.
447 2018-02-23 Jakub Jelinek <jakub@redhat.com>
449 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
450 p->max as pointers rather than using iterative_hash_expr.
452 2018-02-23 Carl Love <cel@us.ibm.com>
454 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
455 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
457 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
458 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
459 P8V_BUILTIN_VEC_VUNSIGNED2.
461 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
464 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
465 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
467 * lra-constraints.c (curr_insn_transform): Set up
468 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
470 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
471 * lra-eliminations.c (spill_pseudos): Ditto.
472 (process_insn_for_elimination): Ditto.
473 * lra-lives.c (reg_early_clobber_p): Use the new macros.
474 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
475 LRA_NON_CLOBBERED_ALT.
477 2018-02-22 Martin Sebor <msebor@redhat.com>
479 PR tree-optimization/84480
480 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
481 to maybe_diag_stxncpy_trunc. Call it.
482 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
483 from gimple_fold_builtin_strcpy. Print inlining stack.
484 (handle_builtin_stxncpy): Print inlining stack.
485 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
487 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
490 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
491 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
492 and -fcheck-pointer-bounds are used together.
493 (indirect_thunk_prefix): New enum.
494 (indirect_thunk_need_prefix): New function.
495 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
496 "_nt" instead of "_bnd" for NOTRACK prefix.
497 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
498 (output_indirect_thunk_function): Likewise.
500 (ix86_code_end): Update output_indirect_thunk_function calls.
501 (ix86_output_indirect_branch_via_reg): Replace
502 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
503 (ix86_output_indirect_branch_via_push): Likewise.
504 (ix86_output_function_return): Likewise.
505 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
506 incompatible with -fcf-protection=branch and
507 -fcheck-pointer-bounds.
509 2018-02-22 Steve Ellcey <sellcey@cavium.com>
512 * config/aarch64/aarch64.c (aarch64_print_address_internal):
513 Change gcc_assert call to output_operand_lossage.
515 2018-02-22 Steve Ellcey <sellcey@cavium.com>
517 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
519 2018-02-22 DJ Delorie <dj@redhat.com>
520 Sebastian Perta <sebastian.perta@renesas.com>
521 Oleg Endo <olegendo@gcc.gnu.org>
523 * config/rx/rx.c (rx_rtx_costs): New function.
524 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
526 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
528 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
530 2018-02-22 Martin Liska <mliska@suse.cz>
533 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
534 Add "native" as a possible value.
536 2018-02-22 Martin Liska <mliska@suse.cz>
539 * config/i386/i386.c (ix86_option_override_internal):
540 Add "native" as a possible value for -march and -mtune.
542 2018-02-22 Jakub Jelinek <jakub@redhat.com>
545 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
546 to all type variants.
548 PR tree-optimization/84503
549 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
550 width as info->bitpos + info->bitsize - start.
551 (merged_store_group::merge_overlapping): Simplify width computation.
552 (check_no_overlap): New function.
553 (imm_store_chain_info::try_coalesce_bswap): Compute expected
554 start + width and last_order of the group, fail if check_no_overlap
556 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
557 to group if check_no_overlap fails.
559 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
561 * config/rs6000/altivec.md: Delete contraint arguments to
562 define_expand, define_split, and define_peephole2, and in
563 define_insn_and_split if always unused.
564 * config/rs6000/darwin.md: Ditto.
565 * config/rs6000/dfp.md: Ditto.
566 * config/rs6000/rs6000.md: Ditto.
567 * config/rs6000/sync.md: Ditto.
568 * config/rs6000/vector.md: Ditto.
569 * config/rs6000/vsx.md: Ditto.
571 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
573 * config/rs6000/altivec.md: Write output control strings as braced
574 blocks instead of double-quoted strings.
575 * config/rs6000/darwin.md: Ditto.
576 * config/rs6000/rs6000.md: Ditto.
577 * config/rs6000/vector.md: Ditto.
578 * config/rs6000/vsx.md: Ditto.
580 2018-02-21 Jason Merrill <jason@redhat.com>
582 PR c++/84314 - ICE with templates and fastcall attribute.
583 * attribs.c (build_type_attribute_qual_variant): Remove assert.
585 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
587 * ipa-cp.c (determine_versionability): Fix comment typos.
589 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
592 * ipa-cp.c (determine_versionability): Do not version functions caling
595 2018-02-21 Martin Liska <mliska@suse.cz>
598 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
599 Add "native" as a possible value.
600 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
601 the macro when native cpu detection is available.
603 2018-02-21 Martin Liska <mliska@suse.cz>
606 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
607 Add "native" as a possible value.
608 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
609 when native cpu detection is available.
611 2018-02-21 Jakub Jelinek <jakub@redhat.com>
612 Martin Sebor <msebor@redhat.com>
614 PR tree-optimization/84478
615 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
617 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
618 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
619 support which is conservatively correct, for 2 only stay conservative
620 for maxlen. Formatting and comment capitalization fixes. Add STRICT
621 argument to the 2 argument get_range_strlen, adjust 6 arg
622 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
624 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
625 (gimple_fold_builtin_strlen): Pass true as last argument to
628 2018-02-20 Martin Sebor <msebor@redhat.com>
631 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
632 (builtin_memref::set_base_and_offset): Same. Handle inner references.
633 (builtin_memref::builtin_memref): Factor out parts into
634 set_base_and_offset and call it.
636 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
639 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
640 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
641 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
642 search at the associated MODE_INT.
644 2018-02-20 Jeff Law <law@redhat.com>
647 PR tree-optimization/81592
649 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
650 for range data rather than using global data.
651 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
652 range data rather than using global data.
653 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
654 pass it to children as needed.
655 (struct directive::fmtresult): Similarly.
656 (struct directive::set_width): Similarly.
657 (struct directive::set_precision): Similarly.
658 (format_integer, format_directive, parse_directive): Similarly.
659 (format_none): Accept unnamed vr_values parameter.
660 (format_percent, format_floating, format_character): Similarly.
661 (format_string, format_plain): Similarly.
662 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
663 the EVRP range analyzer for range data rather than using global data.
664 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
665 gimple-ssa-evrp-analyze.h
666 (class sprintf_dom_walker): Add after_dom_children member function.
667 Add evrp_range_analyzer member.
668 (sprintf_dom_walker::before_dom_children): Call into the EVRP
669 range analyzer as needed.
670 (sprintf_dom_walker::after_dom_children): New member function.
671 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
673 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
674 (evrp_range_analyzer::pop_to_marker): Likewise.
676 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
678 PR tree-optimization/84419
679 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
680 with the required type if its current type is compatible but
683 2018-02-20 Jakub Jelinek <jakub@redhat.com>
686 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
689 2018-02-20 Martin Liska <mliska@suse.cz>
692 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
693 possible values if we don't have a hint.
695 2018-02-20 Martin Liska <mliska@suse.cz>
699 * final.c (shorten_branches): Build align_tab array with one
701 * opts.c (finish_options): Add alignment option limit check.
702 (MAX_CODE_ALIGN): Likewise.
703 (MAX_CODE_ALIGN_VALUE): Likewise.
704 * doc/invoke.texi: Document maximum allowed option value for
705 all -falign-* options.
707 2018-02-19 Jakub Jelinek <jakub@redhat.com>
710 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
711 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
712 * var-tracking.c (emit_note_insn_var_location): Remove all references
713 to NOTE_INSN_CALL_ARG_LOCATION.
714 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
715 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
716 Use copy_rtx_if_shared.
717 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
718 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
719 (dwarf2out_var_location): Remove handling of
720 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
722 * final.c (final_scan_insn): Remove all references to
723 NOTE_INSN_CALL_ARG_LOCATION.
724 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
725 before dumping final insns.
726 * except.c (emit_note_eh_region_end): Remove all references to
727 NOTE_INSN_CALL_ARG_LOCATION.
728 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
729 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
730 * config/arc/arc.c (hwloop_optimize): Likewise.
731 * config/arm/arm.c (create_fix_barrier): Likewise.
732 * config/s390/s390.c (s390_chunkify_start): Likewise.
733 * config/sh/sh.c (find_barrier): Likewise.
734 * config/i386/i386.c (rest_of_insert_endbranch,
735 ix86_seh_fixup_eh_fallthru): Likewise.
736 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
737 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
738 * config/frv/frv.c (frv_function_prologue): Likewise.
739 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
741 (note_outside_basic_block_p): Remove all references to
742 NOTE_INSN_CALL_ARG_LOCATION.
743 * gengtype.c (adjust_field_rtx_def): Likewise.
744 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
746 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
747 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
750 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
753 PR tree-optimization/84452
754 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
755 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
758 2018-02-19 Martin Liska <mliska@suse.cz>
761 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
763 2018-02-19 Martin Liska <mliska@suse.cz>
764 Richard Sandiford <richard.sandiford@linaro.org>
766 PR tree-optimization/82491
767 * gimple-fold.c (get_base_constructor): Make earlier bail out
770 2018-02-19 Carl Love <cel@us.ibm.com>
772 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
773 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
775 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
778 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
780 * config/rl78/rl78.md (movdf): New define expand.
782 2018-02-19 Martin Liska <mliska@suse.cz>
785 * doc/invoke.texi: Fix typo.
786 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
788 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
790 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
791 handle rs6000_single_float and rs6000_double_float specially for
794 2018-02-16 Jeff Law <law@redhat.com>
796 * config/rx/rx.c (add_pop_cfi_notes): New function.;
799 2018-02-16 Jakub Jelinek <jakub@redhat.com>
802 * ipa-inline.c (inline_small_functions): Fix a typo.
804 2018-02-16 Nathan Sidwell <nathan@acm.org>
806 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
808 2018-02-16 Carl Love <cel@us.ibm.com>
810 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
811 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
812 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
813 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
814 expansion to P8V_BUILTIN_VEC_FLOAT2.
816 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
818 PR rtl-optimization/70023
819 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
820 src_regno into account.
822 2018-02-16 Carl Love <cel@us.ibm.com>
824 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
825 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
826 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
827 * config/rs6000/rs6000.c: Remove case statements for
828 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
829 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
830 and P9V_BUILTIN_VEC_VINSERT4B.
831 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
832 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
833 * config/rs6000/vsx.md:
834 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
837 2018-02-16 Carl Love <cel@us.ibm.com>
839 * config/rs6000/altivec.h: Add builtin names vec_extract4b
841 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
843 * config/rs6000/rs6000-c.c: Add the definitions for
844 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
845 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
846 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
847 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
848 definition for insert4b and define insn *insert3b_internal.
849 * doc/extend.texi: Add documentation for vec_extract4b.
851 2018-02-16 Nathan Sidwell <nathan@acm.org>
853 * doc/extend.texi (Backwards Compatibility): Mention friend
854 injection. Note for-scope is deprecated.
855 * doc/invoke.texi (-ffriend-injection): Deprecate.
857 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
859 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
860 that moved to I2, also allow destinations that are a paradoxical
861 subreg (instead of a normal reg).
863 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
866 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
869 2018-02-16 Richard Biener <rguenther@suse.de>
871 PR tree-optimization/84037
872 PR tree-optimization/84016
874 * config/i386/i386.c (ix86_builtin_vectorization_cost):
875 Adjust vec_construct for the fact we need additional higher latency
876 128bit inserts for AVX256 and AVX512 vector builds.
877 (ix86_add_stmt_cost): Scale vector construction cost for
880 2018-02-16 Richard Biener <rguenther@suse.de>
882 PR tree-optimization/84417
883 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
884 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
885 (non_rewritable_lvalue_p): Likewise, use poly-ints.
887 2018-02-16 Martin Liska <mliska@suse.cz>
890 * internal-fn.def (ASAN_CHECK): Set proper flags.
891 (ASAN_MARK): Likewise.
893 2018-02-16 Julia Koval <julia.koval@intel.com>
895 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
898 2018-02-16 Jakub Jelinek <jakub@redhat.com>
901 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
902 Use ++iter rather than iter++ for std::list iterators.
903 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
904 defer deleting them until all nodes in the forest are processed. Do
905 free even leaf nodes. Change to_process into auto_vec.
908 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
909 * vec.h (vec_default_construct): Use memset instead of placement new
910 if BROKEN_VALUE_INITIALIZATION is defined.
911 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
912 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
915 PR rtl-optimization/83723
916 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
917 * lra.c (lra_substitute_pseudo): Likewise. If true, use
918 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
920 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
922 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
924 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
926 PR rtl-optimization/81443
927 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
928 from inner REGs to paradoxical SUBREGs.
930 2018-02-16 Richard Biener <rguenther@suse.de>
932 PR tree-optimization/84399
933 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
934 For operands we can analyze at their definition make sure we can
935 analyze them at each use as well.
937 2018-02-16 Richard Biener <rguenther@suse.de>
939 PR tree-optimization/84190
940 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
941 volatile accesses if the decl isn't volatile.
943 2018-02-15 Jason Merrill <jason@redhat.com>
945 PR c++/84314 - ICE with templates and fastcall attribute.
946 * attribs.c (build_type_attribute_qual_variant): Don't clobber
947 TYPE_CANONICAL on an existing type.
949 2018-02-15 Jakub Jelinek <jakub@redhat.com>
951 PR tree-optimization/84383
952 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
953 dstoff nor call operand_equal_p if dstbase is NULL.
955 PR tree-optimization/84334
956 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
957 also a CONSTANT_CLASS_P, punt.
959 2018-02-14 Jim Wilson <jimw@sifive.com>
961 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
962 first SMALL_OPERAND check. New local min_second_step. Move assert
963 to where locals are set. Add TARGET_RVC support.
964 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
966 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
968 * doc/invoke.texi: Correct -Wformat-overflow code sample.
970 2018-02-14 Martin Sebor <msebor@redhat.com>
972 PR tree-optimization/83698
973 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
974 arrays constrain the offset range to their bounds.
975 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
976 (builtin_access::overlap): Avoid setting the size of overlap if it's
978 (maybe_diag_overlap): Also consider arrays when deciding what values
979 of offsets to include in diagnostics.
981 2018-02-14 Martin Sebor <msebor@redhat.com>
984 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
985 that correspond to the kind of a declaration.
987 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
990 * config/pa/pa.md: Load address of PIC label using the linkage table
991 if the label is nonlocal.
993 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
995 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
996 warning message if user requests -maltivec=be.
997 * doc/invoke.texi: Document deprecation of -maltivec=be.
999 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
1002 * config/rs6000/rs6000-c.c: Update definitions for
1003 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
1004 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
1006 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1009 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
1010 add _get_ssp intrinsics. Remove argument from
1011 __builtin_ia32_rdssp[d|q].
1012 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
1013 * config/i386/i386-builtin.def: Remove argument from
1014 __builtin_ia32_rdssp[d|q].
1015 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
1016 ix86_expand_special_args_builtin for _rdssp[d|q].
1017 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
1018 Clear register before usage.
1019 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
1020 Add documentation for new _get_ssp and _inc_ssp intrinsics.
1022 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
1024 PR tree-optimization/84357
1025 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
1026 operand 1 of an ARRAY_REF too.
1028 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
1031 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
1032 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1034 (set_of_reg): New struct.
1035 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
1036 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
1037 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1039 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
1040 Split into bitclr, bitset, bitinvert patterns if appropriate.
1041 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
1042 use rx_fuse_in_memory_bitop.
1043 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
1044 to named insn, correct maximum insn length.
1046 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
1049 * machmode.def: Define a complex mode for PARTIAL_INT.
1050 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
1052 * doc/rtl.texi: Document CSPImode.
1053 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
1055 (msp430_hard_regno_nregs_with_padding): Likewise.
1057 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
1060 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
1062 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
1064 PR rtl-optimization/84169
1065 * combine.c (try_combine): New variable split_i2i3. Set it to true if
1066 we generated a parallel as new i3 and we split that to new i2 and i3
1067 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
1068 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
1069 those to i2, not i1. Partially rewrite this scan code.
1071 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1074 * stor-layout.c (place_field): For variable length fields, adjust
1075 offset_align afterwards not just based on the field's alignment,
1076 but also on the size.
1079 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
1080 of exps and logs in the use_exp2 case.
1082 2018-02-13 Jeff Law <law@redhat.com>
1084 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
1087 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
1090 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
1094 * common.opt (gas-loc-support, gas-locview-support): New.
1095 (ginline-points, ginternal-reset-location-views): New.
1096 * doc/invoke.texi: Document them. Use @itemx where intended.
1097 (gvariable-location-views): Adjust.
1098 * target.def (reset_location_view): New.
1099 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
1100 (TARGET_RESET_LOCATION_VIEW): New.
1101 * doc/tm.texi: Rebuilt.
1102 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
1103 (dwarf2out_default_as_locview_support): New.
1104 (output_asm_line_debug_info): Use option variables.
1105 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
1106 (output_loc_list): Likewise.
1107 (add_high_low_attributes): Check option variables.
1108 Don't output entry view attribute in strict mode.
1109 (gen_inlined_subroutine_die): Check option variables.
1110 (dwarf2out_inline_entry): Likewise.
1111 (init_sections_and_labels): Likewise.
1112 (dwarf2out_early_finish): Likewise.
1113 (maybe_reset_location_view): New, from...
1114 (dwarf2out_var_location): ... here. Call it.
1115 * debug.h (dwarf2out_default_as_loc_support): Declare.
1116 (dwarf2out_default_as_locview_support): Declare.
1117 * hooks.c (hook_int_rtx_insn_0): New.
1118 * hooks.h (hook_int_rtx_insn_0): Declare.
1119 * toplev.c (process_options): Take -gas-loc-support and
1120 -gas-locview-support from dwarf2out. Enable
1121 -gvariable-location-views by default only with locview
1122 assembler support. Enable -ginternal-reset-location-views by
1123 default only if the target defines the corresponding hook.
1124 Enable -ginline-points by default if location views are
1125 enabled; force it disabled if statement frontiers are
1127 * tree-inline.c (expand_call_inline): Check option variables.
1128 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1130 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
1132 PR tree-optimization/84321
1133 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
1134 handling. Also check whether the anti-range contains any values
1135 that satisfy the mask; switch to a VR_RANGE if not.
1137 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
1140 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
1142 2018-02-13 Martin Jambor <mjambor@suse.cz>
1145 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
1146 of call statements, also set location of a load to a temporary.
1148 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
1150 * config/rl78/rl78.c (add_vector_labels): New function.
1151 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
1152 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
1153 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
1154 which checks that no arguments are passed.
1155 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
1156 * doc/extend.texi: Documentation for the new attribute.
1158 2018-02-13 Andreas Schwab <schwab@suse.de>
1160 * config/riscv/linux.h (CPP_SPEC): Define.
1162 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1165 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
1166 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
1167 OPTION_MASK_ISA_AES as first argument to def_builtin_const
1168 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
1169 instead of OPTION_MASK_ISA_PCLMUL as first argument to
1170 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
1171 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
1172 temporarily for AES and PCLMUL builtins.
1174 PR tree-optimization/84339
1175 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
1176 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
1180 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
1181 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
1182 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
1184 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
1186 * omp-simd-clone.h: New file.
1187 * omp-simd-clone.c: Include omp-simd-clone.h.
1188 (expand_simd_clones): No longer static.
1189 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
1190 cgraph.h and omp-simd-clone.h.
1191 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
1192 (vect_recog_widen_shift_pattern): Formatting fix.
1193 (vect_pattern_recog_1): Don't check optab for calls.
1196 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
1197 operands[2] into a REG before using gen_lowpart on it.
1199 2018-02-12 Jeff Law <law@redhat.com>
1202 * config/sh/sh.c (find_barrier): Consider a sibling call
1205 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
1206 successfully back substituting a reg.
1208 2018-02-12 Richard Biener <rguenther@suse.de>
1210 PR tree-optimization/84037
1211 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
1212 parameter, move visited init to caller.
1213 (vect_slp_analyze_operations): Separate cost from validity
1214 check, initialize visited once for all instances.
1215 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
1217 * tree-vect-stmts.c (vect_model_simple_cost): Make early
1219 (vect_model_promotion_demotion_cost): Likewise.
1220 (vectorizable_bswap): Guard cost modeling with !slp_node
1221 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
1223 (vectorizable_call): Likewise.
1224 (vectorizable_conversion): Likewise.
1225 (vectorizable_assignment): Likewise.
1226 (vectorizable_shift): Likewise.
1227 (vectorizable_operation): Likewise.
1228 (vectorizable_store): Likewise.
1229 (vectorizable_load): Likewise.
1230 (vectorizable_condition): Likewise.
1231 (vectorizable_comparison): Likewise.
1233 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
1236 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
1237 (ASAN_MARK): Fix fnspec to account for return value, change pointer
1238 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
1240 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1243 * params.def (inline-min-speedup): Increase from 8 to 15.
1244 (max-inline-insns-auto): Decrease from 40 to 30.
1245 * ipa-split.c (consider_split): Add some buffer for function to
1246 be considered inlining candidate.
1247 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
1250 2018-02-12 Richard Biener <rguenther@suse.de>
1252 PR tree-optimization/84037
1253 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
1254 matched stmts if we cannot swap the non-matched ones.
1256 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1258 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
1259 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
1260 _mm_maskz_scalef_round_ss): New intrinsics.
1261 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
1262 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
1263 __builtin_ia32_scalefss_round): Remove.
1264 (__builtin_ia32_scalefsd_mask_round,
1265 __builtin_ia32_scalefss_mask_round): New intrinsics.
1266 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
1267 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
1268 ((match_operand:VF_128 2 "<round_nimm_predicate>"
1269 "<round_constraint>")): Changed to ...
1270 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
1271 "<round_scalar_constraint>")): ... this.
1272 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
1273 %0, %1, %2<round_op3>}"): Changed to ...
1274 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
1275 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
1276 %2<round_scalar_mask_op3>}"): ... this.
1277 * config/i386/subst.md (round_scalar_nimm_predicate): New.
1279 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1281 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
1282 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
1283 (_mm_maskz_sqrt_round_ss): New intrinsics.
1284 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
1285 (__builtin_ia32_sqrtsd_mask_round)
1286 (__builtin_ia32_sqrtss_mask_round): New builtins.
1287 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
1288 (__builtin_ia32_sqrtss_round): Remove.
1289 (__builtin_ia32_sqrtsd_mask_round)
1290 (__builtin_ia32_sqrtss_mask_round): New builtins.
1291 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
1292 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
1293 ((match_operand:VF_128 1 "vector_operand"
1294 "xBm,<round_constraint>")): Changed to ...
1295 ((match_operand:VF_128 1 "vector_operand"
1296 "xBm,<round_scalar_constraint>")): ... this.
1297 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
1298 %0, %2, %<iptr>1<round_op3>}): Changed to ...
1299 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
1300 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
1301 %<iptr>1<round_scalar_mask_op3>}): ... this.
1302 ((set_attr "prefix" "<round_prefix>")): Changed to ...
1303 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
1305 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
1308 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
1309 Cast vec_cmpeq result to correct type.
1310 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
1311 Cast vec_cmpgt result to correct type.
1313 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
1315 * final.c (final_scan_insn_1): Renamed from...
1316 (final_scan_insn): ... this. New wrapper, to recover
1317 seen from the outermost call in recursive ones.
1318 * config/sparc/sparc.c (output_return): Drop seen from call.
1319 (output_sibcall): Likewise.
1320 * config/visium/visium.c (output_branch): Likewise.
1322 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
1324 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
1327 2018-02-10 Alan Modra <amodra@gmail.com>
1330 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
1331 Specify LR as an input.
1333 2018-02-10 Jakub Jelinek <jakub@redhat.com>
1336 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
1337 remove_member_access_dummy_vars): New functions.
1338 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
1339 lower_omp_1, execute_lower_omp): Use them.
1341 PR rtl-optimization/84308
1342 * shrink-wrap.c (spread_components): Release todo vector.
1344 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
1346 PR rtl-optimization/57193
1347 * ira-color.c (struct allocno_color_data): Add member
1348 conflict_allocno_hard_prefs.
1349 (update_conflict_allocno_hard_prefs): New.
1350 (bucket_allocno_compare_func): Add a preference based on
1351 conflict_allocno_hard_prefs.
1352 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
1353 (color_allocnos): Remove a dead code. Initiate
1354 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
1356 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1359 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
1360 constraint from =wa to wa. Avoid a subreg on the output operand,
1361 instead use a pseudo and subreg it in a move.
1362 (p9_xxbrd_<mode>): Changed to ...
1363 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
1364 (p9_xxbrd_v2df): New expander.
1365 (p9_xxbrw_<mode>): Changed to ...
1366 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
1367 (p9_xxbrw_v4sf): New expander.
1369 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1371 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
1373 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
1376 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
1377 multiply in 32-bit mode.
1378 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
1379 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
1382 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1384 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
1385 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
1386 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
1387 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
1389 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
1392 * dwarf2out.c (is_trivial_indirect_ref): New function.
1393 (dwarf2out_late_global_decl): Do not generate a location
1394 attribute for variables that have a non-trivial DECL_VALUE_EXPR
1395 and that are not defined in the current unit.
1397 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1399 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
1400 instead of a libcall for UNORDERED.
1402 2018-02-09 Tamar Christina <tamar.christina@arm.com>
1405 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
1406 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
1408 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1411 * config/s390/s390.c (s390_set_current_function): Invoke
1412 s390_indirect_branch_settings also if fndecl didn't change.
1414 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1416 * config/rs6000/rs6000.md (blockage): Set length to zero.
1418 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1420 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
1422 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1425 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
1426 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
1430 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
1431 PARALLEL incoming that failed vt_get_decl_and_offset check.
1434 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
1435 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
1437 (get_variable_section): For decls in named .bss* sections pass true as
1438 second argument to bss_initializer_p.
1440 2018-02-09 Marek Polacek <polacek@redhat.com>
1441 Jakub Jelinek <jakub@redhat.com>
1444 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
1445 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
1446 Sync some changes from cxx_fold_indirect_ref.
1448 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1450 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
1452 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
1453 (BLOCK_INLINE_ENTRY_LABEL): New.
1454 (dwarf2out_var_location): Disregard inline entry markers.
1455 (inline_entry_data): New struct.
1456 (inline_entry_data_hasher): New hashtable type.
1457 (inline_entry_data_hasher::hash): New.
1458 (inline_entry_data_hasher::equal): New.
1459 (inline_entry_data_table): New variable.
1460 (add_high_low_attributes): Add DW_AT_entry_pc and
1461 DW_AT_GNU_entry_view attributes if a pending entry is found
1462 in inline_entry_data_table. Add old entry_pc attribute only
1463 if debug nonbinding markers are disabled.
1464 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
1465 markers are enabled.
1466 (block_within_block_p, dwarf2out_inline_entry): New.
1467 (dwarf2out_finish): Check that no entries remained in
1468 inline_entry_data_table.
1469 * final.c (reemit_insn_block_notes): Handle inline entry notes.
1470 (final_scan_insn, notice_source_line): Likewise.
1471 (rest_of_clean_state): Skip inline entry markers.
1472 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
1474 * gimple.c (gimple_build_debug_inline_entry): New.
1475 * gimple.h (enum gimple_debug_subcode): Add
1476 GIMPLE_DEBUG_INLINE_ENTRY.
1477 (gimple_build_debug_inline_entry): Declare.
1478 (gimple_debug_inline_entry_p): New.
1479 (gimple_debug_nonbind_marker_p): Adjust.
1480 * insn-notes.def (INLINE_ENTRY): New.
1481 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
1482 inline entry marker notes.
1483 (print_insn): Likewise.
1484 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
1485 (INSN_DEBUG_MARKER_KIND): Likewise.
1486 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
1487 * tree-inline.c (expand_call_inline): Build and insert
1488 debug_inline_entry stmt.
1489 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
1490 inline entry blocks early, if nonbind markers are enabled.
1491 (dump_scope_block): Dump fragment info.
1492 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
1493 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
1494 (gimple_build_debug_inline_entry): New.
1495 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
1496 Enable/disable inline entry points too.
1497 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
1498 (DEBUG_INSN): Describe inline entry markers.
1500 * common.opt (gvariable-location-views): New.
1501 (gvariable-location-views=incompat5): New.
1502 * config.in: Rebuilt.
1503 * configure: Rebuilt.
1504 * configure.ac: Test assembler for view support.
1505 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
1506 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
1507 * dwarf2out.c (var_loc_view): New typedef.
1508 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
1509 (dwarf2out_locviews_in_attribute): New.
1510 (dwarf2out_locviews_in_loclist): New.
1511 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
1512 (enum dw_line_info_opcode): Add LI_adv_address.
1513 (struct dw_line_info_table): Add view.
1514 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
1515 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
1516 (zero_view_p): New variable.
1517 (ZERO_VIEW_P): New macro.
1518 (output_asm_line_debug_info): New.
1519 (struct var_loc_node): Add view.
1520 (add_AT_view_list, AT_loc_list): New.
1521 (add_var_loc_to_decl): Add view param. Test it against last.
1522 (new_loc_list): Add view params. Record them.
1523 (AT_loc_list_ptr): Handle loc and view lists.
1524 (view_list_to_loc_list_val_node): New.
1525 (print_dw_val): Handle dw_val_class_view_list.
1526 (size_of_die): Likewise.
1527 (value_format): Likewise.
1528 (loc_list_has_views): New.
1529 (gen_llsym): Set vl_symbol too.
1530 (maybe_gen_llsym, skip_loc_list_entry): New.
1531 (dwarf2out_maybe_output_loclist_view_pair): New.
1532 (output_loc_list): Output view list or entries too.
1533 (output_view_list_offset): New.
1534 (output_die): Handle dw_val_class_view_list.
1535 (output_dwarf_version): New.
1536 (output_compilation_unit_header): Use it.
1537 (output_skeleton_debug_sections): Likewise.
1538 (output_rnglists, output_line_info): Likewise.
1539 (output_pubnames, output_aranges): Update version comments.
1540 (output_one_line_info_table): Output view numbers in asm comments.
1541 (dw_loc_list): Determine current endview, pass it to new_loc_list.
1542 Call maybe_gen_llsym.
1543 (loc_list_from_tree_1): Adjust.
1544 (add_AT_location_description): Create view list attribute if
1545 needed, check it's absent otherwise.
1546 (convert_cfa_to_fb_loc_list): Adjust.
1547 (maybe_emit_file): Call output_asm_line_debug_info for test.
1548 (dwarf2out_var_location): Reset views as needed. Precompute
1549 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
1550 attribute. Set view.
1551 (new_line_info_table): Reset next view.
1552 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
1553 (dwarf2out_source_line): Likewise. Output view resets and labels to
1554 the assembler, or select appropriate line info opcodes.
1555 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
1556 (optimize_string_length): Catch it. Adjust.
1557 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
1558 dw_val_class_view_list, and remove it if no longer needed.
1559 (hash_loc_list): Hash view numbers.
1560 (loc_list_hasher::equal): Compare them.
1561 (optimize_location_lists): Check whether a view list symbol is
1562 needed, and whether the locview attribute is present, and
1563 whether they match. Remove the locview attribute if no longer
1565 (index_location_lists): Call skip_loc_list_entry for test.
1566 (dwarf2out_finish): Call output_asm_line_debug_info for test.
1567 Use output_dwarf_version.
1568 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
1569 (struct dw_val_node): Add val_view_list.
1570 * final.c (SEEN_NEXT_VIEW): New.
1571 (set_next_view_needed): New.
1572 (clear_next_view_needed): New.
1573 (maybe_output_next_view): New.
1574 (final_start_function): Rename to...
1575 (final_start_function_1): ... this. Take pointer to FIRST,
1576 add SEEN parameter. Emit param bindings in the initial view.
1577 (final_start_function): Reintroduce SEEN-less interface.
1578 (final): Rename to...
1579 (final_1): ... this. Take SEEN parameter. Output final pending
1580 next view at the end.
1581 (final): Reintroduce seen-less interface.
1582 (final_scan_insn): Output pending next view before switching
1583 sections or ending a block. Mark the next view as needed when
1584 outputting variable locations. Notify debug backend of section
1585 changes, and of location view changes.
1586 (rest_of_handle_final): Adjust.
1587 * toplev.c (process_options): Autodetect value for debug variable
1588 location views option. Warn on incompat5 without -gdwarf-5.
1589 * doc/invoke.texi (gvariable-location-views): New.
1590 (gvariable-location-views=incompat5): New.
1591 (gno-variable-location-views): New.
1593 2018-02-08 David Malcolm <dmalcolm@redhat.com>
1595 PR tree-optimization/84136
1596 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
1597 that the result of find_edge is non-NULL.
1599 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1602 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
1603 storing integer register in SImode. Fix cost of 256 and 512
1604 byte aligned SSE register store.
1606 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1608 * config/i386/i386.c (ix86_multiplication_cost): Fix
1609 multiplication cost for TARGET_AVX512DQ.
1611 2018-02-08 Marek Polacek <polacek@redhat.com>
1613 PR tree-optimization/84238
1614 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
1617 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1619 PR tree-optimization/84265
1620 * tree-vect-stmts.c (vectorizable_store): Don't treat
1621 VMAT_CONTIGUOUS accesses as grouped.
1622 (vectorizable_load): Likewise.
1624 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1626 PR tree-optimization/81635
1627 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
1628 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
1629 (test_round_for_mask): New functions.
1630 (wide_int_cc_tests): Call test_round_for_mask.
1631 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
1632 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
1633 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
1634 range returned by get_range_info.
1636 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1639 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
1640 * symtab.c: Include builtins.h
1641 (symtab_node::output_to_lto_symbol_table_p): Move here
1642 from lto-streamer-out.c:output_symbol_p.
1643 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
1644 (output_symbol_p): Move all logic to symtab.c
1645 (produce_symtab): Update.
1647 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1649 * config/s390/s390-opts.h (enum indirect_branch): Define.
1650 * config/s390/s390-protos.h (s390_return_addr_from_memory)
1651 (s390_indirect_branch_via_thunk)
1652 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
1653 (enum s390_indirect_branch_type): Define.
1654 * config/s390/s390.c (struct s390_frame_layout, struct
1655 machine_function): Remove.
1656 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
1657 (indirect_branch_table_label_no, indirect_branch_table_name):
1659 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
1660 (enum s390_indirect_branch_option): Define.
1661 (s390_return_addr_from_memory): New function.
1662 (s390_handle_string_attribute): New function.
1663 (s390_attribute_table): Add new attribute handler.
1664 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
1665 (s390_indirect_branch_via_thunk): New function.
1666 (s390_indirect_branch_via_inline_thunk): New function.
1667 (s390_function_ok_for_sibcall): When jumping via thunk disallow
1668 sibling call optimization for non z10 compiles.
1669 (s390_emit_call): Force indirect branch target to be a single
1670 register. Add r1 clobber for non-z10 compiles.
1671 (s390_emit_epilogue): Emit return jump via return_use expander.
1672 (s390_reorg): Handle JUMP_INSNs as execute targets.
1673 (s390_option_override_internal): Perform validity checks for the
1674 new command line options.
1675 (s390_indirect_branch_attrvalue): New function.
1676 (s390_indirect_branch_settings): New function.
1677 (s390_set_current_function): Invoke s390_indirect_branch_settings.
1678 (s390_output_indirect_thunk_function): New function.
1679 (s390_code_end): Implement target hook.
1680 (s390_case_values_threshold): Implement target hook.
1681 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
1683 * config/s390/s390.h (struct s390_frame_layout)
1684 (struct machine_function): Move here from s390.c.
1685 (TARGET_INDIRECT_BRANCH_NOBP_RET)
1686 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
1687 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
1688 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
1689 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
1690 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
1691 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
1692 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
1693 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
1694 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
1695 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
1696 (mnemonic attribute): Add values which aren't recognized
1698 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
1699 pattern for branch conversion. Fix mnemonic attribute.
1700 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
1701 indirect branch via thunk if requested.
1702 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
1703 ("*indirect_jump"): Disable for branch conversion using out of
1705 ("indirect_jump_via_thunk<mode>_z10")
1706 ("indirect_jump_via_thunk<mode>")
1707 ("indirect_jump_via_inlinethunk<mode>_z10")
1708 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
1709 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
1710 ("casesi_jump_via_inlinethunk<mode>_z10")
1711 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
1712 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
1713 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
1714 ("*indirect2_jump"): Disable for branch conversion.
1715 ("casesi_jump"): Turn into expander and expand patterns for branch
1717 ("return_use"): New expander.
1718 ("*return"): Emit return via thunk and rename it to ...
1719 ("*return<mode>"): ... this one.
1720 * config/s390/s390.opt: Add new options and and enum for the
1723 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1725 * lra-constraints.c (match_reload): Unconditionally use
1726 gen_lowpart_SUBREG, rather than selecting between that
1727 and equivalent gen_rtx_SUBREG code.
1729 2018-02-08 Richard Biener <rguenther@suse.de>
1731 PR tree-optimization/84233
1732 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
1733 changed flag instead of boguously re-using phi_inserted.
1735 2018-02-08 Martin Jambor <mjambor@suse.cz>
1737 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
1738 static local variables.
1740 2018-02-08 Richard Biener <rguenther@suse.de>
1742 PR tree-optimization/84278
1743 * tree-vect-stmts.c (vectorizable_store): When looking for
1744 smaller vector types to perform grouped strided loads/stores
1745 make sure the mode is supported by the target.
1746 (vectorizable_load): Likewise.
1748 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1750 * config/aarch64/aarch64.c (aarch64_components_for_bb):
1751 Increase LDP/STP opportunities by adding adjacent callee-saves.
1753 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1755 PR rtl-optimization/84068
1756 PR rtl-optimization/83459
1757 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
1759 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
1761 PR tree-optimization/84224
1762 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
1763 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
1766 2018-02-07 Iain Sandoe <iain@codesourcery.com>
1769 * config/rs6000/altivec.md (*restore_world): Remove LR use.
1770 * config/rs6000/predicates.md (restore_world_operation): Adjust op
1771 count, remove one USE.
1773 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
1775 * doc/install.texi (Configuration): Document the
1776 --with-long-double-format={ibm,ieee} PowerPC configuration
1780 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
1781 Convert from define_expand to be define_insn_and_split. Rework
1782 float/double/_Float128 conversions to QI/HI/SImode to work with
1783 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
1784 conversions to QI/HImode types did a store and then a load to
1785 truncate the value. For conversions to VSX registers, don't split
1786 the insn, instead emit the code directly. Use the code iterator
1787 any_fix to combine signed and unsigned conversions.
1788 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
1789 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
1790 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
1791 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
1792 (fix_<mode>di2_hw): Likewise.
1793 (fixuns_<mode>di2_hw): Likewise.
1794 (fix_<mode>si2_hw): Likewise.
1795 (fixuns_<mode>si2_hw): Likewise.
1796 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
1797 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
1798 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
1799 fix<uns>_trunc<SFDF:mode>si2_p8.
1800 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
1802 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
1803 (fix<uns>_<mode>_mem): Likewise.
1804 (fctiw<u>z_<mode>_mem): Likewise.
1805 (fix<uns>_<mode>_mem): Likewise.
1806 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
1807 the register allocator from doing a direct move to the GPRs to do
1808 a store, and instead use the ISA 3.0 store byte/half-word from
1809 vector register instruction. For IEEE 128-bit floating point,
1810 also optimize stores of 32-bit ints.
1811 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
1813 2018-02-07 Alan Hayward <alan.hayward@arm.com>
1815 * genextract.c (push_pathstr_operand): New function to support
1817 (walk_rtx): Call push_pathstr_operand.
1818 (print_path): Support [a-zA-Z].
1820 2018-02-07 Richard Biener <rguenther@suse.de>
1822 PR tree-optimization/84037
1823 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
1824 (cse_and_gimplify_to_preheader): Declare.
1825 (vect_get_place_in_interleaving_chain): Likewise.
1826 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
1828 (_loop_vec_info::~_loop_vec_info): Delete it.
1829 (cse_and_gimplify_to_preheader): New function.
1830 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
1831 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
1832 (vectorizable_load): Likewise. For grouped stores always base
1833 the IV on the first element.
1834 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
1835 condition before gimplifying.
1837 2018-02-07 Jakub Jelinek <jakub@redhat.com>
1839 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
1840 *DIV_EXPR and *MOD_EXPR.
1842 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
1845 * config/i386/i386.c (ix86_option_override_internal): Mask out
1846 the CF_SET bit when checking -fcf-protection.
1848 2018-02-07 Tom de Vries <tom@codesourcery.com>
1851 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
1854 2018-02-07 Richard Biener <rguenther@suse.de>
1856 PR tree-optimization/84204
1857 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
1860 PR tree-optimization/84205
1861 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
1862 special-case isl_ast_op_zdiv_r.
1864 PR tree-optimization/84223
1865 * graphite-scop-detection.c (gather_bbs::before_dom_children):
1866 Only add conditions from within the region.
1867 (gather_bbs::after_dom_children): Adjust.
1869 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
1872 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
1873 * config/avr/avr.md: Only post-reload split REG-REG moves if
1874 either register is REGERAL_REG_P.
1876 2018-02-07 Jakub Jelinek <jakub@redhat.com>
1878 PR tree-optimization/84235
1879 * tree-ssa-scopedtables.c
1880 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
1881 if the subtraction is performed in floating point type where NaNs are
1882 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
1883 build 1. Formatting fix.
1885 2018-02-06 Jakub Jelinek <jakub@redhat.com>
1888 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
1889 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
1890 and skip it regardless of bb boundaries. Use CALL_P macro,
1891 don't test INSN_P (insn) together with CALL_P or JUMP_P check
1892 unnecessarily, formatting fix.
1894 2018-02-06 Michael Collison <michael.collison@arm.com>
1896 * config/arm/thumb2.md:
1897 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
1898 (*thumb_mov_notscc): Ditto.
1900 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
1903 * config/rs6000/rs6000.md (su code attribute): Use "u" for
1904 unsigned_fix, not "s".
1906 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1908 * configure.ac (gcc_fn_eh_frame_ro): New function.
1909 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
1910 correct .eh_frame permissions.
1911 * configure: Regenerate.
1913 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
1915 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
1918 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1920 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1921 Display warning message for -mno-speculate-indirect-jumps.
1923 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
1925 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
1927 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
1929 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
1931 PR tree-optimization/84225
1932 * tree-eh.c (find_trapping_overflow): Only call
1933 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
1935 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1938 * config/i386/i386.c: Reimplement the check of possible options
1939 -mibt/-mshstk conbination. Change error messages.
1940 * doc/invoke.texi: Fix a typo: remove extra '='.
1942 2018-02-06 Marek Polacek <polacek@redhat.com>
1944 PR tree-optimization/84228
1945 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
1947 2018-02-06 Tamar Christina <tamar.christina@arm.com>
1950 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
1951 emitted arch directives.
1952 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
1953 __ARM_FEATURE_COPROC before changing architectures.
1955 2018-02-06 Richard Biener <rguenther@suse.de>
1957 * config/i386/i386.c (print_reg): Fix typo.
1958 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
1960 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
1962 * configure: Regenerate.
1964 2018-02-05 Martin Sebor <msebor@redhat.com>
1966 PR tree-optimization/83369
1967 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
1970 2018-02-05 Martin Liska <mliska@suse.cz>
1972 * doc/invoke.texi: Cherry-pick upstream r323995.
1974 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
1976 * ira.c (ira_init_register_move_cost): Adjust comment.
1978 2018-02-05 Martin Liska <mliska@suse.cz>
1980 PR gcov-profile/84137
1981 * doc/gcov.texi: Fix typo in documentation.
1983 2018-02-05 Martin Liska <mliska@suse.cz>
1985 PR gcov-profile/83879
1986 * doc/gcov.texi: Document necessity of --dynamic-list-data when
1987 using dlopen functionality.
1989 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
1991 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
1992 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
1993 _mm_maskz_range_ss, _mm_mask_range_round_ss,
1994 _mm_maskz_range_round_ss): New intrinsics.
1995 (__builtin_ia32_rangesd128_round)
1996 (__builtin_ia32_rangess128_round): Remove.
1997 (__builtin_ia32_rangesd128_mask_round,
1998 __builtin_ia32_rangess128_mask_round): New builtins.
1999 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
2000 __builtin_ia32_rangess128_round): Remove.
2001 (__builtin_ia32_rangesd128_mask_round,
2002 __builtin_ia32_rangess128_mask_round): New builtins.
2003 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
2004 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
2005 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
2006 "<round_saeonly_constraint>")): Changed to ...
2007 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
2008 "<round_saeonly_scalar_constraint>")): ... this.
2009 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
2010 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
2011 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
2012 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
2013 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
2015 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
2017 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
2019 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
2020 Remove all values except native, 8540 and 8548.
2022 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
2024 * config/i386/i386.c (ix86_output_function_return): Pass
2025 INVALID_REGNUM, instead of -1, as invalid register number to
2026 indirect_thunk_name and output_indirect_thunk.
2028 2018-02-02 Julia Koval <julia.koval@intel.com>
2030 * config.gcc: Add -march=icelake.
2031 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
2032 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
2033 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
2034 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
2035 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
2036 (processor_target_table): Add icelake.
2037 (ix86_option_override_internal): Handle new PTAs.
2038 (get_builtin_code_for_version): Handle icelake.
2039 (M_INTEL_COREI7_ICELAKE): New.
2040 (fold_builtin_cpu): Handle icelake.
2041 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
2042 * doc/invoke.texi: Add -march=icelake.
2044 2018-02-02 Julia Koval <julia.koval@intel.com>
2046 * config/i386/i386.c (ix86_option_override_internal): Change flags type
2047 to wide_int_bitmask.
2048 * wide-int-bitmask.h: New.
2050 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2053 * config/i386/i386.md: Replace Pmode with word_mode in
2054 builtin_setjmp_setup and builtin_longjmp to support x32.
2056 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
2060 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
2062 (rs6000_supported_cpu_names): New static variable.
2063 (linux_cpu_translation_table): Likewise.
2064 (elf_platform) <cpu>: Define new static variable and use it.
2065 Translate kernel AT_PLATFORM name to canonical name if needed.
2066 Error if platform name is unknown.
2068 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
2071 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
2073 2018-02-01 Jeff Law <law@redhat.com>
2076 * config/i386/i386.c (release_scratch_register_on_entry): Add new
2077 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
2078 the scratch if RELEASE_VIA_POP is false.
2079 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
2080 If we have to save a temporary register, decrement SIZE appropriately.
2081 Pass new arguments to release_scratch_register_on_entry.
2082 (ix86_adjust_stack_and_probe): Likewise.
2083 (ix86_emit_probe_stack_range): Pass new arguments to
2084 release_scratch_register_on_entry.
2086 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
2088 PR rtl-optimization/84157
2089 * combine.c (change_zero_ext): Use REG_P predicate in
2090 front of HARD_REGISTER_P predicate.
2092 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
2094 * config/avr/avr.c (avr_option_override): Move disabling of
2095 -fdelete-null-pointer-checks to...
2096 * common/config/avr/avr-common.c (avr_option_optimization_table):
2099 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2101 PR tree-optimization/81635
2102 * tree-data-ref.c (split_constant_offset_1): For types that
2103 wrap on overflow, try to use range info to prove that wrapping
2106 2018-02-01 Renlin Li <renlin.li@arm.com>
2109 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
2111 (aarch64_register_move_cost): Likewise.
2112 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
2114 (REG_CLASS_NAMES): Likewise.
2115 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
2116 TAILCALL_ADDR_REGS. Remove IP registers.
2117 * config/aarch64/aarch64.md (Ucs): Update register constraint.
2119 2018-02-01 Richard Biener <rguenther@suse.de>
2121 * domwalk.h (dom_walker::dom_walker): Add additional constructor
2122 for specifying RPO order and allow NULL for that.
2123 * domwalk.c (dom_walker::dom_walker): Likewise.
2124 (dom_walker::walk): Handle NULL RPO order.
2125 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
2127 (rewrite_update_dom_walker): Likewise.
2128 (mark_def_dom_walker): Likewise.
2130 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2132 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
2133 (aarch64_maybe_expand_sve_subreg_move): Declare.
2134 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
2135 * config/aarch64/predicates.md (aarch64_any_register_operand): New
2137 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
2138 that are semantically a reverse operation.
2139 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
2140 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
2141 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
2143 (aarch64_can_change_mode_class): For big-endian, forbid changes
2144 between two SVE modes if they have different element sizes.
2146 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2148 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
2149 the TImode handling for big-endian targets.
2151 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2153 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
2154 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
2156 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
2157 Remove BSWAP handing for big-endian targets and use the form of
2158 LD1RQ appropariate for the mode.
2160 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2162 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
2163 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
2166 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2169 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
2170 check for operands that need to go through aarch64_sve_reload_be.
2172 2018-02-01 Jakub Jelinek <jakub@redhat.com>
2174 PR tree-optimization/81661
2175 PR tree-optimization/84117
2176 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
2177 * tree-eh.c: Include gimplify.h.
2178 (find_trapping_overflow, replace_trapping_overflow,
2179 rewrite_to_non_trapping_overflow): New functions.
2180 * tree-vect-loop.c: Include tree-eh.h.
2181 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
2182 * tree-data-ref.c: Include tree-eh.h.
2183 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
2185 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
2187 PR rtl-optimization/84123
2188 * combine.c (change_zero_ext): Check if hard register satisfies
2189 can_change_dest_mode before calling gen_lowpart_SUBREG.
2191 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
2194 * ira.c (ira_init_register_move_cost): Remove assert.
2196 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2198 PR rtl-optimization/84071
2199 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
2200 * doc/tm.texi: Regenerate.
2202 2018-01-31 Richard Biener <rguenther@suse.de>
2204 PR tree-optimization/84132
2205 * tree-data-ref.c (analyze_miv_subscript): Properly
2206 check whether evolution_function_is_affine_multivariate_p
2207 before calling gcd_of_steps_may_divide_p.
2209 2018-01-31 Julia Koval <julia.koval@intel.com>
2212 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
2213 * config/i386/i386.md (rdpid_rex64) New.
2214 (rdpid): Make 32bit only.
2216 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
2219 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
2220 an IDENTIFIER_NODE for FUNCTION_TYPE's.
2222 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2225 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
2227 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
2229 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2231 PR rtl-optimization/84071
2232 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
2233 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
2235 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2237 * config/arc/arc.c (arc_handle_aux_attribute): New function.
2238 (arc_attribute_table): Add 'aux' attribute.
2239 (arc_in_small_data_p): Consider aux like variables.
2240 (arc_is_aux_reg_p): New function.
2241 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
2242 (arc_get_aux_arg): New function.
2243 (prepare_move_operands): Handle aux-register access.
2244 (arc_handle_aux_attribute): New function.
2245 * doc/extend.texi (ARC Variable attributes): Add subsection.
2247 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2249 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
2250 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
2251 (arc_attribute_table): Add 'uncached' attribute.
2252 (arc_print_operand): Print '.di' flag for uncached memory
2254 (arc_in_small_data_p): Do not consider for small data the uncached
2256 (arc_is_uncached_mem_p): New function.
2257 * config/arc/predicates.md (compact_store_memory_operand): Check
2258 for uncached memory accesses.
2259 (nonvol_nonimm_operand): Likewise.
2260 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
2262 2018-01-31 Jakub Jelinek <jakub@redhat.com>
2265 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
2266 falign-loops=): Add Optimization flag.
2268 2018-01-30 Jeff Law <law@redhat.com>
2271 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
2272 INT_REGISTERS_SAVED. Check it prior to calling
2273 get_scratch_register_on_entry.
2274 (ix86_adjust_stack_and_probe): Similarly.
2275 (ix86_emit_probe_stack_range): Similarly.
2276 (ix86_expand_prologue): Corresponding changes.
2278 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2281 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
2282 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
2284 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
2287 * lra-constraints.c (curr_insn_transform): Process AND in the
2290 2018-01-30 Jakub Jelinek <jakub@redhat.com>
2292 PR rtl-optimization/83986
2293 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
2294 dependence against last_pending_memory_flush in addition to
2297 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
2299 PR tree-optimization/81611
2300 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
2303 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2306 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
2309 2018-01-30 Richard Biener <rguenther@suse.de>
2310 Jakub Jelinek <jakub@redhat.com>
2312 PR tree-optimization/84111
2313 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
2314 inner loops added during recursion, as they don't have up-to-date
2317 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2320 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
2321 (can_inline_edge_by_limits_p): ... here.
2322 (can_early_inline_edge_p, check_callers,
2323 update_caller_keys, update_callee_keys, recursive_inlining,
2324 add_new_edges_to_heap, speculation_useful_p,
2325 inline_small_functions,
2326 inline_small_functions, flatten_function,
2327 inline_to_all_callers_1): Update.
2329 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2331 * profile-count.c (profile_count::combine_with_ipa_count): Handle
2334 2018-01-30 Richard Biener <rguenther@suse.de>
2336 PR tree-optimization/83008
2337 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
2338 invariant and constant vector uses in stmts when they need
2341 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2344 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
2345 * configure: Regenerate.
2347 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2349 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
2351 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
2352 Use gen_rtx_REG rather than gen_lowpart.
2354 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2356 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
2357 rather than 0 when creating partial subregs.
2359 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2361 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
2364 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
2367 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
2368 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
2369 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
2370 flags. This restores the settings used before the 2017-07-24.
2371 Turning off pre increment/decrement/modify allows IVOPTS to
2372 optimize DF/SF loops where the index is an int.
2374 2018-01-29 Richard Biener <rguenther@suse.de>
2375 Kelvin Nilsen <kelvin@gcc.gnu.org>
2378 * tree-vect-stmts.c (vectorizable_call): Don't call
2379 targetm.vectorize_builtin_md_vectorized_function if callee is
2382 2018-01-22 Carl Love <cel@us.ibm.com>
2384 * doc/extend.tex: Fix typo in second arg in
2385 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
2387 2018-01-29 Richard Biener <rguenther@suse.de>
2389 PR tree-optimization/84086
2390 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
2391 (flush_ssaname_freelist): When SSA names were released reset
2392 the SCEV hash table.
2394 2018-01-29 Richard Biener <rguenther@suse.de>
2396 PR tree-optimization/84057
2397 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
2398 removed paths when removing edges.
2400 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
2402 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
2403 -mfunction-return=@var{choice}.
2405 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
2408 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
2410 (layout::print_source_line): Likewise.
2411 (test_get_line_width_without_trailing_whitespace): Add test cases.
2413 2018-01-27 Jakub Jelinek <jakub@redhat.com>
2416 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
2419 2018-01-26 Jim Wilson <jimw@sifive.com>
2421 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
2423 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
2426 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2428 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
2429 and CMP + SUB-immediate -> SUBS.
2431 2018-01-26 Martin Sebor <msebor@redhat.com>
2433 PR tree-optimization/83896
2434 * tree-ssa-strlen.c (get_string_len): Rename...
2435 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
2436 Avoid assuming length is constant.
2437 (handle_char_store): Use HOST_WIDE_INT for string length.
2439 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
2442 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
2443 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
2445 2018-01-26 Richard Biener <rguenther@suse.de>
2447 PR rtl-optimization/84003
2448 * dse.c (record_store): Only record redundant stores when
2449 the earlier store aliases at least all accesses the later one does.
2451 2018-01-26 Jakub Jelinek <jakub@redhat.com>
2453 PR rtl-optimization/83985
2454 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
2455 REG_CFA_RESTORE insns.
2456 (delete_unmarked_insns): Don't ignore separate shrink wrapping
2457 REG_CFA_RESTORE insns here.
2460 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
2461 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
2463 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2465 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
2466 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
2467 (arc_init): Likewise.
2468 (arc_override_options): Likewise.
2469 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
2471 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
2473 * config/arc/arc.h (TARGET_DBNZ): Define.
2474 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
2475 properly set the tune attribute.
2476 (dbnz): Use TARGET_DBNZ guard.
2477 * config/arc/arc.opt (mtune): Add core3 option.
2479 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2481 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
2482 recognize new pic like addresses.
2483 (arc_delegitimize_address): Clean up.
2485 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2487 * config/arc/arc-arches.def: Option mrf16 valid for all
2489 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
2490 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
2491 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
2492 * config/arc/arc-tables.opt: Regenerate.
2493 * config/arc/arc.c (arc_conditional_register_usage): Handle
2494 reduced register file case.
2495 (arc_file_start): Set must have build attributes.
2496 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
2498 * config/arc/arc.opt (mrf16): Add new option.
2499 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
2500 * config/arc/genmultilib.awk: Handle new mrf16 option.
2501 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
2502 * config/arc/t-multilib: Regenerate.
2503 * doc/invoke.texi (ARC Options): Document mrf16 option.
2505 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2507 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
2508 * config/arc/arc.c (arc_handle_secure_attribute): New function.
2509 (arc_attribute_table): Add 'secure_call' attribute.
2510 (arc_print_operand): Print secure call operand.
2511 (arc_function_ok_for_sibcall): Don't optimize tail calls when
2513 (arc_is_secure_call_p): New function. * config/arc/arc.md
2514 (call_i): Add support for sjli instruction.
2515 (call_value_i): Likewise.
2516 * config/arc/constraints.md (Csc): New constraint.
2518 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2519 John Eric Martin <John.Martin@emmicro-us.com>
2521 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
2522 * config/arc/arc.c (_arc_jli_section): New struct.
2523 (arc_jli_section): New type.
2524 (rc_jli_sections): New static variable.
2525 (arc_handle_jli_attribute): New function.
2526 (arc_attribute_table): Add jli_always and jli_fixed attribute.
2527 (arc_file_end): New function.
2528 (TARGET_ASM_FILE_END): Define.
2529 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
2530 (arc_add_jli_section): New function.
2531 (jli_call_scan): Likewise.
2532 (arc_reorg): Call jli_call_scan.
2533 (arc_output_addsi): Remove 'S' from printing asm operand.
2534 (arc_is_jli_call_p): New function.
2535 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
2537 (movhi_insn): Likewise.
2538 (movsi_insn): Likewise.
2539 (movsi_set_cc_insn): Likewise.
2540 (loadqi_update): Likewise.
2541 (load_zeroextendqisi_update): Likewise.
2542 (load_signextendqisi_update): Likewise.
2543 (loadhi_update): Likewise.
2544 (load_zeroextendhisi_update): Likewise.
2545 (load_signextendhisi_update): Likewise.
2546 (loadsi_update): Likewise.
2547 (loadsf_update): Likewise.
2548 (movsicc_insn): Likewise.
2549 (bset_insn): Likewise.
2550 (bxor_insn): Likewise.
2551 (bclr_insn): Likewise.
2552 (bmsk_insn): Likewise.
2553 (bicsi3_insn): Likewise.
2554 (cmpsi_cc_c_insn): Likewise.
2555 (movsi_ne): Likewise.
2556 (movsi_cond_exec): Likewise.
2557 (clrsbsi2): Likewise.
2568 (call_i): Remove 'S' asm letter, add jli instruction.
2569 (call_value_i): Likewise.
2570 * config/arc/arc.op (mjli-always): New option.
2571 * config/arc/constraints.md (Cji): New constraint.
2572 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
2574 (subsf3_fpx): Likewise.
2575 (mulsf3_fpx): Likewise.
2576 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
2578 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
2580 * doc/invoke.texi (ARC): Document mjli-always option.
2582 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
2584 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
2585 avoid addition with 0 and use incw and decw where possible.
2587 2018-01-26 Richard Biener <rguenther@suse.de>
2589 PR tree-optimization/81082
2590 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
2591 association if it requires casting to unsigned.
2592 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
2593 from fold_plusminus_mult_expr to catch important cases late when
2594 range info is available.
2596 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2598 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
2599 * configure.ac (hidden_linkonce): New test.
2600 * configure: Regenerate.
2601 * config.in: Regenerate.
2603 2018-01-26 Julia Koval <julia.koval@intel.com>
2605 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
2606 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
2607 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
2608 _mm_mask_bitshuffle_epi64_mask): Fix type.
2609 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
2610 USI_FTYPE_V4DI_V4DI_USI): Remove.
2611 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
2612 __builtin_ia32_vpshufbitqmb256_mask,
2613 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
2614 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
2615 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
2617 2018-01-26 Alan Modra <amodra@gmail.com>
2620 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
2621 UNSPEC_VBPERMQ. Sort other unspecs.
2623 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
2625 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
2627 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
2630 * predict.c (drop_profile): Do not push/pop cfun; update also
2632 (handle_missing_profiles): Fix logic looking for zero profiles.
2634 2018-01-25 Jakub Jelinek <jakub@redhat.com>
2637 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
2638 on functions with #pragma omp declare simd or functions with simd
2640 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
2641 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
2642 Remove trailing \n from warning_at calls.
2644 2018-01-25 Tom de Vries <tom@codesourcery.com>
2647 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2648 for neutered workers.
2650 2018-01-24 Joseph Myers <joseph@codesourcery.com>
2653 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
2654 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
2656 2018-01-24 Jeff Law <law@redhat.com>
2659 * i386.c (get_probe_interval): Move to earlier point.
2660 (ix86_compute_frame_layout): If -fstack-clash-protection and
2661 the frame is larger than the probe interval, then use pushes
2662 to save registers rather than reg->mem moves.
2663 (ix86_expand_prologue): Remove conditional for int_registers_saved
2666 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
2669 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
2670 min/max for never referenced object.
2672 2018-01-24 Jakub Jelinek <jakub@redhat.com>
2675 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
2677 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
2678 attributes from DECL_ATTRIBUTES (decl) without affecting
2679 DECL_ATTRIBUTES (current_function_decl).
2680 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
2681 functions with non-NULL DECL_ABSTRACT_ORIGIN.
2683 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
2685 PR tree-optimization/83979
2686 * fold-const.c (fold_comparison): Use constant_boolean_node
2687 instead of boolean_{true,false}_node.
2689 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
2691 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
2694 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2696 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
2697 Simplify the clause that sets the length attribute.
2698 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
2699 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
2700 clause that sets the length attribute.
2701 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
2703 2018-01-24 Tom de Vries <tom@codesourcery.com>
2706 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
2707 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
2708 Add strict parameter.
2709 (prevent_branch_around_nothing): Insert dummy insn between branch to
2710 label and label with no ptx insn inbetween.
2711 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
2713 2018-01-24 Tom de Vries <tom@codesourcery.com>
2716 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2717 for neutered threads in warp.
2718 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
2720 2018-01-24 Richard Biener <rguenther@suse.de>
2722 PR tree-optimization/83176
2723 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
2726 2018-01-24 Richard Biener <rguenther@suse.de>
2728 PR tree-optimization/82819
2729 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
2730 code generating pluses that are no-ops in the target precision.
2732 2018-01-24 Richard Biener <rguenther@suse.de>
2735 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
2737 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2739 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
2740 to merge probabilities.
2741 * predict.c (probably_never_executed): Also mark as cold functions
2742 with global 0 profile and guessed local profile.
2743 * profile-count.c (profile_probability::combine_with_count): New
2745 * profile-count.h (profile_probability::operator*,
2746 profile_probability::operator*=, profile_probability::operator/,
2747 profile_probability::operator/=): Reduce precision to adjusted
2748 and set value to guessed on contradictory divisions.
2749 (profile_probability::combine_with_freq): Remove.
2750 (profile_probability::combine_wiht_count): Declare.
2751 (profile_count::force_nonzero):: Set to adjusted.
2752 (profile_count::probability_in):: Set quality to adjusted.
2753 * tree-ssa-tail-merge.c (replace_block_by): Use
2756 2018-01-23 Andrew Waterman <andrew@sifive.com>
2757 Jim Wilson <jimw@sifive.com>
2759 * config/riscv/riscv.c (riscv_stack_boundary): New.
2760 (riscv_option_override): Set riscv_stack_boundary. Handle
2761 riscv_preferred_stack_boundary_arg.
2762 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
2763 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
2764 (STACK_BOUNDARY): Set to riscv_stack_boundary.
2765 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
2766 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
2767 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
2769 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
2772 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
2773 of struct ix86_frame.
2774 (ix86_expand_epilogue): Likewise. Add a local variable for
2775 the reg_save_offset field in struct ix86_frame.
2777 2018-01-23 Bin Cheng <bin.cheng@arm.com>
2779 PR tree-optimization/82604
2780 * tree-loop-distribution.c (enum partition_kind): New enum item
2781 PKIND_PARTIAL_MEMSET.
2782 (partition_builtin_p): Support above new enum item.
2783 (generate_code_for_partition): Ditto.
2784 (compute_access_range): Differentiate cases that equality can be
2785 proven at all loops, the innermost loops or no loops.
2786 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
2787 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
2788 (finalize_partitions, distribute_loop): Don't fuse partition of
2789 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
2790 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
2793 2018-01-23 Martin Liska <mliska@suse.cz>
2795 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
2796 order to ignore the predictor.
2797 (PRED_POLYMORPHIC_CALL): Likewise.
2798 (PRED_RECURSIVE_CALL): Likewise.
2800 2018-01-23 Martin Liska <mliska@suse.cz>
2802 * tree-profile.c (tree_profiling): Print function header to
2803 aware reader which function we are working on.
2804 * value-prof.c (gimple_find_values_to_profile): Do not print
2805 not interesting value histograms.
2807 2018-01-23 Martin Liska <mliska@suse.cz>
2809 * profile-count.h (enum profile_quality): Add
2810 profile_uninitialized as the first value. Do not number values
2811 as they are zero based.
2812 (profile_count::verify): Update sanity check.
2813 (profile_probability::verify): Likewise.
2815 2018-01-23 Nathan Sidwell <nathan@acm.org>
2817 * doc/invoke.texi (ffor-scope): Deprecate.
2819 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2821 PR tree-optimization/83510
2822 * domwalk.c (set_all_edges_as_executable): New function.
2823 (dom_walker::dom_walker): Convert bool param
2824 "skip_unreachable_blocks" to enum reachability. Move setup of
2825 edge flags to set_all_edges_as_executable and only do it when
2826 reachability is REACHABLE_BLOCKS.
2827 * domwalk.h (enum dom_walker::reachability): New enum.
2828 (dom_walker::dom_walker): Convert bool param
2829 "skip_unreachable_blocks" to enum reachability.
2830 (set_all_edges_as_executable): New decl.
2831 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
2832 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
2834 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
2835 but converting true to REACHABLE_BLOCKS.
2836 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
2838 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
2839 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
2840 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
2842 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
2843 if check_all_array_refs will be called.
2845 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2847 * tree.c (selftest::test_location_wrappers): Add more test
2850 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2852 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
2853 (selftest::test_bit_in_range): Likewise.
2855 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
2858 * doc/sourcebuild.texi (vect_float): Say that the selector
2859 only describes the situation when -funsafe-math-optimizations is on.
2860 (vect_float_strict): Document.
2862 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
2864 PR tree-optimization/83965
2865 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
2866 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
2867 instead of checking only for a reduction.
2868 (vect_recog_widen_sum_pattern): Likewise.
2870 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2872 * predict.c (probably_never_executed): Only use precise profile info.
2873 (compute_function_frequency): Skip after inlining hack since we now
2874 have quality checking.
2876 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2878 * profile-count.h (profile_probability::very_unlikely,
2879 profile_probability::unlikely, profile_probability::even): Set
2880 precision to guessed.
2882 2018-01-23 Richard Biener <rguenther@suse.de>
2884 PR tree-optimization/83963
2885 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
2886 Properly terminate dominator walk when crossing the exit edge not
2887 when visiting its source block.
2889 2018-01-23 Jakub Jelinek <jakub@redhat.com>
2892 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
2893 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
2895 2018-01-22 Jakub Jelinek <jakub@redhat.com>
2897 PR tree-optimization/83957
2898 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
2899 semicolon after for body surrounded by braces.
2901 PR tree-optimization/83081
2902 * profile-count.h (profile_probability::split): New method.
2903 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
2904 Use profile_probability::split.
2905 (do_compare_rtx_and_jump): Fix adjustment of probabilities
2906 when splitting a single conditional jump into 2.
2908 2018-01-22 David Malcolm <dmalcolm@redhat.com>
2910 PR tree-optimization/69452
2911 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
2914 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2916 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
2917 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
2918 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
2920 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2922 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
2923 * config/rl78/rl78.md (movdi): New define_expand.
2924 * config/rl78/rl78.c (rl78_split_movdi): New function.
2926 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
2929 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
2931 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
2932 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
2933 128-bit to produce an UNSPEC move to get the double word with the
2934 signbit and then a shift directly to do signbit.
2935 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
2936 implementation with a new version that just does either a direct
2937 move or a regular move. Move memory interface to separate insns.
2938 Move insns so they are next to the expander.
2939 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
2940 with signbit move. Split big and little endian case.
2941 (signbit<mode>2_dm_mem_le): Likewise.
2942 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
2943 (signbit<mode>2_dm2): Likewise.
2945 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2947 * config/rl78/rl78.md (anddi3): New define_expand.
2949 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2951 * config/rl78/rl78.md (umindi3): New define_expand.
2953 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2955 * config/rl78/rl78.md (smindi3): New define_expand.
2957 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2959 * config/rl78/rl78.md (smaxdi3): New define_expand.
2961 2018-01-22 Carl Love <cel@us.ibm.com>
2963 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
2964 LVX_V1TI): Add macro expansion.
2965 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
2966 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
2967 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
2968 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
2969 Change check to determine if the instruction is a byte reversing
2970 entry. Fix typo in comment.
2971 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
2972 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
2973 Add def_builtin calls for new builtins.
2974 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
2975 Add define_insn expansion.
2977 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2979 * config/rl78/rl78.md (umaxdi3): New define_expand.
2981 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2983 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
2984 for non-QImode registers.
2986 2018-01-22 Richard Biener <rguenther@suse.de>
2988 PR tree-optimization/83963
2989 * graphite-scop-detection.c (scop_detection::get_sese): Delay
2990 including the loop exit block.
2991 (scop_detection::merge_sese): Likewise.
2992 (scop_detection::add_scop): Do it here instead.
2994 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2996 * doc/sourcebuild.texi (arm_softfloat): Document.
2998 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
3001 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
3002 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
3003 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
3005 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3006 David Edelsohn <dje.gcc@gmail.com>
3009 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3010 Change "crset eq" to "crset 2".
3011 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3012 (*call_indirect_aix<mode>_nospec): Likewise.
3013 (*call_value_indirect_aix<mode>_nospec): Likewise.
3014 (*call_indirect_elfv2<mode>_nospec): Likewise.
3015 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
3016 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
3017 change assembly output from . to $.
3018 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3019 (indirect_jump<mode>_nospec): Change assembly output from . to $.
3020 (*tablejump<mode>_internal1_nospec): Likewise.
3022 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
3025 * config/sh/sh_optimize_sett_clrt.cc:
3026 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
3028 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
3030 PR tree-optimization/83940
3031 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
3032 offset_dt to vect_constant_def rather than vect_unknown_def_type.
3033 (vect_check_load_store_mask): Add a mask_dt_out parameter and
3034 use it to pass back the definition type.
3035 (vect_check_store_rhs): Likewise rhs_dt_out.
3036 (vect_build_gather_load_calls): Add a mask_dt argument and use
3037 it instead of a call to vect_is_simple_use.
3038 (vectorizable_store): Update calls to vect_check_load_store_mask
3039 and vect_check_store_rhs. Use the dt returned by the latter instead
3040 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
3041 instead of calls to vect_is_simple_use. Pass the scalar rather
3042 than the vector operand to vect_is_simple_use when handling
3043 second and subsequent copies of an rhs value.
3044 (vectorizable_load): Update calls to vect_check_load_store_mask
3045 and vect_build_gather_load_calls. Use the cached mask_dt and
3046 gs_info.offset_dt instead of calls to vect_is_simple_use.
3048 2018-01-20 Jakub Jelinek <jakub@redhat.com>
3051 * tree-emutls.c: Include gimplify.h.
3052 (lower_emutls_2): New function.
3053 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
3054 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
3055 it before further processing.
3058 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
3059 UINTVAL (trueop1) instead of INTVAL (op1).
3061 2018-01-19 Jakub Jelinek <jakub@redhat.com>
3065 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
3066 INCOMING_FRAME_SP_OFFSET if not defined.
3067 (scan_trace): Add ENTRY argument. If true and
3068 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
3069 emit a note to adjust the CFA offset.
3070 (create_cfi_notes): Adjust scan_trace callers.
3071 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
3072 INCOMING_FRAME_SP_OFFSET in the CIE.
3073 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
3074 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
3076 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
3077 * doc/tm.texi: Regenerated.
3079 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3081 PR rtl-optimization/83147
3082 * lra-constraints.c (remove_inheritance_pseudos): Use
3083 lra_substitute_pseudo_within_insn.
3085 2018-01-19 Tom de Vries <tom@codesourcery.com>
3086 Cesar Philippidis <cesar@codesourcery.com>
3089 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
3091 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
3094 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
3095 spaces for function labels.
3097 2018-01-19 Martin Liska <mliska@suse.cz>
3099 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
3100 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
3101 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
3102 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
3103 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
3104 (PRED_CONST_RETURN): Change from 69 to 65.
3105 (PRED_NULL_RETURN): Change from 91 to 71.
3106 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
3107 (PRED_LOOP_GUARD): Change from 66 to 73.
3109 2018-01-19 Martin Liska <mliska@suse.cz>
3111 * predict.c (predict_insn_def): Add new assert.
3112 (struct branch_predictor): Change type to signed integer.
3113 (test_prediction_value_range): Amend test to cover
3115 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
3116 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
3117 (PRED_LOOP_ITERATIONS_MAX): Likewise.
3118 (PRED_LOOP_IV_COMPARE): Likewise.
3119 * predict.h (PROB_UNINITIALIZED): Define new constant.
3121 2018-01-19 Martin Liska <mliska@suse.cz>
3123 * predict.c (dump_prediction): Add new format for
3124 analyze_brprob.py script which is enabled with -details
3126 * profile-count.h (precise_p): New function.
3128 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3130 PR tree-optimization/83922
3131 * tree-vect-loop.c (vect_verify_full_masking): Return false if
3132 there are no statements that need masking.
3133 (vect_active_double_reduction_p): New function.
3134 (vect_analyze_loop_operations): Use it when handling phis that
3135 are not in the loop header.
3137 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3139 PR tree-optimization/83914
3140 * tree-vect-loop.c (vectorizable_induction): Don't convert
3141 init_expr or apply the peeling adjustment for inductions
3142 that are nested within the vectorized loop.
3144 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3146 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
3149 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3153 * function.h (gimplify_parameters): Add gimple_seq * argument.
3154 * function.c: Include gimple.h and options.h.
3155 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
3156 for the added local temporaries if needed.
3157 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
3158 if there are any parameter cleanups, wrap whole body into a
3159 try/finally with the cleanups.
3161 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
3164 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
3165 Use GET_MODE_CLASS for scalar floating point.
3167 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3171 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3172 Fix call of call_cgraph_insertion_hooks.
3174 2018-01-18 Martin Sebor <msebor@redhat.com>
3176 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
3178 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3181 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
3184 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
3187 * common.opt: (-ffile-prefix-map): New option.
3188 * opts.c (common_handle_option): Defer it.
3189 * opts-global.c (handle_common_deferred_options): Handle it.
3190 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
3191 * file-prefix-map.h: New file.
3192 (remap_debug_filename, add_debug_prefix_map): ...here.
3193 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
3194 * final.c (debug_prefix_map, add_debug_prefix_map
3195 remap_debug_filename): Move to...
3196 * file-prefix-map.c: New file.
3197 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
3198 generalize, get rid of alloca(), use strrchr() instead of strchr().
3199 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
3200 Implement in terms of add_prefix_map().
3201 (remap_macro_filename, remap_debug_filename): Implement in term of
3203 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
3204 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
3205 * dbxout.c: Include file-prefix-map.h.
3206 * varasm.c: Likewise.
3207 * vmsdbgout.c: Likewise.
3208 * xcoffout.c: Likewise.
3209 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
3210 * doc/cppopts.texi (-fmacro-prefix-map): Document.
3211 * doc/invoke.texi (-ffile-prefix-map): Document.
3212 (-fdebug-prefix-map): Update description.
3214 2018-01-18 Martin Liska <mliska@suse.cz>
3216 * config/i386/i386.c (indirect_thunk_name): Document that also
3218 (output_indirect_thunk): Document why both instructions
3219 (pause and lfence) are generated.
3221 2018-01-18 Richard Biener <rguenther@suse.de>
3223 PR tree-optimization/83887
3224 * graphite-scop-detection.c
3225 (scop_detection::get_nearest_dom_with_single_entry): Remove.
3226 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
3227 (scop_detection::merge_sese): Re-implement with a flood-fill
3228 algorithm that properly finds a SESE region if it exists.
3230 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3233 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
3234 pointer_diff optimizations use view_convert instead of convert.
3236 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3238 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3239 Generate different code for -mno-speculate-indirect-jumps.
3240 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3241 (*call_indirect_aix<mode>): Disable for
3242 -mno-speculate-indirect-jumps.
3243 (*call_indirect_aix<mode>_nospec): New define_insn.
3244 (*call_value_indirect_aix<mode>): Disable for
3245 -mno-speculate-indirect-jumps.
3246 (*call_value_indirect_aix<mode>_nospec): New define_insn.
3247 (*sibcall_nonlocal_sysv<mode>): Generate different code for
3248 -mno-speculate-indirect-jumps.
3249 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3251 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3253 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
3254 long double type, set the flags for noting the default long double
3255 type, even if we don't pass or return a long double type.
3257 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
3260 * ipa-inline.c (flatten_function): Do not overwrite final inlining
3263 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
3265 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
3266 support for merge[hl].
3267 (fold_mergehl_helper): New helper function.
3268 (tree-vector-builder.h): New #include for tree_vector_builder usage.
3269 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
3270 (altivec_vmrglw_direct): Add xxmrglw insn.
3272 2018-01-17 Andrew Waterman <andrew@sifive.com>
3274 * config/riscv/riscv.c (riscv_conditional_register_usage): If
3275 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
3277 2018-01-17 David Malcolm <dmalcolm@redhat.com>
3280 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
3281 call the lto_location_cache before reading the
3282 DECL_SOURCE_LOCATION of the types.
3284 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3285 Richard Sandiford <richard.sandiford@linaro.org>
3287 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
3288 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
3289 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
3290 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
3291 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
3293 * config/aarch64/constraints.md (aarch64_movti_operand):
3295 * config/aarch64/predicates.md (Uti): Add new constraint.
3297 2018-01-17 Carl Love <cel@us.ibm.com>
3299 * config/rs6000/vsx.md (define_expand xl_len_r,
3300 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
3301 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
3303 (define_expand, define_insn): Move the shift left from the
3304 define_insn to the define_expand for lxvl and stxvl instructions.
3305 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
3306 and XL_LEN_R definitions to PURE.
3308 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
3310 * config/i386/i386.c (indirect_thunk_name): Declare regno
3311 as unsigned int. Compare regno with INVALID_REGNUM.
3312 (output_indirect_thunk): Ditto.
3313 (output_indirect_thunk_function): Ditto.
3314 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
3315 in the call to output_indirect_thunk_function.
3317 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
3320 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
3321 rather than the size of inner_type to determine the stack slot size
3322 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
3324 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
3327 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
3330 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
3332 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
3333 endian Linux systems to optionally enable multilibs for selecting
3334 the long double type if the user configured an explicit type.
3335 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
3336 have no long double multilibs if not defined.
3337 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
3338 warn if the user used -mabi={ieee,ibm}longdouble and we built
3339 multilibs for long double.
3340 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
3341 appropriate multilib option.
3342 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
3344 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
3345 for building long double multilibs.
3346 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
3348 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
3350 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
3353 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
3355 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
3358 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
3361 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
3364 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3366 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
3367 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
3369 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3371 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
3372 different rtl trees depending on TARGET_64BIT.
3373 (rs6000_gen_lvx): Likewise.
3375 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3377 * config/visium/visium.md (nop): Tweak comment.
3378 (hazard_nop): Likewise.
3380 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3382 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
3383 -mspeculate-indirect-jumps.
3384 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
3385 for -mno-speculate-indirect-jumps.
3386 (*call_indirect_elfv2<mode>_nospec): New define_insn.
3387 (*call_value_indirect_elfv2<mode>): Disable for
3388 -mno-speculate-indirect-jumps.
3389 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
3390 (indirect_jump): Emit different RTL for
3391 -mno-speculate-indirect-jumps.
3392 (*indirect_jump<mode>): Disable for
3393 -mno-speculate-indirect-jumps.
3394 (*indirect_jump<mode>_nospec): New define_insn.
3395 (tablejump): Emit different RTL for
3396 -mno-speculate-indirect-jumps.
3397 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
3398 (tablejumpsi_nospec): New define_expand.
3399 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
3400 (tablejumpdi_nospec): New define_expand.
3401 (*tablejump<mode>_internal1): Disable for
3402 -mno-speculate-indirect-jumps.
3403 (*tablejump<mode>_internal1_nospec): New define_insn.
3404 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
3407 2018-01-16 Artyom Skrobov tyomitch@gmail.com
3409 * caller-save.c (insert_save): Drop unnecessary parameter. All
3412 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3413 Richard Biener <rguenth@suse.de>
3416 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
3417 return early, inline manually is_gimple_sizepos. Make sure if we
3418 call gimplify_expr we don't end up with a gimple constant.
3419 * tree.c (variably_modified_type_p): Don't return true for
3420 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
3421 * gimplify.h (is_gimple_sizepos): Remove.
3423 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3425 PR tree-optimization/83857
3426 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
3427 vectorizable_live_operation for pure SLP statements.
3428 (vectorizable_live_operation): Handle PHIs.
3430 2018-01-16 Richard Biener <rguenther@suse.de>
3432 PR tree-optimization/83867
3433 * tree-vect-stmts.c (vect_transform_stmt): Precompute
3434 nested_in_vect_loop_p since the scalar stmt may get invalidated.
3436 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3439 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
3440 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
3441 If off is not INTEGER_CST, issue a may not be aligned warning
3442 rather than isn't aligned. Use isn%'t rather than isn't.
3443 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
3445 <case MULT_EXPR>: Improve the case when bottom and one of the
3446 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
3447 operand, in that case check if the other operand is multiple of
3448 bottom divided by the INTEGER_CST operand.
3450 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3453 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
3454 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
3455 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
3456 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
3457 * config/pa/pa.c (pa_function_arg_advance): Likewise.
3458 (pa_function_arg, pa_arg_partial_bytes): Likewise.
3459 (pa_function_arg_size): New function.
3461 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3463 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
3464 in a separate statement.
3466 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3468 PR tree-optimization/83847
3469 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
3470 group gathers and scatters.
3472 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3474 PR rtl-optimization/86620
3475 * params.def (max-sched-ready-insns): Bump minimum value to 1.
3477 PR rtl-optimization/83213
3478 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
3479 to last if both are JUMP_INSNs.
3481 PR tree-optimization/83843
3482 * gimple-ssa-store-merging.c
3483 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
3484 store_immediate_info for bswap/nop orig_stores.
3486 2018-01-15 Andrew Waterman <andrew@sifive.com>
3488 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
3490 <UDIV>: Increase cost if !TARGET_DIV.
3492 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
3494 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
3495 (define_attr "cr_logical_3op"): New.
3496 (cceq_ior_compare): Adjust.
3497 (cceq_ior_compare_complement): Adjust.
3498 (*cceq_rev_compare): Adjust.
3499 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
3500 (is_cracked_insn): Adjust.
3501 (insn_must_be_first_in_group): Adjust.
3502 * config/rs6000/40x.md: Adjust.
3503 * config/rs6000/440.md: Adjust.
3504 * config/rs6000/476.md: Adjust.
3505 * config/rs6000/601.md: Adjust.
3506 * config/rs6000/603.md: Adjust.
3507 * config/rs6000/6xx.md: Adjust.
3508 * config/rs6000/7450.md: Adjust.
3509 * config/rs6000/7xx.md: Adjust.
3510 * config/rs6000/8540.md: Adjust.
3511 * config/rs6000/cell.md: Adjust.
3512 * config/rs6000/e300c2c3.md: Adjust.
3513 * config/rs6000/e500mc.md: Adjust.
3514 * config/rs6000/e500mc64.md: Adjust.
3515 * config/rs6000/e5500.md: Adjust.
3516 * config/rs6000/e6500.md: Adjust.
3517 * config/rs6000/mpc.md: Adjust.
3518 * config/rs6000/power4.md: Adjust.
3519 * config/rs6000/power5.md: Adjust.
3520 * config/rs6000/power6.md: Adjust.
3521 * config/rs6000/power7.md: Adjust.
3522 * config/rs6000/power8.md: Adjust.
3523 * config/rs6000/power9.md: Adjust.
3524 * config/rs6000/rs64.md: Adjust.
3525 * config/rs6000/titan.md: Adjust.
3527 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3529 * config/i386/predicates.md (indirect_branch_operand): Rewrite
3530 ix86_indirect_branch_register logic.
3532 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3534 * config/i386/constraints.md (Bs): Update
3535 ix86_indirect_branch_register check. Don't check
3536 ix86_indirect_branch_register with GOT_memory_operand.
3538 * config/i386/predicates.md (GOT_memory_operand): Don't check
3539 ix86_indirect_branch_register here.
3540 (GOT32_symbol_operand): Likewise.
3542 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3544 * config/i386/predicates.md (constant_call_address_operand):
3545 Rewrite ix86_indirect_branch_register logic.
3546 (sibcall_insn_operand): Likewise.
3548 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3550 * config/i386/constraints.md (Bs): Replace
3551 ix86_indirect_branch_thunk_register with
3552 ix86_indirect_branch_register.
3554 * config/i386/i386.md (indirect_jump): Likewise.
3555 (tablejump): Likewise.
3556 (*sibcall_memory): Likewise.
3557 (*sibcall_value_memory): Likewise.
3558 Peepholes of indirect call and jump via memory: Likewise.
3559 * config/i386/i386.opt: Likewise.
3560 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3561 (GOT_memory_operand): Likewise.
3562 (call_insn_operand): Likewise.
3563 (sibcall_insn_operand): Likewise.
3564 (GOT32_symbol_operand): Likewise.
3566 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3569 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
3570 type rather than type addr's type points to.
3571 (expand_omp_atomic_mutex): Likewise.
3572 (expand_omp_atomic): Likewise.
3574 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3577 * config/i386/i386.c (output_indirect_thunk_function): Use
3578 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
3579 for __x86_return_thunk.
3581 2018-01-15 Richard Biener <rguenther@suse.de>
3584 * expmed.c (extract_bit_field_1): Fix typo.
3586 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3589 * config/arm/iterators.md (VF): New mode iterator.
3590 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
3591 Remove integer-related logic from pattern.
3592 (neon_vabd<mode>_3): Likewise.
3594 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3597 * common.opt (fstrict-overflow): No longer an alias.
3598 (fwrapv-pointer): New option.
3599 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
3600 also for pointer types based on flag_wrapv_pointer.
3601 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
3602 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
3603 opts->x_flag_wrapv got set.
3604 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
3605 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
3606 POINTER_TYPE_OVERFLOW_UNDEFINED.
3607 * match.pd: Likewise in address comparison pattern.
3608 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
3610 2018-01-15 Richard Biener <rguenther@suse.de>
3613 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
3614 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
3615 Reset type names to their identifier if their TYPE_DECL doesn't
3616 have linkage (and thus is used for ODR and devirt).
3617 (save_debug_info_for_decl): Remove.
3618 (save_debug_info_for_type): Likewise.
3619 (add_tree_to_fld_list): Adjust.
3620 * tree-pretty-print.c (dump_generic_node): Make dumping of
3621 type names more robust.
3623 2018-01-15 Richard Biener <rguenther@suse.de>
3625 * BASE-VER: Bump to 8.0.1.
3627 2018-01-14 Martin Sebor <msebor@redhat.com>
3630 * builtins.c (check_access): Avoid warning when the no-warning bit
3633 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
3635 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
3636 * ira-color (allocno_hard_regs_compare): Likewise.
3638 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
3641 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
3642 Use .pushsection/.popsection.
3644 2018-01-14 Martin Sebor <msebor@redhat.com>
3647 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
3649 2018-01-14 Jakub Jelinek <jakub@redhat.com>
3651 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
3652 entry from extra_headers.
3653 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
3654 extra_headers, make the list bitwise identical to the i?86-*-* one.
3656 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3658 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
3659 -mcmodel=large with -mindirect-branch=thunk,
3660 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
3661 -mfunction-return=thunk-extern.
3662 * doc/invoke.texi: Document -mcmodel=large is incompatible with
3663 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
3664 -mfunction-return=thunk and -mfunction-return=thunk-extern.
3666 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3668 * config/i386/i386.c (print_reg): Print the name of the full
3669 integer register without '%'.
3670 (ix86_print_operand): Handle 'V'.
3671 * doc/extend.texi: Document 'V' modifier.
3673 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3675 * config/i386/constraints.md (Bs): Disallow memory operand for
3676 -mindirect-branch-register.
3678 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3679 (GOT_memory_operand): Likewise.
3680 (call_insn_operand): Likewise.
3681 (sibcall_insn_operand): Likewise.
3682 (GOT32_symbol_operand): Likewise.
3683 * config/i386/i386.md (indirect_jump): Call convert_memory_address
3684 for -mindirect-branch-register.
3685 (tablejump): Likewise.
3686 (*sibcall_memory): Likewise.
3687 (*sibcall_value_memory): Likewise.
3688 Disallow peepholes of indirect call and jump via memory for
3689 -mindirect-branch-register.
3690 (*call_pop): Replace m with Bw.
3691 (*call_value_pop): Likewise.
3692 (*sibcall_pop_memory): Replace m with Bs.
3693 * config/i386/i386.opt (mindirect-branch-register): New option.
3694 * doc/invoke.texi: Document -mindirect-branch-register option.
3696 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3698 * config/i386/i386-protos.h (ix86_output_function_return): New.
3699 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
3700 set function_return_type.
3701 (indirect_thunk_name): Add ret_p to indicate thunk for function
3703 (output_indirect_thunk_function): Pass false to
3704 indirect_thunk_name.
3705 (ix86_output_indirect_branch_via_reg): Likewise.
3706 (ix86_output_indirect_branch_via_push): Likewise.
3707 (output_indirect_thunk_function): Create alias for function
3708 return thunk if regno < 0.
3709 (ix86_output_function_return): New function.
3710 (ix86_handle_fndecl_attribute): Handle function_return.
3711 (ix86_attribute_table): Add function_return.
3712 * config/i386/i386.h (machine_function): Add
3713 function_return_type.
3714 * config/i386/i386.md (simple_return_internal): Use
3715 ix86_output_function_return.
3716 (simple_return_internal_long): Likewise.
3717 * config/i386/i386.opt (mfunction-return=): New option.
3718 (indirect_branch): Mention -mfunction-return=.
3719 * doc/extend.texi: Document function_return function attribute.
3720 * doc/invoke.texi: Document -mfunction-return= option.
3722 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3724 * config/i386/i386-opts.h (indirect_branch): New.
3725 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
3726 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
3727 with local indirect jump when converting indirect call and jump.
3728 (ix86_set_indirect_branch_type): New.
3729 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
3730 (indirectlabelno): New.
3731 (indirect_thunk_needed): Likewise.
3732 (indirect_thunk_bnd_needed): Likewise.
3733 (indirect_thunks_used): Likewise.
3734 (indirect_thunks_bnd_used): Likewise.
3735 (INDIRECT_LABEL): Likewise.
3736 (indirect_thunk_name): Likewise.
3737 (output_indirect_thunk): Likewise.
3738 (output_indirect_thunk_function): Likewise.
3739 (ix86_output_indirect_branch_via_reg): Likewise.
3740 (ix86_output_indirect_branch_via_push): Likewise.
3741 (ix86_output_indirect_branch): Likewise.
3742 (ix86_output_indirect_jmp): Likewise.
3743 (ix86_code_end): Call output_indirect_thunk_function if needed.
3744 (ix86_output_call_insn): Call ix86_output_indirect_branch if
3746 (ix86_handle_fndecl_attribute): Handle indirect_branch.
3747 (ix86_attribute_table): Add indirect_branch.
3748 * config/i386/i386.h (machine_function): Add indirect_branch_type
3749 and has_local_indirect_jump.
3750 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
3752 (tablejump): Likewise.
3753 (*indirect_jump): Use ix86_output_indirect_jmp.
3754 (*tablejump_1): Likewise.
3755 (simple_return_indirect_internal): Likewise.
3756 * config/i386/i386.opt (mindirect-branch=): New option.
3757 (indirect_branch): New.
3760 (thunk-inline): Likewise.
3761 (thunk-extern): Likewise.
3762 * doc/extend.texi: Document indirect_branch function attribute.
3763 * doc/invoke.texi: Document -mindirect-branch= option.
3765 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
3768 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
3770 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
3772 * ipa-inline.c (want_inline_small_function_p): Return false if
3773 inlining has already failed with CIF_FINAL_ERROR.
3774 (update_caller_keys): Call want_inline_small_function_p before
3776 (update_callee_keys): Likewise.
3778 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3780 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
3782 (rs6000_quadword_masked_address_p): Likewise.
3783 (quad_aligned_load_p): Likewise.
3784 (quad_aligned_store_p): Likewise.
3785 (const_load_sequence_p): Add comment to describe the outer-most loop.
3786 (mimic_memory_attributes_and_flags): New function.
3787 (rs6000_gen_stvx): Likewise.
3788 (replace_swapped_aligned_store): Likewise.
3789 (rs6000_gen_lvx): Likewise.
3790 (replace_swapped_aligned_load): Likewise.
3791 (replace_swapped_load_constant): Capitalize argument name in
3792 comment describing this function.
3793 (rs6000_analyze_swaps): Add a third pass to search for vector loads
3794 and stores that access quad-word aligned addresses and replace
3795 with stvx or lvx instructions when appropriate.
3796 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
3797 New function prototype.
3798 (rs6000_quadword_masked_address_p): Likewise.
3799 (rs6000_gen_lvx): Likewise.
3800 (rs6000_gen_stvx): Likewise.
3801 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
3802 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
3803 when memory address is aligned.
3804 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
3805 this split to select lvx instruction when memory address is aligned.
3806 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
3807 instruction when memory address is aligned.
3808 (*vsx_le_perm_load_v16qi): Likewise.
3809 (four unnamed splitters): Modify to select the stvx instruction
3810 when memory is aligned.
3812 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
3814 * predict.c (determine_unlikely_bbs): Handle correctly BBs
3815 which appears in the queue multiple times.
3817 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3818 Alan Hayward <alan.hayward@arm.com>
3819 David Sherwood <david.sherwood@arm.com>
3821 * tree-vectorizer.h (vec_lower_bound): New structure.
3822 (_loop_vec_info): Add check_nonzero and lower_bounds.
3823 (LOOP_VINFO_CHECK_NONZERO): New macro.
3824 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
3825 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
3826 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
3827 fields. Make seg_len the distance travelled, not including the
3829 (dr_direction_indicator): Declare.
3830 (dr_zero_step_indicator): Likewise.
3831 (dr_known_forward_stride_p): Likewise.
3832 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
3834 (runtime_alias_check_p): Allow runtime alias checks with
3836 (operator ==): Compare access_size and align.
3837 (prune_runtime_alias_test_list): Rework for new distinction between
3838 the access_size and seg_len.
3839 (create_intersect_range_checks_index): Likewise. Cope with polynomial
3841 (get_segment_min_max): New function.
3842 (create_intersect_range_checks): Use it.
3843 (dr_step_indicator): New function.
3844 (dr_direction_indicator): Likewise.
3845 (dr_zero_step_indicator): Likewise.
3846 (dr_known_forward_stride_p): Likewise.
3847 * tree-loop-distribution.c (data_ref_segment_size): Return
3848 DR_STEP * (niters - 1).
3849 (compute_alias_check_pairs): Update call to the dr_with_seg_len
3851 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
3852 (vect_preserves_scalar_order_p): New function, split out from...
3853 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
3854 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
3855 (vect_vfa_access_size): New function.
3856 (vect_vfa_align): Likewise.
3857 (vect_compile_time_alias): Take access_size_a and access_b arguments.
3858 (dump_lower_bound): New function.
3859 (vect_check_lower_bound): Likewise.
3860 (vect_small_gap_p): Likewise.
3861 (vectorizable_with_step_bound_p): Likewise.
3862 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
3863 depencies if the vectorization factor is 1. Convert the checks
3864 for nonzero steps into checks on the bounds of DR_STEP. Try using
3865 a bunds check for variable steps if the minimum required step is
3866 relatively small. Update calls to the dr_with_seg_len
3867 constructor and to vect_compile_time_alias.
3868 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
3870 (vect_loop_versioning): Call it.
3871 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
3873 (vect_estimate_min_profitable_iters): Account for any bounds checks.
3875 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3876 Alan Hayward <alan.hayward@arm.com>
3877 David Sherwood <david.sherwood@arm.com>
3879 * doc/sourcebuild.texi (vect_scatter_store): Document.
3880 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
3882 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
3884 * genopinit.c (main): Add supports_vec_scatter_store and
3885 supports_vec_scatter_store_cached to target_optabs.
3886 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
3887 IFN_MASK_SCATTER_STORE.
3888 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
3890 * internal-fn.h (internal_store_fn_p): Declare.
3891 (internal_fn_stored_value_index): Likewise.
3892 * internal-fn.c (scatter_store_direct): New macro.
3893 (expand_scatter_store_optab_fn): New function.
3894 (direct_scatter_store_optab_supported_p): New macro.
3895 (internal_store_fn_p): New function.
3896 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
3897 IFN_MASK_SCATTER_STORE.
3898 (internal_fn_mask_index): Likewise.
3899 (internal_fn_stored_value_index): New function.
3900 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
3902 * optabs-query.h (supports_vec_scatter_store_p): Declare.
3903 * optabs-query.c (supports_vec_scatter_store_p): New function.
3904 * tree-vectorizer.h (vect_get_store_rhs): Declare.
3905 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
3906 true for scatter stores.
3907 (vect_gather_scatter_fn_p): Handle scatter stores too.
3908 (vect_check_gather_scatter): Consider using scatter stores if
3909 supports_vec_scatter_store_p.
3910 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
3912 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
3913 internal_fn_stored_value_index.
3914 (check_load_store_masking): Handle scatter stores too.
3915 (vect_get_store_rhs): Make public.
3916 (vectorizable_call): Use internal_store_fn_p.
3917 (vectorizable_store): Handle scatter store internal functions.
3918 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
3919 when deciding whether the end of the group has been reached.
3920 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
3921 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
3922 (mask_scatter_store<mode>): New insns.
3924 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3925 Alan Hayward <alan.hayward@arm.com>
3926 David Sherwood <david.sherwood@arm.com>
3928 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
3929 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
3930 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
3932 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
3933 Use vect_truncate_gather_scatter_offset if we can't treat the
3934 operation as a normal gather load or scatter store.
3935 (get_group_load_store_type): Take the gather_scatter_info
3936 as argument. Try using a gather load or scatter store for
3937 single-element groups.
3938 (get_load_store_type): Update calls to get_group_load_store_type
3939 and vect_use_strided_gather_scatters_p.
3941 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3942 Alan Hayward <alan.hayward@arm.com>
3943 David Sherwood <david.sherwood@arm.com>
3945 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
3946 optional tree argument.
3947 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
3949 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
3950 but continue to use the current value as a fallback.
3951 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
3952 to compare the updates.
3953 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
3954 (get_load_store_type): Use it when handling a strided access.
3955 (vect_get_strided_load_store_ops): New function.
3956 (vect_get_data_ptr_increment): Likewise.
3957 (vectorizable_load): Handle strided gather loads. Always pass
3958 a step to vect_create_data_ref_ptr and bump_vector_ptr.
3960 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3961 Alan Hayward <alan.hayward@arm.com>
3962 David Sherwood <david.sherwood@arm.com>
3964 * doc/md.texi (gather_load@var{m}): Document.
3965 (mask_gather_load@var{m}): Likewise.
3966 * genopinit.c (main): Add supports_vec_gather_load and
3967 supports_vec_gather_load_cached to target_optabs.
3968 * optabs-tree.c (init_tree_optimization_optabs): Use
3969 ggc_cleared_alloc to allocate target_optabs.
3970 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
3971 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
3973 * internal-fn.h (internal_load_fn_p): Declare.
3974 (internal_gather_scatter_fn_p): Likewise.
3975 (internal_fn_mask_index): Likewise.
3976 (internal_gather_scatter_fn_supported_p): Likewise.
3977 * internal-fn.c (gather_load_direct): New macro.
3978 (expand_gather_load_optab_fn): New function.
3979 (direct_gather_load_optab_supported_p): New macro.
3980 (direct_internal_fn_optab): New function.
3981 (internal_load_fn_p): Likewise.
3982 (internal_gather_scatter_fn_p): Likewise.
3983 (internal_fn_mask_index): Likewise.
3984 (internal_gather_scatter_fn_supported_p): Likewise.
3985 * optabs-query.c (supports_at_least_one_mode_p): New function.
3986 (supports_vec_gather_load_p): Likewise.
3987 * optabs-query.h (supports_vec_gather_load_p): Declare.
3988 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
3989 and memory_type field.
3990 (NUM_PATTERNS): Bump to 15.
3991 * tree-vect-data-refs.c: Include internal-fn.h.
3992 (vect_gather_scatter_fn_p): New function.
3993 (vect_describe_gather_scatter_call): Likewise.
3994 (vect_check_gather_scatter): Try using internal functions for
3995 gather loads. Recognize existing calls to a gather load function.
3996 (vect_analyze_data_refs): Consider using gather loads if
3997 supports_vec_gather_load_p.
3998 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
3999 (vect_get_gather_scatter_offset_type): Likewise.
4000 (vect_convert_mask_for_vectype): Likewise.
4001 (vect_add_conversion_to_patterm): Likewise.
4002 (vect_try_gather_scatter_pattern): Likewise.
4003 (vect_recog_gather_scatter_pattern): New pattern recognizer.
4004 (vect_vect_recog_func_ptrs): Add it.
4005 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4006 internal_fn_mask_index and internal_gather_scatter_fn_p.
4007 (check_load_store_masking): Take the gather_scatter_info as an
4008 argument and handle gather loads.
4009 (vect_get_gather_scatter_ops): New function.
4010 (vectorizable_call): Check internal_load_fn_p.
4011 (vectorizable_load): Likewise. Handle gather load internal
4013 (vectorizable_store): Update call to check_load_store_masking.
4014 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
4015 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
4016 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
4017 (aarch64_gather_scale_operand_d): New predicates.
4018 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
4019 (mask_gather_load<mode>): New insns.
4021 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4022 Alan Hayward <alan.hayward@arm.com>
4023 David Sherwood <david.sherwood@arm.com>
4025 * optabs.def (fold_left_plus_optab): New optab.
4026 * doc/md.texi (fold_left_plus_@var{m}): Document.
4027 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
4028 * internal-fn.c (fold_left_direct): Define.
4029 (expand_fold_left_optab_fn): Likewise.
4030 (direct_fold_left_optab_supported_p): Likewise.
4031 * fold-const-call.c (fold_const_fold_left): New function.
4032 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
4033 * tree-parloops.c (valid_reduction_p): New function.
4034 (gather_scalar_reductions): Use it.
4035 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
4036 (vect_finish_replace_stmt): Declare.
4037 * tree-vect-loop.c (fold_left_reduction_fn): New function.
4038 (needs_fold_left_reduction_p): New function, split out from...
4039 (vect_is_simple_reduction): ...here. Accept reductions that
4040 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
4041 (vect_force_simple_reduction): Also store the reduction type in
4042 the assignment's STMT_VINFO_REDUC_TYPE.
4043 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
4044 (merge_with_identity): New function.
4045 (vect_expand_fold_left): Likewise.
4046 (vectorize_fold_left_reduction): Likewise.
4047 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
4048 scalar phi in place for it. Check for target support and reject
4049 cases that would reassociate the operation. Defer the transform
4050 phase to vectorize_fold_left_reduction.
4051 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
4052 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
4053 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
4055 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4057 * tree-if-conv.c (predicate_mem_writes): Remove redundant
4058 call to ifc_temp_var.
4060 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4061 Alan Hayward <alan.hayward@arm.com>
4062 David Sherwood <david.sherwood@arm.com>
4064 * target.def (legitimize_address_displacement): Take the original
4065 offset as a poly_int.
4066 * targhooks.h (default_legitimize_address_displacement): Update
4068 * targhooks.c (default_legitimize_address_displacement): Likewise.
4069 * doc/tm.texi: Regenerate.
4070 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
4071 as an argument, moving assert of ad->disp == ad->disp_term to...
4072 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
4073 Try calling targetm.legitimize_address_displacement before expanding
4074 the address rather than afterwards, and adjust for the new interface.
4075 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
4076 Match the new hook interface. Handle SVE addresses.
4077 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
4080 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4082 * Makefile.in (OBJS): Add early-remat.o.
4083 * target.def (select_early_remat_modes): New hook.
4084 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
4085 * doc/tm.texi: Regenerate.
4086 * targhooks.h (default_select_early_remat_modes): Declare.
4087 * targhooks.c (default_select_early_remat_modes): New function.
4088 * timevar.def (TV_EARLY_REMAT): New timevar.
4089 * passes.def (pass_early_remat): New pass.
4090 * tree-pass.h (make_pass_early_remat): Declare.
4091 * early-remat.c: New file.
4092 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
4094 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
4096 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4097 Alan Hayward <alan.hayward@arm.com>
4098 David Sherwood <david.sherwood@arm.com>
4100 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
4101 vfm1 with a bound_epilog parameter.
4102 (vect_do_peeling): Update calls accordingly, and move the prologue
4103 call earlier in the function. Treat the base bound_epilog as 0 for
4104 fully-masked loops and retain vf - 1 for other loops. Add 1 to
4105 this base when peeling for gaps.
4106 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
4107 with fully-masked loops.
4108 (vect_estimate_min_profitable_iters): Handle the single peeled
4109 iteration in that case.
4111 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4112 Alan Hayward <alan.hayward@arm.com>
4113 David Sherwood <david.sherwood@arm.com>
4115 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
4116 single-element interleaving even if the size is not a power of 2.
4117 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
4118 accesses for single-element interleaving if the group size is
4121 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4122 Alan Hayward <alan.hayward@arm.com>
4123 David Sherwood <david.sherwood@arm.com>
4125 * doc/md.texi (fold_extract_last_@var{m}): Document.
4126 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
4127 * optabs.def (fold_extract_last_optab): New optab.
4128 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
4129 * internal-fn.c (fold_extract_direct): New macro.
4130 (expand_fold_extract_optab_fn): Likewise.
4131 (direct_fold_extract_optab_supported_p): Likewise.
4132 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
4133 * tree-vect-loop.c (vect_model_reduction_cost): Handle
4134 EXTRACT_LAST_REDUCTION.
4135 (get_initial_def_for_reduction): Do not create an initial vector
4136 for EXTRACT_LAST_REDUCTION reductions.
4137 (vectorizable_reduction): Leave the scalar phi in place for
4138 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
4139 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
4140 epilogue code for EXTRACT_LAST_REDUCTION and defer the
4141 transform phase to vectorizable_condition.
4142 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
4144 (vect_finish_stmt_generation): ...here.
4145 (vect_finish_replace_stmt): New function.
4146 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
4147 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
4149 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
4151 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4152 Alan Hayward <alan.hayward@arm.com>
4153 David Sherwood <david.sherwood@arm.com>
4155 * doc/md.texi (extract_last_@var{m}): Document.
4156 * optabs.def (extract_last_optab): New optab.
4157 * internal-fn.def (EXTRACT_LAST): New internal function.
4158 * internal-fn.c (cond_unary_direct): New macro.
4159 (expand_cond_unary_optab_fn): Likewise.
4160 (direct_cond_unary_optab_supported_p): Likewise.
4161 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
4162 loops using EXTRACT_LAST.
4163 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
4164 (extract_last_<mode>): ...this optab.
4165 (vec_extract<mode><Vel>): Update accordingly.
4167 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4168 Alan Hayward <alan.hayward@arm.com>
4169 David Sherwood <david.sherwood@arm.com>
4171 * target.def (empty_mask_is_expensive): New hook.
4172 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
4173 * doc/tm.texi: Regenerate.
4174 * targhooks.h (default_empty_mask_is_expensive): Declare.
4175 * targhooks.c (default_empty_mask_is_expensive): New function.
4176 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
4177 if the target says that empty masks are expensive.
4178 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
4180 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
4182 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4183 Alan Hayward <alan.hayward@arm.com>
4184 David Sherwood <david.sherwood@arm.com>
4186 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
4187 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
4188 (vect_use_loop_mask_for_alignment_p): New function.
4189 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
4190 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
4191 niters_skip argument. Make sure that the first niters_skip elements
4192 of the first iteration are inactive.
4193 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
4194 Update call to vect_set_loop_masks_directly.
4195 (get_misalign_in_elems): New function, split out from...
4196 (vect_gen_prolog_loop_niters): ...here.
4197 (vect_update_init_of_dr): Take a code argument that specifies whether
4198 the adjustment should be added or subtracted.
4199 (vect_update_init_of_drs): Likewise.
4200 (vect_prepare_for_masked_peels): New function.
4201 (vect_do_peeling): Skip prologue peeling if we're using a mask
4202 instead. Update call to vect_update_inits_of_drs.
4203 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4205 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
4206 alignment. Do not include the number of peeled iterations in
4207 the minimum threshold in that case.
4208 (vectorizable_induction): Adjust the start value down by
4209 LOOP_VINFO_MASK_SKIP_NITERS iterations.
4210 (vect_transform_loop): Call vect_prepare_for_masked_peels.
4211 Take the number of skipped iterations into account when calculating
4213 * tree-vect-stmts.c (vect_gen_while_not): New function.
4215 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4216 Alan Hayward <alan.hayward@arm.com>
4217 David Sherwood <david.sherwood@arm.com>
4219 * doc/sourcebuild.texi (vect_fully_masked): Document.
4220 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
4222 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
4224 (vect_analyze_loop_2): ...here. Don't check the vectorization
4225 factor against the number of loop iterations if the loop is
4228 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4229 Alan Hayward <alan.hayward@arm.com>
4230 David Sherwood <david.sherwood@arm.com>
4232 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
4233 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
4234 (dump_groups): Update accordingly.
4235 (iv_use::mem_type): New member variable.
4236 (address_p): New function.
4237 (record_use): Add a mem_type argument and initialize the new
4239 (record_group_use): Add a mem_type argument. Use address_p.
4240 Remove obsolete null checks of base_object. Update call to record_use.
4241 (find_interesting_uses_op): Update call to record_group_use.
4242 (find_interesting_uses_cond): Likewise.
4243 (find_interesting_uses_address): Likewise.
4244 (get_mem_type_for_internal_fn): New function.
4245 (find_address_like_use): Likewise.
4246 (find_interesting_uses_stmt): Try find_address_like_use before
4247 calling find_interesting_uses_op.
4248 (addr_offset_valid_p): Use the iv mem_type field as the type
4249 of the addressed memory.
4250 (add_autoinc_candidates): Likewise.
4251 (get_address_cost): Likewise.
4252 (split_small_address_groups_p): Use address_p.
4253 (split_address_groups): Likewise.
4254 (add_iv_candidate_for_use): Likewise.
4255 (autoinc_possible_for_pair): Likewise.
4256 (rewrite_groups): Likewise.
4257 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
4258 (determine_group_iv_cost): Update after split of USE_ADDRESS.
4259 (get_alias_ptr_type_for_ptr_address): New function.
4260 (rewrite_use_address): Rewrite address uses in calls that were
4261 identified by find_address_like_use.
4263 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4264 Alan Hayward <alan.hayward@arm.com>
4265 David Sherwood <david.sherwood@arm.com>
4267 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
4269 * gimple-expr.h (is_gimple_addressable: Likewise.
4270 * gimple-expr.c (is_gimple_address): Likewise.
4271 * internal-fn.c (expand_call_mem_ref): New function.
4272 (expand_mask_load_optab_fn): Use it.
4273 (expand_mask_store_optab_fn): Likewise.
4275 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4276 Alan Hayward <alan.hayward@arm.com>
4277 David Sherwood <david.sherwood@arm.com>
4279 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
4280 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
4281 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
4282 (cond_umax@var{mode}): Document.
4283 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
4284 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
4285 (cond_umin_optab, cond_umax_optab): New optabs.
4286 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
4287 (COND_IOR, COND_XOR): New internal functions.
4288 * internal-fn.h (get_conditional_internal_fn): Declare.
4289 * internal-fn.c (cond_binary_direct): New macro.
4290 (expand_cond_binary_optab_fn): Likewise.
4291 (direct_cond_binary_optab_supported_p): Likewise.
4292 (get_conditional_internal_fn): New function.
4293 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
4294 Cope with reduction statements that are vectorized as calls rather
4296 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
4297 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
4298 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
4299 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4300 (UNSPEC_COND_EOR): New unspecs.
4301 (optab): Add mappings for them.
4302 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
4303 (sve_int_op, sve_fp_op): New int attributes.
4305 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4306 Alan Hayward <alan.hayward@arm.com>
4307 David Sherwood <david.sherwood@arm.com>
4309 * optabs.def (while_ult_optab): New optab.
4310 * doc/md.texi (while_ult@var{m}@var{n}): Document.
4311 * internal-fn.def (WHILE_ULT): New internal function.
4312 * internal-fn.h (direct_internal_fn_supported_p): New override
4313 that takes two types as argument.
4314 * internal-fn.c (while_direct): New macro.
4315 (expand_while_optab_fn): New function.
4316 (convert_optab_supported_p): Likewise.
4317 (direct_while_optab_supported_p): New macro.
4318 * wide-int.h (wi::udiv_ceil): New function.
4319 * tree-vectorizer.h (rgroup_masks): New structure.
4320 (vec_loop_masks): New typedef.
4321 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
4323 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
4324 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
4325 (vect_max_vf): New function.
4326 (slpeel_make_loop_iterate_ntimes): Delete.
4327 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
4328 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
4329 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
4330 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
4331 internal-fn.h, stor-layout.h and optabs-query.h.
4332 (vect_set_loop_mask): New function.
4333 (add_preheader_seq): Likewise.
4334 (add_header_seq): Likewise.
4335 (interleave_supported_p): Likewise.
4336 (vect_maybe_permute_loop_masks): Likewise.
4337 (vect_set_loop_masks_directly): Likewise.
4338 (vect_set_loop_condition_masked): Likewise.
4339 (vect_set_loop_condition_unmasked): New function, split out from
4340 slpeel_make_loop_iterate_ntimes.
4341 (slpeel_make_loop_iterate_ntimes): Rename to..
4342 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
4343 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
4344 (vect_do_peeling): Update call accordingly.
4345 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
4347 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4348 mask_compare_type, can_fully_mask_p and fully_masked_p.
4349 (release_vec_loop_masks): New function.
4350 (_loop_vec_info): Use it to free the loop masks.
4351 (can_produce_all_loop_masks_p): New function.
4352 (vect_get_max_nscalars_per_iter): Likewise.
4353 (vect_verify_full_masking): Likewise.
4354 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
4355 retries, and free the mask rgroups before retrying. Check loop-wide
4356 reasons for disallowing fully-masked loops. Make the final decision
4357 about whether use a fully-masked loop or not.
4358 (vect_estimate_min_profitable_iters): Do not assume that peeling
4359 for the number of iterations will be needed for fully-masked loops.
4360 (vectorizable_reduction): Disable fully-masked loops.
4361 (vectorizable_live_operation): Likewise.
4362 (vect_halve_mask_nunits): New function.
4363 (vect_double_mask_nunits): Likewise.
4364 (vect_record_loop_mask): Likewise.
4365 (vect_get_loop_mask): Likewise.
4366 (vect_transform_loop): Handle the case in which the final loop
4367 iteration might handle a partial vector. Call vect_set_loop_condition
4368 instead of slpeel_make_loop_iterate_ntimes.
4369 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
4370 (check_load_store_masking): New function.
4371 (prepare_load_store_mask): Likewise.
4372 (vectorizable_store): Handle fully-masked loops.
4373 (vectorizable_load): Likewise.
4374 (supportable_widening_operation): Use vect_halve_mask_nunits for
4376 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
4377 (vect_gen_while): New function.
4378 * config/aarch64/aarch64.md (umax<mode>3): New expander.
4379 (aarch64_uqdec<mode>): New insn.
4381 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4382 Alan Hayward <alan.hayward@arm.com>
4383 David Sherwood <david.sherwood@arm.com>
4385 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
4386 (reduc_xor_scal_optab): New optabs.
4387 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
4388 (reduc_xor_scal_@var{m}): Document.
4389 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
4390 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
4392 * fold-const-call.c (fold_const_call): Handle them.
4393 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
4394 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
4395 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
4396 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
4397 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
4398 (UNSPEC_XORV): New unspecs.
4399 (optab): Add entries for them.
4400 (BITWISEV): New int iterator.
4401 (bit_reduc_op): New int attributes.
4403 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4404 Alan Hayward <alan.hayward@arm.com>
4405 David Sherwood <david.sherwood@arm.com>
4407 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
4408 * internal-fn.def (VEC_SHL_INSERT): New internal function.
4409 * optabs.def (vec_shl_insert_optab): New optab.
4410 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
4411 (duplicate_and_interleave): Likewise.
4412 * tree-vect-loop.c: Include internal-fn.h.
4413 (neutral_op_for_slp_reduction): New function, split out from
4414 get_initial_defs_for_reduction.
4415 (get_initial_def_for_reduction): Handle option 2 for variable-length
4416 vectors by loading the neutral value into a vector and then shifting
4417 the initial value into element 0.
4418 (get_initial_defs_for_reduction): Replace the code argument with
4419 the neutral value calculated by neutral_op_for_slp_reduction.
4420 Use gimple_build_vector for constant-length vectors.
4421 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
4422 but the first group_size elements have a neutral value.
4423 Use duplicate_and_interleave otherwise.
4424 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
4425 Update call to get_initial_defs_for_reduction. Handle SLP
4426 reductions for variable-length vectors by creating one vector
4427 result for each scalar result, with the elements associated
4428 with other scalar results stubbed out with the neutral value.
4429 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
4430 Require IFN_VEC_SHL_INSERT for double reductions on
4431 variable-length vectors, or SLP reductions that have
4432 a neutral value. Require can_duplicate_and_interleave_p
4433 support for variable-length unchained SLP reductions if there
4434 is no neutral value, such as for MIN/MAX reductions. Also require
4435 the number of vector elements to be a multiple of the number of
4436 SLP statements when doing variable-length unchained SLP reductions.
4437 Update call to vect_create_epilog_for_reduction.
4438 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
4439 and remove initial values.
4440 (duplicate_and_interleave): Make public.
4441 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
4442 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
4444 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4445 Alan Hayward <alan.hayward@arm.com>
4446 David Sherwood <david.sherwood@arm.com>
4448 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
4449 (can_duplicate_and_interleave_p): New function.
4450 (vect_get_and_check_slp_defs): Take the vector of statements
4451 rather than just the current one. Remove excess parentheses.
4452 Restriction rejectinon of vect_constant_def and vect_external_def
4453 for variable-length vectors to boolean types, or types for which
4454 can_duplicate_and_interleave_p is false.
4455 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
4456 (duplicate_and_interleave): New function.
4457 (vect_get_constant_vectors): Use gimple_build_vector for
4458 constant-length vectors and suitable variable-length constant
4459 vectors. Use duplicate_and_interleave for other variable-length
4460 vectors. Don't defer the update when inserting new statements.
4462 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4463 Alan Hayward <alan.hayward@arm.com>
4464 David Sherwood <david.sherwood@arm.com>
4466 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
4467 min_profitable_iters doesn't go negative.
4469 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4470 Alan Hayward <alan.hayward@arm.com>
4471 David Sherwood <david.sherwood@arm.com>
4473 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
4474 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
4475 * optabs.def (vec_mask_load_lanes_optab): New optab.
4476 (vec_mask_store_lanes_optab): Likewise.
4477 * internal-fn.def (MASK_LOAD_LANES): New internal function.
4478 (MASK_STORE_LANES): Likewise.
4479 * internal-fn.c (mask_load_lanes_direct): New macro.
4480 (mask_store_lanes_direct): Likewise.
4481 (expand_mask_load_optab_fn): Handle masked operations.
4482 (expand_mask_load_lanes_optab_fn): New macro.
4483 (expand_mask_store_optab_fn): Handle masked operations.
4484 (expand_mask_store_lanes_optab_fn): New macro.
4485 (direct_mask_load_lanes_optab_supported_p): Likewise.
4486 (direct_mask_store_lanes_optab_supported_p): Likewise.
4487 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
4489 (vect_load_lanes_supported): Likewise.
4490 * tree-vect-data-refs.c (strip_conversion): New function.
4491 (can_group_stmts_p): Likewise.
4492 (vect_analyze_data_ref_accesses): Use it instead of checking
4493 for a pair of assignments.
4494 (vect_store_lanes_supported): Take a masked_p parameter.
4495 (vect_load_lanes_supported): Likewise.
4496 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
4497 vect_store_lanes_supported and vect_load_lanes_supported.
4498 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4499 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
4500 parameter. Don't allow gaps for masked accesses.
4501 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
4502 and vect_load_lanes_supported.
4503 (get_load_store_type): Take a masked_p parameter and update
4504 call to get_group_load_store_type.
4505 (vectorizable_store): Update call to get_load_store_type.
4506 Handle IFN_MASK_STORE_LANES.
4507 (vectorizable_load): Update call to get_load_store_type.
4508 Handle IFN_MASK_LOAD_LANES.
4510 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4511 Alan Hayward <alan.hayward@arm.com>
4512 David Sherwood <david.sherwood@arm.com>
4514 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
4516 * config/aarch64/aarch64-protos.h
4517 (aarch64_sve_struct_memory_operand_p): Declare.
4518 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
4519 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
4520 (VPRED, vpred): Handle SVE structure modes.
4521 * config/aarch64/constraints.md (Utx): New constraint.
4522 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
4523 (aarch64_sve_struct_nonimmediate_operand): New predicates.
4524 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
4525 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
4526 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
4527 structure modes. Split into pieces after RA.
4528 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
4529 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
4531 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
4532 SVE structure modes.
4533 (aarch64_classify_address): Likewise.
4534 (sizetochar): Move earlier in file.
4535 (aarch64_print_operand): Handle SVE register lists.
4536 (aarch64_array_mode): New function.
4537 (aarch64_sve_struct_memory_operand_p): Likewise.
4538 (TARGET_ARRAY_MODE): Redefine.
4540 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4541 Alan Hayward <alan.hayward@arm.com>
4542 David Sherwood <david.sherwood@arm.com>
4544 * target.def (array_mode): New target hook.
4545 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
4546 * doc/tm.texi: Regenerate.
4547 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
4548 * hooks.c (hook_optmode_mode_uhwi_none): New function.
4549 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
4551 * stor-layout.c (mode_for_array): Likewise. Support polynomial
4554 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4555 Alan Hayward <alan.hayward@arm.com>
4556 David Sherwood <david.sherwood@arm.com>
4558 * fold-const.c (fold_binary_loc): Check the argument types
4559 rather than the result type when testing for a vector operation.
4561 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4563 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
4564 * doc/tm.texi: Regenerate.
4566 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4567 Alan Hayward <alan.hayward@arm.com>
4568 David Sherwood <david.sherwood@arm.com>
4570 * doc/invoke.texi (-msve-vector-bits=): Document new option.
4571 (sve): Document new AArch64 extension.
4572 * doc/md.texi (w): Extend the description of the AArch64
4573 constraint to include SVE vectors.
4574 (Upl, Upa): Document new AArch64 predicate constraints.
4575 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
4577 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
4578 (msve-vector-bits=): New option.
4579 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
4580 SVE when these are disabled.
4581 (sve): New extension.
4582 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
4583 modes. Adjust their number of units based on aarch64_sve_vg.
4584 (MAX_BITSIZE_MODE_ANY_MODE): Define.
4585 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
4586 aarch64_addr_query_type.
4587 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
4588 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
4589 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
4590 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
4591 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
4592 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
4593 (aarch64_simd_imm_zero_p): Delete.
4594 (aarch64_check_zero_based_sve_index_immediate): Declare.
4595 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4596 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4597 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4598 (aarch64_sve_float_mul_immediate_p): Likewise.
4599 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4601 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
4602 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
4603 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
4604 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
4605 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
4606 (aarch64_regmode_natural_size): Likewise.
4607 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
4608 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
4610 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
4611 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
4612 for VG and the SVE predicate registers.
4613 (V_ALIASES): Add a "z"-prefixed alias.
4614 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
4615 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
4616 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
4617 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
4618 (REG_CLASS_NAMES): Add entries for them.
4619 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
4620 and the predicate registers.
4621 (aarch64_sve_vg): Declare.
4622 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
4623 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
4624 (REGMODE_NATURAL_SIZE): Define.
4625 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
4627 * config/aarch64/aarch64.c: Include cfgrtl.h.
4628 (simd_immediate_info): Add a constructor for series vectors,
4629 and an associated step field.
4630 (aarch64_sve_vg): New variable.
4631 (aarch64_dbx_register_number): Handle VG and the predicate registers.
4632 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
4633 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
4634 (VEC_ANY_DATA, VEC_STRUCT): New constants.
4635 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
4636 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
4637 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
4638 (aarch64_get_mask_mode): New functions.
4639 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
4640 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
4641 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
4642 predicate modes and predicate registers. Explicitly restrict
4643 GPRs to modes of 16 bytes or smaller. Only allow FP registers
4644 to store a vector mode if it is recognized by
4645 aarch64_classify_vector_mode.
4646 (aarch64_regmode_natural_size): New function.
4647 (aarch64_hard_regno_caller_save_mode): Return the original mode
4649 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
4650 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
4651 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
4652 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
4654 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
4655 does not overlap dest if the function is frame-related. Handle
4657 (aarch64_split_add_offset): New function.
4658 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
4659 them aarch64_add_offset.
4660 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
4661 and update call to aarch64_sub_sp.
4662 (aarch64_add_cfa_expression): New function.
4663 (aarch64_expand_prologue): Pass extra temporary registers to the
4664 functions above. Handle the case in which we need to emit new
4665 DW_CFA_expressions for registers that were originally saved
4666 relative to the stack pointer, but now have to be expressed
4667 relative to the frame pointer.
4668 (aarch64_output_mi_thunk): Pass extra temporary registers to the
4670 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
4671 IP0 and IP1 values for SVE frames.
4672 (aarch64_expand_vec_series): New function.
4673 (aarch64_expand_sve_widened_duplicate): Likewise.
4674 (aarch64_expand_sve_const_vector): Likewise.
4675 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
4676 Handle SVE constants. Use emit_move_insn to move a force_const_mem
4677 into the register, rather than emitting a SET directly.
4678 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
4679 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
4680 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
4681 (offset_9bit_signed_scaled_p): New functions.
4682 (aarch64_replicate_bitmask_imm): New function.
4683 (aarch64_bitmask_imm): Use it.
4684 (aarch64_cannot_force_const_mem): Reject expressions involving
4685 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
4686 (aarch64_classify_index): Handle SVE indices, by requiring
4687 a plain register index with a scale that matches the element size.
4688 (aarch64_classify_address): Handle SVE addresses. Assert that
4689 the mode of the address is VOIDmode or an integer mode.
4690 Update call to aarch64_classify_symbol.
4691 (aarch64_classify_symbolic_expression): Update call to
4692 aarch64_classify_symbol.
4693 (aarch64_const_vec_all_in_range_p): New function.
4694 (aarch64_print_vector_float_operand): Likewise.
4695 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
4696 "vN" for FP registers with SVE modes. Handle (const ...) vectors
4697 and the FP immediates 1.0 and 0.5.
4698 (aarch64_print_address_internal): Handle SVE addresses.
4699 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
4700 (aarch64_regno_regclass): Handle predicate registers.
4701 (aarch64_secondary_reload): Handle big-endian reloads of SVE
4703 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
4704 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
4705 (aarch64_convert_sve_vector_bits): New function.
4706 (aarch64_override_options): Use it to handle -msve-vector-bits=.
4707 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4709 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
4710 Handle SVE vector and predicate modes. Accept VL-based constants
4711 that need only one temporary register, and VL offsets that require
4712 no temporary registers.
4713 (aarch64_conditional_register_usage): Mark the predicate registers
4714 as fixed if SVE isn't available.
4715 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
4716 Return true for SVE vector and predicate modes.
4717 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
4718 rather than an unsigned int. Handle SVE modes.
4719 (aarch64_preferred_simd_mode): Update call accordingly. Handle
4721 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
4723 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4724 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4725 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4726 (aarch64_sve_float_mul_immediate_p): New functions.
4727 (aarch64_sve_valid_immediate): New function.
4728 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
4729 Explicitly reject structure modes. Check for INDEX constants.
4730 Handle PTRUE and PFALSE constants.
4731 (aarch64_check_zero_based_sve_index_immediate): New function.
4732 (aarch64_simd_imm_zero_p): Delete.
4733 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
4734 vector modes. Accept constants in the range of CNT[BHWD].
4735 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
4736 ask for an Advanced SIMD mode.
4737 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
4738 (aarch64_simd_vector_alignment): Handle SVE predicates.
4739 (aarch64_vectorize_preferred_vector_alignment): New function.
4740 (aarch64_simd_vector_alignment_reachable): Use it instead of
4742 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
4743 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
4745 (MAX_VECT_LEN): Delete.
4746 (expand_vec_perm_d): Add a vec_flags field.
4747 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
4748 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
4749 (aarch64_evpc_ext): Don't apply a big-endian lane correction
4751 (aarch64_evpc_rev): Rename to...
4752 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
4753 (aarch64_evpc_rev_global): New function.
4754 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
4755 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
4757 (aarch64_evpc_sve_tbl): New function.
4758 (aarch64_expand_vec_perm_const_1): Update after rename of
4759 aarch64_evpc_rev. Handle SVE permutes too, trying
4760 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
4761 than aarch64_evpc_tbl.
4762 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
4763 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
4764 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
4765 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
4766 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
4767 (aarch64_expand_sve_vcond): New functions.
4768 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
4769 of aarch64_vector_mode_p.
4770 (aarch64_dwarf_poly_indeterminate_value): New function.
4771 (aarch64_compute_pressure_classes): Likewise.
4772 (aarch64_can_change_mode_class): Likewise.
4773 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
4774 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
4775 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
4776 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
4777 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
4778 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
4779 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
4780 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
4782 (Dn, Dl, Dr): Accept const as well as const_vector.
4783 (Dz): Likewise. Compare against CONST0_RTX.
4784 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
4785 of "vector" where appropriate.
4786 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
4787 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
4788 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
4789 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
4790 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
4791 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
4792 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
4793 (v_int_equiv): Extend to SVE modes.
4794 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
4796 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
4797 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
4798 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
4799 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
4800 (SVE_COND_FP_CMP): New int iterators.
4801 (perm_hilo): Handle the new unpack unspecs.
4802 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
4804 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
4805 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
4806 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
4807 (aarch64_equality_operator, aarch64_constant_vector_operand)
4808 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
4809 (aarch64_sve_nonimmediate_operand): Likewise.
4810 (aarch64_sve_general_operand): Likewise.
4811 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
4812 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
4813 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
4814 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
4815 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
4816 (aarch64_sve_float_arith_immediate): Likewise.
4817 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
4818 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
4819 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
4820 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
4821 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
4822 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
4823 (aarch64_sve_float_arith_operand): Likewise.
4824 (aarch64_sve_float_arith_with_sub_operand): Likewise.
4825 (aarch64_sve_float_mul_operand): Likewise.
4826 (aarch64_sve_vec_perm_operand): Likewise.
4827 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
4828 (aarch64_mov_operand): Accept const_poly_int and const_vector.
4829 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
4830 as well as const_vector.
4831 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
4832 in file. Use CONST0_RTX and CONSTM1_RTX.
4833 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
4834 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
4835 Use aarch64_simd_imm_zero.
4836 * config/aarch64/aarch64-sve.md: New file.
4837 * config/aarch64/aarch64.md: Include it.
4838 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
4839 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
4840 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
4841 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
4842 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
4843 (sve): New attribute.
4844 (enabled): Disable instructions with the sve attribute unless
4846 (movqi, movhi): Pass CONST_POLY_INT operaneds through
4847 aarch64_expand_mov_immediate.
4848 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
4849 CNT[BHSD] immediates.
4850 (movti): Split CONST_POLY_INT moves into two halves.
4851 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
4852 Split additions that need a temporary here if the destination
4853 is the stack pointer.
4854 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
4855 (*add<mode>3_poly_1): New instruction.
4856 (set_clobber_cc): New expander.
4858 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4860 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
4861 parameter and use it instead of GET_MODE_SIZE (innermode). Use
4862 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
4863 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
4864 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
4865 Change innermode from fixed_mode_size to machine_mode.
4866 (simplify_subreg): Update call accordingly. Handle a constant-sized
4867 subreg of a variable-length CONST_VECTOR.
4869 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4870 Alan Hayward <alan.hayward@arm.com>
4871 David Sherwood <david.sherwood@arm.com>
4873 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
4874 (add_offset_to_base): New function, split out from...
4875 (create_mem_ref): ...here. When handling a scale other than 1,
4876 check first whether the address is valid without the offset.
4877 Add it into the base if so, leaving the index and scale as-is.
4879 2018-01-12 Jakub Jelinek <jakub@redhat.com>
4882 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
4883 fold_for_warn before checking if arg2 is INTEGER_CST.
4885 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
4887 * config/rs6000/predicates.md (load_multiple_operation): Delete.
4888 (store_multiple_operation): Delete.
4889 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
4890 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
4891 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
4892 guarded by TARGET_STRING.
4893 (rs6000_output_load_multiple): Delete.
4894 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
4895 OPTION_MASK_STRING / TARGET_STRING handling.
4896 (print_operand) <'N', 'O'>: Add comment that these are unused now.
4897 (const rs6000_opt_masks) <"string">: Change mask to 0.
4898 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
4899 (MASK_STRING): Delete.
4900 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
4902 (load_multiple): Delete.
4909 (store_multiple): Delete.
4916 (movmemsi_8reg): Delete.
4917 (corresponding unnamed define_insn): Delete.
4918 (movmemsi_6reg): Delete.
4919 (corresponding unnamed define_insn): Delete.
4920 (movmemsi_4reg): Delete.
4921 (corresponding unnamed define_insn): Delete.
4922 (movmemsi_2reg): Delete.
4923 (corresponding unnamed define_insn): Delete.
4924 (movmemsi_1reg): Delete.
4925 (corresponding unnamed define_insn): Delete.
4926 * config/rs6000/rs6000.opt (mno-string): New.
4927 (mstring): Replace by deprecation warning stub.
4928 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
4930 2018-01-12 Jakub Jelinek <jakub@redhat.com>
4932 * regrename.c (regrename_do_replace): If replacing the same
4933 reg multiple times, try to reuse last created gen_raw_REG.
4936 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
4937 main to workaround a bug in GDB.
4939 2018-01-12 Tom de Vries <tom@codesourcery.com>
4942 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
4944 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
4946 PR rtl-optimization/80481
4947 * ira-color.c (get_cap_member): New function.
4948 (allocnos_conflict_by_live_ranges_p): Use it.
4949 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
4950 (setup_slot_coalesced_allocno_live_ranges): Ditto.
4952 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
4955 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
4956 (*saddl_se_1): Ditto.
4958 (*ssubl_se_1): Ditto.
4960 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
4962 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
4963 rather than wi::to_widest for DR_INITs.
4964 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
4965 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
4966 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
4968 (vect_analyze_group_access_1): Note that here.
4970 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
4972 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
4973 polynomial type sizes.
4975 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
4977 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
4978 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
4979 (gimple_add_tmp_var): Likewise.
4981 2018-01-12 Martin Liska <mliska@suse.cz>
4983 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
4984 (gimple_alloc_sizes): Likewise.
4985 (dump_gimple_statistics): Use PRIu64 in printf format.
4986 * gimple.h: Change uint64_t to int.
4988 2018-01-12 Martin Liska <mliska@suse.cz>
4990 * tree-core.h: Use uint64_t instead of int.
4991 * tree.c (tree_node_counts): Likewise.
4992 (tree_node_sizes): Likewise.
4993 (dump_tree_statistics): Use PRIu64 in printf format.
4995 2018-01-12 Martin Liska <mliska@suse.cz>
4997 * Makefile.in: As qsort_chk is implemented in vec.c, add
4998 vec.o to linkage of gencfn-macros.
4999 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
5000 passing the info to record_node_allocation_statistics.
5001 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
5003 * ggc-common.c (struct ggc_usage): Add operator== and use
5004 it in operator< and compare function.
5005 * mem-stats.h (struct mem_usage): Likewise.
5006 * vec.c (struct vec_usage): Remove operator< and compare
5007 function. Can be simply inherited.
5009 2018-01-12 Martin Jambor <mjambor@suse.cz>
5012 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
5013 * tree-ssa-math-opts.c: Include domwalk.h.
5014 (convert_mult_to_fma_1): New function.
5015 (fma_transformation_info): New type.
5016 (fma_deferring_state): Likewise.
5017 (cancel_fma_deferring): New function.
5018 (result_of_phi): Likewise.
5019 (last_fma_candidate_feeds_initial_phi): Likewise.
5020 (convert_mult_to_fma): Added deferring logic, split actual
5021 transformation to convert_mult_to_fma_1.
5022 (math_opts_dom_walker): New type.
5023 (math_opts_dom_walker::after_dom_children): New method, body moved
5024 here from pass_optimize_widening_mul::execute, added deferring logic
5026 (pass_optimize_widening_mul::execute): Moved most of code to
5027 math_opts_dom_walker::after_dom_children.
5028 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
5029 * config/i386/i386.c (ix86_option_override_internal): Added
5030 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
5032 2018-01-12 Richard Biener <rguenther@suse.de>
5035 * dwarf2out.c (gen_variable_die): Do not reset old_die for
5036 inline instance vars.
5038 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
5041 * config/rx/rx.c (rx_is_restricted_memory_address):
5044 2018-01-12 Richard Biener <rguenther@suse.de>
5046 PR tree-optimization/80846
5047 * target.def (split_reduction): New target hook.
5048 * targhooks.c (default_split_reduction): New function.
5049 * targhooks.h (default_split_reduction): Declare.
5050 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
5051 target requests first reduce vectors by combining low and high
5053 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
5054 (get_vectype_for_scalar_type_and_size): Export.
5055 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
5056 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
5057 * doc/tm.texi: Regenerate.
5058 * config/i386/i386.c (ix86_split_reduction): Implement
5059 TARGET_VECTORIZE_SPLIT_REDUCTION.
5061 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5064 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
5065 in PIC mode except for TARGET_VXWORKS_RTP.
5066 * config/sparc/sparc.c: Include cfgrtl.h.
5067 (TARGET_INIT_PIC_REG): Define.
5068 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
5069 (sparc_pic_register_p): New predicate.
5070 (sparc_legitimate_address_p): Use it.
5071 (sparc_legitimize_pic_address): Likewise.
5072 (sparc_delegitimize_address): Likewise.
5073 (sparc_mode_dependent_address_p): Likewise.
5074 (gen_load_pcrel_sym): Remove 4th parameter.
5075 (load_got_register): Adjust call to above. Remove obsolete stuff.
5076 (sparc_expand_prologue): Do not call load_got_register here.
5077 (sparc_flat_expand_prologue): Likewise.
5078 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
5079 (sparc_use_pseudo_pic_reg): New function.
5080 (sparc_init_pic_reg): Likewise.
5081 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5082 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
5084 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
5086 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
5087 Add item for branch_cost.
5089 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5091 PR rtl-optimization/83565
5092 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
5093 not extend the result to a larger mode for rotate operations.
5094 (num_sign_bit_copies1): Likewise.
5096 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5099 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
5101 Use values-Xc.o for -pedantic.
5102 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
5104 2018-01-12 Martin Liska <mliska@suse.cz>
5107 * ipa-devirt.c (final_warning_record::grow_type_warnings):
5109 (possible_polymorphic_call_targets): Use it.
5110 (ipa_devirt): Likewise.
5112 2018-01-12 Martin Liska <mliska@suse.cz>
5114 * profile-count.h (enum profile_quality): Use 0 as invalid
5115 enum value of profile_quality.
5117 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
5119 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
5120 -mext-string options.
5122 2018-01-12 Richard Biener <rguenther@suse.de>
5124 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
5125 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
5126 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
5128 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
5130 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5132 * configure.ac (--with-long-double-format): Add support for the
5133 configuration option to change the default long double format on
5135 * config.gcc (powerpc*-linux*-*): Likewise.
5136 * configure: Regenerate.
5137 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
5138 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
5139 used without modification.
5141 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5143 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
5144 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
5145 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
5146 MISC_BUILTIN_SPEC_BARRIER.
5147 (rs6000_init_builtins): Likewise.
5148 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
5150 (speculation_barrier): New define_insn.
5151 * doc/extend.texi: Document __builtin_speculation_barrier.
5153 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5156 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
5157 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
5158 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
5160 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
5161 integral modes instead of "ss" and "sd".
5162 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
5163 vectors with 32-bit and 64-bit elements.
5164 (vecdupssescalarmodesuffix): New mode attribute.
5165 (vec_dup<mode>): Use it.
5167 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
5170 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
5171 frame if argument is passed on stack.
5173 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5176 * ree.c (combine_reaching_defs): Optimize also
5177 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
5178 reg2=any_extend(exp); reg1=reg2;, formatting fix.
5180 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5183 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
5185 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5188 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
5189 after they are computed.
5191 2018-01-11 Bin Cheng <bin.cheng@arm.com>
5193 PR tree-optimization/83695
5194 * gimple-loop-linterchange.cc
5195 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
5196 reset cached scev information after interchange.
5197 (pass_linterchange::execute): Remove call to scev_reset_htab.
5199 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5201 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
5202 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
5203 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
5204 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
5205 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
5206 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
5207 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
5208 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
5209 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
5210 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
5211 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
5212 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
5213 (V_lane_reg): Likewise.
5214 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
5216 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
5217 (vfmal_lane_low<mode>_intrinsic,
5218 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
5219 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
5220 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
5221 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
5222 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
5223 vfmsl_lane_high<mode>_intrinsic): New define_insns.
5225 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5227 * config/arm/arm-cpus.in (fp16fml): New feature.
5228 (ALL_SIMD): Add fp16fml.
5229 (armv8.2-a): Add fp16fml as an option.
5230 (armv8.3-a): Likewise.
5231 (armv8.4-a): Add fp16fml as part of fp16.
5232 * config/arm/arm.h (TARGET_FP16FML): Define.
5233 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
5235 * config/arm/arm-modes.def (V2HF): Define.
5236 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
5237 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
5238 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
5239 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
5240 vfmsl_low, vfmsl_high): New set of builtins.
5241 * config/arm/iterators.md (PLUSMINUS): New code iterator.
5242 (vfml_op): New code attribute.
5243 (VFMLHALVES): New int iterator.
5244 (VFML, VFMLSEL): New mode attributes.
5245 (V_reg): Define mapping for V2HF.
5246 (V_hi, V_lo): New mode attributes.
5247 (VF_constraint): Likewise.
5248 (vfml_half, vfml_half_selector): New int attributes.
5249 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
5251 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
5252 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
5254 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
5255 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
5256 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
5257 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
5259 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
5260 Document new effective target and option set.
5262 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5264 * config/arm/arm-cpus.in (armv8_4): New feature.
5265 (ARMv8_4a): New fgroup.
5266 (armv8.4-a): New arch.
5267 * config/arm/arm-tables.opt: Regenerate.
5268 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
5269 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
5270 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
5271 Add matching rules for -march=armv8.4-a and extensions.
5272 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
5274 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
5277 * config/rx/rx.md (BW): New mode attribute.
5278 (sync_lock_test_and_setsi): Add mode suffix to insn output.
5280 2018-01-11 Richard Biener <rguenther@suse.de>
5282 PR tree-optimization/83435
5283 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
5284 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
5285 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
5287 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5288 Alan Hayward <alan.hayward@arm.com>
5289 David Sherwood <david.sherwood@arm.com>
5291 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
5293 (aarch64_classify_address): Initialize it. Track polynomial offsets.
5294 (aarch64_print_address_internal): Use it to check for a zero offset.
5296 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5297 Alan Hayward <alan.hayward@arm.com>
5298 David Sherwood <david.sherwood@arm.com>
5300 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
5301 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
5302 Return a poly_int64 rather than a HOST_WIDE_INT.
5303 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
5304 rather than a HOST_WIDE_INT.
5305 * config/aarch64/aarch64.h (aarch64_frame): Protect with
5306 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
5307 hard_fp_offset, frame_size, initial_adjust, callee_offset and
5308 final_offset from HOST_WIDE_INT to poly_int64.
5309 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
5310 to_constant when getting the number of units in an Advanced SIMD
5312 (aarch64_builtin_vectorized_function): Check for a constant number
5314 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
5316 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
5317 attribute instead of GET_MODE_NUNITS.
5318 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
5319 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
5320 GET_MODE_SIZE for fixed-size registers.
5321 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
5322 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
5323 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
5324 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
5325 (aarch64_print_operand, aarch64_print_address_internal)
5326 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
5327 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
5328 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
5329 Handle polynomial GET_MODE_SIZE.
5330 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
5331 wider than SImode without modification.
5332 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
5333 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
5334 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
5335 passing and returning SVE modes.
5336 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
5337 rather than GEN_INT.
5338 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
5339 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
5340 (aarch64_allocate_and_probe_stack_space): Likewise.
5341 (aarch64_layout_frame): Cope with polynomial offsets.
5342 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
5343 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
5345 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
5346 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
5347 poly_int64 rather than a HOST_WIDE_INT.
5348 (aarch64_get_separate_components, aarch64_process_components)
5349 (aarch64_expand_prologue, aarch64_expand_epilogue)
5350 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
5351 (aarch64_anchor_offset): New function, split out from...
5352 (aarch64_legitimize_address): ...here.
5353 (aarch64_builtin_vectorization_cost): Handle polynomial
5354 TYPE_VECTOR_SUBPARTS.
5355 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
5357 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
5358 number of elements from the PARALLEL rather than the mode.
5359 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
5360 rather than GET_MODE_BITSIZE.
5361 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
5362 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
5363 (aarch64_expand_vec_perm_const_1): Handle polynomial
5364 d->perm.length () and d->perm elements.
5365 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
5366 Apply to_constant to d->perm elements.
5367 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
5368 polynomial CONST_VECTOR_NUNITS.
5369 (aarch64_move_pointer): Take amount as a poly_int64 rather
5371 (aarch64_progress_pointer): Avoid temporary variable.
5372 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
5373 the mode attribute instead of GET_MODE.
5375 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5376 Alan Hayward <alan.hayward@arm.com>
5377 David Sherwood <david.sherwood@arm.com>
5379 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
5380 x exists before using it.
5381 (aarch64_add_constant_internal): Rename to...
5382 (aarch64_add_offset_1): ...this. Replace regnum with separate
5383 src and dest rtxes. Handle the case in which they're different,
5384 including when the offset is zero. Replace scratchreg with an rtx.
5385 Use 2 additions if there is no spare register into which we can
5386 move a 16-bit constant.
5387 (aarch64_add_constant): Delete.
5388 (aarch64_add_offset): Replace reg with separate src and dest
5389 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
5390 Use aarch64_add_offset_1.
5391 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
5392 an rtx rather than an int. Take the delta as a poly_int64
5393 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
5394 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
5395 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
5396 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
5397 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
5399 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
5400 aarch64_add_constant.
5402 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5404 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
5405 Use scalar_float_mode.
5407 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5409 * config/aarch64/aarch64-simd.md
5410 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
5411 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
5412 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
5413 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
5414 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
5415 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
5416 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
5417 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
5418 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
5419 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
5421 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5424 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
5425 targ_options->x_arm_arch_string is non NULL.
5427 2018-01-11 Tamar Christina <tamar.christina@arm.com>
5429 * config/aarch64/aarch64.h
5430 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
5432 2018-01-11 Sudakshina Das <sudi.das@arm.com>
5435 * expmed.c (emit_store_flag_force): Swap if const op0
5436 and change VOIDmode to mode of op0.
5438 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5440 PR rtl-optimization/83761
5441 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
5442 than bytes to mode_for_size.
5444 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5447 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
5448 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
5451 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5454 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
5455 when in layout mode.
5456 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
5457 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
5460 2018-01-10 Michael Collison <michael.collison@arm.com>
5462 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
5463 * config/aarch64/aarch64-option-extension.def: Add
5464 AARCH64_OPT_EXTENSION of 'fp16fml'.
5465 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5466 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
5467 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
5468 * config/aarch64/constraints.md (Ui7): New constraint.
5469 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
5470 (VFMLA_SEL_W): Ditto.
5473 (VFMLA16_LOW): New int iterator.
5474 (VFMLA16_HIGH): Ditto.
5475 (UNSPEC_FMLAL): New unspec.
5476 (UNSPEC_FMLSL): Ditto.
5477 (UNSPEC_FMLAL2): Ditto.
5478 (UNSPEC_FMLSL2): Ditto.
5479 (f16mac): New code attribute.
5480 * config/aarch64/aarch64-simd-builtins.def
5481 (aarch64_fmlal_lowv2sf): Ditto.
5482 (aarch64_fmlsl_lowv2sf): Ditto.
5483 (aarch64_fmlalq_lowv4sf): Ditto.
5484 (aarch64_fmlslq_lowv4sf): Ditto.
5485 (aarch64_fmlal_highv2sf): Ditto.
5486 (aarch64_fmlsl_highv2sf): Ditto.
5487 (aarch64_fmlalq_highv4sf): Ditto.
5488 (aarch64_fmlslq_highv4sf): Ditto.
5489 (aarch64_fmlal_lane_lowv2sf): Ditto.
5490 (aarch64_fmlsl_lane_lowv2sf): Ditto.
5491 (aarch64_fmlal_laneq_lowv2sf): Ditto.
5492 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
5493 (aarch64_fmlalq_lane_lowv4sf): Ditto.
5494 (aarch64_fmlsl_lane_lowv4sf): Ditto.
5495 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
5496 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
5497 (aarch64_fmlal_lane_highv2sf): Ditto.
5498 (aarch64_fmlsl_lane_highv2sf): Ditto.
5499 (aarch64_fmlal_laneq_highv2sf): Ditto.
5500 (aarch64_fmlsl_laneq_highv2sf): Ditto.
5501 (aarch64_fmlalq_lane_highv4sf): Ditto.
5502 (aarch64_fmlsl_lane_highv4sf): Ditto.
5503 (aarch64_fmlalq_laneq_highv4sf): Ditto.
5504 (aarch64_fmlsl_laneq_highv4sf): Ditto.
5505 * config/aarch64/aarch64-simd.md:
5506 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
5507 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5508 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
5509 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5510 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
5511 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
5512 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
5513 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
5514 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
5515 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
5516 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
5517 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
5518 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
5519 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
5520 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
5521 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
5522 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
5523 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
5524 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
5525 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
5526 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
5527 (vfmlsl_low_u32): Ditto.
5528 (vfmlalq_low_u32): Ditto.
5529 (vfmlslq_low_u32): Ditto.
5530 (vfmlal_high_u32): Ditto.
5531 (vfmlsl_high_u32): Ditto.
5532 (vfmlalq_high_u32): Ditto.
5533 (vfmlslq_high_u32): Ditto.
5534 (vfmlal_lane_low_u32): Ditto.
5535 (vfmlsl_lane_low_u32): Ditto.
5536 (vfmlal_laneq_low_u32): Ditto.
5537 (vfmlsl_laneq_low_u32): Ditto.
5538 (vfmlalq_lane_low_u32): Ditto.
5539 (vfmlslq_lane_low_u32): Ditto.
5540 (vfmlalq_laneq_low_u32): Ditto.
5541 (vfmlslq_laneq_low_u32): Ditto.
5542 (vfmlal_lane_high_u32): Ditto.
5543 (vfmlsl_lane_high_u32): Ditto.
5544 (vfmlal_laneq_high_u32): Ditto.
5545 (vfmlsl_laneq_high_u32): Ditto.
5546 (vfmlalq_lane_high_u32): Ditto.
5547 (vfmlslq_lane_high_u32): Ditto.
5548 (vfmlalq_laneq_high_u32): Ditto.
5549 (vfmlslq_laneq_high_u32): Ditto.
5550 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
5551 (AARCH64_FL_FOR_ARCH8_4): New.
5552 (AARCH64_ISA_F16FML): New ISA flag.
5553 (TARGET_F16FML): New feature flag for fp16fml.
5554 (doc/invoke.texi): Document new fp16fml option.
5556 2018-01-10 Michael Collison <michael.collison@arm.com>
5558 * config/aarch64/aarch64-builtins.c:
5559 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
5560 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5561 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
5562 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
5563 (AARCH64_ISA_SHA3): New ISA flag.
5564 (TARGET_SHA3): New feature flag for sha3.
5565 * config/aarch64/iterators.md (sha512_op): New int attribute.
5566 (CRYPTO_SHA512): New int iterator.
5567 (UNSPEC_SHA512H): New unspec.
5568 (UNSPEC_SHA512H2): Ditto.
5569 (UNSPEC_SHA512SU0): Ditto.
5570 (UNSPEC_SHA512SU1): Ditto.
5571 * config/aarch64/aarch64-simd-builtins.def
5572 (aarch64_crypto_sha512hqv2di): New builtin.
5573 (aarch64_crypto_sha512h2qv2di): Ditto.
5574 (aarch64_crypto_sha512su0qv2di): Ditto.
5575 (aarch64_crypto_sha512su1qv2di): Ditto.
5576 (aarch64_eor3qv8hi): Ditto.
5577 (aarch64_rax1qv2di): Ditto.
5578 (aarch64_xarqv2di): Ditto.
5579 (aarch64_bcaxqv8hi): Ditto.
5580 * config/aarch64/aarch64-simd.md:
5581 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
5582 (aarch64_crypto_sha512su0qv2di): Ditto.
5583 (aarch64_crypto_sha512su1qv2di): Ditto.
5584 (aarch64_eor3qv8hi): Ditto.
5585 (aarch64_rax1qv2di): Ditto.
5586 (aarch64_xarqv2di): Ditto.
5587 (aarch64_bcaxqv8hi): Ditto.
5588 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
5589 (vsha512h2q_u64): Ditto.
5590 (vsha512su0q_u64): Ditto.
5591 (vsha512su1q_u64): Ditto.
5592 (veor3q_u16): Ditto.
5593 (vrax1q_u64): Ditto.
5595 (vbcaxq_u16): Ditto.
5596 * config/arm/types.md (crypto_sha512): New type attribute.
5597 (crypto_sha3): Ditto.
5598 (doc/invoke.texi): Document new sha3 option.
5600 2018-01-10 Michael Collison <michael.collison@arm.com>
5602 * config/aarch64/aarch64-builtins.c:
5603 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
5604 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5605 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
5606 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
5607 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
5608 (AARCH64_ISA_SM4): New ISA flag.
5609 (TARGET_SM4): New feature flag for sm4.
5610 * config/aarch64/aarch64-simd-builtins.def
5611 (aarch64_sm3ss1qv4si): Ditto.
5612 (aarch64_sm3tt1aq4si): Ditto.
5613 (aarch64_sm3tt1bq4si): Ditto.
5614 (aarch64_sm3tt2aq4si): Ditto.
5615 (aarch64_sm3tt2bq4si): Ditto.
5616 (aarch64_sm3partw1qv4si): Ditto.
5617 (aarch64_sm3partw2qv4si): Ditto.
5618 (aarch64_sm4eqv4si): Ditto.
5619 (aarch64_sm4ekeyqv4si): Ditto.
5620 * config/aarch64/aarch64-simd.md:
5621 (aarch64_sm3ss1qv4si): Ditto.
5622 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
5623 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
5624 (aarch64_sm4eqv4si): Ditto.
5625 (aarch64_sm4ekeyqv4si): Ditto.
5626 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
5627 (sm3part_op): Ditto.
5628 (CRYPTO_SM3TT): Ditto.
5629 (CRYPTO_SM3PART): Ditto.
5630 (UNSPEC_SM3SS1): New unspec.
5631 (UNSPEC_SM3TT1A): Ditto.
5632 (UNSPEC_SM3TT1B): Ditto.
5633 (UNSPEC_SM3TT2A): Ditto.
5634 (UNSPEC_SM3TT2B): Ditto.
5635 (UNSPEC_SM3PARTW1): Ditto.
5636 (UNSPEC_SM3PARTW2): Ditto.
5637 (UNSPEC_SM4E): Ditto.
5638 (UNSPEC_SM4EKEY): Ditto.
5639 * config/aarch64/constraints.md (Ui2): New constraint.
5640 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
5641 * config/arm/types.md (crypto_sm3): New type attribute.
5642 (crypto_sm4): Ditto.
5643 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
5644 (vsm3tt1aq_u32): Ditto.
5645 (vsm3tt1bq_u32): Ditto.
5646 (vsm3tt2aq_u32): Ditto.
5647 (vsm3tt2bq_u32): Ditto.
5648 (vsm3partw1q_u32): Ditto.
5649 (vsm3partw2q_u32): Ditto.
5650 (vsm4eq_u32): Ditto.
5651 (vsm4ekeyq_u32): Ditto.
5652 (doc/invoke.texi): Document new sm4 option.
5654 2018-01-10 Michael Collison <michael.collison@arm.com>
5656 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
5657 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
5658 (AARCH64_FL_FOR_ARCH8_4): New.
5659 (AARCH64_FL_V8_4): New flag.
5660 (doc/invoke.texi): Document new armv8.4-a option.
5662 2018-01-10 Michael Collison <michael.collison@arm.com>
5664 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5665 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
5666 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
5667 * config/aarch64/aarch64-option-extension.def: Add
5668 AARCH64_OPT_EXTENSION of 'sha2'.
5669 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
5670 (crypto): Disable sha2 and aes if crypto disabled.
5671 (crypto): Enable aes and sha2 if enabled.
5672 (simd): Disable sha2 and aes if simd disabled.
5673 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
5675 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
5676 (TARGET_SHA2): New feature flag for sha2.
5677 (TARGET_AES): New feature flag for aes.
5678 * config/aarch64/aarch64-simd.md:
5679 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
5680 conditional on TARGET_AES.
5681 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
5682 (aarch64_crypto_sha1hsi): Make pattern conditional
5684 (aarch64_crypto_sha1hv4si): Ditto.
5685 (aarch64_be_crypto_sha1hv4si): Ditto.
5686 (aarch64_crypto_sha1su1v4si): Ditto.
5687 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
5688 (aarch64_crypto_sha1su0v4si): Ditto.
5689 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
5690 (aarch64_crypto_sha256su0v4si): Ditto.
5691 (aarch64_crypto_sha256su1v4si): Ditto.
5692 (doc/invoke.texi): Document new aes and sha2 options.
5694 2018-01-10 Martin Sebor <msebor@redhat.com>
5696 PR tree-optimization/83781
5697 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
5700 2018-01-11 Martin Sebor <msebor@gmail.com>
5701 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5703 PR tree-optimization/83501
5704 PR tree-optimization/81703
5706 * tree-ssa-strlen.c (get_string_cst): Rename...
5707 (get_string_len): ...to this. Handle global constants.
5708 (handle_char_store): Adjust.
5710 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
5711 Jim Wilson <jimw@sifive.com>
5713 * config/riscv/riscv-protos.h (riscv_output_return): New.
5714 * config/riscv/riscv.c (struct machine_function): New naked_p field.
5715 (riscv_attribute_table, riscv_output_return),
5716 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
5717 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
5718 (riscv_compute_frame_info): Only compute frame->mask if not a naked
5720 (riscv_expand_prologue): Add early return for naked function.
5721 (riscv_expand_epilogue): Likewise.
5722 (riscv_function_ok_for_sibcall): Return false for naked function.
5723 (riscv_set_current_function): New.
5724 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
5725 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
5726 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
5727 * doc/extend.texi (RISC-V Function Attributes): New.
5729 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
5731 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
5732 check for 128-bit long double before checking TCmode.
5733 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
5734 128-bit long doubles before checking TFmode or TCmode.
5735 (FLOAT128_IBM_P): Likewise.
5737 2018-01-10 Martin Sebor <msebor@redhat.com>
5739 PR tree-optimization/83671
5740 * builtins.c (c_strlen): Unconditionally return zero for the empty
5742 Use -Warray-bounds for warnings.
5743 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
5744 for non-constant array indices with COMPONENT_REF, arrays of
5745 arrays, and pointers to arrays.
5746 (gimple_fold_builtin_strlen): Determine and set length range for
5747 non-constant character arrays.
5749 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
5752 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
5755 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
5757 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
5759 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5762 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
5763 VECTOR_MEM_ALTIVEC_OR_VSX_P.
5764 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
5765 indexed_or_indirect_operand predicate.
5766 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
5767 (*vsx_le_perm_load_v8hi): Likewise.
5768 (*vsx_le_perm_load_v16qi): Likewise.
5769 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
5770 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
5771 (*vsx_le_perm_store_v8hi): Likewise.
5772 (*vsx_le_perm_store_v16qi): Likewise.
5773 (eight unnamed splitters): Likewise.
5775 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5777 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
5778 * config/rs6000/emmintrin.h: Likewise.
5779 * config/rs6000/mmintrin.h: Likewise.
5780 * config/rs6000/xmmintrin.h: Likewise.
5782 2018-01-10 David Malcolm <dmalcolm@redhat.com>
5785 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
5787 * tree.c (tree_nop_conversion): Return true for location wrapper
5789 (maybe_wrap_with_location): New function.
5790 (selftest::check_strip_nops): New function.
5791 (selftest::test_location_wrappers): New function.
5792 (selftest::tree_c_tests): Call it.
5793 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
5794 (maybe_wrap_with_location): New decl.
5795 (EXPR_LOCATION_WRAPPER_P): New macro.
5796 (location_wrapper_p): New inline function.
5797 (tree_strip_any_location_wrapper): New inline function.
5799 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
5802 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
5803 stack_realign_offset for the largest alignment of stack slot
5805 (ix86_find_max_used_stack_alignment): New function.
5806 (ix86_finalize_stack_frame_flags): Use it. Set
5807 max_used_stack_alignment if we don't realign stack.
5808 * config/i386/i386.h (machine_function): Add
5809 max_used_stack_alignment.
5811 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
5813 * config/arm/arm.opt (-mbranch-cost): New option.
5814 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
5817 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
5820 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
5821 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
5823 2018-01-10 Richard Biener <rguenther@suse.de>
5826 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
5827 early out so it also covers the case where we have a non-NULL
5830 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
5832 PR tree-optimization/83753
5833 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
5834 for non-strided grouped accesses if the number of elements is 1.
5836 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5839 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
5840 * i386.h (TARGET_USE_GATHER): Define.
5841 * x86-tune.def (X86_TUNE_USE_GATHER): New.
5843 2018-01-10 Martin Liska <mliska@suse.cz>
5846 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
5847 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
5849 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
5850 CLEANUP_NO_PARTITIONING is not set.
5852 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
5854 * doc/rtl.texi: Remove documentation of (const ...) wrappers
5855 for vectors, as a partial revert of r254296.
5856 * rtl.h (const_vec_p): Delete.
5857 (const_vec_duplicate_p): Don't test for vector CONSTs.
5858 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
5859 * expmed.c (make_tree): Likewise.
5862 * common.md (E, F): Use CONSTANT_P instead of checking for
5864 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
5865 checking for CONST_VECTOR.
5867 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5870 * predict.c (force_edge_cold): Handle in more sane way edges
5873 2018-01-09 Carl Love <cel@us.ibm.com>
5875 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
5877 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
5878 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
5879 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
5880 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
5881 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
5882 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
5883 * config/rs6000/rs6000-protos.h: Add extern defition for
5884 rs6000_generate_float2_double_code.
5885 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
5887 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
5888 (float2_v2df): Add define_expand.
5890 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
5893 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
5894 op_mode in the force_to_mode call.
5896 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5898 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
5899 instead of checking each element individually.
5900 (aarch64_evpc_uzp): Likewise.
5901 (aarch64_evpc_zip): Likewise.
5902 (aarch64_evpc_ext): Likewise.
5903 (aarch64_evpc_rev): Likewise.
5904 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
5905 instead of checking each element individually. Return true without
5907 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
5908 whether all selected elements come from the same input, instead of
5909 checking each element individually. Remove calls to gen_rtx_REG,
5910 start_sequence and end_sequence and instead assert that no rtl is
5913 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5915 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
5916 order of HIGH and CONST checks.
5918 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5920 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
5921 if the destination isn't an SSA_NAME.
5923 2018-01-09 Richard Biener <rguenther@suse.de>
5925 PR tree-optimization/83668
5926 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
5928 (canonicalize_loop_form): ... here, renamed from ...
5929 (canonicalize_loop_closed_ssa_form): ... this and amended to
5930 swap successor edges for loop exit blocks to make us use
5931 the RPO order we need for initial schedule generation.
5933 2018-01-09 Joseph Myers <joseph@codesourcery.com>
5935 PR tree-optimization/64811
5936 * match.pd: When optimizing comparisons with Inf, avoid
5937 introducing or losing exceptions from comparisons with NaN.
5939 2018-01-09 Martin Liska <mliska@suse.cz>
5942 * asan.c (shadow_mem_size): Add gcc_assert.
5944 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
5946 Don't save registers in main().
5949 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
5950 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
5951 * config/avr/avr.c (avr_set_current_function): Don't error if
5952 naked, OS_task or OS_main are specified at the same time.
5953 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
5955 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
5957 * common/config/avr/avr-common.c (avr_option_optimization_table):
5958 Switch on -mmain-is-OS_task for optimizing compilations.
5960 2018-01-09 Richard Biener <rguenther@suse.de>
5962 PR tree-optimization/83572
5963 * graphite.c: Include cfganal.h.
5964 (graphite_transform_loops): Connect infinite loops to exit
5965 and remove fake edges at the end.
5967 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5969 * ipa-inline.c (edge_badness): Revert accidental checkin.
5971 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5974 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
5975 symbols; not inline clones.
5977 2018-01-09 Jakub Jelinek <jakub@redhat.com>
5980 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
5981 hard registers. Formatting fixes.
5983 PR preprocessor/83722
5984 * gcc.c (try_generate_repro): Pass
5985 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
5986 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
5989 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
5990 Kito Cheng <kito.cheng@gmail.com>
5992 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
5993 (riscv_leaf_function_p): Delete.
5994 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
5996 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5998 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
6000 (do_ifelse): New function.
6001 (do_isel): New function.
6002 (do_sub3): New function.
6003 (do_add3): New function.
6004 (do_load_mask_compare): New function.
6005 (do_overlap_load_compare): New function.
6006 (expand_compare_loop): New function.
6007 (expand_block_compare): Call expand_compare_loop() when appropriate.
6008 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
6010 (-mblock-compare-inline-loop-limit): New option.
6012 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6015 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
6016 Reverse order of second and third operands in first alternative.
6017 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
6018 of first and second elements in UNSPEC_VPERMR vector.
6019 (altivec_expand_vec_perm_le): Likewise.
6021 2018-01-08 Jeff Law <law@redhat.com>
6023 PR rtl-optimizatin/81308
6024 * tree-switch-conversion.c (cfg_altered): New file scoped static.
6025 (process_switch): If group_case_labels makes a change, then set
6027 (pass_convert_switch::execute): If a switch is converted, then
6028 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
6030 PR rtl-optimization/81308
6031 * recog.c (split_all_insns): Conditionally cleanup the CFG after
6034 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
6036 PR target/83663 - Revert r255946
6037 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
6038 generation for cases where splatting a value is not useful.
6039 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
6040 across a vec_duplicate and a paradoxical subreg forming a vector
6041 mode to a vec_concat.
6043 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6045 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
6046 -march=armv8.3-a variants.
6047 * config/arm/t-multilib: Likewise.
6048 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
6050 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6052 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
6054 (cceq_ior_compare_complement): Give it a name so I can use it, and
6055 change boolean_or_operator predicate to boolean_operator so it can
6056 be used to generate a crand.
6057 (eqne): New code iterator.
6058 (bd/bd_neg): New code_attrs.
6059 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
6060 a single define_insn.
6061 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
6062 decrement (bdnzt/bdnzf/bdzt/bdzf).
6063 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
6064 with the new names of the branch decrement patterns, and added the
6065 names of the branch decrement conditional patterns.
6067 2018-01-08 Richard Biener <rguenther@suse.de>
6069 PR tree-optimization/83563
6070 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
6073 2018-01-08 Richard Biener <rguenther@suse.de>
6076 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
6078 2018-01-08 Richard Biener <rguenther@suse.de>
6080 PR tree-optimization/83685
6081 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
6082 references to abnormals.
6084 2018-01-08 Richard Biener <rguenther@suse.de>
6087 * dwarf2out.c (output_indirect_strings): Handle empty
6088 skeleton_debug_str_hash.
6089 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
6091 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6093 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
6094 (emit_store_direct): Likewise.
6095 (arc_trampoline_adjust_address): Likewise.
6096 (arc_asm_trampoline_template): New function.
6097 (arc_initialize_trampoline): Use asm_trampoline_template.
6098 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
6099 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
6100 * config/arc/arc.md (flush_icache): Delete pattern.
6102 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6104 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
6105 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
6108 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6111 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
6112 by not USED_FOR_TARGET.
6113 (make_pass_resolve_sw_modes): Likewise.
6115 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6117 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
6120 2018-01-08 Richard Biener <rguenther@suse.de>
6123 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
6125 2018-01-08 Richard Biener <rguenther@suse.de>
6128 * match.pd ((t * 2) / 2) -> t): Add missing :c.
6130 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
6133 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
6134 basic blocks with a small number of successors.
6135 (convert_control_dep_chain_into_preds): Improve handling of
6137 (dump_predicates): Split apart into...
6138 (dump_pred_chain): ...here...
6139 (dump_pred_info): ...and here.
6140 (can_one_predicate_be_invalidated_p): Add debugging printfs.
6141 (can_chain_union_be_invalidated_p): Improve check for invalidation
6143 (uninit_uses_cannot_happen): Avoid unnecessary if
6144 convert_control_dep_chain_into_preds yielded nothing.
6146 2018-01-06 Martin Sebor <msebor@redhat.com>
6148 PR tree-optimization/83640
6149 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
6150 subtracting negative offset from size.
6151 (builtin_access::overlap): Adjust offset bounds of the access to fall
6152 within the size of the object if possible.
6154 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
6156 PR rtl-optimization/83699
6157 * expmed.c (extract_bit_field_1): Restrict the vector usage of
6158 extract_bit_field_as_subreg to cases in which the extracted
6159 value is also a vector.
6161 * lra-constraints.c (process_alt_operands): Test for the equivalence
6162 substitutions when detecting a possible reload cycle.
6164 2018-01-06 Jakub Jelinek <jakub@redhat.com>
6167 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
6168 by default if flag_selective_schedling{,2}. Formatting fixes.
6170 PR rtl-optimization/83682
6171 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
6172 if it has non-VECTOR_MODE element mode.
6173 (vec_duplicate_p): Likewise.
6176 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
6177 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
6179 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6182 * config/i386/i386-builtin.def
6183 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
6184 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
6185 Require also OPTION_MASK_ISA_AVX512F in addition to
6186 OPTION_MASK_ISA_GFNI.
6187 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
6188 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
6189 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
6190 to OPTION_MASK_ISA_GFNI.
6191 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
6192 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
6193 OPTION_MASK_ISA_AVX512BW.
6194 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
6195 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
6196 addition to OPTION_MASK_ISA_GFNI.
6197 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
6198 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
6199 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
6200 to OPTION_MASK_ISA_GFNI.
6201 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
6202 a requirement for all ISAs rather than any of them with a few
6204 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
6206 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
6207 bitmasks to be enabled with 3 exceptions, instead of requiring any
6208 enabled ISA with lots of exceptions.
6209 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
6210 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
6211 Change avx512bw in isa attribute to avx512f.
6212 * config/i386/sgxintrin.h: Add license boilerplate.
6213 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
6214 to __AVX512F__ and __AVX512VL to __AVX512VL__.
6215 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
6216 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
6218 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
6219 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
6220 temporarily sse2 rather than sse if not enabled already.
6223 * config/i386/sse.md (VI248_VLBW): Rename to ...
6224 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
6225 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
6226 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
6227 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
6228 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
6229 mode iterator instead of VI248_VLBW.
6231 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
6233 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
6234 (record_modified): Skip clobbers; add debug output.
6235 (param_change_prob): Use sreal frequencies.
6237 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6239 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
6240 punt for user-aligned variables.
6242 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6244 * tree-chrec.c (chrec_contains_symbols): Return true for
6247 2018-01-05 Sudakshina Das <sudi.das@arm.com>
6250 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
6251 of (x|y) == x for BICS pattern.
6253 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6255 PR tree-optimization/83605
6256 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
6257 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
6260 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
6262 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
6263 * config/epiphany/rtems.h: New file.
6265 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6266 Uros Bizjak <ubizjak@gmail.com>
6269 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
6270 QIreg_operand instead of register_operand predicate.
6271 * config/i386/i386.c (ix86_rop_should_change_byte_p,
6272 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
6273 comments instead of -fmitigate[-_]rop.
6275 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6278 * cgraphunit.c (symbol_table::compile): Switch to text_section
6279 before calling assembly_start debug hook.
6280 * run-rtl-passes.c (run_rtl_passes): Likewise.
6283 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6285 * tree-vrp.c (extract_range_from_binary_expr_1): Check
6286 range_int_cst_p rather than !symbolic_range_p before calling
6287 extract_range_from_multiplicative_op_1.
6289 2018-01-04 Jeff Law <law@redhat.com>
6291 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
6292 redundant test in assertion.
6294 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6296 * doc/rtl.texi: Document machine_mode wrapper classes.
6298 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6300 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
6303 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6305 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
6306 the VEC_PERM_EXPR fold to fail.
6308 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6311 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
6312 to switched_sections.
6314 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6317 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
6320 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
6323 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
6324 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
6326 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6329 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
6330 is BLKmode and bitpos not zero or mode change is needed.
6332 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6335 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
6338 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
6341 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
6342 instead of MULT rtx. Update all corresponding splitters.
6344 (*ssub<modesuffix>): Ditto.
6346 (*cmp_sadd_di): Update split patterns.
6347 (*cmp_sadd_si): Ditto.
6348 (*cmp_sadd_sidi): Ditto.
6349 (*cmp_ssub_di): Ditto.
6350 (*cmp_ssub_si): Ditto.
6351 (*cmp_ssub_sidi): Ditto.
6352 * config/alpha/predicates.md (const23_operand): New predicate.
6353 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
6354 Look for ASHIFT, not MULT inner operand.
6355 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
6357 2018-01-04 Martin Liska <mliska@suse.cz>
6359 PR gcov-profile/83669
6360 * gcov.c (output_intermediate_file): Add version to intermediate
6362 * doc/gcov.texi: Document new field 'version' in intermediate
6363 file format. Fix location of '-k' option of gcov command.
6365 2018-01-04 Martin Liska <mliska@suse.cz>
6368 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
6370 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6372 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
6374 2018-01-03 Martin Sebor <msebor@redhat.com>
6376 PR tree-optimization/83655
6377 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
6378 checking calls with invalid arguments.
6380 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6382 * tree-vect-stmts.c (vect_get_store_rhs): New function.
6383 (vectorizable_mask_load_store): Delete.
6384 (vectorizable_call): Return false for masked loads and stores.
6385 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
6386 instead of gimple_assign_rhs1.
6387 (vectorizable_load): Handle IFN_MASK_LOAD.
6388 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
6390 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6392 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
6394 (vectorizable_mask_load_store): ...here.
6395 (vectorizable_load): ...and here.
6397 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6399 * tree-vect-stmts.c (vect_build_all_ones_mask)
6400 (vect_build_zero_merge_argument): New functions, split out from...
6401 (vectorizable_load): ...here.
6403 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6405 * tree-vect-stmts.c (vect_check_store_rhs): New function,
6407 (vectorizable_mask_load_store): ...here.
6408 (vectorizable_store): ...and here.
6410 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6412 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
6414 (vectorizable_mask_load_store): ...here.
6416 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6418 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
6419 (vect_model_store_cost): Take a vec_load_store_type instead of a
6421 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
6422 (vect_model_store_cost): Take a vec_load_store_type instead of a
6424 (vectorizable_mask_load_store): Update accordingly.
6425 (vectorizable_store): Likewise.
6426 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
6428 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6430 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
6431 IFN_MASK_LOAD calls here rather than...
6432 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
6434 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6435 Alan Hayward <alan.hayward@arm.com>
6436 David Sherwood <david.sherwood@arm.com>
6438 * expmed.c (extract_bit_field_1): For vector extracts,
6439 fall back to extract_bit_field_as_subreg if vec_extract
6442 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6443 Alan Hayward <alan.hayward@arm.com>
6444 David Sherwood <david.sherwood@arm.com>
6446 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
6447 they are variable or constant sized.
6448 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
6449 slots for constant-sized data.
6451 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6452 Alan Hayward <alan.hayward@arm.com>
6453 David Sherwood <david.sherwood@arm.com>
6455 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
6456 handling COND_EXPRs with boolean comparisons, try to find a better
6457 basis for the mask type than the boolean itself.
6459 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6461 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
6462 is calculated and how it can be overridden.
6463 * genmodes.c (max_bitsize_mode_any_mode): New variable.
6464 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
6466 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
6469 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6470 Alan Hayward <alan.hayward@arm.com>
6471 David Sherwood <david.sherwood@arm.com>
6473 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
6474 Remove the mode argument.
6475 (aarch64_simd_valid_immediate): Remove the mode and inverse
6477 * config/aarch64/iterators.md (bitsize): New iterator.
6478 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
6479 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
6480 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
6481 aarch64_simd_valid_immediate.
6482 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
6483 (aarch64_reg_or_bic_imm): Likewise.
6484 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
6485 with an insn_type enum and msl with a modifier_type enum.
6486 Replace element_width with a scalar_mode. Change the shift
6487 to unsigned int. Add constructors for scalar_float_mode and
6488 scalar_int_mode elements.
6489 (aarch64_vect_float_const_representable_p): Delete.
6490 (aarch64_can_const_movi_rtx_p)
6491 (aarch64_simd_scalar_immediate_valid_for_move)
6492 (aarch64_simd_make_constant): Update call to
6493 aarch64_simd_valid_immediate.
6494 (aarch64_advsimd_valid_immediate_hs): New function.
6495 (aarch64_advsimd_valid_immediate): Likewise.
6496 (aarch64_simd_valid_immediate): Remove mode and inverse
6497 arguments. Rewrite to use the above. Use const_vec_duplicate_p
6498 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
6499 and aarch64_float_const_representable_p on the result.
6500 (aarch64_output_simd_mov_immediate): Remove mode argument.
6501 Update call to aarch64_simd_valid_immediate and use of
6502 simd_immediate_info.
6503 (aarch64_output_scalar_simd_mov_immediate): Update call
6506 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6507 Alan Hayward <alan.hayward@arm.com>
6508 David Sherwood <david.sherwood@arm.com>
6510 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
6511 (mode_nunits): Likewise CONST_MODE_NUNITS.
6512 * machmode.def (ADJUST_NUNITS): Document.
6513 * genmodes.c (mode_data::need_nunits_adj): New field.
6514 (blank_mode): Update accordingly.
6515 (adj_nunits): New variable.
6516 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
6518 (emit_mode_size_inline): Set need_bytesize_adj for all modes
6519 listed in adj_nunits.
6520 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
6521 listed in adj_nunits. Don't emit case statements for such modes.
6522 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
6523 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
6524 nothing if adj_nunits is nonnull.
6525 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
6526 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
6527 (emit_mode_fbit): Update use of print_maybe_const_decl.
6528 (emit_move_size): Likewise. Treat the array as non-const
6530 (emit_mode_adjustments): Handle adj_nunits.
6532 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6534 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
6535 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
6536 (VECTOR_MODES): Use it.
6537 (make_vector_modes): Take the prefix as an argument.
6539 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6540 Alan Hayward <alan.hayward@arm.com>
6541 David Sherwood <david.sherwood@arm.com>
6543 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
6544 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
6545 for MODE_VECTOR_BOOL.
6546 * machmode.def (VECTOR_BOOL_MODE): Document.
6547 * genmodes.c (VECTOR_BOOL_MODE): New macro.
6548 (make_vector_bool_mode): New function.
6549 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
6551 * lto-streamer-in.c (lto_input_mode_table): Likewise.
6552 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
6554 * stor-layout.c (int_mode_for_mode): Likewise.
6555 * tree.c (build_vector_type_for_mode): Likewise.
6556 * varasm.c (output_constant_pool_2): Likewise.
6557 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
6558 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
6559 for MODE_VECTOR_BOOL.
6560 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
6561 of mode class checks.
6562 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
6563 instead of a list of mode class checks.
6564 (expand_vector_scalar_condition): Likewise.
6565 (type_for_widest_vector_mode): Handle BImode as an inner mode.
6567 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6568 Alan Hayward <alan.hayward@arm.com>
6569 David Sherwood <david.sherwood@arm.com>
6571 * machmode.h (mode_size): Change from unsigned short to
6573 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
6574 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6575 or if measurement_type is not polynomial.
6576 (fixed_size_mode::includes_p): Check for constant-sized modes.
6577 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
6578 return a poly_uint16 rather than an unsigned short.
6579 (emit_mode_size): Change the type of mode_size from unsigned short
6580 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
6581 (emit_mode_adjustments): Cope with polynomial vector sizes.
6582 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6584 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6586 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
6587 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
6588 * caller-save.c (setup_save_areas): Likewise.
6589 (replace_reg_with_saved_mem): Likewise.
6590 * calls.c (emit_library_call_value_1): Likewise.
6591 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
6592 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
6593 (gen_lowpart_for_combine): Likewise.
6594 * convert.c (convert_to_integer_1): Likewise.
6595 * cse.c (equiv_constant, cse_insn): Likewise.
6596 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
6597 (cselib_subst_to_values): Likewise.
6598 * dce.c (word_dce_process_block): Likewise.
6599 * df-problems.c (df_word_lr_mark_ref): Likewise.
6600 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
6601 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
6602 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
6603 (rtl_for_decl_location): Likewise.
6604 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
6605 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
6606 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
6607 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
6608 (expand_expr_real_1): Likewise.
6609 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
6610 (pad_below): Likewise.
6611 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6612 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
6613 * ira.c (get_subreg_tracking_sizes): Likewise.
6614 * ira-build.c (ira_create_allocno_objects): Likewise.
6615 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
6616 (ira_sort_regnos_for_alter_reg): Likewise.
6617 * ira-costs.c (record_operand_costs): Likewise.
6618 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
6619 (resolve_simple_move): Likewise.
6620 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
6621 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
6622 (lra_constraints): Likewise.
6623 (CONST_POOL_OK_P): Reject variable-sized modes.
6624 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
6625 (add_pseudo_to_slot, lra_spill): Likewise.
6626 * omp-low.c (omp_clause_aligned_alignment): Likewise.
6627 * optabs-query.c (get_best_extraction_insn): Likewise.
6628 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6629 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
6630 (expand_mult_highpart, valid_multiword_target_p): Likewise.
6631 * recog.c (offsettable_address_addr_space_p): Likewise.
6632 * regcprop.c (maybe_mode_change): Likewise.
6633 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
6634 * regrename.c (build_def_use): Likewise.
6635 * regstat.c (dump_reg_info): Likewise.
6636 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
6637 (find_reloads, find_reloads_subreg_address): Likewise.
6638 * reload1.c (eliminate_regs_1): Likewise.
6639 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
6640 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
6641 (simplify_binary_operation_1, simplify_subreg): Likewise.
6642 * targhooks.c (default_function_arg_padding): Likewise.
6643 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
6644 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
6645 (verify_gimple_assign_ternary): Likewise.
6646 * tree-inline.c (estimate_move_cost): Likewise.
6647 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6648 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
6649 (get_address_cost_ainc): Likewise.
6650 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
6651 (vect_supportable_dr_alignment): Likewise.
6652 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6653 (vectorizable_reduction): Likewise.
6654 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
6655 (vectorizable_operation, vectorizable_load): Likewise.
6656 * tree.c (build_same_sized_truth_vector_type): Likewise.
6657 * valtrack.c (cleanup_auto_inc_dec): Likewise.
6658 * var-tracking.c (emit_note_insn_var_location): Likewise.
6659 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
6660 (ADDR_VEC_ALIGN): Likewise.
6662 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6663 Alan Hayward <alan.hayward@arm.com>
6664 David Sherwood <david.sherwood@arm.com>
6666 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
6668 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6669 or if measurement_type is polynomial.
6670 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
6671 * combine.c (make_extraction): Likewise.
6672 * dse.c (find_shift_sequence): Likewise.
6673 * dwarf2out.c (mem_loc_descriptor): Likewise.
6674 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
6675 (extract_bit_field, extract_low_bits): Likewise.
6676 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
6677 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
6678 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
6679 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
6680 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6681 * reload.c (find_reloads): Likewise.
6682 * reload1.c (alter_reg): Likewise.
6683 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
6684 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
6685 * tree-if-conv.c (predicate_mem_writes): Likewise.
6686 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
6687 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
6688 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
6689 * valtrack.c (dead_debug_insert_temp): Likewise.
6690 * varasm.c (mergeable_constant_section): Likewise.
6691 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
6693 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6694 Alan Hayward <alan.hayward@arm.com>
6695 David Sherwood <david.sherwood@arm.com>
6697 * expr.c (expand_assignment): Cope with polynomial mode sizes
6698 when assigning to a CONCAT.
6700 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6701 Alan Hayward <alan.hayward@arm.com>
6702 David Sherwood <david.sherwood@arm.com>
6704 * machmode.h (mode_precision): Change from unsigned short to
6706 (mode_to_precision): Return a poly_uint16 rather than an unsigned
6708 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
6709 or if measurement_type is not polynomial.
6710 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
6711 in which the mode is already known to be a scalar_int_mode.
6712 * genmodes.c (emit_mode_precision): Change the type of mode_precision
6713 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
6715 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6716 for GET_MODE_PRECISION.
6717 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6718 for GET_MODE_PRECISION.
6719 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
6721 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
6722 (expand_field_assignment, make_extraction): Likewise.
6723 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
6724 (get_last_value): Likewise.
6725 * convert.c (convert_to_integer_1): Likewise.
6726 * cse.c (cse_insn): Likewise.
6727 * expr.c (expand_expr_real_1): Likewise.
6728 * lra-constraints.c (simplify_operand_subreg): Likewise.
6729 * optabs-query.c (can_atomic_load_p): Likewise.
6730 * optabs.c (expand_atomic_load): Likewise.
6731 (expand_atomic_store): Likewise.
6732 * ree.c (combine_reaching_defs): Likewise.
6733 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
6734 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
6735 * tree.h (type_has_mode_precision_p): Likewise.
6736 * ubsan.c (instrument_si_overflow): Likewise.
6738 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6739 Alan Hayward <alan.hayward@arm.com>
6740 David Sherwood <david.sherwood@arm.com>
6742 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
6743 polynomial numbers of units.
6744 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
6745 (valid_vector_subparts_p): New function.
6746 (build_vector_type): Remove temporary shim and take the number
6747 of units as a poly_uint64 rather than an int.
6748 (build_opaque_vector_type): Take the number of units as a
6749 poly_uint64 rather than an int.
6750 * tree.c (build_vector_from_ctor): Handle polynomial
6751 TYPE_VECTOR_SUBPARTS.
6752 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
6753 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
6754 (build_vector_from_val): If the number of units is variable,
6755 use build_vec_duplicate_cst for constant operands and
6756 VEC_DUPLICATE_EXPR otherwise.
6757 (make_vector_type): Remove temporary is_constant ().
6758 (build_vector_type, build_opaque_vector_type): Take the number of
6759 units as a poly_uint64 rather than an int.
6760 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
6762 * cfgexpand.c (expand_debug_expr): Likewise.
6763 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
6764 (store_constructor, expand_expr_real_1): Likewise.
6765 (const_scalar_mask_from_tree): Likewise.
6766 * fold-const-call.c (fold_const_reduction): Likewise.
6767 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
6768 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
6769 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
6770 (fold_relational_const): Likewise.
6771 (native_interpret_vector): Likewise. Change the size from an
6772 int to an unsigned int.
6773 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
6774 TYPE_VECTOR_SUBPARTS.
6775 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
6776 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
6777 duplicating a non-constant operand into a variable-length vector.
6778 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
6779 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
6780 * ipa-icf.c (sem_variable::equals): Likewise.
6781 * match.pd: Likewise.
6782 * omp-simd-clone.c (simd_clone_subparts): Likewise.
6783 * print-tree.c (print_node): Likewise.
6784 * stor-layout.c (layout_type): Likewise.
6785 * targhooks.c (default_builtin_vectorization_cost): Likewise.
6786 * tree-cfg.c (verify_gimple_comparison): Likewise.
6787 (verify_gimple_assign_binary): Likewise.
6788 (verify_gimple_assign_ternary): Likewise.
6789 (verify_gimple_assign_single): Likewise.
6790 * tree-pretty-print.c (dump_generic_node): Likewise.
6791 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6792 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
6793 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
6794 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
6795 (vect_shift_permute_load_chain): Likewise.
6796 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
6797 (expand_vector_condition, optimize_vector_constructor): Likewise.
6798 (lower_vec_perm, get_compute_type): Likewise.
6799 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6800 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
6801 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
6802 (vect_recog_mask_conversion_pattern): Likewise.
6803 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
6804 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
6805 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
6806 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
6807 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
6808 (vectorizable_shift, vectorizable_operation, vectorizable_store)
6809 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
6810 (supportable_widening_operation): Likewise.
6811 (supportable_narrowing_operation): Likewise.
6812 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
6814 * varasm.c (output_constant): Likewise.
6816 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6817 Alan Hayward <alan.hayward@arm.com>
6818 David Sherwood <david.sherwood@arm.com>
6820 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
6821 so that both the length == 3 and length != 3 cases set up their
6822 own permute vectors. Add comments explaining why we know the
6823 number of elements is constant.
6824 (vect_permute_load_chain): Likewise.
6826 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6827 Alan Hayward <alan.hayward@arm.com>
6828 David Sherwood <david.sherwood@arm.com>
6830 * machmode.h (mode_nunits): Change from unsigned char to
6832 (ONLY_FIXED_SIZE_MODES): New macro.
6833 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
6834 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
6835 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
6837 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
6838 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
6839 or if measurement_type is not polynomial.
6840 * genmodes.c (ZERO_COEFFS): New macro.
6841 (emit_mode_nunits_inline): Make mode_nunits_inline return a
6843 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
6844 Use ZERO_COEFFS when emitting initializers.
6845 * data-streamer.h (bp_pack_poly_value): New function.
6846 (bp_unpack_poly_value): Likewise.
6847 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6848 for GET_MODE_NUNITS.
6849 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6850 for GET_MODE_NUNITS.
6851 * tree.c (make_vector_type): Remove temporary shim and make
6852 the real function take the number of units as a poly_uint64
6854 (build_vector_type_for_mode): Handle polynomial nunits.
6855 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
6856 * emit-rtl.c (const_vec_series_p_1): Likewise.
6857 (gen_rtx_CONST_VECTOR): Likewise.
6858 * fold-const.c (test_vec_duplicate_folding): Likewise.
6859 * genrecog.c (validate_pattern): Likewise.
6860 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
6861 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6862 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
6863 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
6864 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
6865 * rtlanal.c (subreg_get_info): Likewise.
6866 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
6867 (vect_grouped_load_supported): Likewise.
6868 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
6869 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
6870 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
6871 (simplify_const_unary_operation, simplify_binary_operation_1)
6872 (simplify_const_binary_operation, simplify_ternary_operation)
6873 (test_vector_ops_duplicate, test_vector_ops): Likewise.
6874 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
6875 instead of CONST_VECTOR_NUNITS.
6876 * varasm.c (output_constant_pool_2): Likewise.
6877 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
6878 explicit-encoded elements in the XVEC for variable-length vectors.
6880 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6882 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
6884 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6885 Alan Hayward <alan.hayward@arm.com>
6886 David Sherwood <david.sherwood@arm.com>
6888 * coretypes.h (fixed_size_mode): Declare.
6889 (fixed_size_mode_pod): New typedef.
6890 * builtins.h (target_builtins::x_apply_args_mode)
6891 (target_builtins::x_apply_result_mode): Change type to
6892 fixed_size_mode_pod.
6893 * builtins.c (apply_args_size, apply_result_size, result_vector)
6894 (expand_builtin_apply_args_1, expand_builtin_apply)
6895 (expand_builtin_return): Update accordingly.
6897 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6899 * cse.c (hash_rtx_cb): Hash only the encoded elements.
6900 * cselib.c (cselib_hash_rtx): Likewise.
6901 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
6902 CONST_VECTOR encoding.
6904 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6905 Jeff Law <law@redhat.com>
6908 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
6909 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
6910 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
6911 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
6914 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
6915 explicitly probe *sp in a noreturn function if there were any callee
6916 register saves or frame pointer is needed.
6918 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6921 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
6922 BLKmode for ternary, binary or unary expressions.
6925 * var-tracking.c (delete_vta_debug_insn): New inline function.
6926 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
6927 insns from get_insns () to NULL instead of each bb separately.
6928 Use delete_vta_debug_insn. No longer static.
6929 (vt_debug_insns_local, variable_tracking_main_1): Adjust
6930 delete_vta_debug_insns callers.
6931 * rtl.h (delete_vta_debug_insns): Declare.
6932 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
6933 instead of variable_tracking_main.
6935 2018-01-03 Martin Sebor <msebor@redhat.com>
6937 PR tree-optimization/83603
6938 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
6939 arguments past the endof the argument list in functions declared
6940 without a prototype.
6941 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
6942 Avoid checking when arguments are null.
6944 2018-01-03 Martin Sebor <msebor@redhat.com>
6947 * doc/extend.texi (attribute const): Fix a typo.
6948 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
6949 issuing -Wsuggest-attribute for void functions.
6951 2018-01-03 Martin Sebor <msebor@redhat.com>
6953 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
6954 offset_int::from instead of wide_int::to_shwi.
6955 (maybe_diag_overlap): Remove assertion.
6956 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
6957 * gimple-ssa-sprintf.c (format_directive): Same.
6958 (parse_directive): Same.
6959 (sprintf_dom_walker::compute_format_length): Same.
6960 (try_substitute_return_value): Same.
6962 2018-01-03 Jeff Law <law@redhat.com>
6965 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
6966 non-constant residual for zero at runtime and avoid probing in
6967 that case. Reorganize code for trailing problem to mirror handling
6970 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6972 PR tree-optimization/83501
6973 * tree-ssa-strlen.c (get_string_cst): New.
6974 (handle_char_store): Call get_string_cst.
6976 2018-01-03 Martin Liska <mliska@suse.cz>
6978 PR tree-optimization/83593
6979 * tree-ssa-strlen.c: Include tree-cfg.h.
6980 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
6981 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
6982 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
6984 (strlen_dom_walker::before_dom_children): Call
6985 gimple_purge_dead_eh_edges. Dump tranformation with details
6987 (strlen_dom_walker::before_dom_children): Update call by adding
6988 new argument cleanup_eh.
6989 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
6991 2018-01-03 Martin Liska <mliska@suse.cz>
6994 * cif-code.def (VARIADIC_THUNK): New enum value.
6995 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
6998 2018-01-03 Jan Beulich <jbeulich@suse.com>
7000 * sse.md (mov<mode>_internal): Tighten condition for when to use
7001 vmovdqu<ssescalarsize> for TI and OI modes.
7003 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7005 Update copyright years.
7007 2018-01-03 Martin Liska <mliska@suse.cz>
7010 * ipa-visibility.c (function_and_variable_visibility): Skip
7011 functions with noipa attribure.
7013 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7015 * gcc.c (process_command): Update copyright notice dates.
7016 * gcov-dump.c (print_version): Ditto.
7017 * gcov.c (print_version): Ditto.
7018 * gcov-tool.c (print_version): Ditto.
7019 * gengtype.c (create_file): Ditto.
7020 * doc/cpp.texi: Bump @copying's copyright year.
7021 * doc/cppinternals.texi: Ditto.
7022 * doc/gcc.texi: Ditto.
7023 * doc/gccint.texi: Ditto.
7024 * doc/gcov.texi: Ditto.
7025 * doc/install.texi: Ditto.
7026 * doc/invoke.texi: Ditto.
7028 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7030 * vector-builder.h (vector_builder::m_full_nelts): Change from
7031 unsigned int to poly_uint64.
7032 (vector_builder::full_nelts): Update prototype accordingly.
7033 (vector_builder::new_vector): Likewise.
7034 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
7035 (vector_builder::operator ==): Likewise.
7036 (vector_builder::finalize): Likewise.
7037 * int-vector-builder.h (int_vector_builder::int_vector_builder):
7038 Take the number of elements as a poly_uint64 rather than an
7040 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
7041 from unsigned int to poly_uint64.
7042 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
7043 (vec_perm_indices::new_vector): Likewise.
7044 (vec_perm_indices::length): Likewise.
7045 (vec_perm_indices::nelts_per_input): Likewise.
7046 (vec_perm_indices::input_nelts): Likewise.
7047 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
7048 number of elements per input as a poly_uint64 rather than an
7049 unsigned int. Use the original encoding for variable-length
7050 vectors, rather than clamping each individual element.
7051 For the second and subsequent elements in each pattern,
7052 clamp the step and base before clamping their sum.
7053 (vec_perm_indices::series_p): Handle polynomial element counts.
7054 (vec_perm_indices::all_in_range_p): Likewise.
7055 (vec_perm_indices_to_tree): Likewise.
7056 (vec_perm_indices_to_rtx): Likewise.
7057 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
7058 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
7059 (tree_vector_builder::new_binary_operation): Handle polynomial
7060 element counts. Return false if we need to know the number
7061 of elements at compile time.
7062 * fold-const.c (fold_vec_perm): Punt if the number of elements
7063 isn't known at compile time.
7065 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7067 * vec-perm-indices.h (vec_perm_builder): Change element type
7068 from HOST_WIDE_INT to poly_int64.
7069 (vec_perm_indices::element_type): Update accordingly.
7070 (vec_perm_indices::clamp): Handle polynomial element_types.
7071 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7072 (vec_perm_indices::all_in_range_p): Likewise.
7073 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
7075 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
7076 polynomial vec_perm_indices element types.
7077 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
7078 * fold-const.c (fold_vec_perm): Likewise.
7079 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
7080 * tree-vect-generic.c (lower_vec_perm): Likewise.
7081 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7082 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
7083 element type to HOST_WIDE_INT.
7085 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7086 Alan Hayward <alan.hayward@arm.com>
7087 David Sherwood <david.sherwood@arm.com>
7089 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
7090 rather than an int. Use plus_constant.
7091 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
7092 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
7094 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7095 Alan Hayward <alan.hayward@arm.com>
7096 David Sherwood <david.sherwood@arm.com>
7098 * calls.c (emit_call_1, expand_call): Change struct_value_size from
7099 a HOST_WIDE_INT to a poly_int64.
7101 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7102 Alan Hayward <alan.hayward@arm.com>
7103 David Sherwood <david.sherwood@arm.com>
7105 * calls.c (load_register_parameters): Cope with polynomial
7106 mode sizes. Require a constant size for BLKmode parameters
7107 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
7108 forces a parameter to be padded at the lsb end in order to
7109 fill a complete number of words, require the parameter size
7110 to be ordered wrt UNITS_PER_WORD.
7112 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7113 Alan Hayward <alan.hayward@arm.com>
7114 David Sherwood <david.sherwood@arm.com>
7116 * reload1.c (spill_stack_slot_width): Change element type
7117 from unsigned int to poly_uint64_pod.
7118 (alter_reg): Treat mode sizes as polynomial.
7120 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7121 Alan Hayward <alan.hayward@arm.com>
7122 David Sherwood <david.sherwood@arm.com>
7124 * reload.c (complex_word_subreg_p): New function.
7125 (reload_inner_reg_of_subreg, push_reload): Use it.
7127 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7128 Alan Hayward <alan.hayward@arm.com>
7129 David Sherwood <david.sherwood@arm.com>
7131 * lra-constraints.c (process_alt_operands): Reject matched
7132 operands whose sizes aren't ordered.
7133 (match_reload): Refer to this check here.
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 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
7140 that the mode size is in the set {1, 2, 4, 8, 16}.
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 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
7147 Use plus_constant instead of gen_rtx_PLUS.
7149 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7150 Alan Hayward <alan.hayward@arm.com>
7151 David Sherwood <david.sherwood@arm.com>
7153 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
7154 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
7155 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
7156 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
7157 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
7158 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
7159 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
7160 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
7161 * config/i386/i386.c (ix86_push_rounding): ...this new function.
7162 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
7164 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
7165 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
7166 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
7167 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
7168 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
7169 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
7170 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
7171 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
7172 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
7173 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
7175 * expr.c (emit_move_resolve_push): Treat the input and result
7176 of PUSH_ROUNDING as a poly_int64.
7177 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
7178 (emit_push_insn): Likewise.
7179 * lra-eliminations.c (mark_not_eliminable): Likewise.
7180 * recog.c (push_operand): Likewise.
7181 * reload1.c (elimination_effects): Likewise.
7182 * rtlanal.c (nonzero_bits1): Likewise.
7183 * calls.c (store_one_arg): Likewise. Require the padding to be
7184 known at compile time.
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 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
7191 Use plus_constant instead of gen_rtx_PLUS.
7193 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7194 Alan Hayward <alan.hayward@arm.com>
7195 David Sherwood <david.sherwood@arm.com>
7197 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
7200 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7201 Alan Hayward <alan.hayward@arm.com>
7202 David Sherwood <david.sherwood@arm.com>
7204 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
7205 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
7206 via stack temporaries. Treat the mode size as polynomial too.
7208 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7209 Alan Hayward <alan.hayward@arm.com>
7210 David Sherwood <david.sherwood@arm.com>
7212 * expr.c (expand_expr_real_2): When handling conversions involving
7213 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
7214 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
7215 as a poly_uint64 too.
7217 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7218 Alan Hayward <alan.hayward@arm.com>
7219 David Sherwood <david.sherwood@arm.com>
7221 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
7223 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7224 Alan Hayward <alan.hayward@arm.com>
7225 David Sherwood <david.sherwood@arm.com>
7227 * combine.c (can_change_dest_mode): Handle polynomial
7228 REGMODE_NATURAL_SIZE.
7229 * expmed.c (store_bit_field_1): Likewise.
7230 * expr.c (store_constructor): Likewise.
7231 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
7232 and polynomial REGMODE_NATURAL_SIZE.
7233 (gen_lowpart_common): Likewise.
7234 * reginfo.c (record_subregs_of_mode): Likewise.
7235 * rtlanal.c (read_modify_subreg_p): Likewise.
7237 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7238 Alan Hayward <alan.hayward@arm.com>
7239 David Sherwood <david.sherwood@arm.com>
7241 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
7242 numbers of elements.
7244 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7245 Alan Hayward <alan.hayward@arm.com>
7246 David Sherwood <david.sherwood@arm.com>
7248 * match.pd: Cope with polynomial numbers of vector elements.
7250 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7251 Alan Hayward <alan.hayward@arm.com>
7252 David Sherwood <david.sherwood@arm.com>
7254 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
7255 in a POINTER_PLUS_EXPR.
7257 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7258 Alan Hayward <alan.hayward@arm.com>
7259 David Sherwood <david.sherwood@arm.com>
7261 * omp-simd-clone.c (simd_clone_subparts): New function.
7262 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
7263 (ipa_simd_modify_function_body): Likewise.
7265 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7266 Alan Hayward <alan.hayward@arm.com>
7267 David Sherwood <david.sherwood@arm.com>
7269 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
7270 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
7271 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
7272 (expand_vector_condition, vector_element): Likewise.
7273 (subparts_gt): New function.
7274 (get_compute_type): Use subparts_gt.
7275 (count_type_subparts): Delete.
7276 (expand_vector_operations_1): Use subparts_gt instead of
7277 count_type_subparts.
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 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
7284 (vect_compile_time_alias): ...this new function. Do the calculation
7285 on poly_ints rather than trees.
7286 (vect_prune_runtime_alias_test_list): Update call accordingly.
7288 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7289 Alan Hayward <alan.hayward@arm.com>
7290 David Sherwood <david.sherwood@arm.com>
7292 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
7294 (vect_schedule_slp_instance): Likewise.
7296 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7297 Alan Hayward <alan.hayward@arm.com>
7298 David Sherwood <david.sherwood@arm.com>
7300 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
7301 constant and extern definitions for variable-length vectors.
7302 (vect_get_constant_vectors): Note that the number of units
7303 is known to be constant.
7305 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7306 Alan Hayward <alan.hayward@arm.com>
7307 David Sherwood <david.sherwood@arm.com>
7309 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
7310 of units as polynomial. Choose between WIDE and NARROW based
7313 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7314 Alan Hayward <alan.hayward@arm.com>
7315 David Sherwood <david.sherwood@arm.com>
7317 * tree-vect-stmts.c (simd_clone_subparts): New function.
7318 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
7320 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7321 Alan Hayward <alan.hayward@arm.com>
7322 David Sherwood <david.sherwood@arm.com>
7324 * tree-vect-stmts.c (vectorizable_call): Treat the number of
7325 vectors as polynomial. Use build_index_vector for
7328 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7329 Alan Hayward <alan.hayward@arm.com>
7330 David Sherwood <david.sherwood@arm.com>
7332 * tree-vect-stmts.c (get_load_store_type): Treat the number of
7333 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
7334 for variable-length vectors.
7335 (vectorizable_mask_load_store): Treat the number of units as
7336 polynomial, asserting that it is constant if the condition has
7337 already been enforced.
7338 (vectorizable_store, vectorizable_load): Likewise.
7340 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7341 Alan Hayward <alan.hayward@arm.com>
7342 David Sherwood <david.sherwood@arm.com>
7344 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
7345 of units as polynomial. Punt if we can't tell at compile time
7346 which vector contains the final result.
7348 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7349 Alan Hayward <alan.hayward@arm.com>
7350 David Sherwood <david.sherwood@arm.com>
7352 * tree-vect-loop.c (vectorizable_induction): Treat the number
7353 of units as polynomial. Punt on SLP inductions. Use an integer
7354 VEC_SERIES_EXPR for variable-length integer reductions. Use a
7355 cast of such a series for variable-length floating-point
7358 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7359 Alan Hayward <alan.hayward@arm.com>
7360 David Sherwood <david.sherwood@arm.com>
7362 * tree.h (build_index_vector): Declare.
7363 * tree.c (build_index_vector): New function.
7364 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
7365 of units as polynomial, forcibly converting it to a constant if
7366 vectorizable_reduction has already enforced the condition.
7367 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
7368 to create a {1,2,3,...} vector.
7369 (vectorizable_reduction): Treat the number of units as polynomial.
7370 Choose vectype_in based on the largest scalar element size rather
7371 than the smallest number of units. Enforce the restrictions
7374 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7375 Alan Hayward <alan.hayward@arm.com>
7376 David Sherwood <david.sherwood@arm.com>
7378 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
7379 number of units as polynomial.
7381 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7382 Alan Hayward <alan.hayward@arm.com>
7383 David Sherwood <david.sherwood@arm.com>
7385 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
7386 * target.def (autovectorize_vector_sizes): Return the vector sizes
7387 by pointer, using vector_sizes rather than a bitmask.
7388 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
7389 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
7390 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
7392 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
7393 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
7394 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
7395 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
7396 * omp-general.c (omp_max_vf): Likewise.
7397 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7398 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
7399 * tree-vect-loop.c (vect_analyze_loop): Likewise.
7400 * tree-vect-slp.c (vect_slp_bb): Likewise.
7401 * doc/tm.texi: Regenerate.
7402 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
7404 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
7405 the vector size as a poly_uint64 rather than an unsigned int.
7406 (current_vector_size): Change from an unsigned int to a poly_uint64.
7407 (get_vectype_for_scalar_type): Update accordingly.
7408 * tree.h (build_truth_vector_type): Take the size and number of
7409 units as a poly_uint64 rather than an unsigned int.
7410 (build_vector_type): Add a temporary overload that takes
7411 the number of units as a poly_uint64 rather than an unsigned int.
7412 * tree.c (make_vector_type): Likewise.
7413 (build_truth_vector_type): Take the number of units as a poly_uint64
7414 rather than an unsigned int.
7416 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7417 Alan Hayward <alan.hayward@arm.com>
7418 David Sherwood <david.sherwood@arm.com>
7420 * target.def (get_mask_mode): Take the number of units and length
7421 as poly_uint64s rather than unsigned ints.
7422 * targhooks.h (default_get_mask_mode): Update accordingly.
7423 * targhooks.c (default_get_mask_mode): Likewise.
7424 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
7425 * doc/tm.texi: Regenerate.
7427 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7428 Alan Hayward <alan.hayward@arm.com>
7429 David Sherwood <david.sherwood@arm.com>
7431 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
7432 * omp-general.c (omp_max_vf): Likewise.
7433 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
7434 (expand_omp_simd): Handle polynomial safelen.
7435 * omp-low.c (omplow_simd_context): Add a default constructor.
7436 (omplow_simd_context::max_vf): Change from int to poly_uint64.
7437 (lower_rec_simd_input_clauses): Update accordingly.
7438 (lower_rec_input_clauses): Likewise.
7440 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7441 Alan Hayward <alan.hayward@arm.com>
7442 David Sherwood <david.sherwood@arm.com>
7444 * tree-vectorizer.h (vect_nunits_for_cost): New function.
7445 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
7446 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
7447 (vect_analyze_slp_cost): Likewise.
7448 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
7449 (vect_model_load_cost): Likewise.
7451 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7452 Alan Hayward <alan.hayward@arm.com>
7453 David Sherwood <david.sherwood@arm.com>
7455 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
7456 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
7457 from an unsigned int * to a poly_uint64_pod *.
7458 (calculate_unrolling_factor): New function.
7459 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
7461 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7462 Alan Hayward <alan.hayward@arm.com>
7463 David Sherwood <david.sherwood@arm.com>
7465 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
7466 from an unsigned int to a poly_uint64.
7467 (_loop_vec_info::slp_unrolling_factor): Likewise.
7468 (_loop_vec_info::vectorization_factor): Change from an int
7470 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
7471 (vect_get_num_vectors): New function.
7472 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
7473 (vect_get_num_copies): Use vect_get_num_vectors.
7474 (vect_analyze_data_ref_dependences): Change max_vf from an int *
7475 to an unsigned int *.
7476 (vect_analyze_data_refs): Change min_vf from an int * to a
7478 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7479 than an unsigned HOST_WIDE_INT.
7480 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7481 (vect_analyze_data_ref_dependence): Change max_vf from an int *
7482 to an unsigned int *.
7483 (vect_analyze_data_ref_dependences): Likewise.
7484 (vect_compute_data_ref_alignment): Handle polynomial vf.
7485 (vect_enhance_data_refs_alignment): Likewise.
7486 (vect_prune_runtime_alias_test_list): Likewise.
7487 (vect_shift_permute_load_chain): Likewise.
7488 (vect_supportable_dr_alignment): Likewise.
7489 (dependence_distance_ge_vf): Take the vectorization factor as a
7490 poly_uint64 rather than an unsigned HOST_WIDE_INT.
7491 (vect_analyze_data_refs): Change min_vf from an int * to a
7493 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
7494 vfm1 as a poly_uint64 rather than an int. Make the same change
7495 for the returned bound_scalar.
7496 (vect_gen_vector_loop_niters): Handle polynomial vf.
7497 (vect_do_peeling): Likewise. Update call to
7498 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
7499 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
7501 * tree-vect-loop.c (vect_determine_vectorization_factor)
7502 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
7503 (vect_get_known_peeling_cost): Likewise.
7504 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
7505 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
7506 (vect_transform_loop): Likewise. Use the lowest possible VF when
7507 updating the upper bounds of the loop.
7508 (vect_min_worthwhile_factor): Make static. Return an unsigned int
7510 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
7511 polynomial unroll factors.
7512 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
7513 (vect_make_slp_decision): Likewise.
7514 (vect_supported_load_permutation_p): Likewise, and polynomial
7516 (vect_analyze_slp_cost): Handle polynomial vf.
7517 (vect_slp_analyze_node_operations): Likewise.
7518 (vect_slp_analyze_bb_1): Likewise.
7519 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7520 than an unsigned HOST_WIDE_INT.
7521 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
7522 (vectorizable_load): Handle polynomial vf.
7523 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
7525 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
7527 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7528 Alan Hayward <alan.hayward@arm.com>
7529 David Sherwood <david.sherwood@arm.com>
7531 * match.pd: Handle bit operations involving three constants
7532 and try to fold one pair.
7534 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7536 * tree-vect-loop-manip.c: Include gimple-fold.h.
7537 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
7538 niters_maybe_zero parameters. Handle other cases besides a step of 1.
7539 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
7540 Add a path that uses a step of VF instead of 1, but disable it
7542 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
7543 and niters_no_overflow parameters. Update calls to
7544 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
7545 Create a new SSA name if the latter choses to use a ste other
7546 than zero, and return it via niters_vector_mult_vf_var.
7547 * tree-vect-loop.c (vect_transform_loop): Update calls to
7548 vect_do_peeling, vect_gen_vector_loop_niters and
7549 slpeel_make_loop_iterate_ntimes.
7550 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
7551 (vect_gen_vector_loop_niters): Update declarations after above changes.
7553 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
7555 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
7556 128-bit round to integer instructions.
7557 (ceil<mode>2): Likewise.
7558 (btrunc<mode>2): Likewise.
7559 (round<mode>2): Likewise.
7561 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7563 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
7564 unaligned VSX load/store on P8/P9.
7565 (expand_block_clear): Allow the use of unaligned VSX
7566 load/store on P8/P9.
7568 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7570 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
7572 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
7573 swap associated with both a load and a store.
7575 2018-01-02 Andrew Waterman <andrew@sifive.com>
7577 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
7578 * config/riscv/riscv.md (clear_cache): Use it.
7580 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
7582 * web.c: Remove out-of-date comment.
7584 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7586 * expr.c (fixup_args_size_notes): Check that any existing
7587 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
7588 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
7589 (emit_single_push_insn): ...here.
7591 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7593 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
7594 (const_vector_encoded_nelts): New function.
7595 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
7596 (const_vector_int_elt, const_vector_elt): Declare.
7597 * emit-rtl.c (const_vector_int_elt_1): New function.
7598 (const_vector_elt): Likewise.
7599 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
7600 of CONST_VECTOR_ELT.
7602 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7604 * expr.c: Include rtx-vector-builder.h.
7605 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
7606 directly on the tree encoding.
7607 (const_vector_from_tree): Likewise.
7608 * optabs.c: Include rtx-vector-builder.h.
7609 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
7610 sequence of "u" values.
7611 * vec-perm-indices.c: Include rtx-vector-builder.h.
7612 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
7613 directly on the vec_perm_indices encoding.
7615 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7617 * doc/rtl.texi (const_vector): Describe new encoding scheme.
7618 * Makefile.in (OBJS): Add rtx-vector-builder.o.
7619 * rtx-vector-builder.h: New file.
7620 * rtx-vector-builder.c: Likewise.
7621 * rtl.h (rtx_def::u2): Add a const_vector field.
7622 (CONST_VECTOR_NPATTERNS): New macro.
7623 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
7624 (CONST_VECTOR_DUPLICATE_P): Likewise.
7625 (CONST_VECTOR_STEPPED_P): Likewise.
7626 (CONST_VECTOR_ENCODED_ELT): Likewise.
7627 (const_vec_duplicate_p): Check for a duplicated vector encoding.
7628 (unwrap_const_vec_duplicate): Likewise.
7629 (const_vec_series_p): Check for a non-duplicated vector encoding.
7630 Say that the function only returns true for integer vectors.
7631 * emit-rtl.c: Include rtx-vector-builder.h.
7632 (gen_const_vec_duplicate_1): Delete.
7633 (gen_const_vector): Call gen_const_vec_duplicate instead of
7634 gen_const_vec_duplicate_1.
7635 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
7636 (gen_const_vec_duplicate): Use rtx_vector_builder.
7637 (gen_const_vec_series): Likewise.
7638 (gen_rtx_CONST_VECTOR): Likewise.
7639 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
7640 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7641 Build a new vector rather than modifying a CONST_VECTOR in-place.
7642 (handle_special_swappables): Update call accordingly.
7643 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
7644 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7645 Build a new vector rather than modifying a CONST_VECTOR in-place.
7646 (handle_special_swappables): Update call accordingly.
7648 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7650 * simplify-rtx.c (simplify_const_binary_operation): Use
7651 CONST_VECTOR_ELT instead of XVECEXP.
7653 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7655 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
7656 the selector elements to be different from the data elements
7657 if the selector is a VECTOR_CST.
7658 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
7659 ssizetype for the selector.
7661 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7663 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
7664 before testing each element individually.
7665 * tree-vect-generic.c (lower_vec_perm): Likewise.
7667 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7669 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
7670 * selftest-run-tests.c (selftest::run_tests): Call it.
7671 * vector-builder.h (vector_builder::operator ==): New function.
7672 (vector_builder::operator !=): Likewise.
7673 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
7674 (vec_perm_indices::all_from_input_p): New function.
7675 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7676 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
7677 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
7678 instead of reading the VECTOR_CST directly. Detect whether both
7679 vector inputs are the same before constructing the vec_perm_indices,
7680 and update the number of inputs argument accordingly. Use the
7681 utility functions added above. Only construct sel2 if we need to.
7683 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7685 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
7686 the broadcast of the low byte.
7687 (expand_mult_highpart): Use an explicit encoding for the permutes.
7688 * optabs-query.c (can_mult_highpart_p): Likewise.
7689 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
7690 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7691 (vectorizable_bswap): Likewise.
7692 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
7693 explicit encoding for the power-of-2 permutes.
7694 (vect_permute_store_chain): Likewise.
7695 (vect_grouped_load_supported): Likewise.
7696 (vect_permute_load_chain): Likewise.
7698 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7700 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
7701 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
7702 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
7703 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7704 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
7705 (vect_gen_perm_mask_any): Likewise.
7707 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7709 * int-vector-builder.h: New file.
7710 * vec-perm-indices.h: Include int-vector-builder.h.
7711 (vec_perm_indices): Redefine as an int_vector_builder.
7712 (auto_vec_perm_indices): Delete.
7713 (vec_perm_builder): Redefine as a stand-alone class.
7714 (vec_perm_indices::vec_perm_indices): New function.
7715 (vec_perm_indices::clamp): Likewise.
7716 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
7717 (vec_perm_indices::new_vector): New function.
7718 (vec_perm_indices::new_expanded_vector): Update for new
7719 vec_perm_indices class.
7720 (vec_perm_indices::rotate_inputs): New function.
7721 (vec_perm_indices::all_in_range_p): Operate directly on the
7722 encoded form, without computing elided elements.
7723 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
7724 encoding. Update for new vec_perm_indices class.
7725 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
7726 the given vec_perm_builder.
7727 (expand_vec_perm_var): Update vec_perm_builder constructor.
7728 (expand_mult_highpart): Use vec_perm_builder instead of
7729 auto_vec_perm_indices.
7730 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
7731 vec_perm_indices instead of auto_vec_perm_indices. Use a single
7732 or double series encoding as appropriate.
7733 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
7734 vec_perm_indices instead of auto_vec_perm_indices.
7735 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7736 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7737 (vect_permute_store_chain): Likewise.
7738 (vect_grouped_load_supported): Likewise.
7739 (vect_permute_load_chain): Likewise.
7740 (vect_shift_permute_load_chain): Likewise.
7741 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
7742 (vect_transform_slp_perm_load): Likewise.
7743 (vect_schedule_slp_instance): Likewise.
7744 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7745 (vectorizable_mask_load_store): Likewise.
7746 (vectorizable_bswap): Likewise.
7747 (vectorizable_store): Likewise.
7748 (vectorizable_load): Likewise.
7749 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
7750 vec_perm_indices instead of auto_vec_perm_indices. Use
7751 tree_to_vec_perm_builder to read the vector from a tree.
7752 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
7753 vec_perm_builder instead of a vec_perm_indices.
7754 (have_whole_vector_shift): Use vec_perm_builder and
7755 vec_perm_indices instead of auto_vec_perm_indices. Leave the
7756 truncation to calc_vec_perm_mask_for_shift.
7757 (vect_create_epilog_for_reduction): Likewise.
7758 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
7759 from auto_vec_perm_indices to vec_perm_indices.
7760 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7761 instead of changing individual elements.
7762 (aarch64_vectorize_vec_perm_const): Use new_vector to install
7763 the vector in d.perm.
7764 * config/arm/arm.c (expand_vec_perm_d::perm): Change
7765 from auto_vec_perm_indices to vec_perm_indices.
7766 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7767 instead of changing individual elements.
7768 (arm_vectorize_vec_perm_const): Use new_vector to install
7769 the vector in d.perm.
7770 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
7771 Update vec_perm_builder constructor.
7772 (rs6000_expand_interleave): Likewise.
7773 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
7774 (rs6000_expand_interleave): Likewise.
7776 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7778 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
7779 to qimode could truncate the indices.
7780 * optabs.c (expand_vec_perm_var): Likewise.
7782 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7784 * Makefile.in (OBJS): Add vec-perm-indices.o.
7785 * vec-perm-indices.h: New file.
7786 * vec-perm-indices.c: Likewise.
7787 * target.h (vec_perm_indices): Replace with a forward class
7789 (auto_vec_perm_indices): Move to vec-perm-indices.h.
7790 * optabs.h: Include vec-perm-indices.h.
7791 (expand_vec_perm): Delete.
7792 (selector_fits_mode_p, expand_vec_perm_var): Declare.
7793 (expand_vec_perm_const): Declare.
7794 * target.def (vec_perm_const_ok): Replace with...
7795 (vec_perm_const): ...this new hook.
7796 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
7797 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
7798 * doc/tm.texi: Regenerate.
7799 * optabs.def (vec_perm_const): Delete.
7800 * doc/md.texi (vec_perm_const): Likewise.
7801 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
7802 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
7803 expand_vec_perm for constant permutation vectors. Assert that
7804 the mode of variable permutation vectors is the integer equivalent
7805 of the mode that is being permuted.
7806 * optabs-query.h (selector_fits_mode_p): Declare.
7807 * optabs-query.c: Include vec-perm-indices.h.
7808 (selector_fits_mode_p): New function.
7809 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
7810 is defined, instead of checking whether the vec_perm_const_optab
7811 exists. Use targetm.vectorize.vec_perm_const instead of
7812 targetm.vectorize.vec_perm_const_ok. Check whether the indices
7813 fit in the vector mode before using a variable permute.
7814 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
7815 vec_perm_indices instead of an rtx.
7816 (expand_vec_perm): Replace with...
7817 (expand_vec_perm_const): ...this new function. Take the selector
7818 as a vec_perm_indices rather than an rtx. Also take the mode of
7819 the selector. Update call to shift_amt_for_vec_perm_mask.
7820 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
7821 Use vec_perm_indices::new_expanded_vector to expand the original
7822 selector into bytes. Check whether the indices fit in the vector
7823 mode before using a variable permute.
7824 (expand_vec_perm_var): Make global.
7825 (expand_mult_highpart): Use expand_vec_perm_const.
7826 * fold-const.c: Includes vec-perm-indices.h.
7827 * tree-ssa-forwprop.c: Likewise.
7828 * tree-vect-data-refs.c: Likewise.
7829 * tree-vect-generic.c: Likewise.
7830 * tree-vect-loop.c: Likewise.
7831 * tree-vect-slp.c: Likewise.
7832 * tree-vect-stmts.c: Likewise.
7833 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
7835 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
7836 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
7837 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
7838 (aarch64_vectorize_vec_perm_const): ...this new function.
7839 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7840 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7841 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
7842 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
7843 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7844 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7845 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
7847 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
7848 check for NEON modes.
7849 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
7850 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
7851 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
7852 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
7854 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
7855 the old VEC_PERM_CONST conditions.
7856 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
7857 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
7858 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
7859 (ia64_vectorize_vec_perm_const_ok): Merge into...
7860 (ia64_vectorize_vec_perm_const): ...this new function.
7861 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
7862 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
7863 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
7864 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
7865 * config/mips/mips.c (mips_expand_vec_perm_const)
7866 (mips_vectorize_vec_perm_const_ok): Merge into...
7867 (mips_vectorize_vec_perm_const): ...this new function.
7868 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
7869 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
7870 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
7871 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
7872 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
7873 (rs6000_expand_vec_perm_const): Delete.
7874 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
7876 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7877 (altivec_expand_vec_perm_const_le): Take each operand individually.
7878 Operate on constant selectors rather than rtxes.
7879 (altivec_expand_vec_perm_const): Likewise. Update call to
7880 altivec_expand_vec_perm_const_le.
7881 (rs6000_expand_vec_perm_const): Delete.
7882 (rs6000_vectorize_vec_perm_const_ok): Delete.
7883 (rs6000_vectorize_vec_perm_const): New function.
7884 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
7885 an element count and rtx array.
7886 (rs6000_expand_extract_even): Update call accordingly.
7887 (rs6000_expand_interleave): Likewise.
7888 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
7889 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
7890 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
7891 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
7892 (rs6000_expand_vec_perm_const): Delete.
7893 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7894 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7895 (altivec_expand_vec_perm_const_le): Take each operand individually.
7896 Operate on constant selectors rather than rtxes.
7897 (altivec_expand_vec_perm_const): Likewise. Update call to
7898 altivec_expand_vec_perm_const_le.
7899 (rs6000_expand_vec_perm_const): Delete.
7900 (rs6000_vectorize_vec_perm_const_ok): Delete.
7901 (rs6000_vectorize_vec_perm_const): New function. Remove stray
7902 reference to the SPE evmerge intructions.
7903 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
7904 an element count and rtx array.
7905 (rs6000_expand_extract_even): Update call accordingly.
7906 (rs6000_expand_interleave): Likewise.
7907 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
7908 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
7910 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7912 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7914 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
7915 vector mode and that that mode matches the mode of the data
7917 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
7918 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
7919 directly using expand_vec_perm_1 when forcing selectors into
7921 (expand_vec_perm_var): New function, split out from expand_vec_perm.
7923 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7925 * optabs-query.h (can_vec_perm_p): Delete.
7926 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
7927 * optabs-query.c (can_vec_perm_p): Split into...
7928 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
7929 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
7930 particular selector is valid.
7931 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7932 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7933 (vect_grouped_load_supported): Likewise.
7934 (vect_shift_permute_load_chain): Likewise.
7935 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
7936 (vect_transform_slp_perm_load): Likewise.
7937 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7938 (vectorizable_bswap): Likewise.
7939 (vect_gen_perm_mask_checked): Likewise.
7940 * fold-const.c (fold_ternary_loc): Likewise. Don't take
7941 implementations of variable permutation vectors into account
7942 when deciding which selector to use.
7943 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
7944 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
7945 with a false third argument.
7946 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
7947 to test whether the constant selector is valid and can_vec_perm_var_p
7948 to test whether a variable selector is valid.
7950 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7952 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
7953 * optabs-query.c (can_vec_perm_p): Likewise.
7954 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
7955 instead of vec_perm_indices.
7956 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
7957 (vect_gen_perm_mask_checked): Likewise,
7958 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
7959 (vect_gen_perm_mask_checked): Likewise,
7961 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7963 * optabs-query.h (qimode_for_vec_perm): Declare.
7964 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
7965 (qimode_for_vec_perm): ...this new function.
7966 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
7968 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7970 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
7971 does not have a conditional at the top.
7973 2018-01-02 Richard Biener <rguenther@suse.de>
7975 * ipa-inline.c (big_speedup_p): Fix expression.
7977 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
7980 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
7983 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
7987 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
7988 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
7989 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
7990 cond_taken_branch_cost 3->4.
7992 2018-01-01 Jakub Jelinek <jakub@redhat.com>
7994 PR tree-optimization/83581
7995 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
7996 TODO_cleanup_cfg if any changes have been made.
7999 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
8000 convert_modes if target mode has the right side, but different mode
8004 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
8005 last argument when extracting from CONCAT. If either from_real or
8006 from_imag is NULL, use expansion through memory. If result is not
8007 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
8008 the parts directly to inner mode, if even that fails, use expansion
8012 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
8013 check for bswap in mode rather than HImode and use that in expand_unop
8016 Copyright (C) 2018 Free Software Foundation, Inc.
8018 Copying and distribution of this file, with or without modification,
8019 are permitted in any medium without royalty provided the copyright
8020 notice and this notice are preserved.