1 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2 Chung-Ju Wu <jasonwucj@gmail.com>
4 *config/nds32/nds32-memory-manipulation.c
5 (nds32_emit_load_store): New.
6 (nds32_emit_post_inc_load_store): New.
7 (nds32_emit_mem_move): New.
8 (nds32_emit_mem_move_block): New.
9 (nds32_expand_movmemsi_loop_unknown_size): New.
10 (nds32_expand_movmemsi_loop_known_size): New.
11 (nds32_expand_movmemsi_loop): New.
12 (nds32_expand_movmemsi_unroll): New.
13 (nds32_expand_movmemqi): Rename ...
14 (nds32_expand_movmemsi): ... to this.
15 *config/nds32/nds32-multiple.md (movmemqi): Rename ...
16 (movmemsi): ... to this.
17 *config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
18 (nds32_expand_movmemsi): ... to this.
20 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
21 Monk Chiang <sh.chiang04@gmail.com>
22 Chung-Ju Wu <jasonwucj@gmail.com>
24 * config/nds32/nds32-protos.h
25 (nds32_expand_load_multiple): New arguments.
26 (nds32_expand_store_multiple): Ditto.
27 (nds32_valid_multiple_load_store): Rename ...
28 (nds32_valid_multiple_load_store_p): ... to this.
29 * config/nds32/nds32-memory-manipulation.c
30 (nds32_expand_load_multiple): Refine implementation.
31 (nds32_expand_store_multiple): Ditto.
32 * config/nds32/nds32-multiple.md
33 (load_multiple): Update nds32_expand_load_multiple interface.
34 (store_multiple): Update nds32_expand_store_multiple interface.
35 * config/nds32/nds32-predicates.c
36 (nds32_valid_multiple_load_store): Rename ...
37 (nds32_valid_multiple_load_store_p): ... to this and refine
39 * config/nds32/predicates.md
40 (nds32_load_multiple_and_update_address_operation): New predicate.
41 (nds32_store_multiple_and_update_address_operation): New predicate.
43 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
44 Chung-Ju Wu <jasonwucj@gmail.com>
46 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
47 (combo): New attribute.
48 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
50 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
52 * config/nds32/nds32.opt: Change -mcmodel= default value.
54 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
55 Monk Chiang <sh.chiang04@gmail.com>
56 Chung-Ju Wu <jasonwucj@gmail.com>
58 * config/nds32/constants.md (unspec_element): New enum.
59 * config/nds32/constraints.md (Umw): New constraint.
60 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
61 * config/nds32/nds32-intrinsic.md: Likewise.
62 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
63 (nds32_valid_smw_lwm_base_p): New.
64 (nds32_output_smw_single_word): New.
65 (nds32_output_lmw_single_word): New.
66 (nds32_expand_unaligned_load): New.
67 (nds32_expand_unaligned_store): New.
68 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
69 (nds32_output_smw_single_word): Declare.
70 (nds32_output_lmw_single_word): Declare.
71 (nds32_expand_unaligned_load): Declare.
72 (nds32_expand_unaligned_store): Declare.
73 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
74 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
75 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
76 NDS32_BUILTIN_UASTORE_DW.
77 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
80 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
81 Kito Cheng <kito.cheng@gmail.com>
82 Chung-Ju Wu <jasonwucj@gmail.com>
84 * config/nds32/nds32-intrinsic.c
85 (nds32_expand_builtin_null_ftype_reg): Delete.
86 (nds32_expand_builtin_reg_ftype_imm): Ditto.
87 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
88 (nds32_read_argument): New.
89 (nds32_legitimize_target): Ditto.
90 (nds32_legitimize_argument): Ditto.
91 (nds32_check_constant_argument): Ditto.
92 (nds32_expand_unop_builtin): Ditto.
93 (nds32_expand_unopimm_builtin): Ditto.
94 (nds32_expand_binop_builtin): Ditto.
95 (nds32_builtin_decl_impl): Ditto.
96 (builtin_description): Ditto.
97 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
98 (nds32_init_builtins_impl): Ditto.
99 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
100 (nds32_builtin_decl): New.
101 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
102 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
104 2018-03-02 Jeff Law <law@redhat.com>
106 * reorg.c (stop_search_p): Handle DEBUG_INSN.
107 (redundant_insn, fill_simple_delay_slots): Likewise.
108 (fill_slots_from_thread): Likewise.
109 * resource.c (mark_referenced_resources): Likewise.
110 (mark_set_resources, find_dead_or_set_registers): Likewise.
112 2018-03-02 Jakub Jelinek <jakub@redhat.com>
114 * substring-locations.h (format_warning_va): Formatting fix for
116 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
118 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
119 * substring-locations.c: Include intl.h.
120 (format_warning_va): Turned into small wrapper around
121 format_warning_n_va, renamed to ...
122 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
123 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
125 (format_warning_at_substring_n): New function.
126 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
127 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
128 format_warning_at_substring with just a shorter name instead of
129 const function pointer.
130 (fmtwarn_n): New function.
131 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
132 appropriate, get rid of all the fmtstr temporaries, move conditionals
133 with G_() wrapped string literals directly into fmtwarn arguments,
134 cast dir.len to (int), formatting fixes.
136 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
138 * doc/invoke.texi: Remove "Cilk Plus" references.
140 2018-03-02 Jakub Jelinek <jakub@redhat.com>
141 Richard Biener <rguenther@suse.de>
144 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
145 for error or warning attributes if CALL_FROM_THUNK_P is set.
148 2018-03-02 Jakub Jelinek <jakub@redhat.com>
151 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
152 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
155 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
156 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
158 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
159 instead of -1U in last predictors element's probability member.
161 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
164 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
165 arguments if they are comparable.
167 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
169 PR tree-optimization/84634
170 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
171 masks and masked_loop_p with a single loop_masks, making sure it's
172 null for bb vectorization.
174 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
176 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
177 (vect_analyze_data_ref_access): Use loop->safe_len rather than
178 loop->force_vectorize to check whether there is no alias.
180 2018-03-02 Jakub Jelinek <jakub@redhat.com>
183 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
185 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
187 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
188 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
189 instead of a loop around prev_real_insn.
190 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
194 * config/i386/i386.c (ix86_print_operand): Use conditional
195 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
198 2018-03-02 Richard Biener <rguenther@suse.de>
200 PR tree-optimization/84427
201 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
202 (bitmap_set_subtract_values): Rewrite to handle multiple
205 (prune_clobbered_mems): Likewise.
206 (phi_translate): Take edge instead of pred/phiblock.
207 (phi_translate_1): Likewise.
208 (phi_translate_set): Likewise. Insert all translated
209 exprs for a value into the set, keeping possibly multiple
210 expressions per value.
211 (compute_antic_aux): Adjust for phi_translate changes.
212 When intersecting union the expressions and prune those
213 not in the final value set, keeping possibly multiple
214 expressions per value. Do not use value-insertion
215 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
216 all expressions. Add verification that the value-sets
217 only shrink during iteration.
218 (compute_partial_antic_aux): Adjust for the phi_translate changes.
219 (do_pre_regular_insertion): Likewise.
220 (do_pre_partial_partial_insertion): Likewise.
222 2018-03-02 Richard Biener <rguenther@suse.de>
225 * config/darwin.c (saved_debug_info_level): New static global.
226 (darwin_asm_lto_start): Disable debug info generation for LTO out.
227 (darwin_asm_lto_end): Restore debug info generation settings.
229 2018-03-01 Martin Liska <mliska@suse.cz>
232 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
235 2018-03-01 Richard Biener <rguenther@suse.de>
238 * dwarf2out.c (gen_variable_die): Properly handle late VLA
239 type annotation with LTO when debug was disabled at compile-time.
241 2018-03-01 Matthew Fortune <mfortune@gmail.com>
243 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
245 (mips_final_postscan_insn): Likewise.
247 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
249 PR rtl-optimization/84528
250 * alias.c (init_alias_target): Add commentary.
251 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
252 a unique base value if the frame pointer is not eliminated
253 to the stack pointer.
255 2018-03-01 Tom de Vries <tom@codesourcery.com>
257 PR rtl-optimization/83327
258 * lra-int.h (hard_regs_spilled_into): Declare.
259 * lra.c (hard_regs_spilled_into): Define.
260 (init_reg_info): Init hard_regs_spilled_into.
261 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
262 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
263 (process_bb_lives): Handle hard_regs_spilled_into.
264 (lra_create_live_ranges_1): Before doing liveness propagation, clear
265 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
267 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
269 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
270 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
271 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
272 * config/rs6000/aix72.h: New file.
274 2018-02-28 Jakub Jelinek <jakub@redhat.com>
276 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
277 instead of warning_at with conditional singular and plural messages
281 * stor-layout.c (update_alignment_for_field): For
282 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
283 && !DECL_PACKED (field), do the alignment update, just use
284 only desired_align instead of MAX (type_align, desired_align)
286 (place_field): Don't do known_align < desired_align handling
287 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
288 is non-NULL, instead do it after rli->prev_field handling and
289 only if not within a bitfield word. For DECL_PACKED (field)
290 use type_align of BITS_PER_UNIT.
292 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
294 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
295 superfluous parentheses and trailing spaces.
297 2018-02-28 Richard Biener <rguenther@suse.de>
299 PR tree-optimization/84584
300 * graphite-scop-detection.c (scop_detection::add_scop): Discard
301 SCoPs with fake exit edge.
303 2018-02-28 Martin Liska <mliska@suse.cz>
306 * timevar.c (timer::print): Fix format to properly print 100%
309 2018-02-28 Richard Biener <rguenther@suse.de>
312 * genmatch.c (capture_info::walk_match): Do not mark
313 captured expressions without operands as expr_p given
314 they act more like predicates and should be subject to
315 "lost tail" side-effect preserving.
317 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
319 PR rtl-optimization/81611
320 * auto-inc-dec.c (attempt_change): Move dead note from
321 mem_insn if it's the next use of regno
322 (find_address): Take address use of reg holding
323 non-incremented value. Add parm to limit search to the named
325 (merge_in_block): Attempt to use a mem insn that is the next
326 use of the original regno.
328 2018-02-27 Martin Sebor <msebor@redhat.com>
331 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
332 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
334 2018-02-27 Martin Sebor <msebor@redhat.com>
337 * diagnostic-core.h (warning_n, error_n, inform_n): Change
338 n argument to unsigned HOST_WIDE_INT.
339 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
340 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
341 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
342 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
344 2018-02-27 Richard Biener <rguenther@suse.de>
346 PR tree-optimization/84512
347 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
348 Do not use the estimate returned from record_stmt_cost for
349 the scalar iteration cost but sum properly using add_stmt_cost.
351 2018-02-27 Richard Biener <rguenther@suse.de>
353 PR tree-optimization/84466
354 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
355 Adjust last change to less strictly validate use operands.
357 2018-02-27 Martin Liska <mliska@suse.cz>
359 PR gcov-profile/84548
360 * gcov.c (process_file): Allow partial overlap and consider it
361 also as group functions.
362 (output_lines): Properly calculate range of lines for a group.
364 2018-02-27 Martin Liska <mliska@suse.cz>
366 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
367 'ggc' suffixes. Change first column width.
368 (timer::print): Fix formatting of the column.
370 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
372 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
373 preserve inline entry blocks for the sake of debug inline
374 entry point markers alone.
375 (remove_unused_locals): Suggest in comments a better place to
376 force the preservation of inline entry blocks that are
377 otherwise unused, but do not preserve them.
379 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
381 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
383 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
386 * config/i386/constraints.md (Bs): Replace
387 ix86_indirect_branch_register with
388 TARGET_INDIRECT_BRANCH_REGISTER.
390 * config/i386/i386.md (indirect_jump): Likewise.
391 (tablejump): Likewise.
392 (*sibcall_memory): Likewise.
393 (*sibcall_value_memory): Likewise.
394 Peepholes of indirect call and jump via memory: Likewise.
395 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
396 (*sibcall_value_GOT_32): Likewise.
397 * config/i386/predicates.md (indirect_branch_operand): Likewise.
398 (GOT_memory_operand): Likewise.
399 (call_insn_operand): Likewise.
400 (sibcall_insn_operand): Likewise.
401 (GOT32_symbol_operand): Likewise.
402 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
404 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
406 PR rtl-optimization/83496
407 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
408 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
409 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
410 redundant insn, if any.
411 (relax_delay_slots): Likewise.
412 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
414 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
416 PR tree-optimization/83965
417 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
418 that grouped statements are part of a reduction chain. Return
419 true if the statement is not marked as a reduction itself but
421 (vect_recog_dot_prod_pattern): Don't check whether the statement
422 is part of a group here.
423 (vect_recog_sad_pattern): Likewise.
424 (vect_recog_widen_sum_pattern): Likewise.
426 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
429 * final.c (rest_of_clean_state): Also look for calls inside sequences.
431 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
434 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
436 (ix86_output_indirect_function_return): New prototype.
437 (ix86_split_simple_return_pop_internal): Likewise.
438 * config/i386/i386.c (indirect_return_via_cx): New.
439 (indirect_return_via_cx_bnd): Likewise.
440 (indirect_thunk_name): Handle return va CX_REG.
441 (output_indirect_thunk_function): Create alias for
442 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
443 (ix86_output_indirect_jmp): Remove the bool argument.
444 (ix86_output_indirect_function_return): New function.
445 (ix86_split_simple_return_pop_internal): Likewise.
446 * config/i386/i386.md (*indirect_jump): Don't pass false
447 to ix86_output_indirect_jmp.
448 (*tablejump_1): Likewise.
449 (simple_return_pop_internal): Change it to define_insn_and_split.
450 Call ix86_split_simple_return_pop_internal to split it for
452 (simple_return_indirect_internal): Call
453 ix86_output_indirect_function_return instead of
454 ix86_output_indirect_jmp.
456 2018-02-26 Jakub Jelinek <jakub@redhat.com>
459 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
460 memset and value initialization afterwards.
462 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
464 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
466 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
469 * common/config/aarch64/aarch64-common.c
470 (aarch_option_optimization_table[]): Switch
471 off fomit-frame-pointer
473 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
474 Chung-Ju Wu <jasonwucj@gmail.com>
476 * config/nds32/nds32-multiple.md (load_multiple): Disallow
478 (store_multiple): Ditto.
480 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
482 * config.gcc: Add --with-cpu support for nds32 target.
483 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
484 * config/nds32/nds32.opt: Add -mcpu= option.
486 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
488 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
489 isel=yes): Warn for these deprecated options.
491 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
493 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
494 ISA_2_5_MASKS_EMBEDDED.
496 2018-02-23 Jakub Jelinek <jakub@redhat.com>
498 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
499 p->max as pointers rather than using iterative_hash_expr.
501 2018-02-23 Carl Love <cel@us.ibm.com>
503 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
504 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
506 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
507 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
508 P8V_BUILTIN_VEC_VUNSIGNED2.
510 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
513 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
514 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
516 * lra-constraints.c (curr_insn_transform): Set up
517 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
519 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
520 * lra-eliminations.c (spill_pseudos): Ditto.
521 (process_insn_for_elimination): Ditto.
522 * lra-lives.c (reg_early_clobber_p): Use the new macros.
523 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
524 LRA_NON_CLOBBERED_ALT.
526 2018-02-22 Martin Sebor <msebor@redhat.com>
528 PR tree-optimization/84480
529 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
530 to maybe_diag_stxncpy_trunc. Call it.
531 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
532 from gimple_fold_builtin_strcpy. Print inlining stack.
533 (handle_builtin_stxncpy): Print inlining stack.
534 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
536 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
539 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
540 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
541 and -fcheck-pointer-bounds are used together.
542 (indirect_thunk_prefix): New enum.
543 (indirect_thunk_need_prefix): New function.
544 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
545 "_nt" instead of "_bnd" for NOTRACK prefix.
546 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
547 (output_indirect_thunk_function): Likewise.
549 (ix86_code_end): Update output_indirect_thunk_function calls.
550 (ix86_output_indirect_branch_via_reg): Replace
551 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
552 (ix86_output_indirect_branch_via_push): Likewise.
553 (ix86_output_function_return): Likewise.
554 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
555 incompatible with -fcf-protection=branch and
556 -fcheck-pointer-bounds.
558 2018-02-22 Steve Ellcey <sellcey@cavium.com>
561 * config/aarch64/aarch64.c (aarch64_print_address_internal):
562 Change gcc_assert call to output_operand_lossage.
564 2018-02-22 Steve Ellcey <sellcey@cavium.com>
566 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
568 2018-02-22 DJ Delorie <dj@redhat.com>
569 Sebastian Perta <sebastian.perta@renesas.com>
570 Oleg Endo <olegendo@gcc.gnu.org>
572 * config/rx/rx.c (rx_rtx_costs): New function.
573 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
575 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
577 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
579 2018-02-22 Martin Liska <mliska@suse.cz>
582 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
583 Add "native" as a possible value.
585 2018-02-22 Martin Liska <mliska@suse.cz>
588 * config/i386/i386.c (ix86_option_override_internal):
589 Add "native" as a possible value for -march and -mtune.
591 2018-02-22 Jakub Jelinek <jakub@redhat.com>
594 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
595 to all type variants.
597 PR tree-optimization/84503
598 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
599 width as info->bitpos + info->bitsize - start.
600 (merged_store_group::merge_overlapping): Simplify width computation.
601 (check_no_overlap): New function.
602 (imm_store_chain_info::try_coalesce_bswap): Compute expected
603 start + width and last_order of the group, fail if check_no_overlap
605 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
606 to group if check_no_overlap fails.
608 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
610 * config/rs6000/altivec.md: Delete contraint arguments to
611 define_expand, define_split, and define_peephole2, and in
612 define_insn_and_split if always unused.
613 * config/rs6000/darwin.md: Ditto.
614 * config/rs6000/dfp.md: Ditto.
615 * config/rs6000/rs6000.md: Ditto.
616 * config/rs6000/sync.md: Ditto.
617 * config/rs6000/vector.md: Ditto.
618 * config/rs6000/vsx.md: Ditto.
620 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
622 * config/rs6000/altivec.md: Write output control strings as braced
623 blocks instead of double-quoted strings.
624 * config/rs6000/darwin.md: Ditto.
625 * config/rs6000/rs6000.md: Ditto.
626 * config/rs6000/vector.md: Ditto.
627 * config/rs6000/vsx.md: Ditto.
629 2018-02-21 Jason Merrill <jason@redhat.com>
631 PR c++/84314 - ICE with templates and fastcall attribute.
632 * attribs.c (build_type_attribute_qual_variant): Remove assert.
634 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
636 * ipa-cp.c (determine_versionability): Fix comment typos.
638 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
641 * ipa-cp.c (determine_versionability): Do not version functions caling
644 2018-02-21 Martin Liska <mliska@suse.cz>
647 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
648 Add "native" as a possible value.
649 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
650 the macro when native cpu detection is available.
652 2018-02-21 Martin Liska <mliska@suse.cz>
655 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
656 Add "native" as a possible value.
657 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
658 when native cpu detection is available.
660 2018-02-21 Jakub Jelinek <jakub@redhat.com>
661 Martin Sebor <msebor@redhat.com>
663 PR tree-optimization/84478
664 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
666 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
667 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
668 support which is conservatively correct, for 2 only stay conservative
669 for maxlen. Formatting and comment capitalization fixes. Add STRICT
670 argument to the 2 argument get_range_strlen, adjust 6 arg
671 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
673 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
674 (gimple_fold_builtin_strlen): Pass true as last argument to
677 2018-02-20 Martin Sebor <msebor@redhat.com>
680 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
681 (builtin_memref::set_base_and_offset): Same. Handle inner references.
682 (builtin_memref::builtin_memref): Factor out parts into
683 set_base_and_offset and call it.
685 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
688 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
689 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
690 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
691 search at the associated MODE_INT.
693 2018-02-20 Jeff Law <law@redhat.com>
696 PR tree-optimization/81592
698 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
699 for range data rather than using global data.
700 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
701 range data rather than using global data.
702 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
703 pass it to children as needed.
704 (struct directive::fmtresult): Similarly.
705 (struct directive::set_width): Similarly.
706 (struct directive::set_precision): Similarly.
707 (format_integer, format_directive, parse_directive): Similarly.
708 (format_none): Accept unnamed vr_values parameter.
709 (format_percent, format_floating, format_character): Similarly.
710 (format_string, format_plain): Similarly.
711 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
712 the EVRP range analyzer for range data rather than using global data.
713 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
714 gimple-ssa-evrp-analyze.h
715 (class sprintf_dom_walker): Add after_dom_children member function.
716 Add evrp_range_analyzer member.
717 (sprintf_dom_walker::before_dom_children): Call into the EVRP
718 range analyzer as needed.
719 (sprintf_dom_walker::after_dom_children): New member function.
720 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
722 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
723 (evrp_range_analyzer::pop_to_marker): Likewise.
725 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
727 PR tree-optimization/84419
728 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
729 with the required type if its current type is compatible but
732 2018-02-20 Jakub Jelinek <jakub@redhat.com>
735 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
738 2018-02-20 Martin Liska <mliska@suse.cz>
741 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
742 possible values if we don't have a hint.
744 2018-02-20 Martin Liska <mliska@suse.cz>
748 * final.c (shorten_branches): Build align_tab array with one
750 * opts.c (finish_options): Add alignment option limit check.
751 (MAX_CODE_ALIGN): Likewise.
752 (MAX_CODE_ALIGN_VALUE): Likewise.
753 * doc/invoke.texi: Document maximum allowed option value for
754 all -falign-* options.
756 2018-02-19 Jakub Jelinek <jakub@redhat.com>
759 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
760 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
761 * var-tracking.c (emit_note_insn_var_location): Remove all references
762 to NOTE_INSN_CALL_ARG_LOCATION.
763 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
764 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
765 Use copy_rtx_if_shared.
766 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
767 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
768 (dwarf2out_var_location): Remove handling of
769 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
771 * final.c (final_scan_insn): Remove all references to
772 NOTE_INSN_CALL_ARG_LOCATION.
773 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
774 before dumping final insns.
775 * except.c (emit_note_eh_region_end): Remove all references to
776 NOTE_INSN_CALL_ARG_LOCATION.
777 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
778 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
779 * config/arc/arc.c (hwloop_optimize): Likewise.
780 * config/arm/arm.c (create_fix_barrier): Likewise.
781 * config/s390/s390.c (s390_chunkify_start): Likewise.
782 * config/sh/sh.c (find_barrier): Likewise.
783 * config/i386/i386.c (rest_of_insert_endbranch,
784 ix86_seh_fixup_eh_fallthru): Likewise.
785 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
786 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
787 * config/frv/frv.c (frv_function_prologue): Likewise.
788 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
790 (note_outside_basic_block_p): Remove all references to
791 NOTE_INSN_CALL_ARG_LOCATION.
792 * gengtype.c (adjust_field_rtx_def): Likewise.
793 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
795 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
796 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
799 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
802 PR tree-optimization/84452
803 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
804 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
807 2018-02-19 Martin Liska <mliska@suse.cz>
810 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
812 2018-02-19 Martin Liska <mliska@suse.cz>
813 Richard Sandiford <richard.sandiford@linaro.org>
815 PR tree-optimization/82491
816 * gimple-fold.c (get_base_constructor): Make earlier bail out
819 2018-02-19 Carl Love <cel@us.ibm.com>
821 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
822 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
824 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
827 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
829 * config/rl78/rl78.md (movdf): New define expand.
831 2018-02-19 Martin Liska <mliska@suse.cz>
834 * doc/invoke.texi: Fix typo.
835 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
837 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
839 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
840 handle rs6000_single_float and rs6000_double_float specially for
843 2018-02-16 Jeff Law <law@redhat.com>
845 * config/rx/rx.c (add_pop_cfi_notes): New function.;
848 2018-02-16 Jakub Jelinek <jakub@redhat.com>
851 * ipa-inline.c (inline_small_functions): Fix a typo.
853 2018-02-16 Nathan Sidwell <nathan@acm.org>
855 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
857 2018-02-16 Carl Love <cel@us.ibm.com>
859 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
860 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
861 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
862 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
863 expansion to P8V_BUILTIN_VEC_FLOAT2.
865 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
867 PR rtl-optimization/70023
868 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
869 src_regno into account.
871 2018-02-16 Carl Love <cel@us.ibm.com>
873 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
874 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
875 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
876 * config/rs6000/rs6000.c: Remove case statements for
877 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
878 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
879 and P9V_BUILTIN_VEC_VINSERT4B.
880 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
881 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
882 * config/rs6000/vsx.md:
883 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
886 2018-02-16 Carl Love <cel@us.ibm.com>
888 * config/rs6000/altivec.h: Add builtin names vec_extract4b
890 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
892 * config/rs6000/rs6000-c.c: Add the definitions for
893 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
894 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
895 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
896 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
897 definition for insert4b and define insn *insert3b_internal.
898 * doc/extend.texi: Add documentation for vec_extract4b.
900 2018-02-16 Nathan Sidwell <nathan@acm.org>
902 * doc/extend.texi (Backwards Compatibility): Mention friend
903 injection. Note for-scope is deprecated.
904 * doc/invoke.texi (-ffriend-injection): Deprecate.
906 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
908 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
909 that moved to I2, also allow destinations that are a paradoxical
910 subreg (instead of a normal reg).
912 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
915 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
918 2018-02-16 Richard Biener <rguenther@suse.de>
920 PR tree-optimization/84037
921 PR tree-optimization/84016
923 * config/i386/i386.c (ix86_builtin_vectorization_cost):
924 Adjust vec_construct for the fact we need additional higher latency
925 128bit inserts for AVX256 and AVX512 vector builds.
926 (ix86_add_stmt_cost): Scale vector construction cost for
929 2018-02-16 Richard Biener <rguenther@suse.de>
931 PR tree-optimization/84417
932 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
933 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
934 (non_rewritable_lvalue_p): Likewise, use poly-ints.
936 2018-02-16 Martin Liska <mliska@suse.cz>
939 * internal-fn.def (ASAN_CHECK): Set proper flags.
940 (ASAN_MARK): Likewise.
942 2018-02-16 Julia Koval <julia.koval@intel.com>
944 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
947 2018-02-16 Jakub Jelinek <jakub@redhat.com>
950 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
951 Use ++iter rather than iter++ for std::list iterators.
952 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
953 defer deleting them until all nodes in the forest are processed. Do
954 free even leaf nodes. Change to_process into auto_vec.
957 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
958 * vec.h (vec_default_construct): Use memset instead of placement new
959 if BROKEN_VALUE_INITIALIZATION is defined.
960 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
961 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
964 PR rtl-optimization/83723
965 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
966 * lra.c (lra_substitute_pseudo): Likewise. If true, use
967 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
969 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
971 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
973 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
975 PR rtl-optimization/81443
976 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
977 from inner REGs to paradoxical SUBREGs.
979 2018-02-16 Richard Biener <rguenther@suse.de>
981 PR tree-optimization/84399
982 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
983 For operands we can analyze at their definition make sure we can
984 analyze them at each use as well.
986 2018-02-16 Richard Biener <rguenther@suse.de>
988 PR tree-optimization/84190
989 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
990 volatile accesses if the decl isn't volatile.
992 2018-02-15 Jason Merrill <jason@redhat.com>
994 PR c++/84314 - ICE with templates and fastcall attribute.
995 * attribs.c (build_type_attribute_qual_variant): Don't clobber
996 TYPE_CANONICAL on an existing type.
998 2018-02-15 Jakub Jelinek <jakub@redhat.com>
1000 PR tree-optimization/84383
1001 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
1002 dstoff nor call operand_equal_p if dstbase is NULL.
1004 PR tree-optimization/84334
1005 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
1006 also a CONSTANT_CLASS_P, punt.
1008 2018-02-14 Jim Wilson <jimw@sifive.com>
1010 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
1011 first SMALL_OPERAND check. New local min_second_step. Move assert
1012 to where locals are set. Add TARGET_RVC support.
1013 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
1015 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
1017 * doc/invoke.texi: Correct -Wformat-overflow code sample.
1019 2018-02-14 Martin Sebor <msebor@redhat.com>
1021 PR tree-optimization/83698
1022 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
1023 arrays constrain the offset range to their bounds.
1024 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
1025 (builtin_access::overlap): Avoid setting the size of overlap if it's
1027 (maybe_diag_overlap): Also consider arrays when deciding what values
1028 of offsets to include in diagnostics.
1030 2018-02-14 Martin Sebor <msebor@redhat.com>
1033 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
1034 that correspond to the kind of a declaration.
1036 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
1039 * config/pa/pa.md: Load address of PIC label using the linkage table
1040 if the label is nonlocal.
1042 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
1044 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
1045 warning message if user requests -maltivec=be.
1046 * doc/invoke.texi: Document deprecation of -maltivec=be.
1048 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
1051 * config/rs6000/rs6000-c.c: Update definitions for
1052 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
1053 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
1055 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1058 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
1059 add _get_ssp intrinsics. Remove argument from
1060 __builtin_ia32_rdssp[d|q].
1061 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
1062 * config/i386/i386-builtin.def: Remove argument from
1063 __builtin_ia32_rdssp[d|q].
1064 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
1065 ix86_expand_special_args_builtin for _rdssp[d|q].
1066 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
1067 Clear register before usage.
1068 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
1069 Add documentation for new _get_ssp and _inc_ssp intrinsics.
1071 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
1073 PR tree-optimization/84357
1074 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
1075 operand 1 of an ARRAY_REF too.
1077 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
1080 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
1081 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1083 (set_of_reg): New struct.
1084 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
1085 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
1086 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1088 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
1089 Split into bitclr, bitset, bitinvert patterns if appropriate.
1090 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
1091 use rx_fuse_in_memory_bitop.
1092 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
1093 to named insn, correct maximum insn length.
1095 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
1098 * machmode.def: Define a complex mode for PARTIAL_INT.
1099 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
1101 * doc/rtl.texi: Document CSPImode.
1102 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
1104 (msp430_hard_regno_nregs_with_padding): Likewise.
1106 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
1109 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
1111 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
1113 PR rtl-optimization/84169
1114 * combine.c (try_combine): New variable split_i2i3. Set it to true if
1115 we generated a parallel as new i3 and we split that to new i2 and i3
1116 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
1117 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
1118 those to i2, not i1. Partially rewrite this scan code.
1120 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1123 * stor-layout.c (place_field): For variable length fields, adjust
1124 offset_align afterwards not just based on the field's alignment,
1125 but also on the size.
1128 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
1129 of exps and logs in the use_exp2 case.
1131 2018-02-13 Jeff Law <law@redhat.com>
1133 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
1136 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
1139 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
1143 * common.opt (gas-loc-support, gas-locview-support): New.
1144 (ginline-points, ginternal-reset-location-views): New.
1145 * doc/invoke.texi: Document them. Use @itemx where intended.
1146 (gvariable-location-views): Adjust.
1147 * target.def (reset_location_view): New.
1148 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
1149 (TARGET_RESET_LOCATION_VIEW): New.
1150 * doc/tm.texi: Rebuilt.
1151 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
1152 (dwarf2out_default_as_locview_support): New.
1153 (output_asm_line_debug_info): Use option variables.
1154 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
1155 (output_loc_list): Likewise.
1156 (add_high_low_attributes): Check option variables.
1157 Don't output entry view attribute in strict mode.
1158 (gen_inlined_subroutine_die): Check option variables.
1159 (dwarf2out_inline_entry): Likewise.
1160 (init_sections_and_labels): Likewise.
1161 (dwarf2out_early_finish): Likewise.
1162 (maybe_reset_location_view): New, from...
1163 (dwarf2out_var_location): ... here. Call it.
1164 * debug.h (dwarf2out_default_as_loc_support): Declare.
1165 (dwarf2out_default_as_locview_support): Declare.
1166 * hooks.c (hook_int_rtx_insn_0): New.
1167 * hooks.h (hook_int_rtx_insn_0): Declare.
1168 * toplev.c (process_options): Take -gas-loc-support and
1169 -gas-locview-support from dwarf2out. Enable
1170 -gvariable-location-views by default only with locview
1171 assembler support. Enable -ginternal-reset-location-views by
1172 default only if the target defines the corresponding hook.
1173 Enable -ginline-points by default if location views are
1174 enabled; force it disabled if statement frontiers are
1176 * tree-inline.c (expand_call_inline): Check option variables.
1177 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1179 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
1181 PR tree-optimization/84321
1182 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
1183 handling. Also check whether the anti-range contains any values
1184 that satisfy the mask; switch to a VR_RANGE if not.
1186 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
1189 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
1191 2018-02-13 Martin Jambor <mjambor@suse.cz>
1194 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
1195 of call statements, also set location of a load to a temporary.
1197 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
1199 * config/rl78/rl78.c (add_vector_labels): New function.
1200 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
1201 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
1202 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
1203 which checks that no arguments are passed.
1204 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
1205 * doc/extend.texi: Documentation for the new attribute.
1207 2018-02-13 Andreas Schwab <schwab@suse.de>
1209 * config/riscv/linux.h (CPP_SPEC): Define.
1211 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1214 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
1215 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
1216 OPTION_MASK_ISA_AES as first argument to def_builtin_const
1217 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
1218 instead of OPTION_MASK_ISA_PCLMUL as first argument to
1219 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
1220 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
1221 temporarily for AES and PCLMUL builtins.
1223 PR tree-optimization/84339
1224 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
1225 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
1229 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
1230 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
1231 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
1233 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
1235 * omp-simd-clone.h: New file.
1236 * omp-simd-clone.c: Include omp-simd-clone.h.
1237 (expand_simd_clones): No longer static.
1238 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
1239 cgraph.h and omp-simd-clone.h.
1240 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
1241 (vect_recog_widen_shift_pattern): Formatting fix.
1242 (vect_pattern_recog_1): Don't check optab for calls.
1245 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
1246 operands[2] into a REG before using gen_lowpart on it.
1248 2018-02-12 Jeff Law <law@redhat.com>
1251 * config/sh/sh.c (find_barrier): Consider a sibling call
1254 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
1255 successfully back substituting a reg.
1257 2018-02-12 Richard Biener <rguenther@suse.de>
1259 PR tree-optimization/84037
1260 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
1261 parameter, move visited init to caller.
1262 (vect_slp_analyze_operations): Separate cost from validity
1263 check, initialize visited once for all instances.
1264 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
1266 * tree-vect-stmts.c (vect_model_simple_cost): Make early
1268 (vect_model_promotion_demotion_cost): Likewise.
1269 (vectorizable_bswap): Guard cost modeling with !slp_node
1270 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
1272 (vectorizable_call): Likewise.
1273 (vectorizable_conversion): Likewise.
1274 (vectorizable_assignment): Likewise.
1275 (vectorizable_shift): Likewise.
1276 (vectorizable_operation): Likewise.
1277 (vectorizable_store): Likewise.
1278 (vectorizable_load): Likewise.
1279 (vectorizable_condition): Likewise.
1280 (vectorizable_comparison): Likewise.
1282 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
1285 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
1286 (ASAN_MARK): Fix fnspec to account for return value, change pointer
1287 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
1289 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1292 * params.def (inline-min-speedup): Increase from 8 to 15.
1293 (max-inline-insns-auto): Decrease from 40 to 30.
1294 * ipa-split.c (consider_split): Add some buffer for function to
1295 be considered inlining candidate.
1296 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
1299 2018-02-12 Richard Biener <rguenther@suse.de>
1301 PR tree-optimization/84037
1302 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
1303 matched stmts if we cannot swap the non-matched ones.
1305 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1307 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
1308 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
1309 _mm_maskz_scalef_round_ss): New intrinsics.
1310 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
1311 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
1312 __builtin_ia32_scalefss_round): Remove.
1313 (__builtin_ia32_scalefsd_mask_round,
1314 __builtin_ia32_scalefss_mask_round): New intrinsics.
1315 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
1316 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
1317 ((match_operand:VF_128 2 "<round_nimm_predicate>"
1318 "<round_constraint>")): Changed to ...
1319 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
1320 "<round_scalar_constraint>")): ... this.
1321 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
1322 %0, %1, %2<round_op3>}"): Changed to ...
1323 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
1324 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
1325 %2<round_scalar_mask_op3>}"): ... this.
1326 * config/i386/subst.md (round_scalar_nimm_predicate): New.
1328 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1330 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
1331 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
1332 (_mm_maskz_sqrt_round_ss): New intrinsics.
1333 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
1334 (__builtin_ia32_sqrtsd_mask_round)
1335 (__builtin_ia32_sqrtss_mask_round): New builtins.
1336 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
1337 (__builtin_ia32_sqrtss_round): Remove.
1338 (__builtin_ia32_sqrtsd_mask_round)
1339 (__builtin_ia32_sqrtss_mask_round): New builtins.
1340 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
1341 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
1342 ((match_operand:VF_128 1 "vector_operand"
1343 "xBm,<round_constraint>")): Changed to ...
1344 ((match_operand:VF_128 1 "vector_operand"
1345 "xBm,<round_scalar_constraint>")): ... this.
1346 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
1347 %0, %2, %<iptr>1<round_op3>}): Changed to ...
1348 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
1349 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
1350 %<iptr>1<round_scalar_mask_op3>}): ... this.
1351 ((set_attr "prefix" "<round_prefix>")): Changed to ...
1352 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
1354 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
1357 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
1358 Cast vec_cmpeq result to correct type.
1359 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
1360 Cast vec_cmpgt result to correct type.
1362 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
1364 * final.c (final_scan_insn_1): Renamed from...
1365 (final_scan_insn): ... this. New wrapper, to recover
1366 seen from the outermost call in recursive ones.
1367 * config/sparc/sparc.c (output_return): Drop seen from call.
1368 (output_sibcall): Likewise.
1369 * config/visium/visium.c (output_branch): Likewise.
1371 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
1373 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
1376 2018-02-10 Alan Modra <amodra@gmail.com>
1379 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
1380 Specify LR as an input.
1382 2018-02-10 Jakub Jelinek <jakub@redhat.com>
1385 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
1386 remove_member_access_dummy_vars): New functions.
1387 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
1388 lower_omp_1, execute_lower_omp): Use them.
1390 PR rtl-optimization/84308
1391 * shrink-wrap.c (spread_components): Release todo vector.
1393 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
1395 PR rtl-optimization/57193
1396 * ira-color.c (struct allocno_color_data): Add member
1397 conflict_allocno_hard_prefs.
1398 (update_conflict_allocno_hard_prefs): New.
1399 (bucket_allocno_compare_func): Add a preference based on
1400 conflict_allocno_hard_prefs.
1401 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
1402 (color_allocnos): Remove a dead code. Initiate
1403 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
1405 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1408 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
1409 constraint from =wa to wa. Avoid a subreg on the output operand,
1410 instead use a pseudo and subreg it in a move.
1411 (p9_xxbrd_<mode>): Changed to ...
1412 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
1413 (p9_xxbrd_v2df): New expander.
1414 (p9_xxbrw_<mode>): Changed to ...
1415 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
1416 (p9_xxbrw_v4sf): New expander.
1418 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1420 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
1422 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
1425 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
1426 multiply in 32-bit mode.
1427 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
1428 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
1431 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1433 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
1434 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
1435 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
1436 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
1438 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
1441 * dwarf2out.c (is_trivial_indirect_ref): New function.
1442 (dwarf2out_late_global_decl): Do not generate a location
1443 attribute for variables that have a non-trivial DECL_VALUE_EXPR
1444 and that are not defined in the current unit.
1446 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1448 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
1449 instead of a libcall for UNORDERED.
1451 2018-02-09 Tamar Christina <tamar.christina@arm.com>
1454 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
1455 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
1457 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1460 * config/s390/s390.c (s390_set_current_function): Invoke
1461 s390_indirect_branch_settings also if fndecl didn't change.
1463 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1465 * config/rs6000/rs6000.md (blockage): Set length to zero.
1467 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1469 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
1471 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1474 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
1475 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
1479 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
1480 PARALLEL incoming that failed vt_get_decl_and_offset check.
1483 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
1484 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
1486 (get_variable_section): For decls in named .bss* sections pass true as
1487 second argument to bss_initializer_p.
1489 2018-02-09 Marek Polacek <polacek@redhat.com>
1490 Jakub Jelinek <jakub@redhat.com>
1493 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
1494 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
1495 Sync some changes from cxx_fold_indirect_ref.
1497 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1499 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
1501 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
1502 (BLOCK_INLINE_ENTRY_LABEL): New.
1503 (dwarf2out_var_location): Disregard inline entry markers.
1504 (inline_entry_data): New struct.
1505 (inline_entry_data_hasher): New hashtable type.
1506 (inline_entry_data_hasher::hash): New.
1507 (inline_entry_data_hasher::equal): New.
1508 (inline_entry_data_table): New variable.
1509 (add_high_low_attributes): Add DW_AT_entry_pc and
1510 DW_AT_GNU_entry_view attributes if a pending entry is found
1511 in inline_entry_data_table. Add old entry_pc attribute only
1512 if debug nonbinding markers are disabled.
1513 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
1514 markers are enabled.
1515 (block_within_block_p, dwarf2out_inline_entry): New.
1516 (dwarf2out_finish): Check that no entries remained in
1517 inline_entry_data_table.
1518 * final.c (reemit_insn_block_notes): Handle inline entry notes.
1519 (final_scan_insn, notice_source_line): Likewise.
1520 (rest_of_clean_state): Skip inline entry markers.
1521 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
1523 * gimple.c (gimple_build_debug_inline_entry): New.
1524 * gimple.h (enum gimple_debug_subcode): Add
1525 GIMPLE_DEBUG_INLINE_ENTRY.
1526 (gimple_build_debug_inline_entry): Declare.
1527 (gimple_debug_inline_entry_p): New.
1528 (gimple_debug_nonbind_marker_p): Adjust.
1529 * insn-notes.def (INLINE_ENTRY): New.
1530 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
1531 inline entry marker notes.
1532 (print_insn): Likewise.
1533 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
1534 (INSN_DEBUG_MARKER_KIND): Likewise.
1535 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
1536 * tree-inline.c (expand_call_inline): Build and insert
1537 debug_inline_entry stmt.
1538 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
1539 inline entry blocks early, if nonbind markers are enabled.
1540 (dump_scope_block): Dump fragment info.
1541 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
1542 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
1543 (gimple_build_debug_inline_entry): New.
1544 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
1545 Enable/disable inline entry points too.
1546 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
1547 (DEBUG_INSN): Describe inline entry markers.
1549 * common.opt (gvariable-location-views): New.
1550 (gvariable-location-views=incompat5): New.
1551 * config.in: Rebuilt.
1552 * configure: Rebuilt.
1553 * configure.ac: Test assembler for view support.
1554 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
1555 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
1556 * dwarf2out.c (var_loc_view): New typedef.
1557 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
1558 (dwarf2out_locviews_in_attribute): New.
1559 (dwarf2out_locviews_in_loclist): New.
1560 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
1561 (enum dw_line_info_opcode): Add LI_adv_address.
1562 (struct dw_line_info_table): Add view.
1563 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
1564 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
1565 (zero_view_p): New variable.
1566 (ZERO_VIEW_P): New macro.
1567 (output_asm_line_debug_info): New.
1568 (struct var_loc_node): Add view.
1569 (add_AT_view_list, AT_loc_list): New.
1570 (add_var_loc_to_decl): Add view param. Test it against last.
1571 (new_loc_list): Add view params. Record them.
1572 (AT_loc_list_ptr): Handle loc and view lists.
1573 (view_list_to_loc_list_val_node): New.
1574 (print_dw_val): Handle dw_val_class_view_list.
1575 (size_of_die): Likewise.
1576 (value_format): Likewise.
1577 (loc_list_has_views): New.
1578 (gen_llsym): Set vl_symbol too.
1579 (maybe_gen_llsym, skip_loc_list_entry): New.
1580 (dwarf2out_maybe_output_loclist_view_pair): New.
1581 (output_loc_list): Output view list or entries too.
1582 (output_view_list_offset): New.
1583 (output_die): Handle dw_val_class_view_list.
1584 (output_dwarf_version): New.
1585 (output_compilation_unit_header): Use it.
1586 (output_skeleton_debug_sections): Likewise.
1587 (output_rnglists, output_line_info): Likewise.
1588 (output_pubnames, output_aranges): Update version comments.
1589 (output_one_line_info_table): Output view numbers in asm comments.
1590 (dw_loc_list): Determine current endview, pass it to new_loc_list.
1591 Call maybe_gen_llsym.
1592 (loc_list_from_tree_1): Adjust.
1593 (add_AT_location_description): Create view list attribute if
1594 needed, check it's absent otherwise.
1595 (convert_cfa_to_fb_loc_list): Adjust.
1596 (maybe_emit_file): Call output_asm_line_debug_info for test.
1597 (dwarf2out_var_location): Reset views as needed. Precompute
1598 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
1599 attribute. Set view.
1600 (new_line_info_table): Reset next view.
1601 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
1602 (dwarf2out_source_line): Likewise. Output view resets and labels to
1603 the assembler, or select appropriate line info opcodes.
1604 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
1605 (optimize_string_length): Catch it. Adjust.
1606 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
1607 dw_val_class_view_list, and remove it if no longer needed.
1608 (hash_loc_list): Hash view numbers.
1609 (loc_list_hasher::equal): Compare them.
1610 (optimize_location_lists): Check whether a view list symbol is
1611 needed, and whether the locview attribute is present, and
1612 whether they match. Remove the locview attribute if no longer
1614 (index_location_lists): Call skip_loc_list_entry for test.
1615 (dwarf2out_finish): Call output_asm_line_debug_info for test.
1616 Use output_dwarf_version.
1617 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
1618 (struct dw_val_node): Add val_view_list.
1619 * final.c (SEEN_NEXT_VIEW): New.
1620 (set_next_view_needed): New.
1621 (clear_next_view_needed): New.
1622 (maybe_output_next_view): New.
1623 (final_start_function): Rename to...
1624 (final_start_function_1): ... this. Take pointer to FIRST,
1625 add SEEN parameter. Emit param bindings in the initial view.
1626 (final_start_function): Reintroduce SEEN-less interface.
1627 (final): Rename to...
1628 (final_1): ... this. Take SEEN parameter. Output final pending
1629 next view at the end.
1630 (final): Reintroduce seen-less interface.
1631 (final_scan_insn): Output pending next view before switching
1632 sections or ending a block. Mark the next view as needed when
1633 outputting variable locations. Notify debug backend of section
1634 changes, and of location view changes.
1635 (rest_of_handle_final): Adjust.
1636 * toplev.c (process_options): Autodetect value for debug variable
1637 location views option. Warn on incompat5 without -gdwarf-5.
1638 * doc/invoke.texi (gvariable-location-views): New.
1639 (gvariable-location-views=incompat5): New.
1640 (gno-variable-location-views): New.
1642 2018-02-08 David Malcolm <dmalcolm@redhat.com>
1644 PR tree-optimization/84136
1645 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
1646 that the result of find_edge is non-NULL.
1648 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1651 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
1652 storing integer register in SImode. Fix cost of 256 and 512
1653 byte aligned SSE register store.
1655 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1657 * config/i386/i386.c (ix86_multiplication_cost): Fix
1658 multiplication cost for TARGET_AVX512DQ.
1660 2018-02-08 Marek Polacek <polacek@redhat.com>
1662 PR tree-optimization/84238
1663 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
1666 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1668 PR tree-optimization/84265
1669 * tree-vect-stmts.c (vectorizable_store): Don't treat
1670 VMAT_CONTIGUOUS accesses as grouped.
1671 (vectorizable_load): Likewise.
1673 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1675 PR tree-optimization/81635
1676 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
1677 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
1678 (test_round_for_mask): New functions.
1679 (wide_int_cc_tests): Call test_round_for_mask.
1680 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
1681 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
1682 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
1683 range returned by get_range_info.
1685 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1688 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
1689 * symtab.c: Include builtins.h
1690 (symtab_node::output_to_lto_symbol_table_p): Move here
1691 from lto-streamer-out.c:output_symbol_p.
1692 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
1693 (output_symbol_p): Move all logic to symtab.c
1694 (produce_symtab): Update.
1696 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1698 * config/s390/s390-opts.h (enum indirect_branch): Define.
1699 * config/s390/s390-protos.h (s390_return_addr_from_memory)
1700 (s390_indirect_branch_via_thunk)
1701 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
1702 (enum s390_indirect_branch_type): Define.
1703 * config/s390/s390.c (struct s390_frame_layout, struct
1704 machine_function): Remove.
1705 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
1706 (indirect_branch_table_label_no, indirect_branch_table_name):
1708 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
1709 (enum s390_indirect_branch_option): Define.
1710 (s390_return_addr_from_memory): New function.
1711 (s390_handle_string_attribute): New function.
1712 (s390_attribute_table): Add new attribute handler.
1713 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
1714 (s390_indirect_branch_via_thunk): New function.
1715 (s390_indirect_branch_via_inline_thunk): New function.
1716 (s390_function_ok_for_sibcall): When jumping via thunk disallow
1717 sibling call optimization for non z10 compiles.
1718 (s390_emit_call): Force indirect branch target to be a single
1719 register. Add r1 clobber for non-z10 compiles.
1720 (s390_emit_epilogue): Emit return jump via return_use expander.
1721 (s390_reorg): Handle JUMP_INSNs as execute targets.
1722 (s390_option_override_internal): Perform validity checks for the
1723 new command line options.
1724 (s390_indirect_branch_attrvalue): New function.
1725 (s390_indirect_branch_settings): New function.
1726 (s390_set_current_function): Invoke s390_indirect_branch_settings.
1727 (s390_output_indirect_thunk_function): New function.
1728 (s390_code_end): Implement target hook.
1729 (s390_case_values_threshold): Implement target hook.
1730 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
1732 * config/s390/s390.h (struct s390_frame_layout)
1733 (struct machine_function): Move here from s390.c.
1734 (TARGET_INDIRECT_BRANCH_NOBP_RET)
1735 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
1736 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
1737 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
1738 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
1739 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
1740 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
1741 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
1742 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
1743 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
1744 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
1745 (mnemonic attribute): Add values which aren't recognized
1747 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
1748 pattern for branch conversion. Fix mnemonic attribute.
1749 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
1750 indirect branch via thunk if requested.
1751 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
1752 ("*indirect_jump"): Disable for branch conversion using out of
1754 ("indirect_jump_via_thunk<mode>_z10")
1755 ("indirect_jump_via_thunk<mode>")
1756 ("indirect_jump_via_inlinethunk<mode>_z10")
1757 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
1758 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
1759 ("casesi_jump_via_inlinethunk<mode>_z10")
1760 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
1761 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
1762 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
1763 ("*indirect2_jump"): Disable for branch conversion.
1764 ("casesi_jump"): Turn into expander and expand patterns for branch
1766 ("return_use"): New expander.
1767 ("*return"): Emit return via thunk and rename it to ...
1768 ("*return<mode>"): ... this one.
1769 * config/s390/s390.opt: Add new options and and enum for the
1772 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1774 * lra-constraints.c (match_reload): Unconditionally use
1775 gen_lowpart_SUBREG, rather than selecting between that
1776 and equivalent gen_rtx_SUBREG code.
1778 2018-02-08 Richard Biener <rguenther@suse.de>
1780 PR tree-optimization/84233
1781 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
1782 changed flag instead of boguously re-using phi_inserted.
1784 2018-02-08 Martin Jambor <mjambor@suse.cz>
1786 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
1787 static local variables.
1789 2018-02-08 Richard Biener <rguenther@suse.de>
1791 PR tree-optimization/84278
1792 * tree-vect-stmts.c (vectorizable_store): When looking for
1793 smaller vector types to perform grouped strided loads/stores
1794 make sure the mode is supported by the target.
1795 (vectorizable_load): Likewise.
1797 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1799 * config/aarch64/aarch64.c (aarch64_components_for_bb):
1800 Increase LDP/STP opportunities by adding adjacent callee-saves.
1802 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1804 PR rtl-optimization/84068
1805 PR rtl-optimization/83459
1806 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
1808 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
1810 PR tree-optimization/84224
1811 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
1812 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
1815 2018-02-07 Iain Sandoe <iain@codesourcery.com>
1818 * config/rs6000/altivec.md (*restore_world): Remove LR use.
1819 * config/rs6000/predicates.md (restore_world_operation): Adjust op
1820 count, remove one USE.
1822 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
1824 * doc/install.texi (Configuration): Document the
1825 --with-long-double-format={ibm,ieee} PowerPC configuration
1829 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
1830 Convert from define_expand to be define_insn_and_split. Rework
1831 float/double/_Float128 conversions to QI/HI/SImode to work with
1832 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
1833 conversions to QI/HImode types did a store and then a load to
1834 truncate the value. For conversions to VSX registers, don't split
1835 the insn, instead emit the code directly. Use the code iterator
1836 any_fix to combine signed and unsigned conversions.
1837 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
1838 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
1839 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
1840 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
1841 (fix_<mode>di2_hw): Likewise.
1842 (fixuns_<mode>di2_hw): Likewise.
1843 (fix_<mode>si2_hw): Likewise.
1844 (fixuns_<mode>si2_hw): Likewise.
1845 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
1846 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
1847 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
1848 fix<uns>_trunc<SFDF:mode>si2_p8.
1849 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
1851 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
1852 (fix<uns>_<mode>_mem): Likewise.
1853 (fctiw<u>z_<mode>_mem): Likewise.
1854 (fix<uns>_<mode>_mem): Likewise.
1855 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
1856 the register allocator from doing a direct move to the GPRs to do
1857 a store, and instead use the ISA 3.0 store byte/half-word from
1858 vector register instruction. For IEEE 128-bit floating point,
1859 also optimize stores of 32-bit ints.
1860 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
1862 2018-02-07 Alan Hayward <alan.hayward@arm.com>
1864 * genextract.c (push_pathstr_operand): New function to support
1866 (walk_rtx): Call push_pathstr_operand.
1867 (print_path): Support [a-zA-Z].
1869 2018-02-07 Richard Biener <rguenther@suse.de>
1871 PR tree-optimization/84037
1872 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
1873 (cse_and_gimplify_to_preheader): Declare.
1874 (vect_get_place_in_interleaving_chain): Likewise.
1875 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
1877 (_loop_vec_info::~_loop_vec_info): Delete it.
1878 (cse_and_gimplify_to_preheader): New function.
1879 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
1880 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
1881 (vectorizable_load): Likewise. For grouped stores always base
1882 the IV on the first element.
1883 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
1884 condition before gimplifying.
1886 2018-02-07 Jakub Jelinek <jakub@redhat.com>
1888 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
1889 *DIV_EXPR and *MOD_EXPR.
1891 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
1894 * config/i386/i386.c (ix86_option_override_internal): Mask out
1895 the CF_SET bit when checking -fcf-protection.
1897 2018-02-07 Tom de Vries <tom@codesourcery.com>
1900 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
1903 2018-02-07 Richard Biener <rguenther@suse.de>
1905 PR tree-optimization/84204
1906 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
1909 PR tree-optimization/84205
1910 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
1911 special-case isl_ast_op_zdiv_r.
1913 PR tree-optimization/84223
1914 * graphite-scop-detection.c (gather_bbs::before_dom_children):
1915 Only add conditions from within the region.
1916 (gather_bbs::after_dom_children): Adjust.
1918 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
1921 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
1922 * config/avr/avr.md: Only post-reload split REG-REG moves if
1923 either register is REGERAL_REG_P.
1925 2018-02-07 Jakub Jelinek <jakub@redhat.com>
1927 PR tree-optimization/84235
1928 * tree-ssa-scopedtables.c
1929 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
1930 if the subtraction is performed in floating point type where NaNs are
1931 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
1932 build 1. Formatting fix.
1934 2018-02-06 Jakub Jelinek <jakub@redhat.com>
1937 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
1938 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
1939 and skip it regardless of bb boundaries. Use CALL_P macro,
1940 don't test INSN_P (insn) together with CALL_P or JUMP_P check
1941 unnecessarily, formatting fix.
1943 2018-02-06 Michael Collison <michael.collison@arm.com>
1945 * config/arm/thumb2.md:
1946 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
1947 (*thumb_mov_notscc): Ditto.
1949 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
1952 * config/rs6000/rs6000.md (su code attribute): Use "u" for
1953 unsigned_fix, not "s".
1955 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1957 * configure.ac (gcc_fn_eh_frame_ro): New function.
1958 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
1959 correct .eh_frame permissions.
1960 * configure: Regenerate.
1962 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
1964 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
1967 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1969 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1970 Display warning message for -mno-speculate-indirect-jumps.
1972 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
1974 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
1976 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
1978 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
1980 PR tree-optimization/84225
1981 * tree-eh.c (find_trapping_overflow): Only call
1982 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
1984 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1987 * config/i386/i386.c: Reimplement the check of possible options
1988 -mibt/-mshstk conbination. Change error messages.
1989 * doc/invoke.texi: Fix a typo: remove extra '='.
1991 2018-02-06 Marek Polacek <polacek@redhat.com>
1993 PR tree-optimization/84228
1994 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
1996 2018-02-06 Tamar Christina <tamar.christina@arm.com>
1999 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
2000 emitted arch directives.
2001 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
2002 __ARM_FEATURE_COPROC before changing architectures.
2004 2018-02-06 Richard Biener <rguenther@suse.de>
2006 * config/i386/i386.c (print_reg): Fix typo.
2007 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
2009 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
2011 * configure: Regenerate.
2013 2018-02-05 Martin Sebor <msebor@redhat.com>
2015 PR tree-optimization/83369
2016 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
2019 2018-02-05 Martin Liska <mliska@suse.cz>
2021 * doc/invoke.texi: Cherry-pick upstream r323995.
2023 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
2025 * ira.c (ira_init_register_move_cost): Adjust comment.
2027 2018-02-05 Martin Liska <mliska@suse.cz>
2029 PR gcov-profile/84137
2030 * doc/gcov.texi: Fix typo in documentation.
2032 2018-02-05 Martin Liska <mliska@suse.cz>
2034 PR gcov-profile/83879
2035 * doc/gcov.texi: Document necessity of --dynamic-list-data when
2036 using dlopen functionality.
2038 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
2040 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
2041 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
2042 _mm_maskz_range_ss, _mm_mask_range_round_ss,
2043 _mm_maskz_range_round_ss): New intrinsics.
2044 (__builtin_ia32_rangesd128_round)
2045 (__builtin_ia32_rangess128_round): Remove.
2046 (__builtin_ia32_rangesd128_mask_round,
2047 __builtin_ia32_rangess128_mask_round): New builtins.
2048 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
2049 __builtin_ia32_rangess128_round): Remove.
2050 (__builtin_ia32_rangesd128_mask_round,
2051 __builtin_ia32_rangess128_mask_round): New builtins.
2052 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
2053 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
2054 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
2055 "<round_saeonly_constraint>")): Changed to ...
2056 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
2057 "<round_saeonly_scalar_constraint>")): ... this.
2058 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
2059 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
2060 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
2061 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
2062 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
2064 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
2066 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
2068 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
2069 Remove all values except native, 8540 and 8548.
2071 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
2073 * config/i386/i386.c (ix86_output_function_return): Pass
2074 INVALID_REGNUM, instead of -1, as invalid register number to
2075 indirect_thunk_name and output_indirect_thunk.
2077 2018-02-02 Julia Koval <julia.koval@intel.com>
2079 * config.gcc: Add -march=icelake.
2080 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
2081 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
2082 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
2083 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
2084 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
2085 (processor_target_table): Add icelake.
2086 (ix86_option_override_internal): Handle new PTAs.
2087 (get_builtin_code_for_version): Handle icelake.
2088 (M_INTEL_COREI7_ICELAKE): New.
2089 (fold_builtin_cpu): Handle icelake.
2090 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
2091 * doc/invoke.texi: Add -march=icelake.
2093 2018-02-02 Julia Koval <julia.koval@intel.com>
2095 * config/i386/i386.c (ix86_option_override_internal): Change flags type
2096 to wide_int_bitmask.
2097 * wide-int-bitmask.h: New.
2099 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2102 * config/i386/i386.md: Replace Pmode with word_mode in
2103 builtin_setjmp_setup and builtin_longjmp to support x32.
2105 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
2109 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
2111 (rs6000_supported_cpu_names): New static variable.
2112 (linux_cpu_translation_table): Likewise.
2113 (elf_platform) <cpu>: Define new static variable and use it.
2114 Translate kernel AT_PLATFORM name to canonical name if needed.
2115 Error if platform name is unknown.
2117 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
2120 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
2122 2018-02-01 Jeff Law <law@redhat.com>
2125 * config/i386/i386.c (release_scratch_register_on_entry): Add new
2126 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
2127 the scratch if RELEASE_VIA_POP is false.
2128 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
2129 If we have to save a temporary register, decrement SIZE appropriately.
2130 Pass new arguments to release_scratch_register_on_entry.
2131 (ix86_adjust_stack_and_probe): Likewise.
2132 (ix86_emit_probe_stack_range): Pass new arguments to
2133 release_scratch_register_on_entry.
2135 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
2137 PR rtl-optimization/84157
2138 * combine.c (change_zero_ext): Use REG_P predicate in
2139 front of HARD_REGISTER_P predicate.
2141 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
2143 * config/avr/avr.c (avr_option_override): Move disabling of
2144 -fdelete-null-pointer-checks to...
2145 * common/config/avr/avr-common.c (avr_option_optimization_table):
2148 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2150 PR tree-optimization/81635
2151 * tree-data-ref.c (split_constant_offset_1): For types that
2152 wrap on overflow, try to use range info to prove that wrapping
2155 2018-02-01 Renlin Li <renlin.li@arm.com>
2158 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
2160 (aarch64_register_move_cost): Likewise.
2161 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
2163 (REG_CLASS_NAMES): Likewise.
2164 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
2165 TAILCALL_ADDR_REGS. Remove IP registers.
2166 * config/aarch64/aarch64.md (Ucs): Update register constraint.
2168 2018-02-01 Richard Biener <rguenther@suse.de>
2170 * domwalk.h (dom_walker::dom_walker): Add additional constructor
2171 for specifying RPO order and allow NULL for that.
2172 * domwalk.c (dom_walker::dom_walker): Likewise.
2173 (dom_walker::walk): Handle NULL RPO order.
2174 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
2176 (rewrite_update_dom_walker): Likewise.
2177 (mark_def_dom_walker): Likewise.
2179 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2181 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
2182 (aarch64_maybe_expand_sve_subreg_move): Declare.
2183 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
2184 * config/aarch64/predicates.md (aarch64_any_register_operand): New
2186 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
2187 that are semantically a reverse operation.
2188 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
2189 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
2190 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
2192 (aarch64_can_change_mode_class): For big-endian, forbid changes
2193 between two SVE modes if they have different element sizes.
2195 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2197 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
2198 the TImode handling for big-endian targets.
2200 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2202 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
2203 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
2205 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
2206 Remove BSWAP handing for big-endian targets and use the form of
2207 LD1RQ appropariate for the mode.
2209 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2211 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
2212 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
2215 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2218 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
2219 check for operands that need to go through aarch64_sve_reload_be.
2221 2018-02-01 Jakub Jelinek <jakub@redhat.com>
2223 PR tree-optimization/81661
2224 PR tree-optimization/84117
2225 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
2226 * tree-eh.c: Include gimplify.h.
2227 (find_trapping_overflow, replace_trapping_overflow,
2228 rewrite_to_non_trapping_overflow): New functions.
2229 * tree-vect-loop.c: Include tree-eh.h.
2230 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
2231 * tree-data-ref.c: Include tree-eh.h.
2232 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
2234 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
2236 PR rtl-optimization/84123
2237 * combine.c (change_zero_ext): Check if hard register satisfies
2238 can_change_dest_mode before calling gen_lowpart_SUBREG.
2240 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
2243 * ira.c (ira_init_register_move_cost): Remove assert.
2245 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2247 PR rtl-optimization/84071
2248 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
2249 * doc/tm.texi: Regenerate.
2251 2018-01-31 Richard Biener <rguenther@suse.de>
2253 PR tree-optimization/84132
2254 * tree-data-ref.c (analyze_miv_subscript): Properly
2255 check whether evolution_function_is_affine_multivariate_p
2256 before calling gcd_of_steps_may_divide_p.
2258 2018-01-31 Julia Koval <julia.koval@intel.com>
2261 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
2262 * config/i386/i386.md (rdpid_rex64) New.
2263 (rdpid): Make 32bit only.
2265 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
2268 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
2269 an IDENTIFIER_NODE for FUNCTION_TYPE's.
2271 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2274 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
2276 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
2278 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2280 PR rtl-optimization/84071
2281 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
2282 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
2284 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2286 * config/arc/arc.c (arc_handle_aux_attribute): New function.
2287 (arc_attribute_table): Add 'aux' attribute.
2288 (arc_in_small_data_p): Consider aux like variables.
2289 (arc_is_aux_reg_p): New function.
2290 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
2291 (arc_get_aux_arg): New function.
2292 (prepare_move_operands): Handle aux-register access.
2293 (arc_handle_aux_attribute): New function.
2294 * doc/extend.texi (ARC Variable attributes): Add subsection.
2296 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2298 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
2299 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
2300 (arc_attribute_table): Add 'uncached' attribute.
2301 (arc_print_operand): Print '.di' flag for uncached memory
2303 (arc_in_small_data_p): Do not consider for small data the uncached
2305 (arc_is_uncached_mem_p): New function.
2306 * config/arc/predicates.md (compact_store_memory_operand): Check
2307 for uncached memory accesses.
2308 (nonvol_nonimm_operand): Likewise.
2309 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
2311 2018-01-31 Jakub Jelinek <jakub@redhat.com>
2314 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
2315 falign-loops=): Add Optimization flag.
2317 2018-01-30 Jeff Law <law@redhat.com>
2320 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
2321 INT_REGISTERS_SAVED. Check it prior to calling
2322 get_scratch_register_on_entry.
2323 (ix86_adjust_stack_and_probe): Similarly.
2324 (ix86_emit_probe_stack_range): Similarly.
2325 (ix86_expand_prologue): Corresponding changes.
2327 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2330 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
2331 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
2333 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
2336 * lra-constraints.c (curr_insn_transform): Process AND in the
2339 2018-01-30 Jakub Jelinek <jakub@redhat.com>
2341 PR rtl-optimization/83986
2342 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
2343 dependence against last_pending_memory_flush in addition to
2346 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
2348 PR tree-optimization/81611
2349 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
2352 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2355 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
2358 2018-01-30 Richard Biener <rguenther@suse.de>
2359 Jakub Jelinek <jakub@redhat.com>
2361 PR tree-optimization/84111
2362 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
2363 inner loops added during recursion, as they don't have up-to-date
2366 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2369 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
2370 (can_inline_edge_by_limits_p): ... here.
2371 (can_early_inline_edge_p, check_callers,
2372 update_caller_keys, update_callee_keys, recursive_inlining,
2373 add_new_edges_to_heap, speculation_useful_p,
2374 inline_small_functions,
2375 inline_small_functions, flatten_function,
2376 inline_to_all_callers_1): Update.
2378 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2380 * profile-count.c (profile_count::combine_with_ipa_count): Handle
2383 2018-01-30 Richard Biener <rguenther@suse.de>
2385 PR tree-optimization/83008
2386 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
2387 invariant and constant vector uses in stmts when they need
2390 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2393 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
2394 * configure: Regenerate.
2396 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2398 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
2400 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
2401 Use gen_rtx_REG rather than gen_lowpart.
2403 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2405 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
2406 rather than 0 when creating partial subregs.
2408 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2410 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
2413 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
2416 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
2417 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
2418 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
2419 flags. This restores the settings used before the 2017-07-24.
2420 Turning off pre increment/decrement/modify allows IVOPTS to
2421 optimize DF/SF loops where the index is an int.
2423 2018-01-29 Richard Biener <rguenther@suse.de>
2424 Kelvin Nilsen <kelvin@gcc.gnu.org>
2427 * tree-vect-stmts.c (vectorizable_call): Don't call
2428 targetm.vectorize_builtin_md_vectorized_function if callee is
2431 2018-01-22 Carl Love <cel@us.ibm.com>
2433 * doc/extend.tex: Fix typo in second arg in
2434 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
2436 2018-01-29 Richard Biener <rguenther@suse.de>
2438 PR tree-optimization/84086
2439 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
2440 (flush_ssaname_freelist): When SSA names were released reset
2441 the SCEV hash table.
2443 2018-01-29 Richard Biener <rguenther@suse.de>
2445 PR tree-optimization/84057
2446 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
2447 removed paths when removing edges.
2449 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
2451 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
2452 -mfunction-return=@var{choice}.
2454 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
2457 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
2459 (layout::print_source_line): Likewise.
2460 (test_get_line_width_without_trailing_whitespace): Add test cases.
2462 2018-01-27 Jakub Jelinek <jakub@redhat.com>
2465 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
2468 2018-01-26 Jim Wilson <jimw@sifive.com>
2470 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
2472 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
2475 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2477 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
2478 and CMP + SUB-immediate -> SUBS.
2480 2018-01-26 Martin Sebor <msebor@redhat.com>
2482 PR tree-optimization/83896
2483 * tree-ssa-strlen.c (get_string_len): Rename...
2484 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
2485 Avoid assuming length is constant.
2486 (handle_char_store): Use HOST_WIDE_INT for string length.
2488 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
2491 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
2492 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
2494 2018-01-26 Richard Biener <rguenther@suse.de>
2496 PR rtl-optimization/84003
2497 * dse.c (record_store): Only record redundant stores when
2498 the earlier store aliases at least all accesses the later one does.
2500 2018-01-26 Jakub Jelinek <jakub@redhat.com>
2502 PR rtl-optimization/83985
2503 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
2504 REG_CFA_RESTORE insns.
2505 (delete_unmarked_insns): Don't ignore separate shrink wrapping
2506 REG_CFA_RESTORE insns here.
2509 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
2510 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
2512 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2514 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
2515 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
2516 (arc_init): Likewise.
2517 (arc_override_options): Likewise.
2518 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
2520 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
2522 * config/arc/arc.h (TARGET_DBNZ): Define.
2523 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
2524 properly set the tune attribute.
2525 (dbnz): Use TARGET_DBNZ guard.
2526 * config/arc/arc.opt (mtune): Add core3 option.
2528 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2530 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
2531 recognize new pic like addresses.
2532 (arc_delegitimize_address): Clean up.
2534 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2536 * config/arc/arc-arches.def: Option mrf16 valid for all
2538 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
2539 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
2540 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
2541 * config/arc/arc-tables.opt: Regenerate.
2542 * config/arc/arc.c (arc_conditional_register_usage): Handle
2543 reduced register file case.
2544 (arc_file_start): Set must have build attributes.
2545 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
2547 * config/arc/arc.opt (mrf16): Add new option.
2548 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
2549 * config/arc/genmultilib.awk: Handle new mrf16 option.
2550 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
2551 * config/arc/t-multilib: Regenerate.
2552 * doc/invoke.texi (ARC Options): Document mrf16 option.
2554 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2556 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
2557 * config/arc/arc.c (arc_handle_secure_attribute): New function.
2558 (arc_attribute_table): Add 'secure_call' attribute.
2559 (arc_print_operand): Print secure call operand.
2560 (arc_function_ok_for_sibcall): Don't optimize tail calls when
2562 (arc_is_secure_call_p): New function. * config/arc/arc.md
2563 (call_i): Add support for sjli instruction.
2564 (call_value_i): Likewise.
2565 * config/arc/constraints.md (Csc): New constraint.
2567 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2568 John Eric Martin <John.Martin@emmicro-us.com>
2570 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
2571 * config/arc/arc.c (_arc_jli_section): New struct.
2572 (arc_jli_section): New type.
2573 (rc_jli_sections): New static variable.
2574 (arc_handle_jli_attribute): New function.
2575 (arc_attribute_table): Add jli_always and jli_fixed attribute.
2576 (arc_file_end): New function.
2577 (TARGET_ASM_FILE_END): Define.
2578 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
2579 (arc_add_jli_section): New function.
2580 (jli_call_scan): Likewise.
2581 (arc_reorg): Call jli_call_scan.
2582 (arc_output_addsi): Remove 'S' from printing asm operand.
2583 (arc_is_jli_call_p): New function.
2584 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
2586 (movhi_insn): Likewise.
2587 (movsi_insn): Likewise.
2588 (movsi_set_cc_insn): Likewise.
2589 (loadqi_update): Likewise.
2590 (load_zeroextendqisi_update): Likewise.
2591 (load_signextendqisi_update): Likewise.
2592 (loadhi_update): Likewise.
2593 (load_zeroextendhisi_update): Likewise.
2594 (load_signextendhisi_update): Likewise.
2595 (loadsi_update): Likewise.
2596 (loadsf_update): Likewise.
2597 (movsicc_insn): Likewise.
2598 (bset_insn): Likewise.
2599 (bxor_insn): Likewise.
2600 (bclr_insn): Likewise.
2601 (bmsk_insn): Likewise.
2602 (bicsi3_insn): Likewise.
2603 (cmpsi_cc_c_insn): Likewise.
2604 (movsi_ne): Likewise.
2605 (movsi_cond_exec): Likewise.
2606 (clrsbsi2): Likewise.
2617 (call_i): Remove 'S' asm letter, add jli instruction.
2618 (call_value_i): Likewise.
2619 * config/arc/arc.op (mjli-always): New option.
2620 * config/arc/constraints.md (Cji): New constraint.
2621 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
2623 (subsf3_fpx): Likewise.
2624 (mulsf3_fpx): Likewise.
2625 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
2627 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
2629 * doc/invoke.texi (ARC): Document mjli-always option.
2631 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
2633 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
2634 avoid addition with 0 and use incw and decw where possible.
2636 2018-01-26 Richard Biener <rguenther@suse.de>
2638 PR tree-optimization/81082
2639 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
2640 association if it requires casting to unsigned.
2641 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
2642 from fold_plusminus_mult_expr to catch important cases late when
2643 range info is available.
2645 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2647 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
2648 * configure.ac (hidden_linkonce): New test.
2649 * configure: Regenerate.
2650 * config.in: Regenerate.
2652 2018-01-26 Julia Koval <julia.koval@intel.com>
2654 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
2655 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
2656 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
2657 _mm_mask_bitshuffle_epi64_mask): Fix type.
2658 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
2659 USI_FTYPE_V4DI_V4DI_USI): Remove.
2660 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
2661 __builtin_ia32_vpshufbitqmb256_mask,
2662 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
2663 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
2664 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
2666 2018-01-26 Alan Modra <amodra@gmail.com>
2669 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
2670 UNSPEC_VBPERMQ. Sort other unspecs.
2672 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
2674 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
2676 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
2679 * predict.c (drop_profile): Do not push/pop cfun; update also
2681 (handle_missing_profiles): Fix logic looking for zero profiles.
2683 2018-01-25 Jakub Jelinek <jakub@redhat.com>
2686 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
2687 on functions with #pragma omp declare simd or functions with simd
2689 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
2690 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
2691 Remove trailing \n from warning_at calls.
2693 2018-01-25 Tom de Vries <tom@codesourcery.com>
2696 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2697 for neutered workers.
2699 2018-01-24 Joseph Myers <joseph@codesourcery.com>
2702 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
2703 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
2705 2018-01-24 Jeff Law <law@redhat.com>
2708 * i386.c (get_probe_interval): Move to earlier point.
2709 (ix86_compute_frame_layout): If -fstack-clash-protection and
2710 the frame is larger than the probe interval, then use pushes
2711 to save registers rather than reg->mem moves.
2712 (ix86_expand_prologue): Remove conditional for int_registers_saved
2715 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
2718 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
2719 min/max for never referenced object.
2721 2018-01-24 Jakub Jelinek <jakub@redhat.com>
2724 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
2726 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
2727 attributes from DECL_ATTRIBUTES (decl) without affecting
2728 DECL_ATTRIBUTES (current_function_decl).
2729 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
2730 functions with non-NULL DECL_ABSTRACT_ORIGIN.
2732 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
2734 PR tree-optimization/83979
2735 * fold-const.c (fold_comparison): Use constant_boolean_node
2736 instead of boolean_{true,false}_node.
2738 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
2740 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
2743 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2745 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
2746 Simplify the clause that sets the length attribute.
2747 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
2748 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
2749 clause that sets the length attribute.
2750 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
2752 2018-01-24 Tom de Vries <tom@codesourcery.com>
2755 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
2756 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
2757 Add strict parameter.
2758 (prevent_branch_around_nothing): Insert dummy insn between branch to
2759 label and label with no ptx insn inbetween.
2760 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
2762 2018-01-24 Tom de Vries <tom@codesourcery.com>
2765 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2766 for neutered threads in warp.
2767 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
2769 2018-01-24 Richard Biener <rguenther@suse.de>
2771 PR tree-optimization/83176
2772 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
2775 2018-01-24 Richard Biener <rguenther@suse.de>
2777 PR tree-optimization/82819
2778 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
2779 code generating pluses that are no-ops in the target precision.
2781 2018-01-24 Richard Biener <rguenther@suse.de>
2784 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
2786 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2788 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
2789 to merge probabilities.
2790 * predict.c (probably_never_executed): Also mark as cold functions
2791 with global 0 profile and guessed local profile.
2792 * profile-count.c (profile_probability::combine_with_count): New
2794 * profile-count.h (profile_probability::operator*,
2795 profile_probability::operator*=, profile_probability::operator/,
2796 profile_probability::operator/=): Reduce precision to adjusted
2797 and set value to guessed on contradictory divisions.
2798 (profile_probability::combine_with_freq): Remove.
2799 (profile_probability::combine_wiht_count): Declare.
2800 (profile_count::force_nonzero):: Set to adjusted.
2801 (profile_count::probability_in):: Set quality to adjusted.
2802 * tree-ssa-tail-merge.c (replace_block_by): Use
2805 2018-01-23 Andrew Waterman <andrew@sifive.com>
2806 Jim Wilson <jimw@sifive.com>
2808 * config/riscv/riscv.c (riscv_stack_boundary): New.
2809 (riscv_option_override): Set riscv_stack_boundary. Handle
2810 riscv_preferred_stack_boundary_arg.
2811 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
2812 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
2813 (STACK_BOUNDARY): Set to riscv_stack_boundary.
2814 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
2815 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
2816 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
2818 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
2821 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
2822 of struct ix86_frame.
2823 (ix86_expand_epilogue): Likewise. Add a local variable for
2824 the reg_save_offset field in struct ix86_frame.
2826 2018-01-23 Bin Cheng <bin.cheng@arm.com>
2828 PR tree-optimization/82604
2829 * tree-loop-distribution.c (enum partition_kind): New enum item
2830 PKIND_PARTIAL_MEMSET.
2831 (partition_builtin_p): Support above new enum item.
2832 (generate_code_for_partition): Ditto.
2833 (compute_access_range): Differentiate cases that equality can be
2834 proven at all loops, the innermost loops or no loops.
2835 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
2836 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
2837 (finalize_partitions, distribute_loop): Don't fuse partition of
2838 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
2839 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
2842 2018-01-23 Martin Liska <mliska@suse.cz>
2844 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
2845 order to ignore the predictor.
2846 (PRED_POLYMORPHIC_CALL): Likewise.
2847 (PRED_RECURSIVE_CALL): Likewise.
2849 2018-01-23 Martin Liska <mliska@suse.cz>
2851 * tree-profile.c (tree_profiling): Print function header to
2852 aware reader which function we are working on.
2853 * value-prof.c (gimple_find_values_to_profile): Do not print
2854 not interesting value histograms.
2856 2018-01-23 Martin Liska <mliska@suse.cz>
2858 * profile-count.h (enum profile_quality): Add
2859 profile_uninitialized as the first value. Do not number values
2860 as they are zero based.
2861 (profile_count::verify): Update sanity check.
2862 (profile_probability::verify): Likewise.
2864 2018-01-23 Nathan Sidwell <nathan@acm.org>
2866 * doc/invoke.texi (ffor-scope): Deprecate.
2868 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2870 PR tree-optimization/83510
2871 * domwalk.c (set_all_edges_as_executable): New function.
2872 (dom_walker::dom_walker): Convert bool param
2873 "skip_unreachable_blocks" to enum reachability. Move setup of
2874 edge flags to set_all_edges_as_executable and only do it when
2875 reachability is REACHABLE_BLOCKS.
2876 * domwalk.h (enum dom_walker::reachability): New enum.
2877 (dom_walker::dom_walker): Convert bool param
2878 "skip_unreachable_blocks" to enum reachability.
2879 (set_all_edges_as_executable): New decl.
2880 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
2881 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
2883 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
2884 but converting true to REACHABLE_BLOCKS.
2885 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
2887 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
2888 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
2889 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
2891 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
2892 if check_all_array_refs will be called.
2894 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2896 * tree.c (selftest::test_location_wrappers): Add more test
2899 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2901 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
2902 (selftest::test_bit_in_range): Likewise.
2904 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
2907 * doc/sourcebuild.texi (vect_float): Say that the selector
2908 only describes the situation when -funsafe-math-optimizations is on.
2909 (vect_float_strict): Document.
2911 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
2913 PR tree-optimization/83965
2914 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
2915 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
2916 instead of checking only for a reduction.
2917 (vect_recog_widen_sum_pattern): Likewise.
2919 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2921 * predict.c (probably_never_executed): Only use precise profile info.
2922 (compute_function_frequency): Skip after inlining hack since we now
2923 have quality checking.
2925 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2927 * profile-count.h (profile_probability::very_unlikely,
2928 profile_probability::unlikely, profile_probability::even): Set
2929 precision to guessed.
2931 2018-01-23 Richard Biener <rguenther@suse.de>
2933 PR tree-optimization/83963
2934 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
2935 Properly terminate dominator walk when crossing the exit edge not
2936 when visiting its source block.
2938 2018-01-23 Jakub Jelinek <jakub@redhat.com>
2941 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
2942 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
2944 2018-01-22 Jakub Jelinek <jakub@redhat.com>
2946 PR tree-optimization/83957
2947 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
2948 semicolon after for body surrounded by braces.
2950 PR tree-optimization/83081
2951 * profile-count.h (profile_probability::split): New method.
2952 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
2953 Use profile_probability::split.
2954 (do_compare_rtx_and_jump): Fix adjustment of probabilities
2955 when splitting a single conditional jump into 2.
2957 2018-01-22 David Malcolm <dmalcolm@redhat.com>
2959 PR tree-optimization/69452
2960 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
2963 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2965 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
2966 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
2967 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
2969 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2971 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
2972 * config/rl78/rl78.md (movdi): New define_expand.
2973 * config/rl78/rl78.c (rl78_split_movdi): New function.
2975 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
2978 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
2980 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
2981 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
2982 128-bit to produce an UNSPEC move to get the double word with the
2983 signbit and then a shift directly to do signbit.
2984 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
2985 implementation with a new version that just does either a direct
2986 move or a regular move. Move memory interface to separate insns.
2987 Move insns so they are next to the expander.
2988 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
2989 with signbit move. Split big and little endian case.
2990 (signbit<mode>2_dm_mem_le): Likewise.
2991 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
2992 (signbit<mode>2_dm2): Likewise.
2994 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2996 * config/rl78/rl78.md (anddi3): New define_expand.
2998 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3000 * config/rl78/rl78.md (umindi3): New define_expand.
3002 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3004 * config/rl78/rl78.md (smindi3): New define_expand.
3006 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3008 * config/rl78/rl78.md (smaxdi3): New define_expand.
3010 2018-01-22 Carl Love <cel@us.ibm.com>
3012 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
3013 LVX_V1TI): Add macro expansion.
3014 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
3015 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
3016 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
3017 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
3018 Change check to determine if the instruction is a byte reversing
3019 entry. Fix typo in comment.
3020 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
3021 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
3022 Add def_builtin calls for new builtins.
3023 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
3024 Add define_insn expansion.
3026 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3028 * config/rl78/rl78.md (umaxdi3): New define_expand.
3030 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3032 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
3033 for non-QImode registers.
3035 2018-01-22 Richard Biener <rguenther@suse.de>
3037 PR tree-optimization/83963
3038 * graphite-scop-detection.c (scop_detection::get_sese): Delay
3039 including the loop exit block.
3040 (scop_detection::merge_sese): Likewise.
3041 (scop_detection::add_scop): Do it here instead.
3043 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3045 * doc/sourcebuild.texi (arm_softfloat): Document.
3047 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
3050 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
3051 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
3052 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
3054 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3055 David Edelsohn <dje.gcc@gmail.com>
3058 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3059 Change "crset eq" to "crset 2".
3060 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3061 (*call_indirect_aix<mode>_nospec): Likewise.
3062 (*call_value_indirect_aix<mode>_nospec): Likewise.
3063 (*call_indirect_elfv2<mode>_nospec): Likewise.
3064 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
3065 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
3066 change assembly output from . to $.
3067 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3068 (indirect_jump<mode>_nospec): Change assembly output from . to $.
3069 (*tablejump<mode>_internal1_nospec): Likewise.
3071 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
3074 * config/sh/sh_optimize_sett_clrt.cc:
3075 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
3077 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
3079 PR tree-optimization/83940
3080 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
3081 offset_dt to vect_constant_def rather than vect_unknown_def_type.
3082 (vect_check_load_store_mask): Add a mask_dt_out parameter and
3083 use it to pass back the definition type.
3084 (vect_check_store_rhs): Likewise rhs_dt_out.
3085 (vect_build_gather_load_calls): Add a mask_dt argument and use
3086 it instead of a call to vect_is_simple_use.
3087 (vectorizable_store): Update calls to vect_check_load_store_mask
3088 and vect_check_store_rhs. Use the dt returned by the latter instead
3089 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
3090 instead of calls to vect_is_simple_use. Pass the scalar rather
3091 than the vector operand to vect_is_simple_use when handling
3092 second and subsequent copies of an rhs value.
3093 (vectorizable_load): Update calls to vect_check_load_store_mask
3094 and vect_build_gather_load_calls. Use the cached mask_dt and
3095 gs_info.offset_dt instead of calls to vect_is_simple_use.
3097 2018-01-20 Jakub Jelinek <jakub@redhat.com>
3100 * tree-emutls.c: Include gimplify.h.
3101 (lower_emutls_2): New function.
3102 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
3103 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
3104 it before further processing.
3107 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
3108 UINTVAL (trueop1) instead of INTVAL (op1).
3110 2018-01-19 Jakub Jelinek <jakub@redhat.com>
3114 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
3115 INCOMING_FRAME_SP_OFFSET if not defined.
3116 (scan_trace): Add ENTRY argument. If true and
3117 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
3118 emit a note to adjust the CFA offset.
3119 (create_cfi_notes): Adjust scan_trace callers.
3120 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
3121 INCOMING_FRAME_SP_OFFSET in the CIE.
3122 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
3123 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
3125 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
3126 * doc/tm.texi: Regenerated.
3128 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3130 PR rtl-optimization/83147
3131 * lra-constraints.c (remove_inheritance_pseudos): Use
3132 lra_substitute_pseudo_within_insn.
3134 2018-01-19 Tom de Vries <tom@codesourcery.com>
3135 Cesar Philippidis <cesar@codesourcery.com>
3138 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
3140 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
3143 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
3144 spaces for function labels.
3146 2018-01-19 Martin Liska <mliska@suse.cz>
3148 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
3149 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
3150 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
3151 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
3152 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
3153 (PRED_CONST_RETURN): Change from 69 to 65.
3154 (PRED_NULL_RETURN): Change from 91 to 71.
3155 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
3156 (PRED_LOOP_GUARD): Change from 66 to 73.
3158 2018-01-19 Martin Liska <mliska@suse.cz>
3160 * predict.c (predict_insn_def): Add new assert.
3161 (struct branch_predictor): Change type to signed integer.
3162 (test_prediction_value_range): Amend test to cover
3164 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
3165 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
3166 (PRED_LOOP_ITERATIONS_MAX): Likewise.
3167 (PRED_LOOP_IV_COMPARE): Likewise.
3168 * predict.h (PROB_UNINITIALIZED): Define new constant.
3170 2018-01-19 Martin Liska <mliska@suse.cz>
3172 * predict.c (dump_prediction): Add new format for
3173 analyze_brprob.py script which is enabled with -details
3175 * profile-count.h (precise_p): New function.
3177 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3179 PR tree-optimization/83922
3180 * tree-vect-loop.c (vect_verify_full_masking): Return false if
3181 there are no statements that need masking.
3182 (vect_active_double_reduction_p): New function.
3183 (vect_analyze_loop_operations): Use it when handling phis that
3184 are not in the loop header.
3186 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3188 PR tree-optimization/83914
3189 * tree-vect-loop.c (vectorizable_induction): Don't convert
3190 init_expr or apply the peeling adjustment for inductions
3191 that are nested within the vectorized loop.
3193 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3195 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
3198 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3202 * function.h (gimplify_parameters): Add gimple_seq * argument.
3203 * function.c: Include gimple.h and options.h.
3204 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
3205 for the added local temporaries if needed.
3206 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
3207 if there are any parameter cleanups, wrap whole body into a
3208 try/finally with the cleanups.
3210 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
3213 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
3214 Use GET_MODE_CLASS for scalar floating point.
3216 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3220 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3221 Fix call of call_cgraph_insertion_hooks.
3223 2018-01-18 Martin Sebor <msebor@redhat.com>
3225 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
3227 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3230 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
3233 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
3236 * common.opt: (-ffile-prefix-map): New option.
3237 * opts.c (common_handle_option): Defer it.
3238 * opts-global.c (handle_common_deferred_options): Handle it.
3239 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
3240 * file-prefix-map.h: New file.
3241 (remap_debug_filename, add_debug_prefix_map): ...here.
3242 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
3243 * final.c (debug_prefix_map, add_debug_prefix_map
3244 remap_debug_filename): Move to...
3245 * file-prefix-map.c: New file.
3246 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
3247 generalize, get rid of alloca(), use strrchr() instead of strchr().
3248 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
3249 Implement in terms of add_prefix_map().
3250 (remap_macro_filename, remap_debug_filename): Implement in term of
3252 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
3253 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
3254 * dbxout.c: Include file-prefix-map.h.
3255 * varasm.c: Likewise.
3256 * vmsdbgout.c: Likewise.
3257 * xcoffout.c: Likewise.
3258 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
3259 * doc/cppopts.texi (-fmacro-prefix-map): Document.
3260 * doc/invoke.texi (-ffile-prefix-map): Document.
3261 (-fdebug-prefix-map): Update description.
3263 2018-01-18 Martin Liska <mliska@suse.cz>
3265 * config/i386/i386.c (indirect_thunk_name): Document that also
3267 (output_indirect_thunk): Document why both instructions
3268 (pause and lfence) are generated.
3270 2018-01-18 Richard Biener <rguenther@suse.de>
3272 PR tree-optimization/83887
3273 * graphite-scop-detection.c
3274 (scop_detection::get_nearest_dom_with_single_entry): Remove.
3275 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
3276 (scop_detection::merge_sese): Re-implement with a flood-fill
3277 algorithm that properly finds a SESE region if it exists.
3279 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3282 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
3283 pointer_diff optimizations use view_convert instead of convert.
3285 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3287 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3288 Generate different code for -mno-speculate-indirect-jumps.
3289 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3290 (*call_indirect_aix<mode>): Disable for
3291 -mno-speculate-indirect-jumps.
3292 (*call_indirect_aix<mode>_nospec): New define_insn.
3293 (*call_value_indirect_aix<mode>): Disable for
3294 -mno-speculate-indirect-jumps.
3295 (*call_value_indirect_aix<mode>_nospec): New define_insn.
3296 (*sibcall_nonlocal_sysv<mode>): Generate different code for
3297 -mno-speculate-indirect-jumps.
3298 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3300 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3302 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
3303 long double type, set the flags for noting the default long double
3304 type, even if we don't pass or return a long double type.
3306 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
3309 * ipa-inline.c (flatten_function): Do not overwrite final inlining
3312 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
3314 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
3315 support for merge[hl].
3316 (fold_mergehl_helper): New helper function.
3317 (tree-vector-builder.h): New #include for tree_vector_builder usage.
3318 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
3319 (altivec_vmrglw_direct): Add xxmrglw insn.
3321 2018-01-17 Andrew Waterman <andrew@sifive.com>
3323 * config/riscv/riscv.c (riscv_conditional_register_usage): If
3324 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
3326 2018-01-17 David Malcolm <dmalcolm@redhat.com>
3329 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
3330 call the lto_location_cache before reading the
3331 DECL_SOURCE_LOCATION of the types.
3333 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3334 Richard Sandiford <richard.sandiford@linaro.org>
3336 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
3337 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
3338 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
3339 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
3340 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
3342 * config/aarch64/constraints.md (aarch64_movti_operand):
3344 * config/aarch64/predicates.md (Uti): Add new constraint.
3346 2018-01-17 Carl Love <cel@us.ibm.com>
3348 * config/rs6000/vsx.md (define_expand xl_len_r,
3349 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
3350 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
3352 (define_expand, define_insn): Move the shift left from the
3353 define_insn to the define_expand for lxvl and stxvl instructions.
3354 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
3355 and XL_LEN_R definitions to PURE.
3357 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
3359 * config/i386/i386.c (indirect_thunk_name): Declare regno
3360 as unsigned int. Compare regno with INVALID_REGNUM.
3361 (output_indirect_thunk): Ditto.
3362 (output_indirect_thunk_function): Ditto.
3363 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
3364 in the call to output_indirect_thunk_function.
3366 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
3369 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
3370 rather than the size of inner_type to determine the stack slot size
3371 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
3373 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
3376 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
3379 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
3381 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
3382 endian Linux systems to optionally enable multilibs for selecting
3383 the long double type if the user configured an explicit type.
3384 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
3385 have no long double multilibs if not defined.
3386 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
3387 warn if the user used -mabi={ieee,ibm}longdouble and we built
3388 multilibs for long double.
3389 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
3390 appropriate multilib option.
3391 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
3393 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
3394 for building long double multilibs.
3395 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
3397 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
3399 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
3402 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
3404 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
3407 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
3410 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
3413 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3415 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
3416 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
3418 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3420 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
3421 different rtl trees depending on TARGET_64BIT.
3422 (rs6000_gen_lvx): Likewise.
3424 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3426 * config/visium/visium.md (nop): Tweak comment.
3427 (hazard_nop): Likewise.
3429 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3431 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
3432 -mspeculate-indirect-jumps.
3433 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
3434 for -mno-speculate-indirect-jumps.
3435 (*call_indirect_elfv2<mode>_nospec): New define_insn.
3436 (*call_value_indirect_elfv2<mode>): Disable for
3437 -mno-speculate-indirect-jumps.
3438 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
3439 (indirect_jump): Emit different RTL for
3440 -mno-speculate-indirect-jumps.
3441 (*indirect_jump<mode>): Disable for
3442 -mno-speculate-indirect-jumps.
3443 (*indirect_jump<mode>_nospec): New define_insn.
3444 (tablejump): Emit different RTL for
3445 -mno-speculate-indirect-jumps.
3446 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
3447 (tablejumpsi_nospec): New define_expand.
3448 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
3449 (tablejumpdi_nospec): New define_expand.
3450 (*tablejump<mode>_internal1): Disable for
3451 -mno-speculate-indirect-jumps.
3452 (*tablejump<mode>_internal1_nospec): New define_insn.
3453 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
3456 2018-01-16 Artyom Skrobov tyomitch@gmail.com
3458 * caller-save.c (insert_save): Drop unnecessary parameter. All
3461 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3462 Richard Biener <rguenth@suse.de>
3465 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
3466 return early, inline manually is_gimple_sizepos. Make sure if we
3467 call gimplify_expr we don't end up with a gimple constant.
3468 * tree.c (variably_modified_type_p): Don't return true for
3469 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
3470 * gimplify.h (is_gimple_sizepos): Remove.
3472 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3474 PR tree-optimization/83857
3475 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
3476 vectorizable_live_operation for pure SLP statements.
3477 (vectorizable_live_operation): Handle PHIs.
3479 2018-01-16 Richard Biener <rguenther@suse.de>
3481 PR tree-optimization/83867
3482 * tree-vect-stmts.c (vect_transform_stmt): Precompute
3483 nested_in_vect_loop_p since the scalar stmt may get invalidated.
3485 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3488 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
3489 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
3490 If off is not INTEGER_CST, issue a may not be aligned warning
3491 rather than isn't aligned. Use isn%'t rather than isn't.
3492 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
3494 <case MULT_EXPR>: Improve the case when bottom and one of the
3495 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
3496 operand, in that case check if the other operand is multiple of
3497 bottom divided by the INTEGER_CST operand.
3499 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3502 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
3503 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
3504 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
3505 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
3506 * config/pa/pa.c (pa_function_arg_advance): Likewise.
3507 (pa_function_arg, pa_arg_partial_bytes): Likewise.
3508 (pa_function_arg_size): New function.
3510 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3512 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
3513 in a separate statement.
3515 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3517 PR tree-optimization/83847
3518 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
3519 group gathers and scatters.
3521 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3523 PR rtl-optimization/86620
3524 * params.def (max-sched-ready-insns): Bump minimum value to 1.
3526 PR rtl-optimization/83213
3527 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
3528 to last if both are JUMP_INSNs.
3530 PR tree-optimization/83843
3531 * gimple-ssa-store-merging.c
3532 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
3533 store_immediate_info for bswap/nop orig_stores.
3535 2018-01-15 Andrew Waterman <andrew@sifive.com>
3537 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
3539 <UDIV>: Increase cost if !TARGET_DIV.
3541 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
3543 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
3544 (define_attr "cr_logical_3op"): New.
3545 (cceq_ior_compare): Adjust.
3546 (cceq_ior_compare_complement): Adjust.
3547 (*cceq_rev_compare): Adjust.
3548 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
3549 (is_cracked_insn): Adjust.
3550 (insn_must_be_first_in_group): Adjust.
3551 * config/rs6000/40x.md: Adjust.
3552 * config/rs6000/440.md: Adjust.
3553 * config/rs6000/476.md: Adjust.
3554 * config/rs6000/601.md: Adjust.
3555 * config/rs6000/603.md: Adjust.
3556 * config/rs6000/6xx.md: Adjust.
3557 * config/rs6000/7450.md: Adjust.
3558 * config/rs6000/7xx.md: Adjust.
3559 * config/rs6000/8540.md: Adjust.
3560 * config/rs6000/cell.md: Adjust.
3561 * config/rs6000/e300c2c3.md: Adjust.
3562 * config/rs6000/e500mc.md: Adjust.
3563 * config/rs6000/e500mc64.md: Adjust.
3564 * config/rs6000/e5500.md: Adjust.
3565 * config/rs6000/e6500.md: Adjust.
3566 * config/rs6000/mpc.md: Adjust.
3567 * config/rs6000/power4.md: Adjust.
3568 * config/rs6000/power5.md: Adjust.
3569 * config/rs6000/power6.md: Adjust.
3570 * config/rs6000/power7.md: Adjust.
3571 * config/rs6000/power8.md: Adjust.
3572 * config/rs6000/power9.md: Adjust.
3573 * config/rs6000/rs64.md: Adjust.
3574 * config/rs6000/titan.md: Adjust.
3576 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3578 * config/i386/predicates.md (indirect_branch_operand): Rewrite
3579 ix86_indirect_branch_register logic.
3581 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3583 * config/i386/constraints.md (Bs): Update
3584 ix86_indirect_branch_register check. Don't check
3585 ix86_indirect_branch_register with GOT_memory_operand.
3587 * config/i386/predicates.md (GOT_memory_operand): Don't check
3588 ix86_indirect_branch_register here.
3589 (GOT32_symbol_operand): Likewise.
3591 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3593 * config/i386/predicates.md (constant_call_address_operand):
3594 Rewrite ix86_indirect_branch_register logic.
3595 (sibcall_insn_operand): Likewise.
3597 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3599 * config/i386/constraints.md (Bs): Replace
3600 ix86_indirect_branch_thunk_register with
3601 ix86_indirect_branch_register.
3603 * config/i386/i386.md (indirect_jump): Likewise.
3604 (tablejump): Likewise.
3605 (*sibcall_memory): Likewise.
3606 (*sibcall_value_memory): Likewise.
3607 Peepholes of indirect call and jump via memory: Likewise.
3608 * config/i386/i386.opt: Likewise.
3609 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3610 (GOT_memory_operand): Likewise.
3611 (call_insn_operand): Likewise.
3612 (sibcall_insn_operand): Likewise.
3613 (GOT32_symbol_operand): Likewise.
3615 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3618 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
3619 type rather than type addr's type points to.
3620 (expand_omp_atomic_mutex): Likewise.
3621 (expand_omp_atomic): Likewise.
3623 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3626 * config/i386/i386.c (output_indirect_thunk_function): Use
3627 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
3628 for __x86_return_thunk.
3630 2018-01-15 Richard Biener <rguenther@suse.de>
3633 * expmed.c (extract_bit_field_1): Fix typo.
3635 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3638 * config/arm/iterators.md (VF): New mode iterator.
3639 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
3640 Remove integer-related logic from pattern.
3641 (neon_vabd<mode>_3): Likewise.
3643 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3646 * common.opt (fstrict-overflow): No longer an alias.
3647 (fwrapv-pointer): New option.
3648 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
3649 also for pointer types based on flag_wrapv_pointer.
3650 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
3651 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
3652 opts->x_flag_wrapv got set.
3653 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
3654 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
3655 POINTER_TYPE_OVERFLOW_UNDEFINED.
3656 * match.pd: Likewise in address comparison pattern.
3657 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
3659 2018-01-15 Richard Biener <rguenther@suse.de>
3662 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
3663 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
3664 Reset type names to their identifier if their TYPE_DECL doesn't
3665 have linkage (and thus is used for ODR and devirt).
3666 (save_debug_info_for_decl): Remove.
3667 (save_debug_info_for_type): Likewise.
3668 (add_tree_to_fld_list): Adjust.
3669 * tree-pretty-print.c (dump_generic_node): Make dumping of
3670 type names more robust.
3672 2018-01-15 Richard Biener <rguenther@suse.de>
3674 * BASE-VER: Bump to 8.0.1.
3676 2018-01-14 Martin Sebor <msebor@redhat.com>
3679 * builtins.c (check_access): Avoid warning when the no-warning bit
3682 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
3684 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
3685 * ira-color (allocno_hard_regs_compare): Likewise.
3687 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
3690 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
3691 Use .pushsection/.popsection.
3693 2018-01-14 Martin Sebor <msebor@redhat.com>
3696 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
3698 2018-01-14 Jakub Jelinek <jakub@redhat.com>
3700 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
3701 entry from extra_headers.
3702 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
3703 extra_headers, make the list bitwise identical to the i?86-*-* one.
3705 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3707 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
3708 -mcmodel=large with -mindirect-branch=thunk,
3709 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
3710 -mfunction-return=thunk-extern.
3711 * doc/invoke.texi: Document -mcmodel=large is incompatible with
3712 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
3713 -mfunction-return=thunk and -mfunction-return=thunk-extern.
3715 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3717 * config/i386/i386.c (print_reg): Print the name of the full
3718 integer register without '%'.
3719 (ix86_print_operand): Handle 'V'.
3720 * doc/extend.texi: Document 'V' modifier.
3722 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3724 * config/i386/constraints.md (Bs): Disallow memory operand for
3725 -mindirect-branch-register.
3727 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3728 (GOT_memory_operand): Likewise.
3729 (call_insn_operand): Likewise.
3730 (sibcall_insn_operand): Likewise.
3731 (GOT32_symbol_operand): Likewise.
3732 * config/i386/i386.md (indirect_jump): Call convert_memory_address
3733 for -mindirect-branch-register.
3734 (tablejump): Likewise.
3735 (*sibcall_memory): Likewise.
3736 (*sibcall_value_memory): Likewise.
3737 Disallow peepholes of indirect call and jump via memory for
3738 -mindirect-branch-register.
3739 (*call_pop): Replace m with Bw.
3740 (*call_value_pop): Likewise.
3741 (*sibcall_pop_memory): Replace m with Bs.
3742 * config/i386/i386.opt (mindirect-branch-register): New option.
3743 * doc/invoke.texi: Document -mindirect-branch-register option.
3745 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3747 * config/i386/i386-protos.h (ix86_output_function_return): New.
3748 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
3749 set function_return_type.
3750 (indirect_thunk_name): Add ret_p to indicate thunk for function
3752 (output_indirect_thunk_function): Pass false to
3753 indirect_thunk_name.
3754 (ix86_output_indirect_branch_via_reg): Likewise.
3755 (ix86_output_indirect_branch_via_push): Likewise.
3756 (output_indirect_thunk_function): Create alias for function
3757 return thunk if regno < 0.
3758 (ix86_output_function_return): New function.
3759 (ix86_handle_fndecl_attribute): Handle function_return.
3760 (ix86_attribute_table): Add function_return.
3761 * config/i386/i386.h (machine_function): Add
3762 function_return_type.
3763 * config/i386/i386.md (simple_return_internal): Use
3764 ix86_output_function_return.
3765 (simple_return_internal_long): Likewise.
3766 * config/i386/i386.opt (mfunction-return=): New option.
3767 (indirect_branch): Mention -mfunction-return=.
3768 * doc/extend.texi: Document function_return function attribute.
3769 * doc/invoke.texi: Document -mfunction-return= option.
3771 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3773 * config/i386/i386-opts.h (indirect_branch): New.
3774 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
3775 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
3776 with local indirect jump when converting indirect call and jump.
3777 (ix86_set_indirect_branch_type): New.
3778 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
3779 (indirectlabelno): New.
3780 (indirect_thunk_needed): Likewise.
3781 (indirect_thunk_bnd_needed): Likewise.
3782 (indirect_thunks_used): Likewise.
3783 (indirect_thunks_bnd_used): Likewise.
3784 (INDIRECT_LABEL): Likewise.
3785 (indirect_thunk_name): Likewise.
3786 (output_indirect_thunk): Likewise.
3787 (output_indirect_thunk_function): Likewise.
3788 (ix86_output_indirect_branch_via_reg): Likewise.
3789 (ix86_output_indirect_branch_via_push): Likewise.
3790 (ix86_output_indirect_branch): Likewise.
3791 (ix86_output_indirect_jmp): Likewise.
3792 (ix86_code_end): Call output_indirect_thunk_function if needed.
3793 (ix86_output_call_insn): Call ix86_output_indirect_branch if
3795 (ix86_handle_fndecl_attribute): Handle indirect_branch.
3796 (ix86_attribute_table): Add indirect_branch.
3797 * config/i386/i386.h (machine_function): Add indirect_branch_type
3798 and has_local_indirect_jump.
3799 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
3801 (tablejump): Likewise.
3802 (*indirect_jump): Use ix86_output_indirect_jmp.
3803 (*tablejump_1): Likewise.
3804 (simple_return_indirect_internal): Likewise.
3805 * config/i386/i386.opt (mindirect-branch=): New option.
3806 (indirect_branch): New.
3809 (thunk-inline): Likewise.
3810 (thunk-extern): Likewise.
3811 * doc/extend.texi: Document indirect_branch function attribute.
3812 * doc/invoke.texi: Document -mindirect-branch= option.
3814 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
3817 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
3819 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
3821 * ipa-inline.c (want_inline_small_function_p): Return false if
3822 inlining has already failed with CIF_FINAL_ERROR.
3823 (update_caller_keys): Call want_inline_small_function_p before
3825 (update_callee_keys): Likewise.
3827 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3829 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
3831 (rs6000_quadword_masked_address_p): Likewise.
3832 (quad_aligned_load_p): Likewise.
3833 (quad_aligned_store_p): Likewise.
3834 (const_load_sequence_p): Add comment to describe the outer-most loop.
3835 (mimic_memory_attributes_and_flags): New function.
3836 (rs6000_gen_stvx): Likewise.
3837 (replace_swapped_aligned_store): Likewise.
3838 (rs6000_gen_lvx): Likewise.
3839 (replace_swapped_aligned_load): Likewise.
3840 (replace_swapped_load_constant): Capitalize argument name in
3841 comment describing this function.
3842 (rs6000_analyze_swaps): Add a third pass to search for vector loads
3843 and stores that access quad-word aligned addresses and replace
3844 with stvx or lvx instructions when appropriate.
3845 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
3846 New function prototype.
3847 (rs6000_quadword_masked_address_p): Likewise.
3848 (rs6000_gen_lvx): Likewise.
3849 (rs6000_gen_stvx): Likewise.
3850 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
3851 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
3852 when memory address is aligned.
3853 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
3854 this split to select lvx instruction when memory address is aligned.
3855 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
3856 instruction when memory address is aligned.
3857 (*vsx_le_perm_load_v16qi): Likewise.
3858 (four unnamed splitters): Modify to select the stvx instruction
3859 when memory is aligned.
3861 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
3863 * predict.c (determine_unlikely_bbs): Handle correctly BBs
3864 which appears in the queue multiple times.
3866 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3867 Alan Hayward <alan.hayward@arm.com>
3868 David Sherwood <david.sherwood@arm.com>
3870 * tree-vectorizer.h (vec_lower_bound): New structure.
3871 (_loop_vec_info): Add check_nonzero and lower_bounds.
3872 (LOOP_VINFO_CHECK_NONZERO): New macro.
3873 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
3874 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
3875 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
3876 fields. Make seg_len the distance travelled, not including the
3878 (dr_direction_indicator): Declare.
3879 (dr_zero_step_indicator): Likewise.
3880 (dr_known_forward_stride_p): Likewise.
3881 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
3883 (runtime_alias_check_p): Allow runtime alias checks with
3885 (operator ==): Compare access_size and align.
3886 (prune_runtime_alias_test_list): Rework for new distinction between
3887 the access_size and seg_len.
3888 (create_intersect_range_checks_index): Likewise. Cope with polynomial
3890 (get_segment_min_max): New function.
3891 (create_intersect_range_checks): Use it.
3892 (dr_step_indicator): New function.
3893 (dr_direction_indicator): Likewise.
3894 (dr_zero_step_indicator): Likewise.
3895 (dr_known_forward_stride_p): Likewise.
3896 * tree-loop-distribution.c (data_ref_segment_size): Return
3897 DR_STEP * (niters - 1).
3898 (compute_alias_check_pairs): Update call to the dr_with_seg_len
3900 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
3901 (vect_preserves_scalar_order_p): New function, split out from...
3902 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
3903 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
3904 (vect_vfa_access_size): New function.
3905 (vect_vfa_align): Likewise.
3906 (vect_compile_time_alias): Take access_size_a and access_b arguments.
3907 (dump_lower_bound): New function.
3908 (vect_check_lower_bound): Likewise.
3909 (vect_small_gap_p): Likewise.
3910 (vectorizable_with_step_bound_p): Likewise.
3911 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
3912 depencies if the vectorization factor is 1. Convert the checks
3913 for nonzero steps into checks on the bounds of DR_STEP. Try using
3914 a bunds check for variable steps if the minimum required step is
3915 relatively small. Update calls to the dr_with_seg_len
3916 constructor and to vect_compile_time_alias.
3917 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
3919 (vect_loop_versioning): Call it.
3920 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
3922 (vect_estimate_min_profitable_iters): Account for any bounds checks.
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 * doc/sourcebuild.texi (vect_scatter_store): Document.
3929 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
3931 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
3933 * genopinit.c (main): Add supports_vec_scatter_store and
3934 supports_vec_scatter_store_cached to target_optabs.
3935 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
3936 IFN_MASK_SCATTER_STORE.
3937 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
3939 * internal-fn.h (internal_store_fn_p): Declare.
3940 (internal_fn_stored_value_index): Likewise.
3941 * internal-fn.c (scatter_store_direct): New macro.
3942 (expand_scatter_store_optab_fn): New function.
3943 (direct_scatter_store_optab_supported_p): New macro.
3944 (internal_store_fn_p): New function.
3945 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
3946 IFN_MASK_SCATTER_STORE.
3947 (internal_fn_mask_index): Likewise.
3948 (internal_fn_stored_value_index): New function.
3949 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
3951 * optabs-query.h (supports_vec_scatter_store_p): Declare.
3952 * optabs-query.c (supports_vec_scatter_store_p): New function.
3953 * tree-vectorizer.h (vect_get_store_rhs): Declare.
3954 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
3955 true for scatter stores.
3956 (vect_gather_scatter_fn_p): Handle scatter stores too.
3957 (vect_check_gather_scatter): Consider using scatter stores if
3958 supports_vec_scatter_store_p.
3959 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
3961 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
3962 internal_fn_stored_value_index.
3963 (check_load_store_masking): Handle scatter stores too.
3964 (vect_get_store_rhs): Make public.
3965 (vectorizable_call): Use internal_store_fn_p.
3966 (vectorizable_store): Handle scatter store internal functions.
3967 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
3968 when deciding whether the end of the group has been reached.
3969 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
3970 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
3971 (mask_scatter_store<mode>): New insns.
3973 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3974 Alan Hayward <alan.hayward@arm.com>
3975 David Sherwood <david.sherwood@arm.com>
3977 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
3978 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
3979 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
3981 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
3982 Use vect_truncate_gather_scatter_offset if we can't treat the
3983 operation as a normal gather load or scatter store.
3984 (get_group_load_store_type): Take the gather_scatter_info
3985 as argument. Try using a gather load or scatter store for
3986 single-element groups.
3987 (get_load_store_type): Update calls to get_group_load_store_type
3988 and vect_use_strided_gather_scatters_p.
3990 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3991 Alan Hayward <alan.hayward@arm.com>
3992 David Sherwood <david.sherwood@arm.com>
3994 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
3995 optional tree argument.
3996 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
3998 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
3999 but continue to use the current value as a fallback.
4000 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
4001 to compare the updates.
4002 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
4003 (get_load_store_type): Use it when handling a strided access.
4004 (vect_get_strided_load_store_ops): New function.
4005 (vect_get_data_ptr_increment): Likewise.
4006 (vectorizable_load): Handle strided gather loads. Always pass
4007 a step to vect_create_data_ref_ptr and bump_vector_ptr.
4009 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4010 Alan Hayward <alan.hayward@arm.com>
4011 David Sherwood <david.sherwood@arm.com>
4013 * doc/md.texi (gather_load@var{m}): Document.
4014 (mask_gather_load@var{m}): Likewise.
4015 * genopinit.c (main): Add supports_vec_gather_load and
4016 supports_vec_gather_load_cached to target_optabs.
4017 * optabs-tree.c (init_tree_optimization_optabs): Use
4018 ggc_cleared_alloc to allocate target_optabs.
4019 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
4020 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
4022 * internal-fn.h (internal_load_fn_p): Declare.
4023 (internal_gather_scatter_fn_p): Likewise.
4024 (internal_fn_mask_index): Likewise.
4025 (internal_gather_scatter_fn_supported_p): Likewise.
4026 * internal-fn.c (gather_load_direct): New macro.
4027 (expand_gather_load_optab_fn): New function.
4028 (direct_gather_load_optab_supported_p): New macro.
4029 (direct_internal_fn_optab): New function.
4030 (internal_load_fn_p): Likewise.
4031 (internal_gather_scatter_fn_p): Likewise.
4032 (internal_fn_mask_index): Likewise.
4033 (internal_gather_scatter_fn_supported_p): Likewise.
4034 * optabs-query.c (supports_at_least_one_mode_p): New function.
4035 (supports_vec_gather_load_p): Likewise.
4036 * optabs-query.h (supports_vec_gather_load_p): Declare.
4037 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
4038 and memory_type field.
4039 (NUM_PATTERNS): Bump to 15.
4040 * tree-vect-data-refs.c: Include internal-fn.h.
4041 (vect_gather_scatter_fn_p): New function.
4042 (vect_describe_gather_scatter_call): Likewise.
4043 (vect_check_gather_scatter): Try using internal functions for
4044 gather loads. Recognize existing calls to a gather load function.
4045 (vect_analyze_data_refs): Consider using gather loads if
4046 supports_vec_gather_load_p.
4047 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
4048 (vect_get_gather_scatter_offset_type): Likewise.
4049 (vect_convert_mask_for_vectype): Likewise.
4050 (vect_add_conversion_to_patterm): Likewise.
4051 (vect_try_gather_scatter_pattern): Likewise.
4052 (vect_recog_gather_scatter_pattern): New pattern recognizer.
4053 (vect_vect_recog_func_ptrs): Add it.
4054 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4055 internal_fn_mask_index and internal_gather_scatter_fn_p.
4056 (check_load_store_masking): Take the gather_scatter_info as an
4057 argument and handle gather loads.
4058 (vect_get_gather_scatter_ops): New function.
4059 (vectorizable_call): Check internal_load_fn_p.
4060 (vectorizable_load): Likewise. Handle gather load internal
4062 (vectorizable_store): Update call to check_load_store_masking.
4063 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
4064 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
4065 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
4066 (aarch64_gather_scale_operand_d): New predicates.
4067 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
4068 (mask_gather_load<mode>): New insns.
4070 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4071 Alan Hayward <alan.hayward@arm.com>
4072 David Sherwood <david.sherwood@arm.com>
4074 * optabs.def (fold_left_plus_optab): New optab.
4075 * doc/md.texi (fold_left_plus_@var{m}): Document.
4076 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
4077 * internal-fn.c (fold_left_direct): Define.
4078 (expand_fold_left_optab_fn): Likewise.
4079 (direct_fold_left_optab_supported_p): Likewise.
4080 * fold-const-call.c (fold_const_fold_left): New function.
4081 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
4082 * tree-parloops.c (valid_reduction_p): New function.
4083 (gather_scalar_reductions): Use it.
4084 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
4085 (vect_finish_replace_stmt): Declare.
4086 * tree-vect-loop.c (fold_left_reduction_fn): New function.
4087 (needs_fold_left_reduction_p): New function, split out from...
4088 (vect_is_simple_reduction): ...here. Accept reductions that
4089 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
4090 (vect_force_simple_reduction): Also store the reduction type in
4091 the assignment's STMT_VINFO_REDUC_TYPE.
4092 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
4093 (merge_with_identity): New function.
4094 (vect_expand_fold_left): Likewise.
4095 (vectorize_fold_left_reduction): Likewise.
4096 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
4097 scalar phi in place for it. Check for target support and reject
4098 cases that would reassociate the operation. Defer the transform
4099 phase to vectorize_fold_left_reduction.
4100 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
4101 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
4102 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
4104 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4106 * tree-if-conv.c (predicate_mem_writes): Remove redundant
4107 call to ifc_temp_var.
4109 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4110 Alan Hayward <alan.hayward@arm.com>
4111 David Sherwood <david.sherwood@arm.com>
4113 * target.def (legitimize_address_displacement): Take the original
4114 offset as a poly_int.
4115 * targhooks.h (default_legitimize_address_displacement): Update
4117 * targhooks.c (default_legitimize_address_displacement): Likewise.
4118 * doc/tm.texi: Regenerate.
4119 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
4120 as an argument, moving assert of ad->disp == ad->disp_term to...
4121 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
4122 Try calling targetm.legitimize_address_displacement before expanding
4123 the address rather than afterwards, and adjust for the new interface.
4124 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
4125 Match the new hook interface. Handle SVE addresses.
4126 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
4129 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4131 * Makefile.in (OBJS): Add early-remat.o.
4132 * target.def (select_early_remat_modes): New hook.
4133 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
4134 * doc/tm.texi: Regenerate.
4135 * targhooks.h (default_select_early_remat_modes): Declare.
4136 * targhooks.c (default_select_early_remat_modes): New function.
4137 * timevar.def (TV_EARLY_REMAT): New timevar.
4138 * passes.def (pass_early_remat): New pass.
4139 * tree-pass.h (make_pass_early_remat): Declare.
4140 * early-remat.c: New file.
4141 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
4143 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
4145 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4146 Alan Hayward <alan.hayward@arm.com>
4147 David Sherwood <david.sherwood@arm.com>
4149 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
4150 vfm1 with a bound_epilog parameter.
4151 (vect_do_peeling): Update calls accordingly, and move the prologue
4152 call earlier in the function. Treat the base bound_epilog as 0 for
4153 fully-masked loops and retain vf - 1 for other loops. Add 1 to
4154 this base when peeling for gaps.
4155 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
4156 with fully-masked loops.
4157 (vect_estimate_min_profitable_iters): Handle the single peeled
4158 iteration in that case.
4160 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4161 Alan Hayward <alan.hayward@arm.com>
4162 David Sherwood <david.sherwood@arm.com>
4164 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
4165 single-element interleaving even if the size is not a power of 2.
4166 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
4167 accesses for single-element interleaving if the group size is
4170 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4171 Alan Hayward <alan.hayward@arm.com>
4172 David Sherwood <david.sherwood@arm.com>
4174 * doc/md.texi (fold_extract_last_@var{m}): Document.
4175 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
4176 * optabs.def (fold_extract_last_optab): New optab.
4177 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
4178 * internal-fn.c (fold_extract_direct): New macro.
4179 (expand_fold_extract_optab_fn): Likewise.
4180 (direct_fold_extract_optab_supported_p): Likewise.
4181 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
4182 * tree-vect-loop.c (vect_model_reduction_cost): Handle
4183 EXTRACT_LAST_REDUCTION.
4184 (get_initial_def_for_reduction): Do not create an initial vector
4185 for EXTRACT_LAST_REDUCTION reductions.
4186 (vectorizable_reduction): Leave the scalar phi in place for
4187 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
4188 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
4189 epilogue code for EXTRACT_LAST_REDUCTION and defer the
4190 transform phase to vectorizable_condition.
4191 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
4193 (vect_finish_stmt_generation): ...here.
4194 (vect_finish_replace_stmt): New function.
4195 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
4196 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
4198 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
4200 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4201 Alan Hayward <alan.hayward@arm.com>
4202 David Sherwood <david.sherwood@arm.com>
4204 * doc/md.texi (extract_last_@var{m}): Document.
4205 * optabs.def (extract_last_optab): New optab.
4206 * internal-fn.def (EXTRACT_LAST): New internal function.
4207 * internal-fn.c (cond_unary_direct): New macro.
4208 (expand_cond_unary_optab_fn): Likewise.
4209 (direct_cond_unary_optab_supported_p): Likewise.
4210 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
4211 loops using EXTRACT_LAST.
4212 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
4213 (extract_last_<mode>): ...this optab.
4214 (vec_extract<mode><Vel>): Update accordingly.
4216 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4217 Alan Hayward <alan.hayward@arm.com>
4218 David Sherwood <david.sherwood@arm.com>
4220 * target.def (empty_mask_is_expensive): New hook.
4221 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
4222 * doc/tm.texi: Regenerate.
4223 * targhooks.h (default_empty_mask_is_expensive): Declare.
4224 * targhooks.c (default_empty_mask_is_expensive): New function.
4225 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
4226 if the target says that empty masks are expensive.
4227 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
4229 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
4231 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4232 Alan Hayward <alan.hayward@arm.com>
4233 David Sherwood <david.sherwood@arm.com>
4235 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
4236 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
4237 (vect_use_loop_mask_for_alignment_p): New function.
4238 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
4239 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
4240 niters_skip argument. Make sure that the first niters_skip elements
4241 of the first iteration are inactive.
4242 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
4243 Update call to vect_set_loop_masks_directly.
4244 (get_misalign_in_elems): New function, split out from...
4245 (vect_gen_prolog_loop_niters): ...here.
4246 (vect_update_init_of_dr): Take a code argument that specifies whether
4247 the adjustment should be added or subtracted.
4248 (vect_update_init_of_drs): Likewise.
4249 (vect_prepare_for_masked_peels): New function.
4250 (vect_do_peeling): Skip prologue peeling if we're using a mask
4251 instead. Update call to vect_update_inits_of_drs.
4252 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4254 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
4255 alignment. Do not include the number of peeled iterations in
4256 the minimum threshold in that case.
4257 (vectorizable_induction): Adjust the start value down by
4258 LOOP_VINFO_MASK_SKIP_NITERS iterations.
4259 (vect_transform_loop): Call vect_prepare_for_masked_peels.
4260 Take the number of skipped iterations into account when calculating
4262 * tree-vect-stmts.c (vect_gen_while_not): New function.
4264 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4265 Alan Hayward <alan.hayward@arm.com>
4266 David Sherwood <david.sherwood@arm.com>
4268 * doc/sourcebuild.texi (vect_fully_masked): Document.
4269 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
4271 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
4273 (vect_analyze_loop_2): ...here. Don't check the vectorization
4274 factor against the number of loop iterations if the loop is
4277 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4278 Alan Hayward <alan.hayward@arm.com>
4279 David Sherwood <david.sherwood@arm.com>
4281 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
4282 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
4283 (dump_groups): Update accordingly.
4284 (iv_use::mem_type): New member variable.
4285 (address_p): New function.
4286 (record_use): Add a mem_type argument and initialize the new
4288 (record_group_use): Add a mem_type argument. Use address_p.
4289 Remove obsolete null checks of base_object. Update call to record_use.
4290 (find_interesting_uses_op): Update call to record_group_use.
4291 (find_interesting_uses_cond): Likewise.
4292 (find_interesting_uses_address): Likewise.
4293 (get_mem_type_for_internal_fn): New function.
4294 (find_address_like_use): Likewise.
4295 (find_interesting_uses_stmt): Try find_address_like_use before
4296 calling find_interesting_uses_op.
4297 (addr_offset_valid_p): Use the iv mem_type field as the type
4298 of the addressed memory.
4299 (add_autoinc_candidates): Likewise.
4300 (get_address_cost): Likewise.
4301 (split_small_address_groups_p): Use address_p.
4302 (split_address_groups): Likewise.
4303 (add_iv_candidate_for_use): Likewise.
4304 (autoinc_possible_for_pair): Likewise.
4305 (rewrite_groups): Likewise.
4306 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
4307 (determine_group_iv_cost): Update after split of USE_ADDRESS.
4308 (get_alias_ptr_type_for_ptr_address): New function.
4309 (rewrite_use_address): Rewrite address uses in calls that were
4310 identified by find_address_like_use.
4312 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4313 Alan Hayward <alan.hayward@arm.com>
4314 David Sherwood <david.sherwood@arm.com>
4316 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
4318 * gimple-expr.h (is_gimple_addressable: Likewise.
4319 * gimple-expr.c (is_gimple_address): Likewise.
4320 * internal-fn.c (expand_call_mem_ref): New function.
4321 (expand_mask_load_optab_fn): Use it.
4322 (expand_mask_store_optab_fn): Likewise.
4324 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4325 Alan Hayward <alan.hayward@arm.com>
4326 David Sherwood <david.sherwood@arm.com>
4328 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
4329 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
4330 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
4331 (cond_umax@var{mode}): Document.
4332 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
4333 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
4334 (cond_umin_optab, cond_umax_optab): New optabs.
4335 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
4336 (COND_IOR, COND_XOR): New internal functions.
4337 * internal-fn.h (get_conditional_internal_fn): Declare.
4338 * internal-fn.c (cond_binary_direct): New macro.
4339 (expand_cond_binary_optab_fn): Likewise.
4340 (direct_cond_binary_optab_supported_p): Likewise.
4341 (get_conditional_internal_fn): New function.
4342 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
4343 Cope with reduction statements that are vectorized as calls rather
4345 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
4346 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
4347 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
4348 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4349 (UNSPEC_COND_EOR): New unspecs.
4350 (optab): Add mappings for them.
4351 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
4352 (sve_int_op, sve_fp_op): New int attributes.
4354 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4355 Alan Hayward <alan.hayward@arm.com>
4356 David Sherwood <david.sherwood@arm.com>
4358 * optabs.def (while_ult_optab): New optab.
4359 * doc/md.texi (while_ult@var{m}@var{n}): Document.
4360 * internal-fn.def (WHILE_ULT): New internal function.
4361 * internal-fn.h (direct_internal_fn_supported_p): New override
4362 that takes two types as argument.
4363 * internal-fn.c (while_direct): New macro.
4364 (expand_while_optab_fn): New function.
4365 (convert_optab_supported_p): Likewise.
4366 (direct_while_optab_supported_p): New macro.
4367 * wide-int.h (wi::udiv_ceil): New function.
4368 * tree-vectorizer.h (rgroup_masks): New structure.
4369 (vec_loop_masks): New typedef.
4370 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
4372 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
4373 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
4374 (vect_max_vf): New function.
4375 (slpeel_make_loop_iterate_ntimes): Delete.
4376 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
4377 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
4378 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
4379 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
4380 internal-fn.h, stor-layout.h and optabs-query.h.
4381 (vect_set_loop_mask): New function.
4382 (add_preheader_seq): Likewise.
4383 (add_header_seq): Likewise.
4384 (interleave_supported_p): Likewise.
4385 (vect_maybe_permute_loop_masks): Likewise.
4386 (vect_set_loop_masks_directly): Likewise.
4387 (vect_set_loop_condition_masked): Likewise.
4388 (vect_set_loop_condition_unmasked): New function, split out from
4389 slpeel_make_loop_iterate_ntimes.
4390 (slpeel_make_loop_iterate_ntimes): Rename to..
4391 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
4392 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
4393 (vect_do_peeling): Update call accordingly.
4394 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
4396 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4397 mask_compare_type, can_fully_mask_p and fully_masked_p.
4398 (release_vec_loop_masks): New function.
4399 (_loop_vec_info): Use it to free the loop masks.
4400 (can_produce_all_loop_masks_p): New function.
4401 (vect_get_max_nscalars_per_iter): Likewise.
4402 (vect_verify_full_masking): Likewise.
4403 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
4404 retries, and free the mask rgroups before retrying. Check loop-wide
4405 reasons for disallowing fully-masked loops. Make the final decision
4406 about whether use a fully-masked loop or not.
4407 (vect_estimate_min_profitable_iters): Do not assume that peeling
4408 for the number of iterations will be needed for fully-masked loops.
4409 (vectorizable_reduction): Disable fully-masked loops.
4410 (vectorizable_live_operation): Likewise.
4411 (vect_halve_mask_nunits): New function.
4412 (vect_double_mask_nunits): Likewise.
4413 (vect_record_loop_mask): Likewise.
4414 (vect_get_loop_mask): Likewise.
4415 (vect_transform_loop): Handle the case in which the final loop
4416 iteration might handle a partial vector. Call vect_set_loop_condition
4417 instead of slpeel_make_loop_iterate_ntimes.
4418 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
4419 (check_load_store_masking): New function.
4420 (prepare_load_store_mask): Likewise.
4421 (vectorizable_store): Handle fully-masked loops.
4422 (vectorizable_load): Likewise.
4423 (supportable_widening_operation): Use vect_halve_mask_nunits for
4425 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
4426 (vect_gen_while): New function.
4427 * config/aarch64/aarch64.md (umax<mode>3): New expander.
4428 (aarch64_uqdec<mode>): New insn.
4430 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4431 Alan Hayward <alan.hayward@arm.com>
4432 David Sherwood <david.sherwood@arm.com>
4434 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
4435 (reduc_xor_scal_optab): New optabs.
4436 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
4437 (reduc_xor_scal_@var{m}): Document.
4438 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
4439 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
4441 * fold-const-call.c (fold_const_call): Handle them.
4442 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
4443 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
4444 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
4445 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
4446 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
4447 (UNSPEC_XORV): New unspecs.
4448 (optab): Add entries for them.
4449 (BITWISEV): New int iterator.
4450 (bit_reduc_op): New int attributes.
4452 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4453 Alan Hayward <alan.hayward@arm.com>
4454 David Sherwood <david.sherwood@arm.com>
4456 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
4457 * internal-fn.def (VEC_SHL_INSERT): New internal function.
4458 * optabs.def (vec_shl_insert_optab): New optab.
4459 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
4460 (duplicate_and_interleave): Likewise.
4461 * tree-vect-loop.c: Include internal-fn.h.
4462 (neutral_op_for_slp_reduction): New function, split out from
4463 get_initial_defs_for_reduction.
4464 (get_initial_def_for_reduction): Handle option 2 for variable-length
4465 vectors by loading the neutral value into a vector and then shifting
4466 the initial value into element 0.
4467 (get_initial_defs_for_reduction): Replace the code argument with
4468 the neutral value calculated by neutral_op_for_slp_reduction.
4469 Use gimple_build_vector for constant-length vectors.
4470 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
4471 but the first group_size elements have a neutral value.
4472 Use duplicate_and_interleave otherwise.
4473 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
4474 Update call to get_initial_defs_for_reduction. Handle SLP
4475 reductions for variable-length vectors by creating one vector
4476 result for each scalar result, with the elements associated
4477 with other scalar results stubbed out with the neutral value.
4478 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
4479 Require IFN_VEC_SHL_INSERT for double reductions on
4480 variable-length vectors, or SLP reductions that have
4481 a neutral value. Require can_duplicate_and_interleave_p
4482 support for variable-length unchained SLP reductions if there
4483 is no neutral value, such as for MIN/MAX reductions. Also require
4484 the number of vector elements to be a multiple of the number of
4485 SLP statements when doing variable-length unchained SLP reductions.
4486 Update call to vect_create_epilog_for_reduction.
4487 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
4488 and remove initial values.
4489 (duplicate_and_interleave): Make public.
4490 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
4491 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
4493 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4494 Alan Hayward <alan.hayward@arm.com>
4495 David Sherwood <david.sherwood@arm.com>
4497 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
4498 (can_duplicate_and_interleave_p): New function.
4499 (vect_get_and_check_slp_defs): Take the vector of statements
4500 rather than just the current one. Remove excess parentheses.
4501 Restriction rejectinon of vect_constant_def and vect_external_def
4502 for variable-length vectors to boolean types, or types for which
4503 can_duplicate_and_interleave_p is false.
4504 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
4505 (duplicate_and_interleave): New function.
4506 (vect_get_constant_vectors): Use gimple_build_vector for
4507 constant-length vectors and suitable variable-length constant
4508 vectors. Use duplicate_and_interleave for other variable-length
4509 vectors. Don't defer the update when inserting new statements.
4511 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4512 Alan Hayward <alan.hayward@arm.com>
4513 David Sherwood <david.sherwood@arm.com>
4515 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
4516 min_profitable_iters doesn't go negative.
4518 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4519 Alan Hayward <alan.hayward@arm.com>
4520 David Sherwood <david.sherwood@arm.com>
4522 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
4523 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
4524 * optabs.def (vec_mask_load_lanes_optab): New optab.
4525 (vec_mask_store_lanes_optab): Likewise.
4526 * internal-fn.def (MASK_LOAD_LANES): New internal function.
4527 (MASK_STORE_LANES): Likewise.
4528 * internal-fn.c (mask_load_lanes_direct): New macro.
4529 (mask_store_lanes_direct): Likewise.
4530 (expand_mask_load_optab_fn): Handle masked operations.
4531 (expand_mask_load_lanes_optab_fn): New macro.
4532 (expand_mask_store_optab_fn): Handle masked operations.
4533 (expand_mask_store_lanes_optab_fn): New macro.
4534 (direct_mask_load_lanes_optab_supported_p): Likewise.
4535 (direct_mask_store_lanes_optab_supported_p): Likewise.
4536 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
4538 (vect_load_lanes_supported): Likewise.
4539 * tree-vect-data-refs.c (strip_conversion): New function.
4540 (can_group_stmts_p): Likewise.
4541 (vect_analyze_data_ref_accesses): Use it instead of checking
4542 for a pair of assignments.
4543 (vect_store_lanes_supported): Take a masked_p parameter.
4544 (vect_load_lanes_supported): Likewise.
4545 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
4546 vect_store_lanes_supported and vect_load_lanes_supported.
4547 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4548 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
4549 parameter. Don't allow gaps for masked accesses.
4550 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
4551 and vect_load_lanes_supported.
4552 (get_load_store_type): Take a masked_p parameter and update
4553 call to get_group_load_store_type.
4554 (vectorizable_store): Update call to get_load_store_type.
4555 Handle IFN_MASK_STORE_LANES.
4556 (vectorizable_load): Update call to get_load_store_type.
4557 Handle IFN_MASK_LOAD_LANES.
4559 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4560 Alan Hayward <alan.hayward@arm.com>
4561 David Sherwood <david.sherwood@arm.com>
4563 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
4565 * config/aarch64/aarch64-protos.h
4566 (aarch64_sve_struct_memory_operand_p): Declare.
4567 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
4568 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
4569 (VPRED, vpred): Handle SVE structure modes.
4570 * config/aarch64/constraints.md (Utx): New constraint.
4571 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
4572 (aarch64_sve_struct_nonimmediate_operand): New predicates.
4573 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
4574 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
4575 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
4576 structure modes. Split into pieces after RA.
4577 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
4578 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
4580 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
4581 SVE structure modes.
4582 (aarch64_classify_address): Likewise.
4583 (sizetochar): Move earlier in file.
4584 (aarch64_print_operand): Handle SVE register lists.
4585 (aarch64_array_mode): New function.
4586 (aarch64_sve_struct_memory_operand_p): Likewise.
4587 (TARGET_ARRAY_MODE): Redefine.
4589 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4590 Alan Hayward <alan.hayward@arm.com>
4591 David Sherwood <david.sherwood@arm.com>
4593 * target.def (array_mode): New target hook.
4594 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
4595 * doc/tm.texi: Regenerate.
4596 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
4597 * hooks.c (hook_optmode_mode_uhwi_none): New function.
4598 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
4600 * stor-layout.c (mode_for_array): Likewise. Support polynomial
4603 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4604 Alan Hayward <alan.hayward@arm.com>
4605 David Sherwood <david.sherwood@arm.com>
4607 * fold-const.c (fold_binary_loc): Check the argument types
4608 rather than the result type when testing for a vector operation.
4610 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4612 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
4613 * doc/tm.texi: Regenerate.
4615 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4616 Alan Hayward <alan.hayward@arm.com>
4617 David Sherwood <david.sherwood@arm.com>
4619 * doc/invoke.texi (-msve-vector-bits=): Document new option.
4620 (sve): Document new AArch64 extension.
4621 * doc/md.texi (w): Extend the description of the AArch64
4622 constraint to include SVE vectors.
4623 (Upl, Upa): Document new AArch64 predicate constraints.
4624 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
4626 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
4627 (msve-vector-bits=): New option.
4628 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
4629 SVE when these are disabled.
4630 (sve): New extension.
4631 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
4632 modes. Adjust their number of units based on aarch64_sve_vg.
4633 (MAX_BITSIZE_MODE_ANY_MODE): Define.
4634 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
4635 aarch64_addr_query_type.
4636 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
4637 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
4638 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
4639 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
4640 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
4641 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
4642 (aarch64_simd_imm_zero_p): Delete.
4643 (aarch64_check_zero_based_sve_index_immediate): Declare.
4644 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4645 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4646 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4647 (aarch64_sve_float_mul_immediate_p): Likewise.
4648 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4650 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
4651 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
4652 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
4653 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
4654 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
4655 (aarch64_regmode_natural_size): Likewise.
4656 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
4657 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
4659 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
4660 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
4661 for VG and the SVE predicate registers.
4662 (V_ALIASES): Add a "z"-prefixed alias.
4663 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
4664 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
4665 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
4666 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
4667 (REG_CLASS_NAMES): Add entries for them.
4668 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
4669 and the predicate registers.
4670 (aarch64_sve_vg): Declare.
4671 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
4672 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
4673 (REGMODE_NATURAL_SIZE): Define.
4674 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
4676 * config/aarch64/aarch64.c: Include cfgrtl.h.
4677 (simd_immediate_info): Add a constructor for series vectors,
4678 and an associated step field.
4679 (aarch64_sve_vg): New variable.
4680 (aarch64_dbx_register_number): Handle VG and the predicate registers.
4681 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
4682 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
4683 (VEC_ANY_DATA, VEC_STRUCT): New constants.
4684 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
4685 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
4686 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
4687 (aarch64_get_mask_mode): New functions.
4688 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
4689 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
4690 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
4691 predicate modes and predicate registers. Explicitly restrict
4692 GPRs to modes of 16 bytes or smaller. Only allow FP registers
4693 to store a vector mode if it is recognized by
4694 aarch64_classify_vector_mode.
4695 (aarch64_regmode_natural_size): New function.
4696 (aarch64_hard_regno_caller_save_mode): Return the original mode
4698 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
4699 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
4700 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
4701 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
4703 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
4704 does not overlap dest if the function is frame-related. Handle
4706 (aarch64_split_add_offset): New function.
4707 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
4708 them aarch64_add_offset.
4709 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
4710 and update call to aarch64_sub_sp.
4711 (aarch64_add_cfa_expression): New function.
4712 (aarch64_expand_prologue): Pass extra temporary registers to the
4713 functions above. Handle the case in which we need to emit new
4714 DW_CFA_expressions for registers that were originally saved
4715 relative to the stack pointer, but now have to be expressed
4716 relative to the frame pointer.
4717 (aarch64_output_mi_thunk): Pass extra temporary registers to the
4719 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
4720 IP0 and IP1 values for SVE frames.
4721 (aarch64_expand_vec_series): New function.
4722 (aarch64_expand_sve_widened_duplicate): Likewise.
4723 (aarch64_expand_sve_const_vector): Likewise.
4724 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
4725 Handle SVE constants. Use emit_move_insn to move a force_const_mem
4726 into the register, rather than emitting a SET directly.
4727 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
4728 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
4729 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
4730 (offset_9bit_signed_scaled_p): New functions.
4731 (aarch64_replicate_bitmask_imm): New function.
4732 (aarch64_bitmask_imm): Use it.
4733 (aarch64_cannot_force_const_mem): Reject expressions involving
4734 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
4735 (aarch64_classify_index): Handle SVE indices, by requiring
4736 a plain register index with a scale that matches the element size.
4737 (aarch64_classify_address): Handle SVE addresses. Assert that
4738 the mode of the address is VOIDmode or an integer mode.
4739 Update call to aarch64_classify_symbol.
4740 (aarch64_classify_symbolic_expression): Update call to
4741 aarch64_classify_symbol.
4742 (aarch64_const_vec_all_in_range_p): New function.
4743 (aarch64_print_vector_float_operand): Likewise.
4744 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
4745 "vN" for FP registers with SVE modes. Handle (const ...) vectors
4746 and the FP immediates 1.0 and 0.5.
4747 (aarch64_print_address_internal): Handle SVE addresses.
4748 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
4749 (aarch64_regno_regclass): Handle predicate registers.
4750 (aarch64_secondary_reload): Handle big-endian reloads of SVE
4752 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
4753 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
4754 (aarch64_convert_sve_vector_bits): New function.
4755 (aarch64_override_options): Use it to handle -msve-vector-bits=.
4756 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4758 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
4759 Handle SVE vector and predicate modes. Accept VL-based constants
4760 that need only one temporary register, and VL offsets that require
4761 no temporary registers.
4762 (aarch64_conditional_register_usage): Mark the predicate registers
4763 as fixed if SVE isn't available.
4764 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
4765 Return true for SVE vector and predicate modes.
4766 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
4767 rather than an unsigned int. Handle SVE modes.
4768 (aarch64_preferred_simd_mode): Update call accordingly. Handle
4770 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
4772 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4773 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4774 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4775 (aarch64_sve_float_mul_immediate_p): New functions.
4776 (aarch64_sve_valid_immediate): New function.
4777 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
4778 Explicitly reject structure modes. Check for INDEX constants.
4779 Handle PTRUE and PFALSE constants.
4780 (aarch64_check_zero_based_sve_index_immediate): New function.
4781 (aarch64_simd_imm_zero_p): Delete.
4782 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
4783 vector modes. Accept constants in the range of CNT[BHWD].
4784 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
4785 ask for an Advanced SIMD mode.
4786 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
4787 (aarch64_simd_vector_alignment): Handle SVE predicates.
4788 (aarch64_vectorize_preferred_vector_alignment): New function.
4789 (aarch64_simd_vector_alignment_reachable): Use it instead of
4791 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
4792 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
4794 (MAX_VECT_LEN): Delete.
4795 (expand_vec_perm_d): Add a vec_flags field.
4796 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
4797 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
4798 (aarch64_evpc_ext): Don't apply a big-endian lane correction
4800 (aarch64_evpc_rev): Rename to...
4801 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
4802 (aarch64_evpc_rev_global): New function.
4803 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
4804 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
4806 (aarch64_evpc_sve_tbl): New function.
4807 (aarch64_expand_vec_perm_const_1): Update after rename of
4808 aarch64_evpc_rev. Handle SVE permutes too, trying
4809 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
4810 than aarch64_evpc_tbl.
4811 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
4812 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
4813 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
4814 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
4815 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
4816 (aarch64_expand_sve_vcond): New functions.
4817 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
4818 of aarch64_vector_mode_p.
4819 (aarch64_dwarf_poly_indeterminate_value): New function.
4820 (aarch64_compute_pressure_classes): Likewise.
4821 (aarch64_can_change_mode_class): Likewise.
4822 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
4823 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
4824 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
4825 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
4826 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
4827 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
4828 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
4829 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
4831 (Dn, Dl, Dr): Accept const as well as const_vector.
4832 (Dz): Likewise. Compare against CONST0_RTX.
4833 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
4834 of "vector" where appropriate.
4835 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
4836 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
4837 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
4838 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
4839 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
4840 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
4841 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
4842 (v_int_equiv): Extend to SVE modes.
4843 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
4845 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
4846 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
4847 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
4848 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
4849 (SVE_COND_FP_CMP): New int iterators.
4850 (perm_hilo): Handle the new unpack unspecs.
4851 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
4853 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
4854 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
4855 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
4856 (aarch64_equality_operator, aarch64_constant_vector_operand)
4857 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
4858 (aarch64_sve_nonimmediate_operand): Likewise.
4859 (aarch64_sve_general_operand): Likewise.
4860 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
4861 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
4862 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
4863 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
4864 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
4865 (aarch64_sve_float_arith_immediate): Likewise.
4866 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
4867 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
4868 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
4869 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
4870 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
4871 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
4872 (aarch64_sve_float_arith_operand): Likewise.
4873 (aarch64_sve_float_arith_with_sub_operand): Likewise.
4874 (aarch64_sve_float_mul_operand): Likewise.
4875 (aarch64_sve_vec_perm_operand): Likewise.
4876 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
4877 (aarch64_mov_operand): Accept const_poly_int and const_vector.
4878 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
4879 as well as const_vector.
4880 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
4881 in file. Use CONST0_RTX and CONSTM1_RTX.
4882 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
4883 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
4884 Use aarch64_simd_imm_zero.
4885 * config/aarch64/aarch64-sve.md: New file.
4886 * config/aarch64/aarch64.md: Include it.
4887 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
4888 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
4889 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
4890 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
4891 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
4892 (sve): New attribute.
4893 (enabled): Disable instructions with the sve attribute unless
4895 (movqi, movhi): Pass CONST_POLY_INT operaneds through
4896 aarch64_expand_mov_immediate.
4897 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
4898 CNT[BHSD] immediates.
4899 (movti): Split CONST_POLY_INT moves into two halves.
4900 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
4901 Split additions that need a temporary here if the destination
4902 is the stack pointer.
4903 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
4904 (*add<mode>3_poly_1): New instruction.
4905 (set_clobber_cc): New expander.
4907 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4909 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
4910 parameter and use it instead of GET_MODE_SIZE (innermode). Use
4911 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
4912 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
4913 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
4914 Change innermode from fixed_mode_size to machine_mode.
4915 (simplify_subreg): Update call accordingly. Handle a constant-sized
4916 subreg of a variable-length CONST_VECTOR.
4918 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4919 Alan Hayward <alan.hayward@arm.com>
4920 David Sherwood <david.sherwood@arm.com>
4922 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
4923 (add_offset_to_base): New function, split out from...
4924 (create_mem_ref): ...here. When handling a scale other than 1,
4925 check first whether the address is valid without the offset.
4926 Add it into the base if so, leaving the index and scale as-is.
4928 2018-01-12 Jakub Jelinek <jakub@redhat.com>
4931 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
4932 fold_for_warn before checking if arg2 is INTEGER_CST.
4934 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
4936 * config/rs6000/predicates.md (load_multiple_operation): Delete.
4937 (store_multiple_operation): Delete.
4938 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
4939 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
4940 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
4941 guarded by TARGET_STRING.
4942 (rs6000_output_load_multiple): Delete.
4943 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
4944 OPTION_MASK_STRING / TARGET_STRING handling.
4945 (print_operand) <'N', 'O'>: Add comment that these are unused now.
4946 (const rs6000_opt_masks) <"string">: Change mask to 0.
4947 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
4948 (MASK_STRING): Delete.
4949 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
4951 (load_multiple): Delete.
4958 (store_multiple): Delete.
4965 (movmemsi_8reg): Delete.
4966 (corresponding unnamed define_insn): Delete.
4967 (movmemsi_6reg): Delete.
4968 (corresponding unnamed define_insn): Delete.
4969 (movmemsi_4reg): Delete.
4970 (corresponding unnamed define_insn): Delete.
4971 (movmemsi_2reg): Delete.
4972 (corresponding unnamed define_insn): Delete.
4973 (movmemsi_1reg): Delete.
4974 (corresponding unnamed define_insn): Delete.
4975 * config/rs6000/rs6000.opt (mno-string): New.
4976 (mstring): Replace by deprecation warning stub.
4977 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
4979 2018-01-12 Jakub Jelinek <jakub@redhat.com>
4981 * regrename.c (regrename_do_replace): If replacing the same
4982 reg multiple times, try to reuse last created gen_raw_REG.
4985 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
4986 main to workaround a bug in GDB.
4988 2018-01-12 Tom de Vries <tom@codesourcery.com>
4991 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
4993 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
4995 PR rtl-optimization/80481
4996 * ira-color.c (get_cap_member): New function.
4997 (allocnos_conflict_by_live_ranges_p): Use it.
4998 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
4999 (setup_slot_coalesced_allocno_live_ranges): Ditto.
5001 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
5004 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
5005 (*saddl_se_1): Ditto.
5007 (*ssubl_se_1): Ditto.
5009 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5011 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
5012 rather than wi::to_widest for DR_INITs.
5013 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
5014 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
5015 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
5017 (vect_analyze_group_access_1): Note that here.
5019 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5021 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
5022 polynomial type sizes.
5024 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5026 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
5027 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
5028 (gimple_add_tmp_var): Likewise.
5030 2018-01-12 Martin Liska <mliska@suse.cz>
5032 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
5033 (gimple_alloc_sizes): Likewise.
5034 (dump_gimple_statistics): Use PRIu64 in printf format.
5035 * gimple.h: Change uint64_t to int.
5037 2018-01-12 Martin Liska <mliska@suse.cz>
5039 * tree-core.h: Use uint64_t instead of int.
5040 * tree.c (tree_node_counts): Likewise.
5041 (tree_node_sizes): Likewise.
5042 (dump_tree_statistics): Use PRIu64 in printf format.
5044 2018-01-12 Martin Liska <mliska@suse.cz>
5046 * Makefile.in: As qsort_chk is implemented in vec.c, add
5047 vec.o to linkage of gencfn-macros.
5048 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
5049 passing the info to record_node_allocation_statistics.
5050 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
5052 * ggc-common.c (struct ggc_usage): Add operator== and use
5053 it in operator< and compare function.
5054 * mem-stats.h (struct mem_usage): Likewise.
5055 * vec.c (struct vec_usage): Remove operator< and compare
5056 function. Can be simply inherited.
5058 2018-01-12 Martin Jambor <mjambor@suse.cz>
5061 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
5062 * tree-ssa-math-opts.c: Include domwalk.h.
5063 (convert_mult_to_fma_1): New function.
5064 (fma_transformation_info): New type.
5065 (fma_deferring_state): Likewise.
5066 (cancel_fma_deferring): New function.
5067 (result_of_phi): Likewise.
5068 (last_fma_candidate_feeds_initial_phi): Likewise.
5069 (convert_mult_to_fma): Added deferring logic, split actual
5070 transformation to convert_mult_to_fma_1.
5071 (math_opts_dom_walker): New type.
5072 (math_opts_dom_walker::after_dom_children): New method, body moved
5073 here from pass_optimize_widening_mul::execute, added deferring logic
5075 (pass_optimize_widening_mul::execute): Moved most of code to
5076 math_opts_dom_walker::after_dom_children.
5077 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
5078 * config/i386/i386.c (ix86_option_override_internal): Added
5079 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
5081 2018-01-12 Richard Biener <rguenther@suse.de>
5084 * dwarf2out.c (gen_variable_die): Do not reset old_die for
5085 inline instance vars.
5087 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
5090 * config/rx/rx.c (rx_is_restricted_memory_address):
5093 2018-01-12 Richard Biener <rguenther@suse.de>
5095 PR tree-optimization/80846
5096 * target.def (split_reduction): New target hook.
5097 * targhooks.c (default_split_reduction): New function.
5098 * targhooks.h (default_split_reduction): Declare.
5099 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
5100 target requests first reduce vectors by combining low and high
5102 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
5103 (get_vectype_for_scalar_type_and_size): Export.
5104 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
5105 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
5106 * doc/tm.texi: Regenerate.
5107 * config/i386/i386.c (ix86_split_reduction): Implement
5108 TARGET_VECTORIZE_SPLIT_REDUCTION.
5110 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5113 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
5114 in PIC mode except for TARGET_VXWORKS_RTP.
5115 * config/sparc/sparc.c: Include cfgrtl.h.
5116 (TARGET_INIT_PIC_REG): Define.
5117 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
5118 (sparc_pic_register_p): New predicate.
5119 (sparc_legitimate_address_p): Use it.
5120 (sparc_legitimize_pic_address): Likewise.
5121 (sparc_delegitimize_address): Likewise.
5122 (sparc_mode_dependent_address_p): Likewise.
5123 (gen_load_pcrel_sym): Remove 4th parameter.
5124 (load_got_register): Adjust call to above. Remove obsolete stuff.
5125 (sparc_expand_prologue): Do not call load_got_register here.
5126 (sparc_flat_expand_prologue): Likewise.
5127 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
5128 (sparc_use_pseudo_pic_reg): New function.
5129 (sparc_init_pic_reg): Likewise.
5130 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5131 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
5133 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
5135 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
5136 Add item for branch_cost.
5138 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5140 PR rtl-optimization/83565
5141 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
5142 not extend the result to a larger mode for rotate operations.
5143 (num_sign_bit_copies1): Likewise.
5145 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5148 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
5150 Use values-Xc.o for -pedantic.
5151 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
5153 2018-01-12 Martin Liska <mliska@suse.cz>
5156 * ipa-devirt.c (final_warning_record::grow_type_warnings):
5158 (possible_polymorphic_call_targets): Use it.
5159 (ipa_devirt): Likewise.
5161 2018-01-12 Martin Liska <mliska@suse.cz>
5163 * profile-count.h (enum profile_quality): Use 0 as invalid
5164 enum value of profile_quality.
5166 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
5168 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
5169 -mext-string options.
5171 2018-01-12 Richard Biener <rguenther@suse.de>
5173 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
5174 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
5175 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
5177 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
5179 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5181 * configure.ac (--with-long-double-format): Add support for the
5182 configuration option to change the default long double format on
5184 * config.gcc (powerpc*-linux*-*): Likewise.
5185 * configure: Regenerate.
5186 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
5187 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
5188 used without modification.
5190 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5192 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
5193 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
5194 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
5195 MISC_BUILTIN_SPEC_BARRIER.
5196 (rs6000_init_builtins): Likewise.
5197 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
5199 (speculation_barrier): New define_insn.
5200 * doc/extend.texi: Document __builtin_speculation_barrier.
5202 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5205 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
5206 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
5207 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
5209 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
5210 integral modes instead of "ss" and "sd".
5211 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
5212 vectors with 32-bit and 64-bit elements.
5213 (vecdupssescalarmodesuffix): New mode attribute.
5214 (vec_dup<mode>): Use it.
5216 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
5219 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
5220 frame if argument is passed on stack.
5222 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5225 * ree.c (combine_reaching_defs): Optimize also
5226 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
5227 reg2=any_extend(exp); reg1=reg2;, formatting fix.
5229 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5232 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
5234 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5237 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
5238 after they are computed.
5240 2018-01-11 Bin Cheng <bin.cheng@arm.com>
5242 PR tree-optimization/83695
5243 * gimple-loop-linterchange.cc
5244 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
5245 reset cached scev information after interchange.
5246 (pass_linterchange::execute): Remove call to scev_reset_htab.
5248 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5250 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
5251 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
5252 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
5253 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
5254 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
5255 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
5256 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
5257 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
5258 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
5259 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
5260 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
5261 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
5262 (V_lane_reg): Likewise.
5263 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
5265 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
5266 (vfmal_lane_low<mode>_intrinsic,
5267 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
5268 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
5269 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
5270 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
5271 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
5272 vfmsl_lane_high<mode>_intrinsic): New define_insns.
5274 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5276 * config/arm/arm-cpus.in (fp16fml): New feature.
5277 (ALL_SIMD): Add fp16fml.
5278 (armv8.2-a): Add fp16fml as an option.
5279 (armv8.3-a): Likewise.
5280 (armv8.4-a): Add fp16fml as part of fp16.
5281 * config/arm/arm.h (TARGET_FP16FML): Define.
5282 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
5284 * config/arm/arm-modes.def (V2HF): Define.
5285 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
5286 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
5287 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
5288 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
5289 vfmsl_low, vfmsl_high): New set of builtins.
5290 * config/arm/iterators.md (PLUSMINUS): New code iterator.
5291 (vfml_op): New code attribute.
5292 (VFMLHALVES): New int iterator.
5293 (VFML, VFMLSEL): New mode attributes.
5294 (V_reg): Define mapping for V2HF.
5295 (V_hi, V_lo): New mode attributes.
5296 (VF_constraint): Likewise.
5297 (vfml_half, vfml_half_selector): New int attributes.
5298 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
5300 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
5301 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
5303 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
5304 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
5305 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
5306 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
5308 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
5309 Document new effective target and option set.
5311 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5313 * config/arm/arm-cpus.in (armv8_4): New feature.
5314 (ARMv8_4a): New fgroup.
5315 (armv8.4-a): New arch.
5316 * config/arm/arm-tables.opt: Regenerate.
5317 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
5318 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
5319 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
5320 Add matching rules for -march=armv8.4-a and extensions.
5321 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
5323 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
5326 * config/rx/rx.md (BW): New mode attribute.
5327 (sync_lock_test_and_setsi): Add mode suffix to insn output.
5329 2018-01-11 Richard Biener <rguenther@suse.de>
5331 PR tree-optimization/83435
5332 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
5333 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
5334 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
5336 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5337 Alan Hayward <alan.hayward@arm.com>
5338 David Sherwood <david.sherwood@arm.com>
5340 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
5342 (aarch64_classify_address): Initialize it. Track polynomial offsets.
5343 (aarch64_print_address_internal): Use it to check for a zero offset.
5345 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5346 Alan Hayward <alan.hayward@arm.com>
5347 David Sherwood <david.sherwood@arm.com>
5349 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
5350 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
5351 Return a poly_int64 rather than a HOST_WIDE_INT.
5352 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
5353 rather than a HOST_WIDE_INT.
5354 * config/aarch64/aarch64.h (aarch64_frame): Protect with
5355 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
5356 hard_fp_offset, frame_size, initial_adjust, callee_offset and
5357 final_offset from HOST_WIDE_INT to poly_int64.
5358 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
5359 to_constant when getting the number of units in an Advanced SIMD
5361 (aarch64_builtin_vectorized_function): Check for a constant number
5363 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
5365 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
5366 attribute instead of GET_MODE_NUNITS.
5367 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
5368 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
5369 GET_MODE_SIZE for fixed-size registers.
5370 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
5371 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
5372 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
5373 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
5374 (aarch64_print_operand, aarch64_print_address_internal)
5375 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
5376 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
5377 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
5378 Handle polynomial GET_MODE_SIZE.
5379 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
5380 wider than SImode without modification.
5381 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
5382 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
5383 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
5384 passing and returning SVE modes.
5385 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
5386 rather than GEN_INT.
5387 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
5388 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
5389 (aarch64_allocate_and_probe_stack_space): Likewise.
5390 (aarch64_layout_frame): Cope with polynomial offsets.
5391 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
5392 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
5394 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
5395 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
5396 poly_int64 rather than a HOST_WIDE_INT.
5397 (aarch64_get_separate_components, aarch64_process_components)
5398 (aarch64_expand_prologue, aarch64_expand_epilogue)
5399 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
5400 (aarch64_anchor_offset): New function, split out from...
5401 (aarch64_legitimize_address): ...here.
5402 (aarch64_builtin_vectorization_cost): Handle polynomial
5403 TYPE_VECTOR_SUBPARTS.
5404 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
5406 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
5407 number of elements from the PARALLEL rather than the mode.
5408 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
5409 rather than GET_MODE_BITSIZE.
5410 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
5411 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
5412 (aarch64_expand_vec_perm_const_1): Handle polynomial
5413 d->perm.length () and d->perm elements.
5414 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
5415 Apply to_constant to d->perm elements.
5416 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
5417 polynomial CONST_VECTOR_NUNITS.
5418 (aarch64_move_pointer): Take amount as a poly_int64 rather
5420 (aarch64_progress_pointer): Avoid temporary variable.
5421 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
5422 the mode attribute instead of GET_MODE.
5424 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5425 Alan Hayward <alan.hayward@arm.com>
5426 David Sherwood <david.sherwood@arm.com>
5428 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
5429 x exists before using it.
5430 (aarch64_add_constant_internal): Rename to...
5431 (aarch64_add_offset_1): ...this. Replace regnum with separate
5432 src and dest rtxes. Handle the case in which they're different,
5433 including when the offset is zero. Replace scratchreg with an rtx.
5434 Use 2 additions if there is no spare register into which we can
5435 move a 16-bit constant.
5436 (aarch64_add_constant): Delete.
5437 (aarch64_add_offset): Replace reg with separate src and dest
5438 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
5439 Use aarch64_add_offset_1.
5440 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
5441 an rtx rather than an int. Take the delta as a poly_int64
5442 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
5443 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
5444 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
5445 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
5446 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
5448 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
5449 aarch64_add_constant.
5451 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5453 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
5454 Use scalar_float_mode.
5456 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5458 * config/aarch64/aarch64-simd.md
5459 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
5460 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
5461 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
5462 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
5463 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
5464 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
5465 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
5466 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
5467 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
5468 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
5470 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5473 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
5474 targ_options->x_arm_arch_string is non NULL.
5476 2018-01-11 Tamar Christina <tamar.christina@arm.com>
5478 * config/aarch64/aarch64.h
5479 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
5481 2018-01-11 Sudakshina Das <sudi.das@arm.com>
5484 * expmed.c (emit_store_flag_force): Swap if const op0
5485 and change VOIDmode to mode of op0.
5487 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5489 PR rtl-optimization/83761
5490 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
5491 than bytes to mode_for_size.
5493 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5496 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
5497 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
5500 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5503 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
5504 when in layout mode.
5505 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
5506 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
5509 2018-01-10 Michael Collison <michael.collison@arm.com>
5511 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
5512 * config/aarch64/aarch64-option-extension.def: Add
5513 AARCH64_OPT_EXTENSION of 'fp16fml'.
5514 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5515 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
5516 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
5517 * config/aarch64/constraints.md (Ui7): New constraint.
5518 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
5519 (VFMLA_SEL_W): Ditto.
5522 (VFMLA16_LOW): New int iterator.
5523 (VFMLA16_HIGH): Ditto.
5524 (UNSPEC_FMLAL): New unspec.
5525 (UNSPEC_FMLSL): Ditto.
5526 (UNSPEC_FMLAL2): Ditto.
5527 (UNSPEC_FMLSL2): Ditto.
5528 (f16mac): New code attribute.
5529 * config/aarch64/aarch64-simd-builtins.def
5530 (aarch64_fmlal_lowv2sf): Ditto.
5531 (aarch64_fmlsl_lowv2sf): Ditto.
5532 (aarch64_fmlalq_lowv4sf): Ditto.
5533 (aarch64_fmlslq_lowv4sf): Ditto.
5534 (aarch64_fmlal_highv2sf): Ditto.
5535 (aarch64_fmlsl_highv2sf): Ditto.
5536 (aarch64_fmlalq_highv4sf): Ditto.
5537 (aarch64_fmlslq_highv4sf): Ditto.
5538 (aarch64_fmlal_lane_lowv2sf): Ditto.
5539 (aarch64_fmlsl_lane_lowv2sf): Ditto.
5540 (aarch64_fmlal_laneq_lowv2sf): Ditto.
5541 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
5542 (aarch64_fmlalq_lane_lowv4sf): Ditto.
5543 (aarch64_fmlsl_lane_lowv4sf): Ditto.
5544 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
5545 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
5546 (aarch64_fmlal_lane_highv2sf): Ditto.
5547 (aarch64_fmlsl_lane_highv2sf): Ditto.
5548 (aarch64_fmlal_laneq_highv2sf): Ditto.
5549 (aarch64_fmlsl_laneq_highv2sf): Ditto.
5550 (aarch64_fmlalq_lane_highv4sf): Ditto.
5551 (aarch64_fmlsl_lane_highv4sf): Ditto.
5552 (aarch64_fmlalq_laneq_highv4sf): Ditto.
5553 (aarch64_fmlsl_laneq_highv4sf): Ditto.
5554 * config/aarch64/aarch64-simd.md:
5555 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
5556 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5557 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
5558 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5559 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
5560 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
5561 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
5562 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
5563 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
5564 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
5565 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
5566 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
5567 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
5568 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
5569 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
5570 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
5571 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
5572 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
5573 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
5574 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
5575 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
5576 (vfmlsl_low_u32): Ditto.
5577 (vfmlalq_low_u32): Ditto.
5578 (vfmlslq_low_u32): Ditto.
5579 (vfmlal_high_u32): Ditto.
5580 (vfmlsl_high_u32): Ditto.
5581 (vfmlalq_high_u32): Ditto.
5582 (vfmlslq_high_u32): Ditto.
5583 (vfmlal_lane_low_u32): Ditto.
5584 (vfmlsl_lane_low_u32): Ditto.
5585 (vfmlal_laneq_low_u32): Ditto.
5586 (vfmlsl_laneq_low_u32): Ditto.
5587 (vfmlalq_lane_low_u32): Ditto.
5588 (vfmlslq_lane_low_u32): Ditto.
5589 (vfmlalq_laneq_low_u32): Ditto.
5590 (vfmlslq_laneq_low_u32): Ditto.
5591 (vfmlal_lane_high_u32): Ditto.
5592 (vfmlsl_lane_high_u32): Ditto.
5593 (vfmlal_laneq_high_u32): Ditto.
5594 (vfmlsl_laneq_high_u32): Ditto.
5595 (vfmlalq_lane_high_u32): Ditto.
5596 (vfmlslq_lane_high_u32): Ditto.
5597 (vfmlalq_laneq_high_u32): Ditto.
5598 (vfmlslq_laneq_high_u32): Ditto.
5599 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
5600 (AARCH64_FL_FOR_ARCH8_4): New.
5601 (AARCH64_ISA_F16FML): New ISA flag.
5602 (TARGET_F16FML): New feature flag for fp16fml.
5603 (doc/invoke.texi): Document new fp16fml option.
5605 2018-01-10 Michael Collison <michael.collison@arm.com>
5607 * config/aarch64/aarch64-builtins.c:
5608 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
5609 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5610 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
5611 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
5612 (AARCH64_ISA_SHA3): New ISA flag.
5613 (TARGET_SHA3): New feature flag for sha3.
5614 * config/aarch64/iterators.md (sha512_op): New int attribute.
5615 (CRYPTO_SHA512): New int iterator.
5616 (UNSPEC_SHA512H): New unspec.
5617 (UNSPEC_SHA512H2): Ditto.
5618 (UNSPEC_SHA512SU0): Ditto.
5619 (UNSPEC_SHA512SU1): Ditto.
5620 * config/aarch64/aarch64-simd-builtins.def
5621 (aarch64_crypto_sha512hqv2di): New builtin.
5622 (aarch64_crypto_sha512h2qv2di): Ditto.
5623 (aarch64_crypto_sha512su0qv2di): Ditto.
5624 (aarch64_crypto_sha512su1qv2di): Ditto.
5625 (aarch64_eor3qv8hi): Ditto.
5626 (aarch64_rax1qv2di): Ditto.
5627 (aarch64_xarqv2di): Ditto.
5628 (aarch64_bcaxqv8hi): Ditto.
5629 * config/aarch64/aarch64-simd.md:
5630 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
5631 (aarch64_crypto_sha512su0qv2di): Ditto.
5632 (aarch64_crypto_sha512su1qv2di): Ditto.
5633 (aarch64_eor3qv8hi): Ditto.
5634 (aarch64_rax1qv2di): Ditto.
5635 (aarch64_xarqv2di): Ditto.
5636 (aarch64_bcaxqv8hi): Ditto.
5637 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
5638 (vsha512h2q_u64): Ditto.
5639 (vsha512su0q_u64): Ditto.
5640 (vsha512su1q_u64): Ditto.
5641 (veor3q_u16): Ditto.
5642 (vrax1q_u64): Ditto.
5644 (vbcaxq_u16): Ditto.
5645 * config/arm/types.md (crypto_sha512): New type attribute.
5646 (crypto_sha3): Ditto.
5647 (doc/invoke.texi): Document new sha3 option.
5649 2018-01-10 Michael Collison <michael.collison@arm.com>
5651 * config/aarch64/aarch64-builtins.c:
5652 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
5653 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5654 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
5655 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
5656 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
5657 (AARCH64_ISA_SM4): New ISA flag.
5658 (TARGET_SM4): New feature flag for sm4.
5659 * config/aarch64/aarch64-simd-builtins.def
5660 (aarch64_sm3ss1qv4si): Ditto.
5661 (aarch64_sm3tt1aq4si): Ditto.
5662 (aarch64_sm3tt1bq4si): Ditto.
5663 (aarch64_sm3tt2aq4si): Ditto.
5664 (aarch64_sm3tt2bq4si): Ditto.
5665 (aarch64_sm3partw1qv4si): Ditto.
5666 (aarch64_sm3partw2qv4si): Ditto.
5667 (aarch64_sm4eqv4si): Ditto.
5668 (aarch64_sm4ekeyqv4si): Ditto.
5669 * config/aarch64/aarch64-simd.md:
5670 (aarch64_sm3ss1qv4si): Ditto.
5671 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
5672 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
5673 (aarch64_sm4eqv4si): Ditto.
5674 (aarch64_sm4ekeyqv4si): Ditto.
5675 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
5676 (sm3part_op): Ditto.
5677 (CRYPTO_SM3TT): Ditto.
5678 (CRYPTO_SM3PART): Ditto.
5679 (UNSPEC_SM3SS1): New unspec.
5680 (UNSPEC_SM3TT1A): Ditto.
5681 (UNSPEC_SM3TT1B): Ditto.
5682 (UNSPEC_SM3TT2A): Ditto.
5683 (UNSPEC_SM3TT2B): Ditto.
5684 (UNSPEC_SM3PARTW1): Ditto.
5685 (UNSPEC_SM3PARTW2): Ditto.
5686 (UNSPEC_SM4E): Ditto.
5687 (UNSPEC_SM4EKEY): Ditto.
5688 * config/aarch64/constraints.md (Ui2): New constraint.
5689 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
5690 * config/arm/types.md (crypto_sm3): New type attribute.
5691 (crypto_sm4): Ditto.
5692 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
5693 (vsm3tt1aq_u32): Ditto.
5694 (vsm3tt1bq_u32): Ditto.
5695 (vsm3tt2aq_u32): Ditto.
5696 (vsm3tt2bq_u32): Ditto.
5697 (vsm3partw1q_u32): Ditto.
5698 (vsm3partw2q_u32): Ditto.
5699 (vsm4eq_u32): Ditto.
5700 (vsm4ekeyq_u32): Ditto.
5701 (doc/invoke.texi): Document new sm4 option.
5703 2018-01-10 Michael Collison <michael.collison@arm.com>
5705 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
5706 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
5707 (AARCH64_FL_FOR_ARCH8_4): New.
5708 (AARCH64_FL_V8_4): New flag.
5709 (doc/invoke.texi): Document new armv8.4-a option.
5711 2018-01-10 Michael Collison <michael.collison@arm.com>
5713 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5714 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
5715 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
5716 * config/aarch64/aarch64-option-extension.def: Add
5717 AARCH64_OPT_EXTENSION of 'sha2'.
5718 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
5719 (crypto): Disable sha2 and aes if crypto disabled.
5720 (crypto): Enable aes and sha2 if enabled.
5721 (simd): Disable sha2 and aes if simd disabled.
5722 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
5724 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
5725 (TARGET_SHA2): New feature flag for sha2.
5726 (TARGET_AES): New feature flag for aes.
5727 * config/aarch64/aarch64-simd.md:
5728 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
5729 conditional on TARGET_AES.
5730 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
5731 (aarch64_crypto_sha1hsi): Make pattern conditional
5733 (aarch64_crypto_sha1hv4si): Ditto.
5734 (aarch64_be_crypto_sha1hv4si): Ditto.
5735 (aarch64_crypto_sha1su1v4si): Ditto.
5736 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
5737 (aarch64_crypto_sha1su0v4si): Ditto.
5738 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
5739 (aarch64_crypto_sha256su0v4si): Ditto.
5740 (aarch64_crypto_sha256su1v4si): Ditto.
5741 (doc/invoke.texi): Document new aes and sha2 options.
5743 2018-01-10 Martin Sebor <msebor@redhat.com>
5745 PR tree-optimization/83781
5746 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
5749 2018-01-11 Martin Sebor <msebor@gmail.com>
5750 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5752 PR tree-optimization/83501
5753 PR tree-optimization/81703
5755 * tree-ssa-strlen.c (get_string_cst): Rename...
5756 (get_string_len): ...to this. Handle global constants.
5757 (handle_char_store): Adjust.
5759 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
5760 Jim Wilson <jimw@sifive.com>
5762 * config/riscv/riscv-protos.h (riscv_output_return): New.
5763 * config/riscv/riscv.c (struct machine_function): New naked_p field.
5764 (riscv_attribute_table, riscv_output_return),
5765 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
5766 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
5767 (riscv_compute_frame_info): Only compute frame->mask if not a naked
5769 (riscv_expand_prologue): Add early return for naked function.
5770 (riscv_expand_epilogue): Likewise.
5771 (riscv_function_ok_for_sibcall): Return false for naked function.
5772 (riscv_set_current_function): New.
5773 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
5774 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
5775 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
5776 * doc/extend.texi (RISC-V Function Attributes): New.
5778 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
5780 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
5781 check for 128-bit long double before checking TCmode.
5782 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
5783 128-bit long doubles before checking TFmode or TCmode.
5784 (FLOAT128_IBM_P): Likewise.
5786 2018-01-10 Martin Sebor <msebor@redhat.com>
5788 PR tree-optimization/83671
5789 * builtins.c (c_strlen): Unconditionally return zero for the empty
5791 Use -Warray-bounds for warnings.
5792 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
5793 for non-constant array indices with COMPONENT_REF, arrays of
5794 arrays, and pointers to arrays.
5795 (gimple_fold_builtin_strlen): Determine and set length range for
5796 non-constant character arrays.
5798 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
5801 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
5804 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
5806 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
5808 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5811 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
5812 VECTOR_MEM_ALTIVEC_OR_VSX_P.
5813 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
5814 indexed_or_indirect_operand predicate.
5815 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
5816 (*vsx_le_perm_load_v8hi): Likewise.
5817 (*vsx_le_perm_load_v16qi): Likewise.
5818 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
5819 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
5820 (*vsx_le_perm_store_v8hi): Likewise.
5821 (*vsx_le_perm_store_v16qi): Likewise.
5822 (eight unnamed splitters): Likewise.
5824 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5826 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
5827 * config/rs6000/emmintrin.h: Likewise.
5828 * config/rs6000/mmintrin.h: Likewise.
5829 * config/rs6000/xmmintrin.h: Likewise.
5831 2018-01-10 David Malcolm <dmalcolm@redhat.com>
5834 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
5836 * tree.c (tree_nop_conversion): Return true for location wrapper
5838 (maybe_wrap_with_location): New function.
5839 (selftest::check_strip_nops): New function.
5840 (selftest::test_location_wrappers): New function.
5841 (selftest::tree_c_tests): Call it.
5842 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
5843 (maybe_wrap_with_location): New decl.
5844 (EXPR_LOCATION_WRAPPER_P): New macro.
5845 (location_wrapper_p): New inline function.
5846 (tree_strip_any_location_wrapper): New inline function.
5848 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
5851 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
5852 stack_realign_offset for the largest alignment of stack slot
5854 (ix86_find_max_used_stack_alignment): New function.
5855 (ix86_finalize_stack_frame_flags): Use it. Set
5856 max_used_stack_alignment if we don't realign stack.
5857 * config/i386/i386.h (machine_function): Add
5858 max_used_stack_alignment.
5860 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
5862 * config/arm/arm.opt (-mbranch-cost): New option.
5863 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
5866 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
5869 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
5870 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
5872 2018-01-10 Richard Biener <rguenther@suse.de>
5875 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
5876 early out so it also covers the case where we have a non-NULL
5879 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
5881 PR tree-optimization/83753
5882 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
5883 for non-strided grouped accesses if the number of elements is 1.
5885 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5888 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
5889 * i386.h (TARGET_USE_GATHER): Define.
5890 * x86-tune.def (X86_TUNE_USE_GATHER): New.
5892 2018-01-10 Martin Liska <mliska@suse.cz>
5895 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
5896 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
5898 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
5899 CLEANUP_NO_PARTITIONING is not set.
5901 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
5903 * doc/rtl.texi: Remove documentation of (const ...) wrappers
5904 for vectors, as a partial revert of r254296.
5905 * rtl.h (const_vec_p): Delete.
5906 (const_vec_duplicate_p): Don't test for vector CONSTs.
5907 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
5908 * expmed.c (make_tree): Likewise.
5911 * common.md (E, F): Use CONSTANT_P instead of checking for
5913 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
5914 checking for CONST_VECTOR.
5916 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5919 * predict.c (force_edge_cold): Handle in more sane way edges
5922 2018-01-09 Carl Love <cel@us.ibm.com>
5924 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
5926 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
5927 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
5928 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
5929 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
5930 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
5931 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
5932 * config/rs6000/rs6000-protos.h: Add extern defition for
5933 rs6000_generate_float2_double_code.
5934 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
5936 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
5937 (float2_v2df): Add define_expand.
5939 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
5942 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
5943 op_mode in the force_to_mode call.
5945 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5947 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
5948 instead of checking each element individually.
5949 (aarch64_evpc_uzp): Likewise.
5950 (aarch64_evpc_zip): Likewise.
5951 (aarch64_evpc_ext): Likewise.
5952 (aarch64_evpc_rev): Likewise.
5953 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
5954 instead of checking each element individually. Return true without
5956 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
5957 whether all selected elements come from the same input, instead of
5958 checking each element individually. Remove calls to gen_rtx_REG,
5959 start_sequence and end_sequence and instead assert that no rtl is
5962 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5964 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
5965 order of HIGH and CONST checks.
5967 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5969 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
5970 if the destination isn't an SSA_NAME.
5972 2018-01-09 Richard Biener <rguenther@suse.de>
5974 PR tree-optimization/83668
5975 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
5977 (canonicalize_loop_form): ... here, renamed from ...
5978 (canonicalize_loop_closed_ssa_form): ... this and amended to
5979 swap successor edges for loop exit blocks to make us use
5980 the RPO order we need for initial schedule generation.
5982 2018-01-09 Joseph Myers <joseph@codesourcery.com>
5984 PR tree-optimization/64811
5985 * match.pd: When optimizing comparisons with Inf, avoid
5986 introducing or losing exceptions from comparisons with NaN.
5988 2018-01-09 Martin Liska <mliska@suse.cz>
5991 * asan.c (shadow_mem_size): Add gcc_assert.
5993 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
5995 Don't save registers in main().
5998 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
5999 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
6000 * config/avr/avr.c (avr_set_current_function): Don't error if
6001 naked, OS_task or OS_main are specified at the same time.
6002 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
6004 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
6006 * common/config/avr/avr-common.c (avr_option_optimization_table):
6007 Switch on -mmain-is-OS_task for optimizing compilations.
6009 2018-01-09 Richard Biener <rguenther@suse.de>
6011 PR tree-optimization/83572
6012 * graphite.c: Include cfganal.h.
6013 (graphite_transform_loops): Connect infinite loops to exit
6014 and remove fake edges at the end.
6016 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6018 * ipa-inline.c (edge_badness): Revert accidental checkin.
6020 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6023 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
6024 symbols; not inline clones.
6026 2018-01-09 Jakub Jelinek <jakub@redhat.com>
6029 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
6030 hard registers. Formatting fixes.
6032 PR preprocessor/83722
6033 * gcc.c (try_generate_repro): Pass
6034 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
6035 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
6038 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
6039 Kito Cheng <kito.cheng@gmail.com>
6041 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
6042 (riscv_leaf_function_p): Delete.
6043 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
6045 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6047 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
6049 (do_ifelse): New function.
6050 (do_isel): New function.
6051 (do_sub3): New function.
6052 (do_add3): New function.
6053 (do_load_mask_compare): New function.
6054 (do_overlap_load_compare): New function.
6055 (expand_compare_loop): New function.
6056 (expand_block_compare): Call expand_compare_loop() when appropriate.
6057 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
6059 (-mblock-compare-inline-loop-limit): New option.
6061 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6064 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
6065 Reverse order of second and third operands in first alternative.
6066 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
6067 of first and second elements in UNSPEC_VPERMR vector.
6068 (altivec_expand_vec_perm_le): Likewise.
6070 2018-01-08 Jeff Law <law@redhat.com>
6072 PR rtl-optimizatin/81308
6073 * tree-switch-conversion.c (cfg_altered): New file scoped static.
6074 (process_switch): If group_case_labels makes a change, then set
6076 (pass_convert_switch::execute): If a switch is converted, then
6077 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
6079 PR rtl-optimization/81308
6080 * recog.c (split_all_insns): Conditionally cleanup the CFG after
6083 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
6085 PR target/83663 - Revert r255946
6086 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
6087 generation for cases where splatting a value is not useful.
6088 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
6089 across a vec_duplicate and a paradoxical subreg forming a vector
6090 mode to a vec_concat.
6092 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6094 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
6095 -march=armv8.3-a variants.
6096 * config/arm/t-multilib: Likewise.
6097 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
6099 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6101 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
6103 (cceq_ior_compare_complement): Give it a name so I can use it, and
6104 change boolean_or_operator predicate to boolean_operator so it can
6105 be used to generate a crand.
6106 (eqne): New code iterator.
6107 (bd/bd_neg): New code_attrs.
6108 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
6109 a single define_insn.
6110 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
6111 decrement (bdnzt/bdnzf/bdzt/bdzf).
6112 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
6113 with the new names of the branch decrement patterns, and added the
6114 names of the branch decrement conditional patterns.
6116 2018-01-08 Richard Biener <rguenther@suse.de>
6118 PR tree-optimization/83563
6119 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
6122 2018-01-08 Richard Biener <rguenther@suse.de>
6125 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
6127 2018-01-08 Richard Biener <rguenther@suse.de>
6129 PR tree-optimization/83685
6130 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
6131 references to abnormals.
6133 2018-01-08 Richard Biener <rguenther@suse.de>
6136 * dwarf2out.c (output_indirect_strings): Handle empty
6137 skeleton_debug_str_hash.
6138 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
6140 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6142 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
6143 (emit_store_direct): Likewise.
6144 (arc_trampoline_adjust_address): Likewise.
6145 (arc_asm_trampoline_template): New function.
6146 (arc_initialize_trampoline): Use asm_trampoline_template.
6147 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
6148 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
6149 * config/arc/arc.md (flush_icache): Delete pattern.
6151 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6153 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
6154 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
6157 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6160 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
6161 by not USED_FOR_TARGET.
6162 (make_pass_resolve_sw_modes): Likewise.
6164 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6166 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
6169 2018-01-08 Richard Biener <rguenther@suse.de>
6172 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
6174 2018-01-08 Richard Biener <rguenther@suse.de>
6177 * match.pd ((t * 2) / 2) -> t): Add missing :c.
6179 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
6182 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
6183 basic blocks with a small number of successors.
6184 (convert_control_dep_chain_into_preds): Improve handling of
6186 (dump_predicates): Split apart into...
6187 (dump_pred_chain): ...here...
6188 (dump_pred_info): ...and here.
6189 (can_one_predicate_be_invalidated_p): Add debugging printfs.
6190 (can_chain_union_be_invalidated_p): Improve check for invalidation
6192 (uninit_uses_cannot_happen): Avoid unnecessary if
6193 convert_control_dep_chain_into_preds yielded nothing.
6195 2018-01-06 Martin Sebor <msebor@redhat.com>
6197 PR tree-optimization/83640
6198 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
6199 subtracting negative offset from size.
6200 (builtin_access::overlap): Adjust offset bounds of the access to fall
6201 within the size of the object if possible.
6203 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
6205 PR rtl-optimization/83699
6206 * expmed.c (extract_bit_field_1): Restrict the vector usage of
6207 extract_bit_field_as_subreg to cases in which the extracted
6208 value is also a vector.
6210 * lra-constraints.c (process_alt_operands): Test for the equivalence
6211 substitutions when detecting a possible reload cycle.
6213 2018-01-06 Jakub Jelinek <jakub@redhat.com>
6216 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
6217 by default if flag_selective_schedling{,2}. Formatting fixes.
6219 PR rtl-optimization/83682
6220 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
6221 if it has non-VECTOR_MODE element mode.
6222 (vec_duplicate_p): Likewise.
6225 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
6226 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
6228 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6231 * config/i386/i386-builtin.def
6232 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
6233 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
6234 Require also OPTION_MASK_ISA_AVX512F in addition to
6235 OPTION_MASK_ISA_GFNI.
6236 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
6237 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
6238 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
6239 to OPTION_MASK_ISA_GFNI.
6240 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
6241 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
6242 OPTION_MASK_ISA_AVX512BW.
6243 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
6244 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
6245 addition to OPTION_MASK_ISA_GFNI.
6246 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
6247 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
6248 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
6249 to OPTION_MASK_ISA_GFNI.
6250 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
6251 a requirement for all ISAs rather than any of them with a few
6253 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
6255 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
6256 bitmasks to be enabled with 3 exceptions, instead of requiring any
6257 enabled ISA with lots of exceptions.
6258 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
6259 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
6260 Change avx512bw in isa attribute to avx512f.
6261 * config/i386/sgxintrin.h: Add license boilerplate.
6262 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
6263 to __AVX512F__ and __AVX512VL to __AVX512VL__.
6264 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
6265 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
6267 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
6268 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
6269 temporarily sse2 rather than sse if not enabled already.
6272 * config/i386/sse.md (VI248_VLBW): Rename to ...
6273 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
6274 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
6275 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
6276 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
6277 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
6278 mode iterator instead of VI248_VLBW.
6280 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
6282 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
6283 (record_modified): Skip clobbers; add debug output.
6284 (param_change_prob): Use sreal frequencies.
6286 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6288 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
6289 punt for user-aligned variables.
6291 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6293 * tree-chrec.c (chrec_contains_symbols): Return true for
6296 2018-01-05 Sudakshina Das <sudi.das@arm.com>
6299 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
6300 of (x|y) == x for BICS pattern.
6302 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6304 PR tree-optimization/83605
6305 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
6306 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
6309 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
6311 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
6312 * config/epiphany/rtems.h: New file.
6314 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6315 Uros Bizjak <ubizjak@gmail.com>
6318 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
6319 QIreg_operand instead of register_operand predicate.
6320 * config/i386/i386.c (ix86_rop_should_change_byte_p,
6321 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
6322 comments instead of -fmitigate[-_]rop.
6324 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6327 * cgraphunit.c (symbol_table::compile): Switch to text_section
6328 before calling assembly_start debug hook.
6329 * run-rtl-passes.c (run_rtl_passes): Likewise.
6332 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6334 * tree-vrp.c (extract_range_from_binary_expr_1): Check
6335 range_int_cst_p rather than !symbolic_range_p before calling
6336 extract_range_from_multiplicative_op_1.
6338 2018-01-04 Jeff Law <law@redhat.com>
6340 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
6341 redundant test in assertion.
6343 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6345 * doc/rtl.texi: Document machine_mode wrapper classes.
6347 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6349 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
6352 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6354 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
6355 the VEC_PERM_EXPR fold to fail.
6357 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6360 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
6361 to switched_sections.
6363 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6366 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
6369 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
6372 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
6373 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
6375 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6378 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
6379 is BLKmode and bitpos not zero or mode change is needed.
6381 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6384 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
6387 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
6390 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
6391 instead of MULT rtx. Update all corresponding splitters.
6393 (*ssub<modesuffix>): Ditto.
6395 (*cmp_sadd_di): Update split patterns.
6396 (*cmp_sadd_si): Ditto.
6397 (*cmp_sadd_sidi): Ditto.
6398 (*cmp_ssub_di): Ditto.
6399 (*cmp_ssub_si): Ditto.
6400 (*cmp_ssub_sidi): Ditto.
6401 * config/alpha/predicates.md (const23_operand): New predicate.
6402 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
6403 Look for ASHIFT, not MULT inner operand.
6404 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
6406 2018-01-04 Martin Liska <mliska@suse.cz>
6408 PR gcov-profile/83669
6409 * gcov.c (output_intermediate_file): Add version to intermediate
6411 * doc/gcov.texi: Document new field 'version' in intermediate
6412 file format. Fix location of '-k' option of gcov command.
6414 2018-01-04 Martin Liska <mliska@suse.cz>
6417 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
6419 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6421 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
6423 2018-01-03 Martin Sebor <msebor@redhat.com>
6425 PR tree-optimization/83655
6426 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
6427 checking calls with invalid arguments.
6429 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6431 * tree-vect-stmts.c (vect_get_store_rhs): New function.
6432 (vectorizable_mask_load_store): Delete.
6433 (vectorizable_call): Return false for masked loads and stores.
6434 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
6435 instead of gimple_assign_rhs1.
6436 (vectorizable_load): Handle IFN_MASK_LOAD.
6437 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
6439 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6441 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
6443 (vectorizable_mask_load_store): ...here.
6444 (vectorizable_load): ...and here.
6446 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6448 * tree-vect-stmts.c (vect_build_all_ones_mask)
6449 (vect_build_zero_merge_argument): New functions, split out from...
6450 (vectorizable_load): ...here.
6452 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6454 * tree-vect-stmts.c (vect_check_store_rhs): New function,
6456 (vectorizable_mask_load_store): ...here.
6457 (vectorizable_store): ...and here.
6459 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6461 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
6463 (vectorizable_mask_load_store): ...here.
6465 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6467 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
6468 (vect_model_store_cost): Take a vec_load_store_type instead of a
6470 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
6471 (vect_model_store_cost): Take a vec_load_store_type instead of a
6473 (vectorizable_mask_load_store): Update accordingly.
6474 (vectorizable_store): Likewise.
6475 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
6477 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6479 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
6480 IFN_MASK_LOAD calls here rather than...
6481 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
6483 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6484 Alan Hayward <alan.hayward@arm.com>
6485 David Sherwood <david.sherwood@arm.com>
6487 * expmed.c (extract_bit_field_1): For vector extracts,
6488 fall back to extract_bit_field_as_subreg if vec_extract
6491 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6492 Alan Hayward <alan.hayward@arm.com>
6493 David Sherwood <david.sherwood@arm.com>
6495 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
6496 they are variable or constant sized.
6497 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
6498 slots for constant-sized data.
6500 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6501 Alan Hayward <alan.hayward@arm.com>
6502 David Sherwood <david.sherwood@arm.com>
6504 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
6505 handling COND_EXPRs with boolean comparisons, try to find a better
6506 basis for the mask type than the boolean itself.
6508 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6510 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
6511 is calculated and how it can be overridden.
6512 * genmodes.c (max_bitsize_mode_any_mode): New variable.
6513 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
6515 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
6518 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6519 Alan Hayward <alan.hayward@arm.com>
6520 David Sherwood <david.sherwood@arm.com>
6522 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
6523 Remove the mode argument.
6524 (aarch64_simd_valid_immediate): Remove the mode and inverse
6526 * config/aarch64/iterators.md (bitsize): New iterator.
6527 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
6528 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
6529 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
6530 aarch64_simd_valid_immediate.
6531 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
6532 (aarch64_reg_or_bic_imm): Likewise.
6533 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
6534 with an insn_type enum and msl with a modifier_type enum.
6535 Replace element_width with a scalar_mode. Change the shift
6536 to unsigned int. Add constructors for scalar_float_mode and
6537 scalar_int_mode elements.
6538 (aarch64_vect_float_const_representable_p): Delete.
6539 (aarch64_can_const_movi_rtx_p)
6540 (aarch64_simd_scalar_immediate_valid_for_move)
6541 (aarch64_simd_make_constant): Update call to
6542 aarch64_simd_valid_immediate.
6543 (aarch64_advsimd_valid_immediate_hs): New function.
6544 (aarch64_advsimd_valid_immediate): Likewise.
6545 (aarch64_simd_valid_immediate): Remove mode and inverse
6546 arguments. Rewrite to use the above. Use const_vec_duplicate_p
6547 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
6548 and aarch64_float_const_representable_p on the result.
6549 (aarch64_output_simd_mov_immediate): Remove mode argument.
6550 Update call to aarch64_simd_valid_immediate and use of
6551 simd_immediate_info.
6552 (aarch64_output_scalar_simd_mov_immediate): Update call
6555 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6556 Alan Hayward <alan.hayward@arm.com>
6557 David Sherwood <david.sherwood@arm.com>
6559 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
6560 (mode_nunits): Likewise CONST_MODE_NUNITS.
6561 * machmode.def (ADJUST_NUNITS): Document.
6562 * genmodes.c (mode_data::need_nunits_adj): New field.
6563 (blank_mode): Update accordingly.
6564 (adj_nunits): New variable.
6565 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
6567 (emit_mode_size_inline): Set need_bytesize_adj for all modes
6568 listed in adj_nunits.
6569 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
6570 listed in adj_nunits. Don't emit case statements for such modes.
6571 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
6572 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
6573 nothing if adj_nunits is nonnull.
6574 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
6575 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
6576 (emit_mode_fbit): Update use of print_maybe_const_decl.
6577 (emit_move_size): Likewise. Treat the array as non-const
6579 (emit_mode_adjustments): Handle adj_nunits.
6581 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6583 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
6584 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
6585 (VECTOR_MODES): Use it.
6586 (make_vector_modes): Take the prefix as an argument.
6588 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6589 Alan Hayward <alan.hayward@arm.com>
6590 David Sherwood <david.sherwood@arm.com>
6592 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
6593 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
6594 for MODE_VECTOR_BOOL.
6595 * machmode.def (VECTOR_BOOL_MODE): Document.
6596 * genmodes.c (VECTOR_BOOL_MODE): New macro.
6597 (make_vector_bool_mode): New function.
6598 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
6600 * lto-streamer-in.c (lto_input_mode_table): Likewise.
6601 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
6603 * stor-layout.c (int_mode_for_mode): Likewise.
6604 * tree.c (build_vector_type_for_mode): Likewise.
6605 * varasm.c (output_constant_pool_2): Likewise.
6606 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
6607 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
6608 for MODE_VECTOR_BOOL.
6609 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
6610 of mode class checks.
6611 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
6612 instead of a list of mode class checks.
6613 (expand_vector_scalar_condition): Likewise.
6614 (type_for_widest_vector_mode): Handle BImode as an inner mode.
6616 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6617 Alan Hayward <alan.hayward@arm.com>
6618 David Sherwood <david.sherwood@arm.com>
6620 * machmode.h (mode_size): Change from unsigned short to
6622 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
6623 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6624 or if measurement_type is not polynomial.
6625 (fixed_size_mode::includes_p): Check for constant-sized modes.
6626 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
6627 return a poly_uint16 rather than an unsigned short.
6628 (emit_mode_size): Change the type of mode_size from unsigned short
6629 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
6630 (emit_mode_adjustments): Cope with polynomial vector sizes.
6631 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6633 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6635 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
6636 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
6637 * caller-save.c (setup_save_areas): Likewise.
6638 (replace_reg_with_saved_mem): Likewise.
6639 * calls.c (emit_library_call_value_1): Likewise.
6640 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
6641 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
6642 (gen_lowpart_for_combine): Likewise.
6643 * convert.c (convert_to_integer_1): Likewise.
6644 * cse.c (equiv_constant, cse_insn): Likewise.
6645 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
6646 (cselib_subst_to_values): Likewise.
6647 * dce.c (word_dce_process_block): Likewise.
6648 * df-problems.c (df_word_lr_mark_ref): Likewise.
6649 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
6650 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
6651 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
6652 (rtl_for_decl_location): Likewise.
6653 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
6654 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
6655 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
6656 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
6657 (expand_expr_real_1): Likewise.
6658 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
6659 (pad_below): Likewise.
6660 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6661 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
6662 * ira.c (get_subreg_tracking_sizes): Likewise.
6663 * ira-build.c (ira_create_allocno_objects): Likewise.
6664 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
6665 (ira_sort_regnos_for_alter_reg): Likewise.
6666 * ira-costs.c (record_operand_costs): Likewise.
6667 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
6668 (resolve_simple_move): Likewise.
6669 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
6670 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
6671 (lra_constraints): Likewise.
6672 (CONST_POOL_OK_P): Reject variable-sized modes.
6673 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
6674 (add_pseudo_to_slot, lra_spill): Likewise.
6675 * omp-low.c (omp_clause_aligned_alignment): Likewise.
6676 * optabs-query.c (get_best_extraction_insn): Likewise.
6677 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6678 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
6679 (expand_mult_highpart, valid_multiword_target_p): Likewise.
6680 * recog.c (offsettable_address_addr_space_p): Likewise.
6681 * regcprop.c (maybe_mode_change): Likewise.
6682 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
6683 * regrename.c (build_def_use): Likewise.
6684 * regstat.c (dump_reg_info): Likewise.
6685 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
6686 (find_reloads, find_reloads_subreg_address): Likewise.
6687 * reload1.c (eliminate_regs_1): Likewise.
6688 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
6689 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
6690 (simplify_binary_operation_1, simplify_subreg): Likewise.
6691 * targhooks.c (default_function_arg_padding): Likewise.
6692 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
6693 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
6694 (verify_gimple_assign_ternary): Likewise.
6695 * tree-inline.c (estimate_move_cost): Likewise.
6696 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6697 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
6698 (get_address_cost_ainc): Likewise.
6699 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
6700 (vect_supportable_dr_alignment): Likewise.
6701 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6702 (vectorizable_reduction): Likewise.
6703 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
6704 (vectorizable_operation, vectorizable_load): Likewise.
6705 * tree.c (build_same_sized_truth_vector_type): Likewise.
6706 * valtrack.c (cleanup_auto_inc_dec): Likewise.
6707 * var-tracking.c (emit_note_insn_var_location): Likewise.
6708 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
6709 (ADDR_VEC_ALIGN): Likewise.
6711 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6712 Alan Hayward <alan.hayward@arm.com>
6713 David Sherwood <david.sherwood@arm.com>
6715 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
6717 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6718 or if measurement_type is polynomial.
6719 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
6720 * combine.c (make_extraction): Likewise.
6721 * dse.c (find_shift_sequence): Likewise.
6722 * dwarf2out.c (mem_loc_descriptor): Likewise.
6723 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
6724 (extract_bit_field, extract_low_bits): Likewise.
6725 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
6726 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
6727 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
6728 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
6729 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6730 * reload.c (find_reloads): Likewise.
6731 * reload1.c (alter_reg): Likewise.
6732 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
6733 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
6734 * tree-if-conv.c (predicate_mem_writes): Likewise.
6735 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
6736 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
6737 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
6738 * valtrack.c (dead_debug_insert_temp): Likewise.
6739 * varasm.c (mergeable_constant_section): Likewise.
6740 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
6742 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6743 Alan Hayward <alan.hayward@arm.com>
6744 David Sherwood <david.sherwood@arm.com>
6746 * expr.c (expand_assignment): Cope with polynomial mode sizes
6747 when assigning to a CONCAT.
6749 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6750 Alan Hayward <alan.hayward@arm.com>
6751 David Sherwood <david.sherwood@arm.com>
6753 * machmode.h (mode_precision): Change from unsigned short to
6755 (mode_to_precision): Return a poly_uint16 rather than an unsigned
6757 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
6758 or if measurement_type is not polynomial.
6759 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
6760 in which the mode is already known to be a scalar_int_mode.
6761 * genmodes.c (emit_mode_precision): Change the type of mode_precision
6762 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
6764 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6765 for GET_MODE_PRECISION.
6766 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6767 for GET_MODE_PRECISION.
6768 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
6770 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
6771 (expand_field_assignment, make_extraction): Likewise.
6772 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
6773 (get_last_value): Likewise.
6774 * convert.c (convert_to_integer_1): Likewise.
6775 * cse.c (cse_insn): Likewise.
6776 * expr.c (expand_expr_real_1): Likewise.
6777 * lra-constraints.c (simplify_operand_subreg): Likewise.
6778 * optabs-query.c (can_atomic_load_p): Likewise.
6779 * optabs.c (expand_atomic_load): Likewise.
6780 (expand_atomic_store): Likewise.
6781 * ree.c (combine_reaching_defs): Likewise.
6782 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
6783 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
6784 * tree.h (type_has_mode_precision_p): Likewise.
6785 * ubsan.c (instrument_si_overflow): Likewise.
6787 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6788 Alan Hayward <alan.hayward@arm.com>
6789 David Sherwood <david.sherwood@arm.com>
6791 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
6792 polynomial numbers of units.
6793 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
6794 (valid_vector_subparts_p): New function.
6795 (build_vector_type): Remove temporary shim and take the number
6796 of units as a poly_uint64 rather than an int.
6797 (build_opaque_vector_type): Take the number of units as a
6798 poly_uint64 rather than an int.
6799 * tree.c (build_vector_from_ctor): Handle polynomial
6800 TYPE_VECTOR_SUBPARTS.
6801 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
6802 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
6803 (build_vector_from_val): If the number of units is variable,
6804 use build_vec_duplicate_cst for constant operands and
6805 VEC_DUPLICATE_EXPR otherwise.
6806 (make_vector_type): Remove temporary is_constant ().
6807 (build_vector_type, build_opaque_vector_type): Take the number of
6808 units as a poly_uint64 rather than an int.
6809 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
6811 * cfgexpand.c (expand_debug_expr): Likewise.
6812 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
6813 (store_constructor, expand_expr_real_1): Likewise.
6814 (const_scalar_mask_from_tree): Likewise.
6815 * fold-const-call.c (fold_const_reduction): Likewise.
6816 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
6817 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
6818 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
6819 (fold_relational_const): Likewise.
6820 (native_interpret_vector): Likewise. Change the size from an
6821 int to an unsigned int.
6822 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
6823 TYPE_VECTOR_SUBPARTS.
6824 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
6825 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
6826 duplicating a non-constant operand into a variable-length vector.
6827 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
6828 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
6829 * ipa-icf.c (sem_variable::equals): Likewise.
6830 * match.pd: Likewise.
6831 * omp-simd-clone.c (simd_clone_subparts): Likewise.
6832 * print-tree.c (print_node): Likewise.
6833 * stor-layout.c (layout_type): Likewise.
6834 * targhooks.c (default_builtin_vectorization_cost): Likewise.
6835 * tree-cfg.c (verify_gimple_comparison): Likewise.
6836 (verify_gimple_assign_binary): Likewise.
6837 (verify_gimple_assign_ternary): Likewise.
6838 (verify_gimple_assign_single): Likewise.
6839 * tree-pretty-print.c (dump_generic_node): Likewise.
6840 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6841 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
6842 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
6843 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
6844 (vect_shift_permute_load_chain): Likewise.
6845 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
6846 (expand_vector_condition, optimize_vector_constructor): Likewise.
6847 (lower_vec_perm, get_compute_type): Likewise.
6848 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6849 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
6850 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
6851 (vect_recog_mask_conversion_pattern): Likewise.
6852 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
6853 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
6854 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
6855 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
6856 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
6857 (vectorizable_shift, vectorizable_operation, vectorizable_store)
6858 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
6859 (supportable_widening_operation): Likewise.
6860 (supportable_narrowing_operation): Likewise.
6861 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
6863 * varasm.c (output_constant): Likewise.
6865 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6866 Alan Hayward <alan.hayward@arm.com>
6867 David Sherwood <david.sherwood@arm.com>
6869 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
6870 so that both the length == 3 and length != 3 cases set up their
6871 own permute vectors. Add comments explaining why we know the
6872 number of elements is constant.
6873 (vect_permute_load_chain): Likewise.
6875 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6876 Alan Hayward <alan.hayward@arm.com>
6877 David Sherwood <david.sherwood@arm.com>
6879 * machmode.h (mode_nunits): Change from unsigned char to
6881 (ONLY_FIXED_SIZE_MODES): New macro.
6882 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
6883 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
6884 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
6886 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
6887 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
6888 or if measurement_type is not polynomial.
6889 * genmodes.c (ZERO_COEFFS): New macro.
6890 (emit_mode_nunits_inline): Make mode_nunits_inline return a
6892 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
6893 Use ZERO_COEFFS when emitting initializers.
6894 * data-streamer.h (bp_pack_poly_value): New function.
6895 (bp_unpack_poly_value): Likewise.
6896 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6897 for GET_MODE_NUNITS.
6898 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6899 for GET_MODE_NUNITS.
6900 * tree.c (make_vector_type): Remove temporary shim and make
6901 the real function take the number of units as a poly_uint64
6903 (build_vector_type_for_mode): Handle polynomial nunits.
6904 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
6905 * emit-rtl.c (const_vec_series_p_1): Likewise.
6906 (gen_rtx_CONST_VECTOR): Likewise.
6907 * fold-const.c (test_vec_duplicate_folding): Likewise.
6908 * genrecog.c (validate_pattern): Likewise.
6909 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
6910 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6911 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
6912 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
6913 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
6914 * rtlanal.c (subreg_get_info): Likewise.
6915 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
6916 (vect_grouped_load_supported): Likewise.
6917 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
6918 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
6919 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
6920 (simplify_const_unary_operation, simplify_binary_operation_1)
6921 (simplify_const_binary_operation, simplify_ternary_operation)
6922 (test_vector_ops_duplicate, test_vector_ops): Likewise.
6923 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
6924 instead of CONST_VECTOR_NUNITS.
6925 * varasm.c (output_constant_pool_2): Likewise.
6926 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
6927 explicit-encoded elements in the XVEC for variable-length vectors.
6929 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6931 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
6933 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6934 Alan Hayward <alan.hayward@arm.com>
6935 David Sherwood <david.sherwood@arm.com>
6937 * coretypes.h (fixed_size_mode): Declare.
6938 (fixed_size_mode_pod): New typedef.
6939 * builtins.h (target_builtins::x_apply_args_mode)
6940 (target_builtins::x_apply_result_mode): Change type to
6941 fixed_size_mode_pod.
6942 * builtins.c (apply_args_size, apply_result_size, result_vector)
6943 (expand_builtin_apply_args_1, expand_builtin_apply)
6944 (expand_builtin_return): Update accordingly.
6946 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6948 * cse.c (hash_rtx_cb): Hash only the encoded elements.
6949 * cselib.c (cselib_hash_rtx): Likewise.
6950 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
6951 CONST_VECTOR encoding.
6953 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6954 Jeff Law <law@redhat.com>
6957 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
6958 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
6959 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
6960 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
6963 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
6964 explicitly probe *sp in a noreturn function if there were any callee
6965 register saves or frame pointer is needed.
6967 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6970 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
6971 BLKmode for ternary, binary or unary expressions.
6974 * var-tracking.c (delete_vta_debug_insn): New inline function.
6975 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
6976 insns from get_insns () to NULL instead of each bb separately.
6977 Use delete_vta_debug_insn. No longer static.
6978 (vt_debug_insns_local, variable_tracking_main_1): Adjust
6979 delete_vta_debug_insns callers.
6980 * rtl.h (delete_vta_debug_insns): Declare.
6981 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
6982 instead of variable_tracking_main.
6984 2018-01-03 Martin Sebor <msebor@redhat.com>
6986 PR tree-optimization/83603
6987 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
6988 arguments past the endof the argument list in functions declared
6989 without a prototype.
6990 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
6991 Avoid checking when arguments are null.
6993 2018-01-03 Martin Sebor <msebor@redhat.com>
6996 * doc/extend.texi (attribute const): Fix a typo.
6997 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
6998 issuing -Wsuggest-attribute for void functions.
7000 2018-01-03 Martin Sebor <msebor@redhat.com>
7002 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
7003 offset_int::from instead of wide_int::to_shwi.
7004 (maybe_diag_overlap): Remove assertion.
7005 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
7006 * gimple-ssa-sprintf.c (format_directive): Same.
7007 (parse_directive): Same.
7008 (sprintf_dom_walker::compute_format_length): Same.
7009 (try_substitute_return_value): Same.
7011 2018-01-03 Jeff Law <law@redhat.com>
7014 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
7015 non-constant residual for zero at runtime and avoid probing in
7016 that case. Reorganize code for trailing problem to mirror handling
7019 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7021 PR tree-optimization/83501
7022 * tree-ssa-strlen.c (get_string_cst): New.
7023 (handle_char_store): Call get_string_cst.
7025 2018-01-03 Martin Liska <mliska@suse.cz>
7027 PR tree-optimization/83593
7028 * tree-ssa-strlen.c: Include tree-cfg.h.
7029 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
7030 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
7031 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
7033 (strlen_dom_walker::before_dom_children): Call
7034 gimple_purge_dead_eh_edges. Dump tranformation with details
7036 (strlen_dom_walker::before_dom_children): Update call by adding
7037 new argument cleanup_eh.
7038 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
7040 2018-01-03 Martin Liska <mliska@suse.cz>
7043 * cif-code.def (VARIADIC_THUNK): New enum value.
7044 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
7047 2018-01-03 Jan Beulich <jbeulich@suse.com>
7049 * sse.md (mov<mode>_internal): Tighten condition for when to use
7050 vmovdqu<ssescalarsize> for TI and OI modes.
7052 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7054 Update copyright years.
7056 2018-01-03 Martin Liska <mliska@suse.cz>
7059 * ipa-visibility.c (function_and_variable_visibility): Skip
7060 functions with noipa attribure.
7062 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7064 * gcc.c (process_command): Update copyright notice dates.
7065 * gcov-dump.c (print_version): Ditto.
7066 * gcov.c (print_version): Ditto.
7067 * gcov-tool.c (print_version): Ditto.
7068 * gengtype.c (create_file): Ditto.
7069 * doc/cpp.texi: Bump @copying's copyright year.
7070 * doc/cppinternals.texi: Ditto.
7071 * doc/gcc.texi: Ditto.
7072 * doc/gccint.texi: Ditto.
7073 * doc/gcov.texi: Ditto.
7074 * doc/install.texi: Ditto.
7075 * doc/invoke.texi: Ditto.
7077 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7079 * vector-builder.h (vector_builder::m_full_nelts): Change from
7080 unsigned int to poly_uint64.
7081 (vector_builder::full_nelts): Update prototype accordingly.
7082 (vector_builder::new_vector): Likewise.
7083 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
7084 (vector_builder::operator ==): Likewise.
7085 (vector_builder::finalize): Likewise.
7086 * int-vector-builder.h (int_vector_builder::int_vector_builder):
7087 Take the number of elements as a poly_uint64 rather than an
7089 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
7090 from unsigned int to poly_uint64.
7091 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
7092 (vec_perm_indices::new_vector): Likewise.
7093 (vec_perm_indices::length): Likewise.
7094 (vec_perm_indices::nelts_per_input): Likewise.
7095 (vec_perm_indices::input_nelts): Likewise.
7096 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
7097 number of elements per input as a poly_uint64 rather than an
7098 unsigned int. Use the original encoding for variable-length
7099 vectors, rather than clamping each individual element.
7100 For the second and subsequent elements in each pattern,
7101 clamp the step and base before clamping their sum.
7102 (vec_perm_indices::series_p): Handle polynomial element counts.
7103 (vec_perm_indices::all_in_range_p): Likewise.
7104 (vec_perm_indices_to_tree): Likewise.
7105 (vec_perm_indices_to_rtx): Likewise.
7106 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
7107 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
7108 (tree_vector_builder::new_binary_operation): Handle polynomial
7109 element counts. Return false if we need to know the number
7110 of elements at compile time.
7111 * fold-const.c (fold_vec_perm): Punt if the number of elements
7112 isn't known at compile time.
7114 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7116 * vec-perm-indices.h (vec_perm_builder): Change element type
7117 from HOST_WIDE_INT to poly_int64.
7118 (vec_perm_indices::element_type): Update accordingly.
7119 (vec_perm_indices::clamp): Handle polynomial element_types.
7120 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7121 (vec_perm_indices::all_in_range_p): Likewise.
7122 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
7124 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
7125 polynomial vec_perm_indices element types.
7126 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
7127 * fold-const.c (fold_vec_perm): Likewise.
7128 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
7129 * tree-vect-generic.c (lower_vec_perm): Likewise.
7130 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7131 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
7132 element type to HOST_WIDE_INT.
7134 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7135 Alan Hayward <alan.hayward@arm.com>
7136 David Sherwood <david.sherwood@arm.com>
7138 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
7139 rather than an int. Use plus_constant.
7140 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
7141 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
7143 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7144 Alan Hayward <alan.hayward@arm.com>
7145 David Sherwood <david.sherwood@arm.com>
7147 * calls.c (emit_call_1, expand_call): Change struct_value_size from
7148 a HOST_WIDE_INT to a poly_int64.
7150 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7151 Alan Hayward <alan.hayward@arm.com>
7152 David Sherwood <david.sherwood@arm.com>
7154 * calls.c (load_register_parameters): Cope with polynomial
7155 mode sizes. Require a constant size for BLKmode parameters
7156 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
7157 forces a parameter to be padded at the lsb end in order to
7158 fill a complete number of words, require the parameter size
7159 to be ordered wrt UNITS_PER_WORD.
7161 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7162 Alan Hayward <alan.hayward@arm.com>
7163 David Sherwood <david.sherwood@arm.com>
7165 * reload1.c (spill_stack_slot_width): Change element type
7166 from unsigned int to poly_uint64_pod.
7167 (alter_reg): Treat mode sizes as polynomial.
7169 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7170 Alan Hayward <alan.hayward@arm.com>
7171 David Sherwood <david.sherwood@arm.com>
7173 * reload.c (complex_word_subreg_p): New function.
7174 (reload_inner_reg_of_subreg, push_reload): Use it.
7176 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7177 Alan Hayward <alan.hayward@arm.com>
7178 David Sherwood <david.sherwood@arm.com>
7180 * lra-constraints.c (process_alt_operands): Reject matched
7181 operands whose sizes aren't ordered.
7182 (match_reload): Refer to this check here.
7184 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7185 Alan Hayward <alan.hayward@arm.com>
7186 David Sherwood <david.sherwood@arm.com>
7188 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
7189 that the mode size is in the set {1, 2, 4, 8, 16}.
7191 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7192 Alan Hayward <alan.hayward@arm.com>
7193 David Sherwood <david.sherwood@arm.com>
7195 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
7196 Use plus_constant instead of gen_rtx_PLUS.
7198 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7199 Alan Hayward <alan.hayward@arm.com>
7200 David Sherwood <david.sherwood@arm.com>
7202 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
7203 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
7204 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
7205 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
7206 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
7207 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
7208 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
7209 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
7210 * config/i386/i386.c (ix86_push_rounding): ...this new function.
7211 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
7213 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
7214 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
7215 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
7216 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
7217 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
7218 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
7219 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
7220 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
7221 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
7222 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
7224 * expr.c (emit_move_resolve_push): Treat the input and result
7225 of PUSH_ROUNDING as a poly_int64.
7226 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
7227 (emit_push_insn): Likewise.
7228 * lra-eliminations.c (mark_not_eliminable): Likewise.
7229 * recog.c (push_operand): Likewise.
7230 * reload1.c (elimination_effects): Likewise.
7231 * rtlanal.c (nonzero_bits1): Likewise.
7232 * calls.c (store_one_arg): Likewise. Require the padding to be
7233 known at compile time.
7235 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7236 Alan Hayward <alan.hayward@arm.com>
7237 David Sherwood <david.sherwood@arm.com>
7239 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
7240 Use plus_constant instead of gen_rtx_PLUS.
7242 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7243 Alan Hayward <alan.hayward@arm.com>
7244 David Sherwood <david.sherwood@arm.com>
7246 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
7249 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7250 Alan Hayward <alan.hayward@arm.com>
7251 David Sherwood <david.sherwood@arm.com>
7253 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
7254 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
7255 via stack temporaries. Treat the mode size as polynomial too.
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 * expr.c (expand_expr_real_2): When handling conversions involving
7262 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
7263 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
7264 as a poly_uint64 too.
7266 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7267 Alan Hayward <alan.hayward@arm.com>
7268 David Sherwood <david.sherwood@arm.com>
7270 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
7272 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7273 Alan Hayward <alan.hayward@arm.com>
7274 David Sherwood <david.sherwood@arm.com>
7276 * combine.c (can_change_dest_mode): Handle polynomial
7277 REGMODE_NATURAL_SIZE.
7278 * expmed.c (store_bit_field_1): Likewise.
7279 * expr.c (store_constructor): Likewise.
7280 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
7281 and polynomial REGMODE_NATURAL_SIZE.
7282 (gen_lowpart_common): Likewise.
7283 * reginfo.c (record_subregs_of_mode): Likewise.
7284 * rtlanal.c (read_modify_subreg_p): Likewise.
7286 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7287 Alan Hayward <alan.hayward@arm.com>
7288 David Sherwood <david.sherwood@arm.com>
7290 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
7291 numbers of elements.
7293 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7294 Alan Hayward <alan.hayward@arm.com>
7295 David Sherwood <david.sherwood@arm.com>
7297 * match.pd: Cope with polynomial numbers of vector elements.
7299 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7300 Alan Hayward <alan.hayward@arm.com>
7301 David Sherwood <david.sherwood@arm.com>
7303 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
7304 in a POINTER_PLUS_EXPR.
7306 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7307 Alan Hayward <alan.hayward@arm.com>
7308 David Sherwood <david.sherwood@arm.com>
7310 * omp-simd-clone.c (simd_clone_subparts): New function.
7311 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
7312 (ipa_simd_modify_function_body): Likewise.
7314 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7315 Alan Hayward <alan.hayward@arm.com>
7316 David Sherwood <david.sherwood@arm.com>
7318 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
7319 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
7320 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
7321 (expand_vector_condition, vector_element): Likewise.
7322 (subparts_gt): New function.
7323 (get_compute_type): Use subparts_gt.
7324 (count_type_subparts): Delete.
7325 (expand_vector_operations_1): Use subparts_gt instead of
7326 count_type_subparts.
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-data-refs.c (vect_no_alias_p): Replace with...
7333 (vect_compile_time_alias): ...this new function. Do the calculation
7334 on poly_ints rather than trees.
7335 (vect_prune_runtime_alias_test_list): Update call accordingly.
7337 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7338 Alan Hayward <alan.hayward@arm.com>
7339 David Sherwood <david.sherwood@arm.com>
7341 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
7343 (vect_schedule_slp_instance): Likewise.
7345 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7346 Alan Hayward <alan.hayward@arm.com>
7347 David Sherwood <david.sherwood@arm.com>
7349 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
7350 constant and extern definitions for variable-length vectors.
7351 (vect_get_constant_vectors): Note that the number of units
7352 is known to be constant.
7354 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7355 Alan Hayward <alan.hayward@arm.com>
7356 David Sherwood <david.sherwood@arm.com>
7358 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
7359 of units as polynomial. Choose between WIDE and NARROW based
7362 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7363 Alan Hayward <alan.hayward@arm.com>
7364 David Sherwood <david.sherwood@arm.com>
7366 * tree-vect-stmts.c (simd_clone_subparts): New function.
7367 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
7369 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7370 Alan Hayward <alan.hayward@arm.com>
7371 David Sherwood <david.sherwood@arm.com>
7373 * tree-vect-stmts.c (vectorizable_call): Treat the number of
7374 vectors as polynomial. Use build_index_vector for
7377 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7378 Alan Hayward <alan.hayward@arm.com>
7379 David Sherwood <david.sherwood@arm.com>
7381 * tree-vect-stmts.c (get_load_store_type): Treat the number of
7382 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
7383 for variable-length vectors.
7384 (vectorizable_mask_load_store): Treat the number of units as
7385 polynomial, asserting that it is constant if the condition has
7386 already been enforced.
7387 (vectorizable_store, vectorizable_load): Likewise.
7389 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7390 Alan Hayward <alan.hayward@arm.com>
7391 David Sherwood <david.sherwood@arm.com>
7393 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
7394 of units as polynomial. Punt if we can't tell at compile time
7395 which vector contains the final result.
7397 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7398 Alan Hayward <alan.hayward@arm.com>
7399 David Sherwood <david.sherwood@arm.com>
7401 * tree-vect-loop.c (vectorizable_induction): Treat the number
7402 of units as polynomial. Punt on SLP inductions. Use an integer
7403 VEC_SERIES_EXPR for variable-length integer reductions. Use a
7404 cast of such a series for variable-length floating-point
7407 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7408 Alan Hayward <alan.hayward@arm.com>
7409 David Sherwood <david.sherwood@arm.com>
7411 * tree.h (build_index_vector): Declare.
7412 * tree.c (build_index_vector): New function.
7413 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
7414 of units as polynomial, forcibly converting it to a constant if
7415 vectorizable_reduction has already enforced the condition.
7416 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
7417 to create a {1,2,3,...} vector.
7418 (vectorizable_reduction): Treat the number of units as polynomial.
7419 Choose vectype_in based on the largest scalar element size rather
7420 than the smallest number of units. Enforce the restrictions
7423 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7424 Alan Hayward <alan.hayward@arm.com>
7425 David Sherwood <david.sherwood@arm.com>
7427 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
7428 number of units as polynomial.
7430 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7431 Alan Hayward <alan.hayward@arm.com>
7432 David Sherwood <david.sherwood@arm.com>
7434 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
7435 * target.def (autovectorize_vector_sizes): Return the vector sizes
7436 by pointer, using vector_sizes rather than a bitmask.
7437 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
7438 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
7439 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
7441 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
7442 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
7443 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
7444 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
7445 * omp-general.c (omp_max_vf): Likewise.
7446 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7447 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
7448 * tree-vect-loop.c (vect_analyze_loop): Likewise.
7449 * tree-vect-slp.c (vect_slp_bb): Likewise.
7450 * doc/tm.texi: Regenerate.
7451 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
7453 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
7454 the vector size as a poly_uint64 rather than an unsigned int.
7455 (current_vector_size): Change from an unsigned int to a poly_uint64.
7456 (get_vectype_for_scalar_type): Update accordingly.
7457 * tree.h (build_truth_vector_type): Take the size and number of
7458 units as a poly_uint64 rather than an unsigned int.
7459 (build_vector_type): Add a temporary overload that takes
7460 the number of units as a poly_uint64 rather than an unsigned int.
7461 * tree.c (make_vector_type): Likewise.
7462 (build_truth_vector_type): Take the number of units as a poly_uint64
7463 rather than an unsigned int.
7465 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7466 Alan Hayward <alan.hayward@arm.com>
7467 David Sherwood <david.sherwood@arm.com>
7469 * target.def (get_mask_mode): Take the number of units and length
7470 as poly_uint64s rather than unsigned ints.
7471 * targhooks.h (default_get_mask_mode): Update accordingly.
7472 * targhooks.c (default_get_mask_mode): Likewise.
7473 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
7474 * doc/tm.texi: Regenerate.
7476 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7477 Alan Hayward <alan.hayward@arm.com>
7478 David Sherwood <david.sherwood@arm.com>
7480 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
7481 * omp-general.c (omp_max_vf): Likewise.
7482 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
7483 (expand_omp_simd): Handle polynomial safelen.
7484 * omp-low.c (omplow_simd_context): Add a default constructor.
7485 (omplow_simd_context::max_vf): Change from int to poly_uint64.
7486 (lower_rec_simd_input_clauses): Update accordingly.
7487 (lower_rec_input_clauses): Likewise.
7489 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7490 Alan Hayward <alan.hayward@arm.com>
7491 David Sherwood <david.sherwood@arm.com>
7493 * tree-vectorizer.h (vect_nunits_for_cost): New function.
7494 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
7495 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
7496 (vect_analyze_slp_cost): Likewise.
7497 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
7498 (vect_model_load_cost): Likewise.
7500 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7501 Alan Hayward <alan.hayward@arm.com>
7502 David Sherwood <david.sherwood@arm.com>
7504 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
7505 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
7506 from an unsigned int * to a poly_uint64_pod *.
7507 (calculate_unrolling_factor): New function.
7508 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
7510 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7511 Alan Hayward <alan.hayward@arm.com>
7512 David Sherwood <david.sherwood@arm.com>
7514 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
7515 from an unsigned int to a poly_uint64.
7516 (_loop_vec_info::slp_unrolling_factor): Likewise.
7517 (_loop_vec_info::vectorization_factor): Change from an int
7519 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
7520 (vect_get_num_vectors): New function.
7521 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
7522 (vect_get_num_copies): Use vect_get_num_vectors.
7523 (vect_analyze_data_ref_dependences): Change max_vf from an int *
7524 to an unsigned int *.
7525 (vect_analyze_data_refs): Change min_vf from an int * to a
7527 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7528 than an unsigned HOST_WIDE_INT.
7529 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7530 (vect_analyze_data_ref_dependence): Change max_vf from an int *
7531 to an unsigned int *.
7532 (vect_analyze_data_ref_dependences): Likewise.
7533 (vect_compute_data_ref_alignment): Handle polynomial vf.
7534 (vect_enhance_data_refs_alignment): Likewise.
7535 (vect_prune_runtime_alias_test_list): Likewise.
7536 (vect_shift_permute_load_chain): Likewise.
7537 (vect_supportable_dr_alignment): Likewise.
7538 (dependence_distance_ge_vf): Take the vectorization factor as a
7539 poly_uint64 rather than an unsigned HOST_WIDE_INT.
7540 (vect_analyze_data_refs): Change min_vf from an int * to a
7542 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
7543 vfm1 as a poly_uint64 rather than an int. Make the same change
7544 for the returned bound_scalar.
7545 (vect_gen_vector_loop_niters): Handle polynomial vf.
7546 (vect_do_peeling): Likewise. Update call to
7547 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
7548 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
7550 * tree-vect-loop.c (vect_determine_vectorization_factor)
7551 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
7552 (vect_get_known_peeling_cost): Likewise.
7553 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
7554 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
7555 (vect_transform_loop): Likewise. Use the lowest possible VF when
7556 updating the upper bounds of the loop.
7557 (vect_min_worthwhile_factor): Make static. Return an unsigned int
7559 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
7560 polynomial unroll factors.
7561 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
7562 (vect_make_slp_decision): Likewise.
7563 (vect_supported_load_permutation_p): Likewise, and polynomial
7565 (vect_analyze_slp_cost): Handle polynomial vf.
7566 (vect_slp_analyze_node_operations): Likewise.
7567 (vect_slp_analyze_bb_1): Likewise.
7568 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7569 than an unsigned HOST_WIDE_INT.
7570 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
7571 (vectorizable_load): Handle polynomial vf.
7572 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
7574 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
7576 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7577 Alan Hayward <alan.hayward@arm.com>
7578 David Sherwood <david.sherwood@arm.com>
7580 * match.pd: Handle bit operations involving three constants
7581 and try to fold one pair.
7583 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7585 * tree-vect-loop-manip.c: Include gimple-fold.h.
7586 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
7587 niters_maybe_zero parameters. Handle other cases besides a step of 1.
7588 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
7589 Add a path that uses a step of VF instead of 1, but disable it
7591 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
7592 and niters_no_overflow parameters. Update calls to
7593 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
7594 Create a new SSA name if the latter choses to use a ste other
7595 than zero, and return it via niters_vector_mult_vf_var.
7596 * tree-vect-loop.c (vect_transform_loop): Update calls to
7597 vect_do_peeling, vect_gen_vector_loop_niters and
7598 slpeel_make_loop_iterate_ntimes.
7599 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
7600 (vect_gen_vector_loop_niters): Update declarations after above changes.
7602 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
7604 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
7605 128-bit round to integer instructions.
7606 (ceil<mode>2): Likewise.
7607 (btrunc<mode>2): Likewise.
7608 (round<mode>2): Likewise.
7610 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7612 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
7613 unaligned VSX load/store on P8/P9.
7614 (expand_block_clear): Allow the use of unaligned VSX
7615 load/store on P8/P9.
7617 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7619 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
7621 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
7622 swap associated with both a load and a store.
7624 2018-01-02 Andrew Waterman <andrew@sifive.com>
7626 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
7627 * config/riscv/riscv.md (clear_cache): Use it.
7629 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
7631 * web.c: Remove out-of-date comment.
7633 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7635 * expr.c (fixup_args_size_notes): Check that any existing
7636 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
7637 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
7638 (emit_single_push_insn): ...here.
7640 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7642 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
7643 (const_vector_encoded_nelts): New function.
7644 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
7645 (const_vector_int_elt, const_vector_elt): Declare.
7646 * emit-rtl.c (const_vector_int_elt_1): New function.
7647 (const_vector_elt): Likewise.
7648 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
7649 of CONST_VECTOR_ELT.
7651 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7653 * expr.c: Include rtx-vector-builder.h.
7654 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
7655 directly on the tree encoding.
7656 (const_vector_from_tree): Likewise.
7657 * optabs.c: Include rtx-vector-builder.h.
7658 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
7659 sequence of "u" values.
7660 * vec-perm-indices.c: Include rtx-vector-builder.h.
7661 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
7662 directly on the vec_perm_indices encoding.
7664 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7666 * doc/rtl.texi (const_vector): Describe new encoding scheme.
7667 * Makefile.in (OBJS): Add rtx-vector-builder.o.
7668 * rtx-vector-builder.h: New file.
7669 * rtx-vector-builder.c: Likewise.
7670 * rtl.h (rtx_def::u2): Add a const_vector field.
7671 (CONST_VECTOR_NPATTERNS): New macro.
7672 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
7673 (CONST_VECTOR_DUPLICATE_P): Likewise.
7674 (CONST_VECTOR_STEPPED_P): Likewise.
7675 (CONST_VECTOR_ENCODED_ELT): Likewise.
7676 (const_vec_duplicate_p): Check for a duplicated vector encoding.
7677 (unwrap_const_vec_duplicate): Likewise.
7678 (const_vec_series_p): Check for a non-duplicated vector encoding.
7679 Say that the function only returns true for integer vectors.
7680 * emit-rtl.c: Include rtx-vector-builder.h.
7681 (gen_const_vec_duplicate_1): Delete.
7682 (gen_const_vector): Call gen_const_vec_duplicate instead of
7683 gen_const_vec_duplicate_1.
7684 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
7685 (gen_const_vec_duplicate): Use rtx_vector_builder.
7686 (gen_const_vec_series): Likewise.
7687 (gen_rtx_CONST_VECTOR): Likewise.
7688 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
7689 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7690 Build a new vector rather than modifying a CONST_VECTOR in-place.
7691 (handle_special_swappables): Update call accordingly.
7692 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
7693 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7694 Build a new vector rather than modifying a CONST_VECTOR in-place.
7695 (handle_special_swappables): Update call accordingly.
7697 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7699 * simplify-rtx.c (simplify_const_binary_operation): Use
7700 CONST_VECTOR_ELT instead of XVECEXP.
7702 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7704 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
7705 the selector elements to be different from the data elements
7706 if the selector is a VECTOR_CST.
7707 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
7708 ssizetype for the selector.
7710 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7712 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
7713 before testing each element individually.
7714 * tree-vect-generic.c (lower_vec_perm): Likewise.
7716 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7718 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
7719 * selftest-run-tests.c (selftest::run_tests): Call it.
7720 * vector-builder.h (vector_builder::operator ==): New function.
7721 (vector_builder::operator !=): Likewise.
7722 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
7723 (vec_perm_indices::all_from_input_p): New function.
7724 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7725 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
7726 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
7727 instead of reading the VECTOR_CST directly. Detect whether both
7728 vector inputs are the same before constructing the vec_perm_indices,
7729 and update the number of inputs argument accordingly. Use the
7730 utility functions added above. Only construct sel2 if we need to.
7732 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7734 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
7735 the broadcast of the low byte.
7736 (expand_mult_highpart): Use an explicit encoding for the permutes.
7737 * optabs-query.c (can_mult_highpart_p): Likewise.
7738 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
7739 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7740 (vectorizable_bswap): Likewise.
7741 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
7742 explicit encoding for the power-of-2 permutes.
7743 (vect_permute_store_chain): Likewise.
7744 (vect_grouped_load_supported): Likewise.
7745 (vect_permute_load_chain): Likewise.
7747 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7749 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
7750 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
7751 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
7752 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7753 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
7754 (vect_gen_perm_mask_any): Likewise.
7756 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7758 * int-vector-builder.h: New file.
7759 * vec-perm-indices.h: Include int-vector-builder.h.
7760 (vec_perm_indices): Redefine as an int_vector_builder.
7761 (auto_vec_perm_indices): Delete.
7762 (vec_perm_builder): Redefine as a stand-alone class.
7763 (vec_perm_indices::vec_perm_indices): New function.
7764 (vec_perm_indices::clamp): Likewise.
7765 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
7766 (vec_perm_indices::new_vector): New function.
7767 (vec_perm_indices::new_expanded_vector): Update for new
7768 vec_perm_indices class.
7769 (vec_perm_indices::rotate_inputs): New function.
7770 (vec_perm_indices::all_in_range_p): Operate directly on the
7771 encoded form, without computing elided elements.
7772 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
7773 encoding. Update for new vec_perm_indices class.
7774 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
7775 the given vec_perm_builder.
7776 (expand_vec_perm_var): Update vec_perm_builder constructor.
7777 (expand_mult_highpart): Use vec_perm_builder instead of
7778 auto_vec_perm_indices.
7779 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
7780 vec_perm_indices instead of auto_vec_perm_indices. Use a single
7781 or double series encoding as appropriate.
7782 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
7783 vec_perm_indices instead of auto_vec_perm_indices.
7784 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7785 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7786 (vect_permute_store_chain): Likewise.
7787 (vect_grouped_load_supported): Likewise.
7788 (vect_permute_load_chain): Likewise.
7789 (vect_shift_permute_load_chain): Likewise.
7790 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
7791 (vect_transform_slp_perm_load): Likewise.
7792 (vect_schedule_slp_instance): Likewise.
7793 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7794 (vectorizable_mask_load_store): Likewise.
7795 (vectorizable_bswap): Likewise.
7796 (vectorizable_store): Likewise.
7797 (vectorizable_load): Likewise.
7798 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
7799 vec_perm_indices instead of auto_vec_perm_indices. Use
7800 tree_to_vec_perm_builder to read the vector from a tree.
7801 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
7802 vec_perm_builder instead of a vec_perm_indices.
7803 (have_whole_vector_shift): Use vec_perm_builder and
7804 vec_perm_indices instead of auto_vec_perm_indices. Leave the
7805 truncation to calc_vec_perm_mask_for_shift.
7806 (vect_create_epilog_for_reduction): Likewise.
7807 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
7808 from auto_vec_perm_indices to vec_perm_indices.
7809 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7810 instead of changing individual elements.
7811 (aarch64_vectorize_vec_perm_const): Use new_vector to install
7812 the vector in d.perm.
7813 * config/arm/arm.c (expand_vec_perm_d::perm): Change
7814 from auto_vec_perm_indices to vec_perm_indices.
7815 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7816 instead of changing individual elements.
7817 (arm_vectorize_vec_perm_const): Use new_vector to install
7818 the vector in d.perm.
7819 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
7820 Update vec_perm_builder constructor.
7821 (rs6000_expand_interleave): Likewise.
7822 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
7823 (rs6000_expand_interleave): Likewise.
7825 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7827 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
7828 to qimode could truncate the indices.
7829 * optabs.c (expand_vec_perm_var): Likewise.
7831 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7833 * Makefile.in (OBJS): Add vec-perm-indices.o.
7834 * vec-perm-indices.h: New file.
7835 * vec-perm-indices.c: Likewise.
7836 * target.h (vec_perm_indices): Replace with a forward class
7838 (auto_vec_perm_indices): Move to vec-perm-indices.h.
7839 * optabs.h: Include vec-perm-indices.h.
7840 (expand_vec_perm): Delete.
7841 (selector_fits_mode_p, expand_vec_perm_var): Declare.
7842 (expand_vec_perm_const): Declare.
7843 * target.def (vec_perm_const_ok): Replace with...
7844 (vec_perm_const): ...this new hook.
7845 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
7846 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
7847 * doc/tm.texi: Regenerate.
7848 * optabs.def (vec_perm_const): Delete.
7849 * doc/md.texi (vec_perm_const): Likewise.
7850 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
7851 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
7852 expand_vec_perm for constant permutation vectors. Assert that
7853 the mode of variable permutation vectors is the integer equivalent
7854 of the mode that is being permuted.
7855 * optabs-query.h (selector_fits_mode_p): Declare.
7856 * optabs-query.c: Include vec-perm-indices.h.
7857 (selector_fits_mode_p): New function.
7858 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
7859 is defined, instead of checking whether the vec_perm_const_optab
7860 exists. Use targetm.vectorize.vec_perm_const instead of
7861 targetm.vectorize.vec_perm_const_ok. Check whether the indices
7862 fit in the vector mode before using a variable permute.
7863 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
7864 vec_perm_indices instead of an rtx.
7865 (expand_vec_perm): Replace with...
7866 (expand_vec_perm_const): ...this new function. Take the selector
7867 as a vec_perm_indices rather than an rtx. Also take the mode of
7868 the selector. Update call to shift_amt_for_vec_perm_mask.
7869 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
7870 Use vec_perm_indices::new_expanded_vector to expand the original
7871 selector into bytes. Check whether the indices fit in the vector
7872 mode before using a variable permute.
7873 (expand_vec_perm_var): Make global.
7874 (expand_mult_highpart): Use expand_vec_perm_const.
7875 * fold-const.c: Includes vec-perm-indices.h.
7876 * tree-ssa-forwprop.c: Likewise.
7877 * tree-vect-data-refs.c: Likewise.
7878 * tree-vect-generic.c: Likewise.
7879 * tree-vect-loop.c: Likewise.
7880 * tree-vect-slp.c: Likewise.
7881 * tree-vect-stmts.c: Likewise.
7882 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
7884 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
7885 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
7886 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
7887 (aarch64_vectorize_vec_perm_const): ...this new function.
7888 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7889 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7890 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
7891 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
7892 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7893 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7894 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
7896 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
7897 check for NEON modes.
7898 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
7899 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
7900 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
7901 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
7903 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
7904 the old VEC_PERM_CONST conditions.
7905 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
7906 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
7907 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
7908 (ia64_vectorize_vec_perm_const_ok): Merge into...
7909 (ia64_vectorize_vec_perm_const): ...this new function.
7910 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
7911 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
7912 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
7913 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
7914 * config/mips/mips.c (mips_expand_vec_perm_const)
7915 (mips_vectorize_vec_perm_const_ok): Merge into...
7916 (mips_vectorize_vec_perm_const): ...this new function.
7917 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
7918 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
7919 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
7920 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
7921 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
7922 (rs6000_expand_vec_perm_const): Delete.
7923 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
7925 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7926 (altivec_expand_vec_perm_const_le): Take each operand individually.
7927 Operate on constant selectors rather than rtxes.
7928 (altivec_expand_vec_perm_const): Likewise. Update call to
7929 altivec_expand_vec_perm_const_le.
7930 (rs6000_expand_vec_perm_const): Delete.
7931 (rs6000_vectorize_vec_perm_const_ok): Delete.
7932 (rs6000_vectorize_vec_perm_const): New function.
7933 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
7934 an element count and rtx array.
7935 (rs6000_expand_extract_even): Update call accordingly.
7936 (rs6000_expand_interleave): Likewise.
7937 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
7938 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
7939 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
7940 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
7941 (rs6000_expand_vec_perm_const): Delete.
7942 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7943 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7944 (altivec_expand_vec_perm_const_le): Take each operand individually.
7945 Operate on constant selectors rather than rtxes.
7946 (altivec_expand_vec_perm_const): Likewise. Update call to
7947 altivec_expand_vec_perm_const_le.
7948 (rs6000_expand_vec_perm_const): Delete.
7949 (rs6000_vectorize_vec_perm_const_ok): Delete.
7950 (rs6000_vectorize_vec_perm_const): New function. Remove stray
7951 reference to the SPE evmerge intructions.
7952 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
7953 an element count and rtx array.
7954 (rs6000_expand_extract_even): Update call accordingly.
7955 (rs6000_expand_interleave): Likewise.
7956 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
7957 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
7959 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7961 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7963 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
7964 vector mode and that that mode matches the mode of the data
7966 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
7967 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
7968 directly using expand_vec_perm_1 when forcing selectors into
7970 (expand_vec_perm_var): New function, split out from expand_vec_perm.
7972 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7974 * optabs-query.h (can_vec_perm_p): Delete.
7975 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
7976 * optabs-query.c (can_vec_perm_p): Split into...
7977 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
7978 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
7979 particular selector is valid.
7980 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7981 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7982 (vect_grouped_load_supported): Likewise.
7983 (vect_shift_permute_load_chain): Likewise.
7984 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
7985 (vect_transform_slp_perm_load): Likewise.
7986 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7987 (vectorizable_bswap): Likewise.
7988 (vect_gen_perm_mask_checked): Likewise.
7989 * fold-const.c (fold_ternary_loc): Likewise. Don't take
7990 implementations of variable permutation vectors into account
7991 when deciding which selector to use.
7992 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
7993 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
7994 with a false third argument.
7995 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
7996 to test whether the constant selector is valid and can_vec_perm_var_p
7997 to test whether a variable selector is valid.
7999 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8001 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
8002 * optabs-query.c (can_vec_perm_p): Likewise.
8003 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
8004 instead of vec_perm_indices.
8005 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
8006 (vect_gen_perm_mask_checked): Likewise,
8007 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
8008 (vect_gen_perm_mask_checked): Likewise,
8010 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8012 * optabs-query.h (qimode_for_vec_perm): Declare.
8013 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
8014 (qimode_for_vec_perm): ...this new function.
8015 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
8017 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8019 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
8020 does not have a conditional at the top.
8022 2018-01-02 Richard Biener <rguenther@suse.de>
8024 * ipa-inline.c (big_speedup_p): Fix expression.
8026 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8029 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
8032 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8036 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
8037 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
8038 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
8039 cond_taken_branch_cost 3->4.
8041 2018-01-01 Jakub Jelinek <jakub@redhat.com>
8043 PR tree-optimization/83581
8044 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
8045 TODO_cleanup_cfg if any changes have been made.
8048 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
8049 convert_modes if target mode has the right side, but different mode
8053 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
8054 last argument when extracting from CONCAT. If either from_real or
8055 from_imag is NULL, use expansion through memory. If result is not
8056 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
8057 the parts directly to inner mode, if even that fails, use expansion
8061 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
8062 check for bswap in mode rather than HImode and use that in expand_unop
8065 Copyright (C) 2018 Free Software Foundation, Inc.
8067 Copying and distribution of this file, with or without modification,
8068 are permitted in any medium without royalty provided the copyright
8069 notice and this notice are preserved.