1 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2 Monk Chiang <sh.chiang04@gmail.com>
3 Chung-Ju Wu <jasonwucj@gmail.com>
5 * config/nds32/nds32-protos.h
6 (nds32_expand_load_multiple): New arguments.
7 (nds32_expand_store_multiple): Ditto.
8 (nds32_valid_multiple_load_store): Rename ...
9 (nds32_valid_multiple_load_store_p): ... to this.
10 * config/nds32/nds32-memory-manipulation.c
11 (nds32_expand_load_multiple): Refine implementation.
12 (nds32_expand_store_multiple): Ditto.
13 * config/nds32/nds32-multiple.md
14 (load_multiple): Update nds32_expand_load_multiple interface.
15 (store_multiple): Update nds32_expand_store_multiple interface.
16 * config/nds32/nds32-predicates.c
17 (nds32_valid_multiple_load_store): Rename ...
18 (nds32_valid_multiple_load_store_p): ... to this and refine
20 * config/nds32/predicates.md
21 (nds32_load_multiple_and_update_address_operation): New predicate.
22 (nds32_store_multiple_and_update_address_operation): New predicate.
24 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
25 Chung-Ju Wu <jasonwucj@gmail.com>
27 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
28 (combo): New attribute.
29 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
31 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
33 * config/nds32/nds32.opt: Change -mcmodel= default value.
35 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
36 Monk Chiang <sh.chiang04@gmail.com>
37 Chung-Ju Wu <jasonwucj@gmail.com>
39 * config/nds32/constants.md (unspec_element): New enum.
40 * config/nds32/constraints.md (Umw): New constraint.
41 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
42 * config/nds32/nds32-intrinsic.md: Likewise.
43 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
44 (nds32_valid_smw_lwm_base_p): New.
45 (nds32_output_smw_single_word): New.
46 (nds32_output_lmw_single_word): New.
47 (nds32_expand_unaligned_load): New.
48 (nds32_expand_unaligned_store): New.
49 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
50 (nds32_output_smw_single_word): Declare.
51 (nds32_output_lmw_single_word): Declare.
52 (nds32_expand_unaligned_load): Declare.
53 (nds32_expand_unaligned_store): Declare.
54 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
55 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
56 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
57 NDS32_BUILTIN_UASTORE_DW.
58 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
61 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
62 Kito Cheng <kito.cheng@gmail.com>
63 Chung-Ju Wu <jasonwucj@gmail.com>
65 * config/nds32/nds32-intrinsic.c
66 (nds32_expand_builtin_null_ftype_reg): Delete.
67 (nds32_expand_builtin_reg_ftype_imm): Ditto.
68 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
69 (nds32_read_argument): New.
70 (nds32_legitimize_target): Ditto.
71 (nds32_legitimize_argument): Ditto.
72 (nds32_check_constant_argument): Ditto.
73 (nds32_expand_unop_builtin): Ditto.
74 (nds32_expand_unopimm_builtin): Ditto.
75 (nds32_expand_binop_builtin): Ditto.
76 (nds32_builtin_decl_impl): Ditto.
77 (builtin_description): Ditto.
78 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
79 (nds32_init_builtins_impl): Ditto.
80 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
81 (nds32_builtin_decl): New.
82 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
83 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
85 2018-03-02 Jeff Law <law@redhat.com>
87 * reorg.c (stop_search_p): Handle DEBUG_INSN.
88 (redundant_insn, fill_simple_delay_slots): Likewise.
89 (fill_slots_from_thread): Likewise.
90 * resource.c (mark_referenced_resources): Likewise.
91 (mark_set_resources, find_dead_or_set_registers): Likewise.
93 2018-03-02 Jakub Jelinek <jakub@redhat.com>
95 * substring-locations.h (format_warning_va): Formatting fix for
97 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
99 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
100 * substring-locations.c: Include intl.h.
101 (format_warning_va): Turned into small wrapper around
102 format_warning_n_va, renamed to ...
103 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
104 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
106 (format_warning_at_substring_n): New function.
107 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
108 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
109 format_warning_at_substring with just a shorter name instead of
110 const function pointer.
111 (fmtwarn_n): New function.
112 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
113 appropriate, get rid of all the fmtstr temporaries, move conditionals
114 with G_() wrapped string literals directly into fmtwarn arguments,
115 cast dir.len to (int), formatting fixes.
117 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
119 * doc/invoke.texi: Remove "Cilk Plus" references.
121 2018-03-02 Jakub Jelinek <jakub@redhat.com>
122 Richard Biener <rguenther@suse.de>
125 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
126 for error or warning attributes if CALL_FROM_THUNK_P is set.
129 2018-03-02 Jakub Jelinek <jakub@redhat.com>
132 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
133 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
136 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
137 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
139 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
140 instead of -1U in last predictors element's probability member.
142 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
145 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
146 arguments if they are comparable.
148 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
150 PR tree-optimization/84634
151 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
152 masks and masked_loop_p with a single loop_masks, making sure it's
153 null for bb vectorization.
155 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
157 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
158 (vect_analyze_data_ref_access): Use loop->safe_len rather than
159 loop->force_vectorize to check whether there is no alias.
161 2018-03-02 Jakub Jelinek <jakub@redhat.com>
164 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
166 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
168 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
169 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
170 instead of a loop around prev_real_insn.
171 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
175 * config/i386/i386.c (ix86_print_operand): Use conditional
176 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
179 2018-03-02 Richard Biener <rguenther@suse.de>
181 PR tree-optimization/84427
182 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
183 (bitmap_set_subtract_values): Rewrite to handle multiple
186 (prune_clobbered_mems): Likewise.
187 (phi_translate): Take edge instead of pred/phiblock.
188 (phi_translate_1): Likewise.
189 (phi_translate_set): Likewise. Insert all translated
190 exprs for a value into the set, keeping possibly multiple
191 expressions per value.
192 (compute_antic_aux): Adjust for phi_translate changes.
193 When intersecting union the expressions and prune those
194 not in the final value set, keeping possibly multiple
195 expressions per value. Do not use value-insertion
196 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
197 all expressions. Add verification that the value-sets
198 only shrink during iteration.
199 (compute_partial_antic_aux): Adjust for the phi_translate changes.
200 (do_pre_regular_insertion): Likewise.
201 (do_pre_partial_partial_insertion): Likewise.
203 2018-03-02 Richard Biener <rguenther@suse.de>
206 * config/darwin.c (saved_debug_info_level): New static global.
207 (darwin_asm_lto_start): Disable debug info generation for LTO out.
208 (darwin_asm_lto_end): Restore debug info generation settings.
210 2018-03-01 Martin Liska <mliska@suse.cz>
213 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
216 2018-03-01 Richard Biener <rguenther@suse.de>
219 * dwarf2out.c (gen_variable_die): Properly handle late VLA
220 type annotation with LTO when debug was disabled at compile-time.
222 2018-03-01 Matthew Fortune <mfortune@gmail.com>
224 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
226 (mips_final_postscan_insn): Likewise.
228 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
230 PR rtl-optimization/84528
231 * alias.c (init_alias_target): Add commentary.
232 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
233 a unique base value if the frame pointer is not eliminated
234 to the stack pointer.
236 2018-03-01 Tom de Vries <tom@codesourcery.com>
238 PR rtl-optimization/83327
239 * lra-int.h (hard_regs_spilled_into): Declare.
240 * lra.c (hard_regs_spilled_into): Define.
241 (init_reg_info): Init hard_regs_spilled_into.
242 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
243 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
244 (process_bb_lives): Handle hard_regs_spilled_into.
245 (lra_create_live_ranges_1): Before doing liveness propagation, clear
246 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
248 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
250 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
251 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
252 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
253 * config/rs6000/aix72.h: New file.
255 2018-02-28 Jakub Jelinek <jakub@redhat.com>
257 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
258 instead of warning_at with conditional singular and plural messages
262 * stor-layout.c (update_alignment_for_field): For
263 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
264 && !DECL_PACKED (field), do the alignment update, just use
265 only desired_align instead of MAX (type_align, desired_align)
267 (place_field): Don't do known_align < desired_align handling
268 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
269 is non-NULL, instead do it after rli->prev_field handling and
270 only if not within a bitfield word. For DECL_PACKED (field)
271 use type_align of BITS_PER_UNIT.
273 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
275 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
276 superfluous parentheses and trailing spaces.
278 2018-02-28 Richard Biener <rguenther@suse.de>
280 PR tree-optimization/84584
281 * graphite-scop-detection.c (scop_detection::add_scop): Discard
282 SCoPs with fake exit edge.
284 2018-02-28 Martin Liska <mliska@suse.cz>
287 * timevar.c (timer::print): Fix format to properly print 100%
290 2018-02-28 Richard Biener <rguenther@suse.de>
293 * genmatch.c (capture_info::walk_match): Do not mark
294 captured expressions without operands as expr_p given
295 they act more like predicates and should be subject to
296 "lost tail" side-effect preserving.
298 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
300 PR rtl-optimization/81611
301 * auto-inc-dec.c (attempt_change): Move dead note from
302 mem_insn if it's the next use of regno
303 (find_address): Take address use of reg holding
304 non-incremented value. Add parm to limit search to the named
306 (merge_in_block): Attempt to use a mem insn that is the next
307 use of the original regno.
309 2018-02-27 Martin Sebor <msebor@redhat.com>
312 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
313 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
315 2018-02-27 Martin Sebor <msebor@redhat.com>
318 * diagnostic-core.h (warning_n, error_n, inform_n): Change
319 n argument to unsigned HOST_WIDE_INT.
320 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
321 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
322 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
323 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
325 2018-02-27 Richard Biener <rguenther@suse.de>
327 PR tree-optimization/84512
328 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
329 Do not use the estimate returned from record_stmt_cost for
330 the scalar iteration cost but sum properly using add_stmt_cost.
332 2018-02-27 Richard Biener <rguenther@suse.de>
334 PR tree-optimization/84466
335 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
336 Adjust last change to less strictly validate use operands.
338 2018-02-27 Martin Liska <mliska@suse.cz>
340 PR gcov-profile/84548
341 * gcov.c (process_file): Allow partial overlap and consider it
342 also as group functions.
343 (output_lines): Properly calculate range of lines for a group.
345 2018-02-27 Martin Liska <mliska@suse.cz>
347 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
348 'ggc' suffixes. Change first column width.
349 (timer::print): Fix formatting of the column.
351 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
353 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
354 preserve inline entry blocks for the sake of debug inline
355 entry point markers alone.
356 (remove_unused_locals): Suggest in comments a better place to
357 force the preservation of inline entry blocks that are
358 otherwise unused, but do not preserve them.
360 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
362 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
364 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
367 * config/i386/constraints.md (Bs): Replace
368 ix86_indirect_branch_register with
369 TARGET_INDIRECT_BRANCH_REGISTER.
371 * config/i386/i386.md (indirect_jump): Likewise.
372 (tablejump): Likewise.
373 (*sibcall_memory): Likewise.
374 (*sibcall_value_memory): Likewise.
375 Peepholes of indirect call and jump via memory: Likewise.
376 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
377 (*sibcall_value_GOT_32): Likewise.
378 * config/i386/predicates.md (indirect_branch_operand): Likewise.
379 (GOT_memory_operand): Likewise.
380 (call_insn_operand): Likewise.
381 (sibcall_insn_operand): Likewise.
382 (GOT32_symbol_operand): Likewise.
383 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
385 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
387 PR rtl-optimization/83496
388 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
389 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
390 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
391 redundant insn, if any.
392 (relax_delay_slots): Likewise.
393 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
395 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
397 PR tree-optimization/83965
398 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
399 that grouped statements are part of a reduction chain. Return
400 true if the statement is not marked as a reduction itself but
402 (vect_recog_dot_prod_pattern): Don't check whether the statement
403 is part of a group here.
404 (vect_recog_sad_pattern): Likewise.
405 (vect_recog_widen_sum_pattern): Likewise.
407 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
410 * final.c (rest_of_clean_state): Also look for calls inside sequences.
412 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
415 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
417 (ix86_output_indirect_function_return): New prototype.
418 (ix86_split_simple_return_pop_internal): Likewise.
419 * config/i386/i386.c (indirect_return_via_cx): New.
420 (indirect_return_via_cx_bnd): Likewise.
421 (indirect_thunk_name): Handle return va CX_REG.
422 (output_indirect_thunk_function): Create alias for
423 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
424 (ix86_output_indirect_jmp): Remove the bool argument.
425 (ix86_output_indirect_function_return): New function.
426 (ix86_split_simple_return_pop_internal): Likewise.
427 * config/i386/i386.md (*indirect_jump): Don't pass false
428 to ix86_output_indirect_jmp.
429 (*tablejump_1): Likewise.
430 (simple_return_pop_internal): Change it to define_insn_and_split.
431 Call ix86_split_simple_return_pop_internal to split it for
433 (simple_return_indirect_internal): Call
434 ix86_output_indirect_function_return instead of
435 ix86_output_indirect_jmp.
437 2018-02-26 Jakub Jelinek <jakub@redhat.com>
440 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
441 memset and value initialization afterwards.
443 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
445 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
447 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
450 * common/config/aarch64/aarch64-common.c
451 (aarch_option_optimization_table[]): Switch
452 off fomit-frame-pointer
454 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
455 Chung-Ju Wu <jasonwucj@gmail.com>
457 * config/nds32/nds32-multiple.md (load_multiple): Disallow
459 (store_multiple): Ditto.
461 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
463 * config.gcc: Add --with-cpu support for nds32 target.
464 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
465 * config/nds32/nds32.opt: Add -mcpu= option.
467 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
469 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
470 isel=yes): Warn for these deprecated options.
472 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
474 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
475 ISA_2_5_MASKS_EMBEDDED.
477 2018-02-23 Jakub Jelinek <jakub@redhat.com>
479 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
480 p->max as pointers rather than using iterative_hash_expr.
482 2018-02-23 Carl Love <cel@us.ibm.com>
484 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
485 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
487 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
488 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
489 P8V_BUILTIN_VEC_VUNSIGNED2.
491 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
494 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
495 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
497 * lra-constraints.c (curr_insn_transform): Set up
498 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
500 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
501 * lra-eliminations.c (spill_pseudos): Ditto.
502 (process_insn_for_elimination): Ditto.
503 * lra-lives.c (reg_early_clobber_p): Use the new macros.
504 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
505 LRA_NON_CLOBBERED_ALT.
507 2018-02-22 Martin Sebor <msebor@redhat.com>
509 PR tree-optimization/84480
510 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
511 to maybe_diag_stxncpy_trunc. Call it.
512 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
513 from gimple_fold_builtin_strcpy. Print inlining stack.
514 (handle_builtin_stxncpy): Print inlining stack.
515 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
517 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
520 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
521 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
522 and -fcheck-pointer-bounds are used together.
523 (indirect_thunk_prefix): New enum.
524 (indirect_thunk_need_prefix): New function.
525 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
526 "_nt" instead of "_bnd" for NOTRACK prefix.
527 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
528 (output_indirect_thunk_function): Likewise.
530 (ix86_code_end): Update output_indirect_thunk_function calls.
531 (ix86_output_indirect_branch_via_reg): Replace
532 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
533 (ix86_output_indirect_branch_via_push): Likewise.
534 (ix86_output_function_return): Likewise.
535 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
536 incompatible with -fcf-protection=branch and
537 -fcheck-pointer-bounds.
539 2018-02-22 Steve Ellcey <sellcey@cavium.com>
542 * config/aarch64/aarch64.c (aarch64_print_address_internal):
543 Change gcc_assert call to output_operand_lossage.
545 2018-02-22 Steve Ellcey <sellcey@cavium.com>
547 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
549 2018-02-22 DJ Delorie <dj@redhat.com>
550 Sebastian Perta <sebastian.perta@renesas.com>
551 Oleg Endo <olegendo@gcc.gnu.org>
553 * config/rx/rx.c (rx_rtx_costs): New function.
554 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
556 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
558 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
560 2018-02-22 Martin Liska <mliska@suse.cz>
563 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
564 Add "native" as a possible value.
566 2018-02-22 Martin Liska <mliska@suse.cz>
569 * config/i386/i386.c (ix86_option_override_internal):
570 Add "native" as a possible value for -march and -mtune.
572 2018-02-22 Jakub Jelinek <jakub@redhat.com>
575 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
576 to all type variants.
578 PR tree-optimization/84503
579 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
580 width as info->bitpos + info->bitsize - start.
581 (merged_store_group::merge_overlapping): Simplify width computation.
582 (check_no_overlap): New function.
583 (imm_store_chain_info::try_coalesce_bswap): Compute expected
584 start + width and last_order of the group, fail if check_no_overlap
586 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
587 to group if check_no_overlap fails.
589 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
591 * config/rs6000/altivec.md: Delete contraint arguments to
592 define_expand, define_split, and define_peephole2, and in
593 define_insn_and_split if always unused.
594 * config/rs6000/darwin.md: Ditto.
595 * config/rs6000/dfp.md: Ditto.
596 * config/rs6000/rs6000.md: Ditto.
597 * config/rs6000/sync.md: Ditto.
598 * config/rs6000/vector.md: Ditto.
599 * config/rs6000/vsx.md: Ditto.
601 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
603 * config/rs6000/altivec.md: Write output control strings as braced
604 blocks instead of double-quoted strings.
605 * config/rs6000/darwin.md: Ditto.
606 * config/rs6000/rs6000.md: Ditto.
607 * config/rs6000/vector.md: Ditto.
608 * config/rs6000/vsx.md: Ditto.
610 2018-02-21 Jason Merrill <jason@redhat.com>
612 PR c++/84314 - ICE with templates and fastcall attribute.
613 * attribs.c (build_type_attribute_qual_variant): Remove assert.
615 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
617 * ipa-cp.c (determine_versionability): Fix comment typos.
619 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
622 * ipa-cp.c (determine_versionability): Do not version functions caling
625 2018-02-21 Martin Liska <mliska@suse.cz>
628 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
629 Add "native" as a possible value.
630 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
631 the macro when native cpu detection is available.
633 2018-02-21 Martin Liska <mliska@suse.cz>
636 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
637 Add "native" as a possible value.
638 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
639 when native cpu detection is available.
641 2018-02-21 Jakub Jelinek <jakub@redhat.com>
642 Martin Sebor <msebor@redhat.com>
644 PR tree-optimization/84478
645 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
647 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
648 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
649 support which is conservatively correct, for 2 only stay conservative
650 for maxlen. Formatting and comment capitalization fixes. Add STRICT
651 argument to the 2 argument get_range_strlen, adjust 6 arg
652 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
654 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
655 (gimple_fold_builtin_strlen): Pass true as last argument to
658 2018-02-20 Martin Sebor <msebor@redhat.com>
661 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
662 (builtin_memref::set_base_and_offset): Same. Handle inner references.
663 (builtin_memref::builtin_memref): Factor out parts into
664 set_base_and_offset and call it.
666 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
669 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
670 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
671 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
672 search at the associated MODE_INT.
674 2018-02-20 Jeff Law <law@redhat.com>
677 PR tree-optimization/81592
679 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
680 for range data rather than using global data.
681 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
682 range data rather than using global data.
683 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
684 pass it to children as needed.
685 (struct directive::fmtresult): Similarly.
686 (struct directive::set_width): Similarly.
687 (struct directive::set_precision): Similarly.
688 (format_integer, format_directive, parse_directive): Similarly.
689 (format_none): Accept unnamed vr_values parameter.
690 (format_percent, format_floating, format_character): Similarly.
691 (format_string, format_plain): Similarly.
692 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
693 the EVRP range analyzer for range data rather than using global data.
694 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
695 gimple-ssa-evrp-analyze.h
696 (class sprintf_dom_walker): Add after_dom_children member function.
697 Add evrp_range_analyzer member.
698 (sprintf_dom_walker::before_dom_children): Call into the EVRP
699 range analyzer as needed.
700 (sprintf_dom_walker::after_dom_children): New member function.
701 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
703 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
704 (evrp_range_analyzer::pop_to_marker): Likewise.
706 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
708 PR tree-optimization/84419
709 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
710 with the required type if its current type is compatible but
713 2018-02-20 Jakub Jelinek <jakub@redhat.com>
716 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
719 2018-02-20 Martin Liska <mliska@suse.cz>
722 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
723 possible values if we don't have a hint.
725 2018-02-20 Martin Liska <mliska@suse.cz>
729 * final.c (shorten_branches): Build align_tab array with one
731 * opts.c (finish_options): Add alignment option limit check.
732 (MAX_CODE_ALIGN): Likewise.
733 (MAX_CODE_ALIGN_VALUE): Likewise.
734 * doc/invoke.texi: Document maximum allowed option value for
735 all -falign-* options.
737 2018-02-19 Jakub Jelinek <jakub@redhat.com>
740 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
741 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
742 * var-tracking.c (emit_note_insn_var_location): Remove all references
743 to NOTE_INSN_CALL_ARG_LOCATION.
744 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
745 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
746 Use copy_rtx_if_shared.
747 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
748 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
749 (dwarf2out_var_location): Remove handling of
750 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
752 * final.c (final_scan_insn): Remove all references to
753 NOTE_INSN_CALL_ARG_LOCATION.
754 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
755 before dumping final insns.
756 * except.c (emit_note_eh_region_end): Remove all references to
757 NOTE_INSN_CALL_ARG_LOCATION.
758 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
759 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
760 * config/arc/arc.c (hwloop_optimize): Likewise.
761 * config/arm/arm.c (create_fix_barrier): Likewise.
762 * config/s390/s390.c (s390_chunkify_start): Likewise.
763 * config/sh/sh.c (find_barrier): Likewise.
764 * config/i386/i386.c (rest_of_insert_endbranch,
765 ix86_seh_fixup_eh_fallthru): Likewise.
766 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
767 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
768 * config/frv/frv.c (frv_function_prologue): Likewise.
769 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
771 (note_outside_basic_block_p): Remove all references to
772 NOTE_INSN_CALL_ARG_LOCATION.
773 * gengtype.c (adjust_field_rtx_def): Likewise.
774 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
776 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
777 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
780 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
783 PR tree-optimization/84452
784 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
785 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
788 2018-02-19 Martin Liska <mliska@suse.cz>
791 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
793 2018-02-19 Martin Liska <mliska@suse.cz>
794 Richard Sandiford <richard.sandiford@linaro.org>
796 PR tree-optimization/82491
797 * gimple-fold.c (get_base_constructor): Make earlier bail out
800 2018-02-19 Carl Love <cel@us.ibm.com>
802 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
803 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
805 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
808 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
810 * config/rl78/rl78.md (movdf): New define expand.
812 2018-02-19 Martin Liska <mliska@suse.cz>
815 * doc/invoke.texi: Fix typo.
816 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
818 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
820 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
821 handle rs6000_single_float and rs6000_double_float specially for
824 2018-02-16 Jeff Law <law@redhat.com>
826 * config/rx/rx.c (add_pop_cfi_notes): New function.;
829 2018-02-16 Jakub Jelinek <jakub@redhat.com>
832 * ipa-inline.c (inline_small_functions): Fix a typo.
834 2018-02-16 Nathan Sidwell <nathan@acm.org>
836 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
838 2018-02-16 Carl Love <cel@us.ibm.com>
840 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
841 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
842 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
843 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
844 expansion to P8V_BUILTIN_VEC_FLOAT2.
846 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
848 PR rtl-optimization/70023
849 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
850 src_regno into account.
852 2018-02-16 Carl Love <cel@us.ibm.com>
854 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
855 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
856 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
857 * config/rs6000/rs6000.c: Remove case statements for
858 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
859 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
860 and P9V_BUILTIN_VEC_VINSERT4B.
861 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
862 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
863 * config/rs6000/vsx.md:
864 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
867 2018-02-16 Carl Love <cel@us.ibm.com>
869 * config/rs6000/altivec.h: Add builtin names vec_extract4b
871 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
873 * config/rs6000/rs6000-c.c: Add the definitions for
874 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
875 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
876 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
877 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
878 definition for insert4b and define insn *insert3b_internal.
879 * doc/extend.texi: Add documentation for vec_extract4b.
881 2018-02-16 Nathan Sidwell <nathan@acm.org>
883 * doc/extend.texi (Backwards Compatibility): Mention friend
884 injection. Note for-scope is deprecated.
885 * doc/invoke.texi (-ffriend-injection): Deprecate.
887 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
889 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
890 that moved to I2, also allow destinations that are a paradoxical
891 subreg (instead of a normal reg).
893 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
896 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
899 2018-02-16 Richard Biener <rguenther@suse.de>
901 PR tree-optimization/84037
902 PR tree-optimization/84016
904 * config/i386/i386.c (ix86_builtin_vectorization_cost):
905 Adjust vec_construct for the fact we need additional higher latency
906 128bit inserts for AVX256 and AVX512 vector builds.
907 (ix86_add_stmt_cost): Scale vector construction cost for
910 2018-02-16 Richard Biener <rguenther@suse.de>
912 PR tree-optimization/84417
913 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
914 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
915 (non_rewritable_lvalue_p): Likewise, use poly-ints.
917 2018-02-16 Martin Liska <mliska@suse.cz>
920 * internal-fn.def (ASAN_CHECK): Set proper flags.
921 (ASAN_MARK): Likewise.
923 2018-02-16 Julia Koval <julia.koval@intel.com>
925 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
928 2018-02-16 Jakub Jelinek <jakub@redhat.com>
931 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
932 Use ++iter rather than iter++ for std::list iterators.
933 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
934 defer deleting them until all nodes in the forest are processed. Do
935 free even leaf nodes. Change to_process into auto_vec.
938 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
939 * vec.h (vec_default_construct): Use memset instead of placement new
940 if BROKEN_VALUE_INITIALIZATION is defined.
941 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
942 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
945 PR rtl-optimization/83723
946 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
947 * lra.c (lra_substitute_pseudo): Likewise. If true, use
948 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
950 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
952 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
954 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
956 PR rtl-optimization/81443
957 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
958 from inner REGs to paradoxical SUBREGs.
960 2018-02-16 Richard Biener <rguenther@suse.de>
962 PR tree-optimization/84399
963 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
964 For operands we can analyze at their definition make sure we can
965 analyze them at each use as well.
967 2018-02-16 Richard Biener <rguenther@suse.de>
969 PR tree-optimization/84190
970 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
971 volatile accesses if the decl isn't volatile.
973 2018-02-15 Jason Merrill <jason@redhat.com>
975 PR c++/84314 - ICE with templates and fastcall attribute.
976 * attribs.c (build_type_attribute_qual_variant): Don't clobber
977 TYPE_CANONICAL on an existing type.
979 2018-02-15 Jakub Jelinek <jakub@redhat.com>
981 PR tree-optimization/84383
982 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
983 dstoff nor call operand_equal_p if dstbase is NULL.
985 PR tree-optimization/84334
986 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
987 also a CONSTANT_CLASS_P, punt.
989 2018-02-14 Jim Wilson <jimw@sifive.com>
991 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
992 first SMALL_OPERAND check. New local min_second_step. Move assert
993 to where locals are set. Add TARGET_RVC support.
994 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
996 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
998 * doc/invoke.texi: Correct -Wformat-overflow code sample.
1000 2018-02-14 Martin Sebor <msebor@redhat.com>
1002 PR tree-optimization/83698
1003 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
1004 arrays constrain the offset range to their bounds.
1005 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
1006 (builtin_access::overlap): Avoid setting the size of overlap if it's
1008 (maybe_diag_overlap): Also consider arrays when deciding what values
1009 of offsets to include in diagnostics.
1011 2018-02-14 Martin Sebor <msebor@redhat.com>
1014 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
1015 that correspond to the kind of a declaration.
1017 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
1020 * config/pa/pa.md: Load address of PIC label using the linkage table
1021 if the label is nonlocal.
1023 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
1025 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
1026 warning message if user requests -maltivec=be.
1027 * doc/invoke.texi: Document deprecation of -maltivec=be.
1029 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
1032 * config/rs6000/rs6000-c.c: Update definitions for
1033 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
1034 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
1036 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1039 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
1040 add _get_ssp intrinsics. Remove argument from
1041 __builtin_ia32_rdssp[d|q].
1042 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
1043 * config/i386/i386-builtin.def: Remove argument from
1044 __builtin_ia32_rdssp[d|q].
1045 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
1046 ix86_expand_special_args_builtin for _rdssp[d|q].
1047 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
1048 Clear register before usage.
1049 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
1050 Add documentation for new _get_ssp and _inc_ssp intrinsics.
1052 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
1054 PR tree-optimization/84357
1055 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
1056 operand 1 of an ARRAY_REF too.
1058 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
1061 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
1062 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1064 (set_of_reg): New struct.
1065 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
1066 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
1067 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1069 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
1070 Split into bitclr, bitset, bitinvert patterns if appropriate.
1071 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
1072 use rx_fuse_in_memory_bitop.
1073 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
1074 to named insn, correct maximum insn length.
1076 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
1079 * machmode.def: Define a complex mode for PARTIAL_INT.
1080 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
1082 * doc/rtl.texi: Document CSPImode.
1083 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
1085 (msp430_hard_regno_nregs_with_padding): Likewise.
1087 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
1090 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
1092 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
1094 PR rtl-optimization/84169
1095 * combine.c (try_combine): New variable split_i2i3. Set it to true if
1096 we generated a parallel as new i3 and we split that to new i2 and i3
1097 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
1098 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
1099 those to i2, not i1. Partially rewrite this scan code.
1101 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1104 * stor-layout.c (place_field): For variable length fields, adjust
1105 offset_align afterwards not just based on the field's alignment,
1106 but also on the size.
1109 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
1110 of exps and logs in the use_exp2 case.
1112 2018-02-13 Jeff Law <law@redhat.com>
1114 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
1117 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
1120 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
1124 * common.opt (gas-loc-support, gas-locview-support): New.
1125 (ginline-points, ginternal-reset-location-views): New.
1126 * doc/invoke.texi: Document them. Use @itemx where intended.
1127 (gvariable-location-views): Adjust.
1128 * target.def (reset_location_view): New.
1129 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
1130 (TARGET_RESET_LOCATION_VIEW): New.
1131 * doc/tm.texi: Rebuilt.
1132 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
1133 (dwarf2out_default_as_locview_support): New.
1134 (output_asm_line_debug_info): Use option variables.
1135 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
1136 (output_loc_list): Likewise.
1137 (add_high_low_attributes): Check option variables.
1138 Don't output entry view attribute in strict mode.
1139 (gen_inlined_subroutine_die): Check option variables.
1140 (dwarf2out_inline_entry): Likewise.
1141 (init_sections_and_labels): Likewise.
1142 (dwarf2out_early_finish): Likewise.
1143 (maybe_reset_location_view): New, from...
1144 (dwarf2out_var_location): ... here. Call it.
1145 * debug.h (dwarf2out_default_as_loc_support): Declare.
1146 (dwarf2out_default_as_locview_support): Declare.
1147 * hooks.c (hook_int_rtx_insn_0): New.
1148 * hooks.h (hook_int_rtx_insn_0): Declare.
1149 * toplev.c (process_options): Take -gas-loc-support and
1150 -gas-locview-support from dwarf2out. Enable
1151 -gvariable-location-views by default only with locview
1152 assembler support. Enable -ginternal-reset-location-views by
1153 default only if the target defines the corresponding hook.
1154 Enable -ginline-points by default if location views are
1155 enabled; force it disabled if statement frontiers are
1157 * tree-inline.c (expand_call_inline): Check option variables.
1158 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1160 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
1162 PR tree-optimization/84321
1163 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
1164 handling. Also check whether the anti-range contains any values
1165 that satisfy the mask; switch to a VR_RANGE if not.
1167 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
1170 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
1172 2018-02-13 Martin Jambor <mjambor@suse.cz>
1175 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
1176 of call statements, also set location of a load to a temporary.
1178 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
1180 * config/rl78/rl78.c (add_vector_labels): New function.
1181 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
1182 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
1183 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
1184 which checks that no arguments are passed.
1185 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
1186 * doc/extend.texi: Documentation for the new attribute.
1188 2018-02-13 Andreas Schwab <schwab@suse.de>
1190 * config/riscv/linux.h (CPP_SPEC): Define.
1192 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1195 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
1196 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
1197 OPTION_MASK_ISA_AES as first argument to def_builtin_const
1198 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
1199 instead of OPTION_MASK_ISA_PCLMUL as first argument to
1200 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
1201 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
1202 temporarily for AES and PCLMUL builtins.
1204 PR tree-optimization/84339
1205 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
1206 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
1210 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
1211 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
1212 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
1214 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
1216 * omp-simd-clone.h: New file.
1217 * omp-simd-clone.c: Include omp-simd-clone.h.
1218 (expand_simd_clones): No longer static.
1219 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
1220 cgraph.h and omp-simd-clone.h.
1221 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
1222 (vect_recog_widen_shift_pattern): Formatting fix.
1223 (vect_pattern_recog_1): Don't check optab for calls.
1226 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
1227 operands[2] into a REG before using gen_lowpart on it.
1229 2018-02-12 Jeff Law <law@redhat.com>
1232 * config/sh/sh.c (find_barrier): Consider a sibling call
1235 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
1236 successfully back substituting a reg.
1238 2018-02-12 Richard Biener <rguenther@suse.de>
1240 PR tree-optimization/84037
1241 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
1242 parameter, move visited init to caller.
1243 (vect_slp_analyze_operations): Separate cost from validity
1244 check, initialize visited once for all instances.
1245 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
1247 * tree-vect-stmts.c (vect_model_simple_cost): Make early
1249 (vect_model_promotion_demotion_cost): Likewise.
1250 (vectorizable_bswap): Guard cost modeling with !slp_node
1251 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
1253 (vectorizable_call): Likewise.
1254 (vectorizable_conversion): Likewise.
1255 (vectorizable_assignment): Likewise.
1256 (vectorizable_shift): Likewise.
1257 (vectorizable_operation): Likewise.
1258 (vectorizable_store): Likewise.
1259 (vectorizable_load): Likewise.
1260 (vectorizable_condition): Likewise.
1261 (vectorizable_comparison): Likewise.
1263 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
1266 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
1267 (ASAN_MARK): Fix fnspec to account for return value, change pointer
1268 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
1270 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1273 * params.def (inline-min-speedup): Increase from 8 to 15.
1274 (max-inline-insns-auto): Decrease from 40 to 30.
1275 * ipa-split.c (consider_split): Add some buffer for function to
1276 be considered inlining candidate.
1277 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
1280 2018-02-12 Richard Biener <rguenther@suse.de>
1282 PR tree-optimization/84037
1283 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
1284 matched stmts if we cannot swap the non-matched ones.
1286 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1288 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
1289 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
1290 _mm_maskz_scalef_round_ss): New intrinsics.
1291 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
1292 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
1293 __builtin_ia32_scalefss_round): Remove.
1294 (__builtin_ia32_scalefsd_mask_round,
1295 __builtin_ia32_scalefss_mask_round): New intrinsics.
1296 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
1297 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
1298 ((match_operand:VF_128 2 "<round_nimm_predicate>"
1299 "<round_constraint>")): Changed to ...
1300 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
1301 "<round_scalar_constraint>")): ... this.
1302 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
1303 %0, %1, %2<round_op3>}"): Changed to ...
1304 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
1305 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
1306 %2<round_scalar_mask_op3>}"): ... this.
1307 * config/i386/subst.md (round_scalar_nimm_predicate): New.
1309 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1311 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
1312 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
1313 (_mm_maskz_sqrt_round_ss): New intrinsics.
1314 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
1315 (__builtin_ia32_sqrtsd_mask_round)
1316 (__builtin_ia32_sqrtss_mask_round): New builtins.
1317 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
1318 (__builtin_ia32_sqrtss_round): Remove.
1319 (__builtin_ia32_sqrtsd_mask_round)
1320 (__builtin_ia32_sqrtss_mask_round): New builtins.
1321 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
1322 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
1323 ((match_operand:VF_128 1 "vector_operand"
1324 "xBm,<round_constraint>")): Changed to ...
1325 ((match_operand:VF_128 1 "vector_operand"
1326 "xBm,<round_scalar_constraint>")): ... this.
1327 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
1328 %0, %2, %<iptr>1<round_op3>}): Changed to ...
1329 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
1330 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
1331 %<iptr>1<round_scalar_mask_op3>}): ... this.
1332 ((set_attr "prefix" "<round_prefix>")): Changed to ...
1333 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
1335 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
1338 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
1339 Cast vec_cmpeq result to correct type.
1340 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
1341 Cast vec_cmpgt result to correct type.
1343 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
1345 * final.c (final_scan_insn_1): Renamed from...
1346 (final_scan_insn): ... this. New wrapper, to recover
1347 seen from the outermost call in recursive ones.
1348 * config/sparc/sparc.c (output_return): Drop seen from call.
1349 (output_sibcall): Likewise.
1350 * config/visium/visium.c (output_branch): Likewise.
1352 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
1354 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
1357 2018-02-10 Alan Modra <amodra@gmail.com>
1360 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
1361 Specify LR as an input.
1363 2018-02-10 Jakub Jelinek <jakub@redhat.com>
1366 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
1367 remove_member_access_dummy_vars): New functions.
1368 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
1369 lower_omp_1, execute_lower_omp): Use them.
1371 PR rtl-optimization/84308
1372 * shrink-wrap.c (spread_components): Release todo vector.
1374 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
1376 PR rtl-optimization/57193
1377 * ira-color.c (struct allocno_color_data): Add member
1378 conflict_allocno_hard_prefs.
1379 (update_conflict_allocno_hard_prefs): New.
1380 (bucket_allocno_compare_func): Add a preference based on
1381 conflict_allocno_hard_prefs.
1382 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
1383 (color_allocnos): Remove a dead code. Initiate
1384 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
1386 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1389 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
1390 constraint from =wa to wa. Avoid a subreg on the output operand,
1391 instead use a pseudo and subreg it in a move.
1392 (p9_xxbrd_<mode>): Changed to ...
1393 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
1394 (p9_xxbrd_v2df): New expander.
1395 (p9_xxbrw_<mode>): Changed to ...
1396 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
1397 (p9_xxbrw_v4sf): New expander.
1399 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1401 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
1403 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
1406 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
1407 multiply in 32-bit mode.
1408 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
1409 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
1412 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1414 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
1415 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
1416 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
1417 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
1419 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
1422 * dwarf2out.c (is_trivial_indirect_ref): New function.
1423 (dwarf2out_late_global_decl): Do not generate a location
1424 attribute for variables that have a non-trivial DECL_VALUE_EXPR
1425 and that are not defined in the current unit.
1427 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1429 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
1430 instead of a libcall for UNORDERED.
1432 2018-02-09 Tamar Christina <tamar.christina@arm.com>
1435 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
1436 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
1438 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1441 * config/s390/s390.c (s390_set_current_function): Invoke
1442 s390_indirect_branch_settings also if fndecl didn't change.
1444 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1446 * config/rs6000/rs6000.md (blockage): Set length to zero.
1448 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1450 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
1452 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1455 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
1456 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
1460 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
1461 PARALLEL incoming that failed vt_get_decl_and_offset check.
1464 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
1465 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
1467 (get_variable_section): For decls in named .bss* sections pass true as
1468 second argument to bss_initializer_p.
1470 2018-02-09 Marek Polacek <polacek@redhat.com>
1471 Jakub Jelinek <jakub@redhat.com>
1474 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
1475 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
1476 Sync some changes from cxx_fold_indirect_ref.
1478 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1480 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
1482 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
1483 (BLOCK_INLINE_ENTRY_LABEL): New.
1484 (dwarf2out_var_location): Disregard inline entry markers.
1485 (inline_entry_data): New struct.
1486 (inline_entry_data_hasher): New hashtable type.
1487 (inline_entry_data_hasher::hash): New.
1488 (inline_entry_data_hasher::equal): New.
1489 (inline_entry_data_table): New variable.
1490 (add_high_low_attributes): Add DW_AT_entry_pc and
1491 DW_AT_GNU_entry_view attributes if a pending entry is found
1492 in inline_entry_data_table. Add old entry_pc attribute only
1493 if debug nonbinding markers are disabled.
1494 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
1495 markers are enabled.
1496 (block_within_block_p, dwarf2out_inline_entry): New.
1497 (dwarf2out_finish): Check that no entries remained in
1498 inline_entry_data_table.
1499 * final.c (reemit_insn_block_notes): Handle inline entry notes.
1500 (final_scan_insn, notice_source_line): Likewise.
1501 (rest_of_clean_state): Skip inline entry markers.
1502 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
1504 * gimple.c (gimple_build_debug_inline_entry): New.
1505 * gimple.h (enum gimple_debug_subcode): Add
1506 GIMPLE_DEBUG_INLINE_ENTRY.
1507 (gimple_build_debug_inline_entry): Declare.
1508 (gimple_debug_inline_entry_p): New.
1509 (gimple_debug_nonbind_marker_p): Adjust.
1510 * insn-notes.def (INLINE_ENTRY): New.
1511 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
1512 inline entry marker notes.
1513 (print_insn): Likewise.
1514 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
1515 (INSN_DEBUG_MARKER_KIND): Likewise.
1516 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
1517 * tree-inline.c (expand_call_inline): Build and insert
1518 debug_inline_entry stmt.
1519 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
1520 inline entry blocks early, if nonbind markers are enabled.
1521 (dump_scope_block): Dump fragment info.
1522 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
1523 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
1524 (gimple_build_debug_inline_entry): New.
1525 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
1526 Enable/disable inline entry points too.
1527 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
1528 (DEBUG_INSN): Describe inline entry markers.
1530 * common.opt (gvariable-location-views): New.
1531 (gvariable-location-views=incompat5): New.
1532 * config.in: Rebuilt.
1533 * configure: Rebuilt.
1534 * configure.ac: Test assembler for view support.
1535 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
1536 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
1537 * dwarf2out.c (var_loc_view): New typedef.
1538 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
1539 (dwarf2out_locviews_in_attribute): New.
1540 (dwarf2out_locviews_in_loclist): New.
1541 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
1542 (enum dw_line_info_opcode): Add LI_adv_address.
1543 (struct dw_line_info_table): Add view.
1544 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
1545 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
1546 (zero_view_p): New variable.
1547 (ZERO_VIEW_P): New macro.
1548 (output_asm_line_debug_info): New.
1549 (struct var_loc_node): Add view.
1550 (add_AT_view_list, AT_loc_list): New.
1551 (add_var_loc_to_decl): Add view param. Test it against last.
1552 (new_loc_list): Add view params. Record them.
1553 (AT_loc_list_ptr): Handle loc and view lists.
1554 (view_list_to_loc_list_val_node): New.
1555 (print_dw_val): Handle dw_val_class_view_list.
1556 (size_of_die): Likewise.
1557 (value_format): Likewise.
1558 (loc_list_has_views): New.
1559 (gen_llsym): Set vl_symbol too.
1560 (maybe_gen_llsym, skip_loc_list_entry): New.
1561 (dwarf2out_maybe_output_loclist_view_pair): New.
1562 (output_loc_list): Output view list or entries too.
1563 (output_view_list_offset): New.
1564 (output_die): Handle dw_val_class_view_list.
1565 (output_dwarf_version): New.
1566 (output_compilation_unit_header): Use it.
1567 (output_skeleton_debug_sections): Likewise.
1568 (output_rnglists, output_line_info): Likewise.
1569 (output_pubnames, output_aranges): Update version comments.
1570 (output_one_line_info_table): Output view numbers in asm comments.
1571 (dw_loc_list): Determine current endview, pass it to new_loc_list.
1572 Call maybe_gen_llsym.
1573 (loc_list_from_tree_1): Adjust.
1574 (add_AT_location_description): Create view list attribute if
1575 needed, check it's absent otherwise.
1576 (convert_cfa_to_fb_loc_list): Adjust.
1577 (maybe_emit_file): Call output_asm_line_debug_info for test.
1578 (dwarf2out_var_location): Reset views as needed. Precompute
1579 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
1580 attribute. Set view.
1581 (new_line_info_table): Reset next view.
1582 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
1583 (dwarf2out_source_line): Likewise. Output view resets and labels to
1584 the assembler, or select appropriate line info opcodes.
1585 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
1586 (optimize_string_length): Catch it. Adjust.
1587 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
1588 dw_val_class_view_list, and remove it if no longer needed.
1589 (hash_loc_list): Hash view numbers.
1590 (loc_list_hasher::equal): Compare them.
1591 (optimize_location_lists): Check whether a view list symbol is
1592 needed, and whether the locview attribute is present, and
1593 whether they match. Remove the locview attribute if no longer
1595 (index_location_lists): Call skip_loc_list_entry for test.
1596 (dwarf2out_finish): Call output_asm_line_debug_info for test.
1597 Use output_dwarf_version.
1598 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
1599 (struct dw_val_node): Add val_view_list.
1600 * final.c (SEEN_NEXT_VIEW): New.
1601 (set_next_view_needed): New.
1602 (clear_next_view_needed): New.
1603 (maybe_output_next_view): New.
1604 (final_start_function): Rename to...
1605 (final_start_function_1): ... this. Take pointer to FIRST,
1606 add SEEN parameter. Emit param bindings in the initial view.
1607 (final_start_function): Reintroduce SEEN-less interface.
1608 (final): Rename to...
1609 (final_1): ... this. Take SEEN parameter. Output final pending
1610 next view at the end.
1611 (final): Reintroduce seen-less interface.
1612 (final_scan_insn): Output pending next view before switching
1613 sections or ending a block. Mark the next view as needed when
1614 outputting variable locations. Notify debug backend of section
1615 changes, and of location view changes.
1616 (rest_of_handle_final): Adjust.
1617 * toplev.c (process_options): Autodetect value for debug variable
1618 location views option. Warn on incompat5 without -gdwarf-5.
1619 * doc/invoke.texi (gvariable-location-views): New.
1620 (gvariable-location-views=incompat5): New.
1621 (gno-variable-location-views): New.
1623 2018-02-08 David Malcolm <dmalcolm@redhat.com>
1625 PR tree-optimization/84136
1626 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
1627 that the result of find_edge is non-NULL.
1629 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1632 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
1633 storing integer register in SImode. Fix cost of 256 and 512
1634 byte aligned SSE register store.
1636 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1638 * config/i386/i386.c (ix86_multiplication_cost): Fix
1639 multiplication cost for TARGET_AVX512DQ.
1641 2018-02-08 Marek Polacek <polacek@redhat.com>
1643 PR tree-optimization/84238
1644 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
1647 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1649 PR tree-optimization/84265
1650 * tree-vect-stmts.c (vectorizable_store): Don't treat
1651 VMAT_CONTIGUOUS accesses as grouped.
1652 (vectorizable_load): Likewise.
1654 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1656 PR tree-optimization/81635
1657 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
1658 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
1659 (test_round_for_mask): New functions.
1660 (wide_int_cc_tests): Call test_round_for_mask.
1661 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
1662 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
1663 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
1664 range returned by get_range_info.
1666 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1669 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
1670 * symtab.c: Include builtins.h
1671 (symtab_node::output_to_lto_symbol_table_p): Move here
1672 from lto-streamer-out.c:output_symbol_p.
1673 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
1674 (output_symbol_p): Move all logic to symtab.c
1675 (produce_symtab): Update.
1677 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1679 * config/s390/s390-opts.h (enum indirect_branch): Define.
1680 * config/s390/s390-protos.h (s390_return_addr_from_memory)
1681 (s390_indirect_branch_via_thunk)
1682 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
1683 (enum s390_indirect_branch_type): Define.
1684 * config/s390/s390.c (struct s390_frame_layout, struct
1685 machine_function): Remove.
1686 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
1687 (indirect_branch_table_label_no, indirect_branch_table_name):
1689 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
1690 (enum s390_indirect_branch_option): Define.
1691 (s390_return_addr_from_memory): New function.
1692 (s390_handle_string_attribute): New function.
1693 (s390_attribute_table): Add new attribute handler.
1694 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
1695 (s390_indirect_branch_via_thunk): New function.
1696 (s390_indirect_branch_via_inline_thunk): New function.
1697 (s390_function_ok_for_sibcall): When jumping via thunk disallow
1698 sibling call optimization for non z10 compiles.
1699 (s390_emit_call): Force indirect branch target to be a single
1700 register. Add r1 clobber for non-z10 compiles.
1701 (s390_emit_epilogue): Emit return jump via return_use expander.
1702 (s390_reorg): Handle JUMP_INSNs as execute targets.
1703 (s390_option_override_internal): Perform validity checks for the
1704 new command line options.
1705 (s390_indirect_branch_attrvalue): New function.
1706 (s390_indirect_branch_settings): New function.
1707 (s390_set_current_function): Invoke s390_indirect_branch_settings.
1708 (s390_output_indirect_thunk_function): New function.
1709 (s390_code_end): Implement target hook.
1710 (s390_case_values_threshold): Implement target hook.
1711 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
1713 * config/s390/s390.h (struct s390_frame_layout)
1714 (struct machine_function): Move here from s390.c.
1715 (TARGET_INDIRECT_BRANCH_NOBP_RET)
1716 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
1717 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
1718 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
1719 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
1720 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
1721 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
1722 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
1723 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
1724 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
1725 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
1726 (mnemonic attribute): Add values which aren't recognized
1728 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
1729 pattern for branch conversion. Fix mnemonic attribute.
1730 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
1731 indirect branch via thunk if requested.
1732 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
1733 ("*indirect_jump"): Disable for branch conversion using out of
1735 ("indirect_jump_via_thunk<mode>_z10")
1736 ("indirect_jump_via_thunk<mode>")
1737 ("indirect_jump_via_inlinethunk<mode>_z10")
1738 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
1739 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
1740 ("casesi_jump_via_inlinethunk<mode>_z10")
1741 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
1742 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
1743 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
1744 ("*indirect2_jump"): Disable for branch conversion.
1745 ("casesi_jump"): Turn into expander and expand patterns for branch
1747 ("return_use"): New expander.
1748 ("*return"): Emit return via thunk and rename it to ...
1749 ("*return<mode>"): ... this one.
1750 * config/s390/s390.opt: Add new options and and enum for the
1753 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1755 * lra-constraints.c (match_reload): Unconditionally use
1756 gen_lowpart_SUBREG, rather than selecting between that
1757 and equivalent gen_rtx_SUBREG code.
1759 2018-02-08 Richard Biener <rguenther@suse.de>
1761 PR tree-optimization/84233
1762 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
1763 changed flag instead of boguously re-using phi_inserted.
1765 2018-02-08 Martin Jambor <mjambor@suse.cz>
1767 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
1768 static local variables.
1770 2018-02-08 Richard Biener <rguenther@suse.de>
1772 PR tree-optimization/84278
1773 * tree-vect-stmts.c (vectorizable_store): When looking for
1774 smaller vector types to perform grouped strided loads/stores
1775 make sure the mode is supported by the target.
1776 (vectorizable_load): Likewise.
1778 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1780 * config/aarch64/aarch64.c (aarch64_components_for_bb):
1781 Increase LDP/STP opportunities by adding adjacent callee-saves.
1783 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1785 PR rtl-optimization/84068
1786 PR rtl-optimization/83459
1787 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
1789 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
1791 PR tree-optimization/84224
1792 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
1793 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
1796 2018-02-07 Iain Sandoe <iain@codesourcery.com>
1799 * config/rs6000/altivec.md (*restore_world): Remove LR use.
1800 * config/rs6000/predicates.md (restore_world_operation): Adjust op
1801 count, remove one USE.
1803 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
1805 * doc/install.texi (Configuration): Document the
1806 --with-long-double-format={ibm,ieee} PowerPC configuration
1810 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
1811 Convert from define_expand to be define_insn_and_split. Rework
1812 float/double/_Float128 conversions to QI/HI/SImode to work with
1813 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
1814 conversions to QI/HImode types did a store and then a load to
1815 truncate the value. For conversions to VSX registers, don't split
1816 the insn, instead emit the code directly. Use the code iterator
1817 any_fix to combine signed and unsigned conversions.
1818 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
1819 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
1820 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
1821 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
1822 (fix_<mode>di2_hw): Likewise.
1823 (fixuns_<mode>di2_hw): Likewise.
1824 (fix_<mode>si2_hw): Likewise.
1825 (fixuns_<mode>si2_hw): Likewise.
1826 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
1827 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
1828 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
1829 fix<uns>_trunc<SFDF:mode>si2_p8.
1830 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
1832 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
1833 (fix<uns>_<mode>_mem): Likewise.
1834 (fctiw<u>z_<mode>_mem): Likewise.
1835 (fix<uns>_<mode>_mem): Likewise.
1836 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
1837 the register allocator from doing a direct move to the GPRs to do
1838 a store, and instead use the ISA 3.0 store byte/half-word from
1839 vector register instruction. For IEEE 128-bit floating point,
1840 also optimize stores of 32-bit ints.
1841 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
1843 2018-02-07 Alan Hayward <alan.hayward@arm.com>
1845 * genextract.c (push_pathstr_operand): New function to support
1847 (walk_rtx): Call push_pathstr_operand.
1848 (print_path): Support [a-zA-Z].
1850 2018-02-07 Richard Biener <rguenther@suse.de>
1852 PR tree-optimization/84037
1853 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
1854 (cse_and_gimplify_to_preheader): Declare.
1855 (vect_get_place_in_interleaving_chain): Likewise.
1856 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
1858 (_loop_vec_info::~_loop_vec_info): Delete it.
1859 (cse_and_gimplify_to_preheader): New function.
1860 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
1861 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
1862 (vectorizable_load): Likewise. For grouped stores always base
1863 the IV on the first element.
1864 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
1865 condition before gimplifying.
1867 2018-02-07 Jakub Jelinek <jakub@redhat.com>
1869 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
1870 *DIV_EXPR and *MOD_EXPR.
1872 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
1875 * config/i386/i386.c (ix86_option_override_internal): Mask out
1876 the CF_SET bit when checking -fcf-protection.
1878 2018-02-07 Tom de Vries <tom@codesourcery.com>
1881 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
1884 2018-02-07 Richard Biener <rguenther@suse.de>
1886 PR tree-optimization/84204
1887 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
1890 PR tree-optimization/84205
1891 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
1892 special-case isl_ast_op_zdiv_r.
1894 PR tree-optimization/84223
1895 * graphite-scop-detection.c (gather_bbs::before_dom_children):
1896 Only add conditions from within the region.
1897 (gather_bbs::after_dom_children): Adjust.
1899 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
1902 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
1903 * config/avr/avr.md: Only post-reload split REG-REG moves if
1904 either register is REGERAL_REG_P.
1906 2018-02-07 Jakub Jelinek <jakub@redhat.com>
1908 PR tree-optimization/84235
1909 * tree-ssa-scopedtables.c
1910 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
1911 if the subtraction is performed in floating point type where NaNs are
1912 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
1913 build 1. Formatting fix.
1915 2018-02-06 Jakub Jelinek <jakub@redhat.com>
1918 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
1919 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
1920 and skip it regardless of bb boundaries. Use CALL_P macro,
1921 don't test INSN_P (insn) together with CALL_P or JUMP_P check
1922 unnecessarily, formatting fix.
1924 2018-02-06 Michael Collison <michael.collison@arm.com>
1926 * config/arm/thumb2.md:
1927 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
1928 (*thumb_mov_notscc): Ditto.
1930 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
1933 * config/rs6000/rs6000.md (su code attribute): Use "u" for
1934 unsigned_fix, not "s".
1936 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1938 * configure.ac (gcc_fn_eh_frame_ro): New function.
1939 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
1940 correct .eh_frame permissions.
1941 * configure: Regenerate.
1943 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
1945 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
1948 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1950 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1951 Display warning message for -mno-speculate-indirect-jumps.
1953 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
1955 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
1957 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
1959 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
1961 PR tree-optimization/84225
1962 * tree-eh.c (find_trapping_overflow): Only call
1963 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
1965 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1968 * config/i386/i386.c: Reimplement the check of possible options
1969 -mibt/-mshstk conbination. Change error messages.
1970 * doc/invoke.texi: Fix a typo: remove extra '='.
1972 2018-02-06 Marek Polacek <polacek@redhat.com>
1974 PR tree-optimization/84228
1975 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
1977 2018-02-06 Tamar Christina <tamar.christina@arm.com>
1980 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
1981 emitted arch directives.
1982 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
1983 __ARM_FEATURE_COPROC before changing architectures.
1985 2018-02-06 Richard Biener <rguenther@suse.de>
1987 * config/i386/i386.c (print_reg): Fix typo.
1988 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
1990 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
1992 * configure: Regenerate.
1994 2018-02-05 Martin Sebor <msebor@redhat.com>
1996 PR tree-optimization/83369
1997 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
2000 2018-02-05 Martin Liska <mliska@suse.cz>
2002 * doc/invoke.texi: Cherry-pick upstream r323995.
2004 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
2006 * ira.c (ira_init_register_move_cost): Adjust comment.
2008 2018-02-05 Martin Liska <mliska@suse.cz>
2010 PR gcov-profile/84137
2011 * doc/gcov.texi: Fix typo in documentation.
2013 2018-02-05 Martin Liska <mliska@suse.cz>
2015 PR gcov-profile/83879
2016 * doc/gcov.texi: Document necessity of --dynamic-list-data when
2017 using dlopen functionality.
2019 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
2021 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
2022 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
2023 _mm_maskz_range_ss, _mm_mask_range_round_ss,
2024 _mm_maskz_range_round_ss): New intrinsics.
2025 (__builtin_ia32_rangesd128_round)
2026 (__builtin_ia32_rangess128_round): Remove.
2027 (__builtin_ia32_rangesd128_mask_round,
2028 __builtin_ia32_rangess128_mask_round): New builtins.
2029 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
2030 __builtin_ia32_rangess128_round): Remove.
2031 (__builtin_ia32_rangesd128_mask_round,
2032 __builtin_ia32_rangess128_mask_round): New builtins.
2033 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
2034 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
2035 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
2036 "<round_saeonly_constraint>")): Changed to ...
2037 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
2038 "<round_saeonly_scalar_constraint>")): ... this.
2039 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
2040 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
2041 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
2042 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
2043 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
2045 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
2047 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
2049 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
2050 Remove all values except native, 8540 and 8548.
2052 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
2054 * config/i386/i386.c (ix86_output_function_return): Pass
2055 INVALID_REGNUM, instead of -1, as invalid register number to
2056 indirect_thunk_name and output_indirect_thunk.
2058 2018-02-02 Julia Koval <julia.koval@intel.com>
2060 * config.gcc: Add -march=icelake.
2061 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
2062 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
2063 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
2064 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
2065 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
2066 (processor_target_table): Add icelake.
2067 (ix86_option_override_internal): Handle new PTAs.
2068 (get_builtin_code_for_version): Handle icelake.
2069 (M_INTEL_COREI7_ICELAKE): New.
2070 (fold_builtin_cpu): Handle icelake.
2071 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
2072 * doc/invoke.texi: Add -march=icelake.
2074 2018-02-02 Julia Koval <julia.koval@intel.com>
2076 * config/i386/i386.c (ix86_option_override_internal): Change flags type
2077 to wide_int_bitmask.
2078 * wide-int-bitmask.h: New.
2080 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2083 * config/i386/i386.md: Replace Pmode with word_mode in
2084 builtin_setjmp_setup and builtin_longjmp to support x32.
2086 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
2090 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
2092 (rs6000_supported_cpu_names): New static variable.
2093 (linux_cpu_translation_table): Likewise.
2094 (elf_platform) <cpu>: Define new static variable and use it.
2095 Translate kernel AT_PLATFORM name to canonical name if needed.
2096 Error if platform name is unknown.
2098 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
2101 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
2103 2018-02-01 Jeff Law <law@redhat.com>
2106 * config/i386/i386.c (release_scratch_register_on_entry): Add new
2107 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
2108 the scratch if RELEASE_VIA_POP is false.
2109 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
2110 If we have to save a temporary register, decrement SIZE appropriately.
2111 Pass new arguments to release_scratch_register_on_entry.
2112 (ix86_adjust_stack_and_probe): Likewise.
2113 (ix86_emit_probe_stack_range): Pass new arguments to
2114 release_scratch_register_on_entry.
2116 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
2118 PR rtl-optimization/84157
2119 * combine.c (change_zero_ext): Use REG_P predicate in
2120 front of HARD_REGISTER_P predicate.
2122 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
2124 * config/avr/avr.c (avr_option_override): Move disabling of
2125 -fdelete-null-pointer-checks to...
2126 * common/config/avr/avr-common.c (avr_option_optimization_table):
2129 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2131 PR tree-optimization/81635
2132 * tree-data-ref.c (split_constant_offset_1): For types that
2133 wrap on overflow, try to use range info to prove that wrapping
2136 2018-02-01 Renlin Li <renlin.li@arm.com>
2139 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
2141 (aarch64_register_move_cost): Likewise.
2142 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
2144 (REG_CLASS_NAMES): Likewise.
2145 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
2146 TAILCALL_ADDR_REGS. Remove IP registers.
2147 * config/aarch64/aarch64.md (Ucs): Update register constraint.
2149 2018-02-01 Richard Biener <rguenther@suse.de>
2151 * domwalk.h (dom_walker::dom_walker): Add additional constructor
2152 for specifying RPO order and allow NULL for that.
2153 * domwalk.c (dom_walker::dom_walker): Likewise.
2154 (dom_walker::walk): Handle NULL RPO order.
2155 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
2157 (rewrite_update_dom_walker): Likewise.
2158 (mark_def_dom_walker): Likewise.
2160 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2162 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
2163 (aarch64_maybe_expand_sve_subreg_move): Declare.
2164 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
2165 * config/aarch64/predicates.md (aarch64_any_register_operand): New
2167 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
2168 that are semantically a reverse operation.
2169 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
2170 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
2171 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
2173 (aarch64_can_change_mode_class): For big-endian, forbid changes
2174 between two SVE modes if they have different element sizes.
2176 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2178 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
2179 the TImode handling for big-endian targets.
2181 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2183 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
2184 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
2186 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
2187 Remove BSWAP handing for big-endian targets and use the form of
2188 LD1RQ appropariate for the mode.
2190 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2192 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
2193 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
2196 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2199 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
2200 check for operands that need to go through aarch64_sve_reload_be.
2202 2018-02-01 Jakub Jelinek <jakub@redhat.com>
2204 PR tree-optimization/81661
2205 PR tree-optimization/84117
2206 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
2207 * tree-eh.c: Include gimplify.h.
2208 (find_trapping_overflow, replace_trapping_overflow,
2209 rewrite_to_non_trapping_overflow): New functions.
2210 * tree-vect-loop.c: Include tree-eh.h.
2211 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
2212 * tree-data-ref.c: Include tree-eh.h.
2213 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
2215 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
2217 PR rtl-optimization/84123
2218 * combine.c (change_zero_ext): Check if hard register satisfies
2219 can_change_dest_mode before calling gen_lowpart_SUBREG.
2221 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
2224 * ira.c (ira_init_register_move_cost): Remove assert.
2226 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2228 PR rtl-optimization/84071
2229 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
2230 * doc/tm.texi: Regenerate.
2232 2018-01-31 Richard Biener <rguenther@suse.de>
2234 PR tree-optimization/84132
2235 * tree-data-ref.c (analyze_miv_subscript): Properly
2236 check whether evolution_function_is_affine_multivariate_p
2237 before calling gcd_of_steps_may_divide_p.
2239 2018-01-31 Julia Koval <julia.koval@intel.com>
2242 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
2243 * config/i386/i386.md (rdpid_rex64) New.
2244 (rdpid): Make 32bit only.
2246 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
2249 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
2250 an IDENTIFIER_NODE for FUNCTION_TYPE's.
2252 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2255 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
2257 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
2259 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2261 PR rtl-optimization/84071
2262 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
2263 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
2265 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2267 * config/arc/arc.c (arc_handle_aux_attribute): New function.
2268 (arc_attribute_table): Add 'aux' attribute.
2269 (arc_in_small_data_p): Consider aux like variables.
2270 (arc_is_aux_reg_p): New function.
2271 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
2272 (arc_get_aux_arg): New function.
2273 (prepare_move_operands): Handle aux-register access.
2274 (arc_handle_aux_attribute): New function.
2275 * doc/extend.texi (ARC Variable attributes): Add subsection.
2277 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2279 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
2280 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
2281 (arc_attribute_table): Add 'uncached' attribute.
2282 (arc_print_operand): Print '.di' flag for uncached memory
2284 (arc_in_small_data_p): Do not consider for small data the uncached
2286 (arc_is_uncached_mem_p): New function.
2287 * config/arc/predicates.md (compact_store_memory_operand): Check
2288 for uncached memory accesses.
2289 (nonvol_nonimm_operand): Likewise.
2290 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
2292 2018-01-31 Jakub Jelinek <jakub@redhat.com>
2295 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
2296 falign-loops=): Add Optimization flag.
2298 2018-01-30 Jeff Law <law@redhat.com>
2301 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
2302 INT_REGISTERS_SAVED. Check it prior to calling
2303 get_scratch_register_on_entry.
2304 (ix86_adjust_stack_and_probe): Similarly.
2305 (ix86_emit_probe_stack_range): Similarly.
2306 (ix86_expand_prologue): Corresponding changes.
2308 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2311 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
2312 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
2314 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
2317 * lra-constraints.c (curr_insn_transform): Process AND in the
2320 2018-01-30 Jakub Jelinek <jakub@redhat.com>
2322 PR rtl-optimization/83986
2323 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
2324 dependence against last_pending_memory_flush in addition to
2327 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
2329 PR tree-optimization/81611
2330 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
2333 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2336 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
2339 2018-01-30 Richard Biener <rguenther@suse.de>
2340 Jakub Jelinek <jakub@redhat.com>
2342 PR tree-optimization/84111
2343 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
2344 inner loops added during recursion, as they don't have up-to-date
2347 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2350 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
2351 (can_inline_edge_by_limits_p): ... here.
2352 (can_early_inline_edge_p, check_callers,
2353 update_caller_keys, update_callee_keys, recursive_inlining,
2354 add_new_edges_to_heap, speculation_useful_p,
2355 inline_small_functions,
2356 inline_small_functions, flatten_function,
2357 inline_to_all_callers_1): Update.
2359 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2361 * profile-count.c (profile_count::combine_with_ipa_count): Handle
2364 2018-01-30 Richard Biener <rguenther@suse.de>
2366 PR tree-optimization/83008
2367 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
2368 invariant and constant vector uses in stmts when they need
2371 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2374 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
2375 * configure: Regenerate.
2377 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2379 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
2381 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
2382 Use gen_rtx_REG rather than gen_lowpart.
2384 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2386 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
2387 rather than 0 when creating partial subregs.
2389 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2391 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
2394 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
2397 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
2398 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
2399 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
2400 flags. This restores the settings used before the 2017-07-24.
2401 Turning off pre increment/decrement/modify allows IVOPTS to
2402 optimize DF/SF loops where the index is an int.
2404 2018-01-29 Richard Biener <rguenther@suse.de>
2405 Kelvin Nilsen <kelvin@gcc.gnu.org>
2408 * tree-vect-stmts.c (vectorizable_call): Don't call
2409 targetm.vectorize_builtin_md_vectorized_function if callee is
2412 2018-01-22 Carl Love <cel@us.ibm.com>
2414 * doc/extend.tex: Fix typo in second arg in
2415 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
2417 2018-01-29 Richard Biener <rguenther@suse.de>
2419 PR tree-optimization/84086
2420 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
2421 (flush_ssaname_freelist): When SSA names were released reset
2422 the SCEV hash table.
2424 2018-01-29 Richard Biener <rguenther@suse.de>
2426 PR tree-optimization/84057
2427 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
2428 removed paths when removing edges.
2430 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
2432 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
2433 -mfunction-return=@var{choice}.
2435 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
2438 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
2440 (layout::print_source_line): Likewise.
2441 (test_get_line_width_without_trailing_whitespace): Add test cases.
2443 2018-01-27 Jakub Jelinek <jakub@redhat.com>
2446 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
2449 2018-01-26 Jim Wilson <jimw@sifive.com>
2451 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
2453 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
2456 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2458 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
2459 and CMP + SUB-immediate -> SUBS.
2461 2018-01-26 Martin Sebor <msebor@redhat.com>
2463 PR tree-optimization/83896
2464 * tree-ssa-strlen.c (get_string_len): Rename...
2465 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
2466 Avoid assuming length is constant.
2467 (handle_char_store): Use HOST_WIDE_INT for string length.
2469 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
2472 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
2473 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
2475 2018-01-26 Richard Biener <rguenther@suse.de>
2477 PR rtl-optimization/84003
2478 * dse.c (record_store): Only record redundant stores when
2479 the earlier store aliases at least all accesses the later one does.
2481 2018-01-26 Jakub Jelinek <jakub@redhat.com>
2483 PR rtl-optimization/83985
2484 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
2485 REG_CFA_RESTORE insns.
2486 (delete_unmarked_insns): Don't ignore separate shrink wrapping
2487 REG_CFA_RESTORE insns here.
2490 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
2491 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
2493 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2495 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
2496 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
2497 (arc_init): Likewise.
2498 (arc_override_options): Likewise.
2499 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
2501 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
2503 * config/arc/arc.h (TARGET_DBNZ): Define.
2504 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
2505 properly set the tune attribute.
2506 (dbnz): Use TARGET_DBNZ guard.
2507 * config/arc/arc.opt (mtune): Add core3 option.
2509 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2511 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
2512 recognize new pic like addresses.
2513 (arc_delegitimize_address): Clean up.
2515 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2517 * config/arc/arc-arches.def: Option mrf16 valid for all
2519 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
2520 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
2521 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
2522 * config/arc/arc-tables.opt: Regenerate.
2523 * config/arc/arc.c (arc_conditional_register_usage): Handle
2524 reduced register file case.
2525 (arc_file_start): Set must have build attributes.
2526 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
2528 * config/arc/arc.opt (mrf16): Add new option.
2529 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
2530 * config/arc/genmultilib.awk: Handle new mrf16 option.
2531 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
2532 * config/arc/t-multilib: Regenerate.
2533 * doc/invoke.texi (ARC Options): Document mrf16 option.
2535 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2537 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
2538 * config/arc/arc.c (arc_handle_secure_attribute): New function.
2539 (arc_attribute_table): Add 'secure_call' attribute.
2540 (arc_print_operand): Print secure call operand.
2541 (arc_function_ok_for_sibcall): Don't optimize tail calls when
2543 (arc_is_secure_call_p): New function. * config/arc/arc.md
2544 (call_i): Add support for sjli instruction.
2545 (call_value_i): Likewise.
2546 * config/arc/constraints.md (Csc): New constraint.
2548 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2549 John Eric Martin <John.Martin@emmicro-us.com>
2551 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
2552 * config/arc/arc.c (_arc_jli_section): New struct.
2553 (arc_jli_section): New type.
2554 (rc_jli_sections): New static variable.
2555 (arc_handle_jli_attribute): New function.
2556 (arc_attribute_table): Add jli_always and jli_fixed attribute.
2557 (arc_file_end): New function.
2558 (TARGET_ASM_FILE_END): Define.
2559 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
2560 (arc_add_jli_section): New function.
2561 (jli_call_scan): Likewise.
2562 (arc_reorg): Call jli_call_scan.
2563 (arc_output_addsi): Remove 'S' from printing asm operand.
2564 (arc_is_jli_call_p): New function.
2565 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
2567 (movhi_insn): Likewise.
2568 (movsi_insn): Likewise.
2569 (movsi_set_cc_insn): Likewise.
2570 (loadqi_update): Likewise.
2571 (load_zeroextendqisi_update): Likewise.
2572 (load_signextendqisi_update): Likewise.
2573 (loadhi_update): Likewise.
2574 (load_zeroextendhisi_update): Likewise.
2575 (load_signextendhisi_update): Likewise.
2576 (loadsi_update): Likewise.
2577 (loadsf_update): Likewise.
2578 (movsicc_insn): Likewise.
2579 (bset_insn): Likewise.
2580 (bxor_insn): Likewise.
2581 (bclr_insn): Likewise.
2582 (bmsk_insn): Likewise.
2583 (bicsi3_insn): Likewise.
2584 (cmpsi_cc_c_insn): Likewise.
2585 (movsi_ne): Likewise.
2586 (movsi_cond_exec): Likewise.
2587 (clrsbsi2): Likewise.
2598 (call_i): Remove 'S' asm letter, add jli instruction.
2599 (call_value_i): Likewise.
2600 * config/arc/arc.op (mjli-always): New option.
2601 * config/arc/constraints.md (Cji): New constraint.
2602 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
2604 (subsf3_fpx): Likewise.
2605 (mulsf3_fpx): Likewise.
2606 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
2608 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
2610 * doc/invoke.texi (ARC): Document mjli-always option.
2612 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
2614 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
2615 avoid addition with 0 and use incw and decw where possible.
2617 2018-01-26 Richard Biener <rguenther@suse.de>
2619 PR tree-optimization/81082
2620 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
2621 association if it requires casting to unsigned.
2622 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
2623 from fold_plusminus_mult_expr to catch important cases late when
2624 range info is available.
2626 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2628 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
2629 * configure.ac (hidden_linkonce): New test.
2630 * configure: Regenerate.
2631 * config.in: Regenerate.
2633 2018-01-26 Julia Koval <julia.koval@intel.com>
2635 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
2636 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
2637 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
2638 _mm_mask_bitshuffle_epi64_mask): Fix type.
2639 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
2640 USI_FTYPE_V4DI_V4DI_USI): Remove.
2641 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
2642 __builtin_ia32_vpshufbitqmb256_mask,
2643 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
2644 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
2645 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
2647 2018-01-26 Alan Modra <amodra@gmail.com>
2650 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
2651 UNSPEC_VBPERMQ. Sort other unspecs.
2653 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
2655 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
2657 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
2660 * predict.c (drop_profile): Do not push/pop cfun; update also
2662 (handle_missing_profiles): Fix logic looking for zero profiles.
2664 2018-01-25 Jakub Jelinek <jakub@redhat.com>
2667 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
2668 on functions with #pragma omp declare simd or functions with simd
2670 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
2671 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
2672 Remove trailing \n from warning_at calls.
2674 2018-01-25 Tom de Vries <tom@codesourcery.com>
2677 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2678 for neutered workers.
2680 2018-01-24 Joseph Myers <joseph@codesourcery.com>
2683 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
2684 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
2686 2018-01-24 Jeff Law <law@redhat.com>
2689 * i386.c (get_probe_interval): Move to earlier point.
2690 (ix86_compute_frame_layout): If -fstack-clash-protection and
2691 the frame is larger than the probe interval, then use pushes
2692 to save registers rather than reg->mem moves.
2693 (ix86_expand_prologue): Remove conditional for int_registers_saved
2696 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
2699 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
2700 min/max for never referenced object.
2702 2018-01-24 Jakub Jelinek <jakub@redhat.com>
2705 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
2707 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
2708 attributes from DECL_ATTRIBUTES (decl) without affecting
2709 DECL_ATTRIBUTES (current_function_decl).
2710 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
2711 functions with non-NULL DECL_ABSTRACT_ORIGIN.
2713 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
2715 PR tree-optimization/83979
2716 * fold-const.c (fold_comparison): Use constant_boolean_node
2717 instead of boolean_{true,false}_node.
2719 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
2721 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
2724 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2726 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
2727 Simplify the clause that sets the length attribute.
2728 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
2729 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
2730 clause that sets the length attribute.
2731 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
2733 2018-01-24 Tom de Vries <tom@codesourcery.com>
2736 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
2737 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
2738 Add strict parameter.
2739 (prevent_branch_around_nothing): Insert dummy insn between branch to
2740 label and label with no ptx insn inbetween.
2741 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
2743 2018-01-24 Tom de Vries <tom@codesourcery.com>
2746 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2747 for neutered threads in warp.
2748 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
2750 2018-01-24 Richard Biener <rguenther@suse.de>
2752 PR tree-optimization/83176
2753 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
2756 2018-01-24 Richard Biener <rguenther@suse.de>
2758 PR tree-optimization/82819
2759 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
2760 code generating pluses that are no-ops in the target precision.
2762 2018-01-24 Richard Biener <rguenther@suse.de>
2765 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
2767 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2769 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
2770 to merge probabilities.
2771 * predict.c (probably_never_executed): Also mark as cold functions
2772 with global 0 profile and guessed local profile.
2773 * profile-count.c (profile_probability::combine_with_count): New
2775 * profile-count.h (profile_probability::operator*,
2776 profile_probability::operator*=, profile_probability::operator/,
2777 profile_probability::operator/=): Reduce precision to adjusted
2778 and set value to guessed on contradictory divisions.
2779 (profile_probability::combine_with_freq): Remove.
2780 (profile_probability::combine_wiht_count): Declare.
2781 (profile_count::force_nonzero):: Set to adjusted.
2782 (profile_count::probability_in):: Set quality to adjusted.
2783 * tree-ssa-tail-merge.c (replace_block_by): Use
2786 2018-01-23 Andrew Waterman <andrew@sifive.com>
2787 Jim Wilson <jimw@sifive.com>
2789 * config/riscv/riscv.c (riscv_stack_boundary): New.
2790 (riscv_option_override): Set riscv_stack_boundary. Handle
2791 riscv_preferred_stack_boundary_arg.
2792 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
2793 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
2794 (STACK_BOUNDARY): Set to riscv_stack_boundary.
2795 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
2796 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
2797 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
2799 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
2802 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
2803 of struct ix86_frame.
2804 (ix86_expand_epilogue): Likewise. Add a local variable for
2805 the reg_save_offset field in struct ix86_frame.
2807 2018-01-23 Bin Cheng <bin.cheng@arm.com>
2809 PR tree-optimization/82604
2810 * tree-loop-distribution.c (enum partition_kind): New enum item
2811 PKIND_PARTIAL_MEMSET.
2812 (partition_builtin_p): Support above new enum item.
2813 (generate_code_for_partition): Ditto.
2814 (compute_access_range): Differentiate cases that equality can be
2815 proven at all loops, the innermost loops or no loops.
2816 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
2817 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
2818 (finalize_partitions, distribute_loop): Don't fuse partition of
2819 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
2820 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
2823 2018-01-23 Martin Liska <mliska@suse.cz>
2825 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
2826 order to ignore the predictor.
2827 (PRED_POLYMORPHIC_CALL): Likewise.
2828 (PRED_RECURSIVE_CALL): Likewise.
2830 2018-01-23 Martin Liska <mliska@suse.cz>
2832 * tree-profile.c (tree_profiling): Print function header to
2833 aware reader which function we are working on.
2834 * value-prof.c (gimple_find_values_to_profile): Do not print
2835 not interesting value histograms.
2837 2018-01-23 Martin Liska <mliska@suse.cz>
2839 * profile-count.h (enum profile_quality): Add
2840 profile_uninitialized as the first value. Do not number values
2841 as they are zero based.
2842 (profile_count::verify): Update sanity check.
2843 (profile_probability::verify): Likewise.
2845 2018-01-23 Nathan Sidwell <nathan@acm.org>
2847 * doc/invoke.texi (ffor-scope): Deprecate.
2849 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2851 PR tree-optimization/83510
2852 * domwalk.c (set_all_edges_as_executable): New function.
2853 (dom_walker::dom_walker): Convert bool param
2854 "skip_unreachable_blocks" to enum reachability. Move setup of
2855 edge flags to set_all_edges_as_executable and only do it when
2856 reachability is REACHABLE_BLOCKS.
2857 * domwalk.h (enum dom_walker::reachability): New enum.
2858 (dom_walker::dom_walker): Convert bool param
2859 "skip_unreachable_blocks" to enum reachability.
2860 (set_all_edges_as_executable): New decl.
2861 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
2862 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
2864 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
2865 but converting true to REACHABLE_BLOCKS.
2866 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
2868 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
2869 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
2870 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
2872 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
2873 if check_all_array_refs will be called.
2875 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2877 * tree.c (selftest::test_location_wrappers): Add more test
2880 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2882 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
2883 (selftest::test_bit_in_range): Likewise.
2885 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
2888 * doc/sourcebuild.texi (vect_float): Say that the selector
2889 only describes the situation when -funsafe-math-optimizations is on.
2890 (vect_float_strict): Document.
2892 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
2894 PR tree-optimization/83965
2895 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
2896 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
2897 instead of checking only for a reduction.
2898 (vect_recog_widen_sum_pattern): Likewise.
2900 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2902 * predict.c (probably_never_executed): Only use precise profile info.
2903 (compute_function_frequency): Skip after inlining hack since we now
2904 have quality checking.
2906 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2908 * profile-count.h (profile_probability::very_unlikely,
2909 profile_probability::unlikely, profile_probability::even): Set
2910 precision to guessed.
2912 2018-01-23 Richard Biener <rguenther@suse.de>
2914 PR tree-optimization/83963
2915 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
2916 Properly terminate dominator walk when crossing the exit edge not
2917 when visiting its source block.
2919 2018-01-23 Jakub Jelinek <jakub@redhat.com>
2922 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
2923 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
2925 2018-01-22 Jakub Jelinek <jakub@redhat.com>
2927 PR tree-optimization/83957
2928 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
2929 semicolon after for body surrounded by braces.
2931 PR tree-optimization/83081
2932 * profile-count.h (profile_probability::split): New method.
2933 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
2934 Use profile_probability::split.
2935 (do_compare_rtx_and_jump): Fix adjustment of probabilities
2936 when splitting a single conditional jump into 2.
2938 2018-01-22 David Malcolm <dmalcolm@redhat.com>
2940 PR tree-optimization/69452
2941 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
2944 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2946 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
2947 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
2948 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
2950 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2952 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
2953 * config/rl78/rl78.md (movdi): New define_expand.
2954 * config/rl78/rl78.c (rl78_split_movdi): New function.
2956 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
2959 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
2961 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
2962 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
2963 128-bit to produce an UNSPEC move to get the double word with the
2964 signbit and then a shift directly to do signbit.
2965 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
2966 implementation with a new version that just does either a direct
2967 move or a regular move. Move memory interface to separate insns.
2968 Move insns so they are next to the expander.
2969 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
2970 with signbit move. Split big and little endian case.
2971 (signbit<mode>2_dm_mem_le): Likewise.
2972 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
2973 (signbit<mode>2_dm2): Likewise.
2975 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2977 * config/rl78/rl78.md (anddi3): New define_expand.
2979 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2981 * config/rl78/rl78.md (umindi3): New define_expand.
2983 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2985 * config/rl78/rl78.md (smindi3): New define_expand.
2987 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2989 * config/rl78/rl78.md (smaxdi3): New define_expand.
2991 2018-01-22 Carl Love <cel@us.ibm.com>
2993 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
2994 LVX_V1TI): Add macro expansion.
2995 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
2996 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
2997 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
2998 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
2999 Change check to determine if the instruction is a byte reversing
3000 entry. Fix typo in comment.
3001 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
3002 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
3003 Add def_builtin calls for new builtins.
3004 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
3005 Add define_insn expansion.
3007 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3009 * config/rl78/rl78.md (umaxdi3): New define_expand.
3011 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3013 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
3014 for non-QImode registers.
3016 2018-01-22 Richard Biener <rguenther@suse.de>
3018 PR tree-optimization/83963
3019 * graphite-scop-detection.c (scop_detection::get_sese): Delay
3020 including the loop exit block.
3021 (scop_detection::merge_sese): Likewise.
3022 (scop_detection::add_scop): Do it here instead.
3024 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3026 * doc/sourcebuild.texi (arm_softfloat): Document.
3028 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
3031 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
3032 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
3033 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
3035 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3036 David Edelsohn <dje.gcc@gmail.com>
3039 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3040 Change "crset eq" to "crset 2".
3041 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3042 (*call_indirect_aix<mode>_nospec): Likewise.
3043 (*call_value_indirect_aix<mode>_nospec): Likewise.
3044 (*call_indirect_elfv2<mode>_nospec): Likewise.
3045 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
3046 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
3047 change assembly output from . to $.
3048 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3049 (indirect_jump<mode>_nospec): Change assembly output from . to $.
3050 (*tablejump<mode>_internal1_nospec): Likewise.
3052 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
3055 * config/sh/sh_optimize_sett_clrt.cc:
3056 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
3058 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
3060 PR tree-optimization/83940
3061 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
3062 offset_dt to vect_constant_def rather than vect_unknown_def_type.
3063 (vect_check_load_store_mask): Add a mask_dt_out parameter and
3064 use it to pass back the definition type.
3065 (vect_check_store_rhs): Likewise rhs_dt_out.
3066 (vect_build_gather_load_calls): Add a mask_dt argument and use
3067 it instead of a call to vect_is_simple_use.
3068 (vectorizable_store): Update calls to vect_check_load_store_mask
3069 and vect_check_store_rhs. Use the dt returned by the latter instead
3070 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
3071 instead of calls to vect_is_simple_use. Pass the scalar rather
3072 than the vector operand to vect_is_simple_use when handling
3073 second and subsequent copies of an rhs value.
3074 (vectorizable_load): Update calls to vect_check_load_store_mask
3075 and vect_build_gather_load_calls. Use the cached mask_dt and
3076 gs_info.offset_dt instead of calls to vect_is_simple_use.
3078 2018-01-20 Jakub Jelinek <jakub@redhat.com>
3081 * tree-emutls.c: Include gimplify.h.
3082 (lower_emutls_2): New function.
3083 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
3084 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
3085 it before further processing.
3088 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
3089 UINTVAL (trueop1) instead of INTVAL (op1).
3091 2018-01-19 Jakub Jelinek <jakub@redhat.com>
3095 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
3096 INCOMING_FRAME_SP_OFFSET if not defined.
3097 (scan_trace): Add ENTRY argument. If true and
3098 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
3099 emit a note to adjust the CFA offset.
3100 (create_cfi_notes): Adjust scan_trace callers.
3101 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
3102 INCOMING_FRAME_SP_OFFSET in the CIE.
3103 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
3104 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
3106 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
3107 * doc/tm.texi: Regenerated.
3109 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3111 PR rtl-optimization/83147
3112 * lra-constraints.c (remove_inheritance_pseudos): Use
3113 lra_substitute_pseudo_within_insn.
3115 2018-01-19 Tom de Vries <tom@codesourcery.com>
3116 Cesar Philippidis <cesar@codesourcery.com>
3119 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
3121 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
3124 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
3125 spaces for function labels.
3127 2018-01-19 Martin Liska <mliska@suse.cz>
3129 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
3130 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
3131 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
3132 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
3133 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
3134 (PRED_CONST_RETURN): Change from 69 to 65.
3135 (PRED_NULL_RETURN): Change from 91 to 71.
3136 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
3137 (PRED_LOOP_GUARD): Change from 66 to 73.
3139 2018-01-19 Martin Liska <mliska@suse.cz>
3141 * predict.c (predict_insn_def): Add new assert.
3142 (struct branch_predictor): Change type to signed integer.
3143 (test_prediction_value_range): Amend test to cover
3145 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
3146 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
3147 (PRED_LOOP_ITERATIONS_MAX): Likewise.
3148 (PRED_LOOP_IV_COMPARE): Likewise.
3149 * predict.h (PROB_UNINITIALIZED): Define new constant.
3151 2018-01-19 Martin Liska <mliska@suse.cz>
3153 * predict.c (dump_prediction): Add new format for
3154 analyze_brprob.py script which is enabled with -details
3156 * profile-count.h (precise_p): New function.
3158 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3160 PR tree-optimization/83922
3161 * tree-vect-loop.c (vect_verify_full_masking): Return false if
3162 there are no statements that need masking.
3163 (vect_active_double_reduction_p): New function.
3164 (vect_analyze_loop_operations): Use it when handling phis that
3165 are not in the loop header.
3167 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3169 PR tree-optimization/83914
3170 * tree-vect-loop.c (vectorizable_induction): Don't convert
3171 init_expr or apply the peeling adjustment for inductions
3172 that are nested within the vectorized loop.
3174 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3176 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
3179 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3183 * function.h (gimplify_parameters): Add gimple_seq * argument.
3184 * function.c: Include gimple.h and options.h.
3185 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
3186 for the added local temporaries if needed.
3187 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
3188 if there are any parameter cleanups, wrap whole body into a
3189 try/finally with the cleanups.
3191 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
3194 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
3195 Use GET_MODE_CLASS for scalar floating point.
3197 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3201 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3202 Fix call of call_cgraph_insertion_hooks.
3204 2018-01-18 Martin Sebor <msebor@redhat.com>
3206 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
3208 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3211 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
3214 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
3217 * common.opt: (-ffile-prefix-map): New option.
3218 * opts.c (common_handle_option): Defer it.
3219 * opts-global.c (handle_common_deferred_options): Handle it.
3220 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
3221 * file-prefix-map.h: New file.
3222 (remap_debug_filename, add_debug_prefix_map): ...here.
3223 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
3224 * final.c (debug_prefix_map, add_debug_prefix_map
3225 remap_debug_filename): Move to...
3226 * file-prefix-map.c: New file.
3227 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
3228 generalize, get rid of alloca(), use strrchr() instead of strchr().
3229 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
3230 Implement in terms of add_prefix_map().
3231 (remap_macro_filename, remap_debug_filename): Implement in term of
3233 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
3234 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
3235 * dbxout.c: Include file-prefix-map.h.
3236 * varasm.c: Likewise.
3237 * vmsdbgout.c: Likewise.
3238 * xcoffout.c: Likewise.
3239 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
3240 * doc/cppopts.texi (-fmacro-prefix-map): Document.
3241 * doc/invoke.texi (-ffile-prefix-map): Document.
3242 (-fdebug-prefix-map): Update description.
3244 2018-01-18 Martin Liska <mliska@suse.cz>
3246 * config/i386/i386.c (indirect_thunk_name): Document that also
3248 (output_indirect_thunk): Document why both instructions
3249 (pause and lfence) are generated.
3251 2018-01-18 Richard Biener <rguenther@suse.de>
3253 PR tree-optimization/83887
3254 * graphite-scop-detection.c
3255 (scop_detection::get_nearest_dom_with_single_entry): Remove.
3256 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
3257 (scop_detection::merge_sese): Re-implement with a flood-fill
3258 algorithm that properly finds a SESE region if it exists.
3260 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3263 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
3264 pointer_diff optimizations use view_convert instead of convert.
3266 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3268 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3269 Generate different code for -mno-speculate-indirect-jumps.
3270 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3271 (*call_indirect_aix<mode>): Disable for
3272 -mno-speculate-indirect-jumps.
3273 (*call_indirect_aix<mode>_nospec): New define_insn.
3274 (*call_value_indirect_aix<mode>): Disable for
3275 -mno-speculate-indirect-jumps.
3276 (*call_value_indirect_aix<mode>_nospec): New define_insn.
3277 (*sibcall_nonlocal_sysv<mode>): Generate different code for
3278 -mno-speculate-indirect-jumps.
3279 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3281 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3283 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
3284 long double type, set the flags for noting the default long double
3285 type, even if we don't pass or return a long double type.
3287 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
3290 * ipa-inline.c (flatten_function): Do not overwrite final inlining
3293 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
3295 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
3296 support for merge[hl].
3297 (fold_mergehl_helper): New helper function.
3298 (tree-vector-builder.h): New #include for tree_vector_builder usage.
3299 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
3300 (altivec_vmrglw_direct): Add xxmrglw insn.
3302 2018-01-17 Andrew Waterman <andrew@sifive.com>
3304 * config/riscv/riscv.c (riscv_conditional_register_usage): If
3305 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
3307 2018-01-17 David Malcolm <dmalcolm@redhat.com>
3310 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
3311 call the lto_location_cache before reading the
3312 DECL_SOURCE_LOCATION of the types.
3314 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3315 Richard Sandiford <richard.sandiford@linaro.org>
3317 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
3318 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
3319 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
3320 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
3321 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
3323 * config/aarch64/constraints.md (aarch64_movti_operand):
3325 * config/aarch64/predicates.md (Uti): Add new constraint.
3327 2018-01-17 Carl Love <cel@us.ibm.com>
3329 * config/rs6000/vsx.md (define_expand xl_len_r,
3330 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
3331 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
3333 (define_expand, define_insn): Move the shift left from the
3334 define_insn to the define_expand for lxvl and stxvl instructions.
3335 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
3336 and XL_LEN_R definitions to PURE.
3338 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
3340 * config/i386/i386.c (indirect_thunk_name): Declare regno
3341 as unsigned int. Compare regno with INVALID_REGNUM.
3342 (output_indirect_thunk): Ditto.
3343 (output_indirect_thunk_function): Ditto.
3344 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
3345 in the call to output_indirect_thunk_function.
3347 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
3350 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
3351 rather than the size of inner_type to determine the stack slot size
3352 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
3354 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
3357 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
3360 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
3362 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
3363 endian Linux systems to optionally enable multilibs for selecting
3364 the long double type if the user configured an explicit type.
3365 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
3366 have no long double multilibs if not defined.
3367 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
3368 warn if the user used -mabi={ieee,ibm}longdouble and we built
3369 multilibs for long double.
3370 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
3371 appropriate multilib option.
3372 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
3374 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
3375 for building long double multilibs.
3376 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
3378 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
3380 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
3383 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
3385 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
3388 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
3391 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
3394 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3396 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
3397 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
3399 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3401 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
3402 different rtl trees depending on TARGET_64BIT.
3403 (rs6000_gen_lvx): Likewise.
3405 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3407 * config/visium/visium.md (nop): Tweak comment.
3408 (hazard_nop): Likewise.
3410 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3412 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
3413 -mspeculate-indirect-jumps.
3414 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
3415 for -mno-speculate-indirect-jumps.
3416 (*call_indirect_elfv2<mode>_nospec): New define_insn.
3417 (*call_value_indirect_elfv2<mode>): Disable for
3418 -mno-speculate-indirect-jumps.
3419 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
3420 (indirect_jump): Emit different RTL for
3421 -mno-speculate-indirect-jumps.
3422 (*indirect_jump<mode>): Disable for
3423 -mno-speculate-indirect-jumps.
3424 (*indirect_jump<mode>_nospec): New define_insn.
3425 (tablejump): Emit different RTL for
3426 -mno-speculate-indirect-jumps.
3427 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
3428 (tablejumpsi_nospec): New define_expand.
3429 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
3430 (tablejumpdi_nospec): New define_expand.
3431 (*tablejump<mode>_internal1): Disable for
3432 -mno-speculate-indirect-jumps.
3433 (*tablejump<mode>_internal1_nospec): New define_insn.
3434 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
3437 2018-01-16 Artyom Skrobov tyomitch@gmail.com
3439 * caller-save.c (insert_save): Drop unnecessary parameter. All
3442 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3443 Richard Biener <rguenth@suse.de>
3446 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
3447 return early, inline manually is_gimple_sizepos. Make sure if we
3448 call gimplify_expr we don't end up with a gimple constant.
3449 * tree.c (variably_modified_type_p): Don't return true for
3450 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
3451 * gimplify.h (is_gimple_sizepos): Remove.
3453 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3455 PR tree-optimization/83857
3456 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
3457 vectorizable_live_operation for pure SLP statements.
3458 (vectorizable_live_operation): Handle PHIs.
3460 2018-01-16 Richard Biener <rguenther@suse.de>
3462 PR tree-optimization/83867
3463 * tree-vect-stmts.c (vect_transform_stmt): Precompute
3464 nested_in_vect_loop_p since the scalar stmt may get invalidated.
3466 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3469 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
3470 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
3471 If off is not INTEGER_CST, issue a may not be aligned warning
3472 rather than isn't aligned. Use isn%'t rather than isn't.
3473 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
3475 <case MULT_EXPR>: Improve the case when bottom and one of the
3476 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
3477 operand, in that case check if the other operand is multiple of
3478 bottom divided by the INTEGER_CST operand.
3480 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3483 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
3484 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
3485 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
3486 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
3487 * config/pa/pa.c (pa_function_arg_advance): Likewise.
3488 (pa_function_arg, pa_arg_partial_bytes): Likewise.
3489 (pa_function_arg_size): New function.
3491 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3493 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
3494 in a separate statement.
3496 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3498 PR tree-optimization/83847
3499 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
3500 group gathers and scatters.
3502 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3504 PR rtl-optimization/86620
3505 * params.def (max-sched-ready-insns): Bump minimum value to 1.
3507 PR rtl-optimization/83213
3508 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
3509 to last if both are JUMP_INSNs.
3511 PR tree-optimization/83843
3512 * gimple-ssa-store-merging.c
3513 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
3514 store_immediate_info for bswap/nop orig_stores.
3516 2018-01-15 Andrew Waterman <andrew@sifive.com>
3518 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
3520 <UDIV>: Increase cost if !TARGET_DIV.
3522 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
3524 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
3525 (define_attr "cr_logical_3op"): New.
3526 (cceq_ior_compare): Adjust.
3527 (cceq_ior_compare_complement): Adjust.
3528 (*cceq_rev_compare): Adjust.
3529 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
3530 (is_cracked_insn): Adjust.
3531 (insn_must_be_first_in_group): Adjust.
3532 * config/rs6000/40x.md: Adjust.
3533 * config/rs6000/440.md: Adjust.
3534 * config/rs6000/476.md: Adjust.
3535 * config/rs6000/601.md: Adjust.
3536 * config/rs6000/603.md: Adjust.
3537 * config/rs6000/6xx.md: Adjust.
3538 * config/rs6000/7450.md: Adjust.
3539 * config/rs6000/7xx.md: Adjust.
3540 * config/rs6000/8540.md: Adjust.
3541 * config/rs6000/cell.md: Adjust.
3542 * config/rs6000/e300c2c3.md: Adjust.
3543 * config/rs6000/e500mc.md: Adjust.
3544 * config/rs6000/e500mc64.md: Adjust.
3545 * config/rs6000/e5500.md: Adjust.
3546 * config/rs6000/e6500.md: Adjust.
3547 * config/rs6000/mpc.md: Adjust.
3548 * config/rs6000/power4.md: Adjust.
3549 * config/rs6000/power5.md: Adjust.
3550 * config/rs6000/power6.md: Adjust.
3551 * config/rs6000/power7.md: Adjust.
3552 * config/rs6000/power8.md: Adjust.
3553 * config/rs6000/power9.md: Adjust.
3554 * config/rs6000/rs64.md: Adjust.
3555 * config/rs6000/titan.md: Adjust.
3557 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3559 * config/i386/predicates.md (indirect_branch_operand): Rewrite
3560 ix86_indirect_branch_register logic.
3562 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3564 * config/i386/constraints.md (Bs): Update
3565 ix86_indirect_branch_register check. Don't check
3566 ix86_indirect_branch_register with GOT_memory_operand.
3568 * config/i386/predicates.md (GOT_memory_operand): Don't check
3569 ix86_indirect_branch_register here.
3570 (GOT32_symbol_operand): Likewise.
3572 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3574 * config/i386/predicates.md (constant_call_address_operand):
3575 Rewrite ix86_indirect_branch_register logic.
3576 (sibcall_insn_operand): Likewise.
3578 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3580 * config/i386/constraints.md (Bs): Replace
3581 ix86_indirect_branch_thunk_register with
3582 ix86_indirect_branch_register.
3584 * config/i386/i386.md (indirect_jump): Likewise.
3585 (tablejump): Likewise.
3586 (*sibcall_memory): Likewise.
3587 (*sibcall_value_memory): Likewise.
3588 Peepholes of indirect call and jump via memory: Likewise.
3589 * config/i386/i386.opt: Likewise.
3590 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3591 (GOT_memory_operand): Likewise.
3592 (call_insn_operand): Likewise.
3593 (sibcall_insn_operand): Likewise.
3594 (GOT32_symbol_operand): Likewise.
3596 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3599 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
3600 type rather than type addr's type points to.
3601 (expand_omp_atomic_mutex): Likewise.
3602 (expand_omp_atomic): Likewise.
3604 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3607 * config/i386/i386.c (output_indirect_thunk_function): Use
3608 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
3609 for __x86_return_thunk.
3611 2018-01-15 Richard Biener <rguenther@suse.de>
3614 * expmed.c (extract_bit_field_1): Fix typo.
3616 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3619 * config/arm/iterators.md (VF): New mode iterator.
3620 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
3621 Remove integer-related logic from pattern.
3622 (neon_vabd<mode>_3): Likewise.
3624 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3627 * common.opt (fstrict-overflow): No longer an alias.
3628 (fwrapv-pointer): New option.
3629 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
3630 also for pointer types based on flag_wrapv_pointer.
3631 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
3632 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
3633 opts->x_flag_wrapv got set.
3634 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
3635 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
3636 POINTER_TYPE_OVERFLOW_UNDEFINED.
3637 * match.pd: Likewise in address comparison pattern.
3638 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
3640 2018-01-15 Richard Biener <rguenther@suse.de>
3643 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
3644 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
3645 Reset type names to their identifier if their TYPE_DECL doesn't
3646 have linkage (and thus is used for ODR and devirt).
3647 (save_debug_info_for_decl): Remove.
3648 (save_debug_info_for_type): Likewise.
3649 (add_tree_to_fld_list): Adjust.
3650 * tree-pretty-print.c (dump_generic_node): Make dumping of
3651 type names more robust.
3653 2018-01-15 Richard Biener <rguenther@suse.de>
3655 * BASE-VER: Bump to 8.0.1.
3657 2018-01-14 Martin Sebor <msebor@redhat.com>
3660 * builtins.c (check_access): Avoid warning when the no-warning bit
3663 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
3665 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
3666 * ira-color (allocno_hard_regs_compare): Likewise.
3668 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
3671 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
3672 Use .pushsection/.popsection.
3674 2018-01-14 Martin Sebor <msebor@redhat.com>
3677 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
3679 2018-01-14 Jakub Jelinek <jakub@redhat.com>
3681 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
3682 entry from extra_headers.
3683 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
3684 extra_headers, make the list bitwise identical to the i?86-*-* one.
3686 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3688 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
3689 -mcmodel=large with -mindirect-branch=thunk,
3690 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
3691 -mfunction-return=thunk-extern.
3692 * doc/invoke.texi: Document -mcmodel=large is incompatible with
3693 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
3694 -mfunction-return=thunk and -mfunction-return=thunk-extern.
3696 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3698 * config/i386/i386.c (print_reg): Print the name of the full
3699 integer register without '%'.
3700 (ix86_print_operand): Handle 'V'.
3701 * doc/extend.texi: Document 'V' modifier.
3703 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3705 * config/i386/constraints.md (Bs): Disallow memory operand for
3706 -mindirect-branch-register.
3708 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3709 (GOT_memory_operand): Likewise.
3710 (call_insn_operand): Likewise.
3711 (sibcall_insn_operand): Likewise.
3712 (GOT32_symbol_operand): Likewise.
3713 * config/i386/i386.md (indirect_jump): Call convert_memory_address
3714 for -mindirect-branch-register.
3715 (tablejump): Likewise.
3716 (*sibcall_memory): Likewise.
3717 (*sibcall_value_memory): Likewise.
3718 Disallow peepholes of indirect call and jump via memory for
3719 -mindirect-branch-register.
3720 (*call_pop): Replace m with Bw.
3721 (*call_value_pop): Likewise.
3722 (*sibcall_pop_memory): Replace m with Bs.
3723 * config/i386/i386.opt (mindirect-branch-register): New option.
3724 * doc/invoke.texi: Document -mindirect-branch-register option.
3726 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3728 * config/i386/i386-protos.h (ix86_output_function_return): New.
3729 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
3730 set function_return_type.
3731 (indirect_thunk_name): Add ret_p to indicate thunk for function
3733 (output_indirect_thunk_function): Pass false to
3734 indirect_thunk_name.
3735 (ix86_output_indirect_branch_via_reg): Likewise.
3736 (ix86_output_indirect_branch_via_push): Likewise.
3737 (output_indirect_thunk_function): Create alias for function
3738 return thunk if regno < 0.
3739 (ix86_output_function_return): New function.
3740 (ix86_handle_fndecl_attribute): Handle function_return.
3741 (ix86_attribute_table): Add function_return.
3742 * config/i386/i386.h (machine_function): Add
3743 function_return_type.
3744 * config/i386/i386.md (simple_return_internal): Use
3745 ix86_output_function_return.
3746 (simple_return_internal_long): Likewise.
3747 * config/i386/i386.opt (mfunction-return=): New option.
3748 (indirect_branch): Mention -mfunction-return=.
3749 * doc/extend.texi: Document function_return function attribute.
3750 * doc/invoke.texi: Document -mfunction-return= option.
3752 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3754 * config/i386/i386-opts.h (indirect_branch): New.
3755 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
3756 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
3757 with local indirect jump when converting indirect call and jump.
3758 (ix86_set_indirect_branch_type): New.
3759 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
3760 (indirectlabelno): New.
3761 (indirect_thunk_needed): Likewise.
3762 (indirect_thunk_bnd_needed): Likewise.
3763 (indirect_thunks_used): Likewise.
3764 (indirect_thunks_bnd_used): Likewise.
3765 (INDIRECT_LABEL): Likewise.
3766 (indirect_thunk_name): Likewise.
3767 (output_indirect_thunk): Likewise.
3768 (output_indirect_thunk_function): Likewise.
3769 (ix86_output_indirect_branch_via_reg): Likewise.
3770 (ix86_output_indirect_branch_via_push): Likewise.
3771 (ix86_output_indirect_branch): Likewise.
3772 (ix86_output_indirect_jmp): Likewise.
3773 (ix86_code_end): Call output_indirect_thunk_function if needed.
3774 (ix86_output_call_insn): Call ix86_output_indirect_branch if
3776 (ix86_handle_fndecl_attribute): Handle indirect_branch.
3777 (ix86_attribute_table): Add indirect_branch.
3778 * config/i386/i386.h (machine_function): Add indirect_branch_type
3779 and has_local_indirect_jump.
3780 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
3782 (tablejump): Likewise.
3783 (*indirect_jump): Use ix86_output_indirect_jmp.
3784 (*tablejump_1): Likewise.
3785 (simple_return_indirect_internal): Likewise.
3786 * config/i386/i386.opt (mindirect-branch=): New option.
3787 (indirect_branch): New.
3790 (thunk-inline): Likewise.
3791 (thunk-extern): Likewise.
3792 * doc/extend.texi: Document indirect_branch function attribute.
3793 * doc/invoke.texi: Document -mindirect-branch= option.
3795 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
3798 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
3800 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
3802 * ipa-inline.c (want_inline_small_function_p): Return false if
3803 inlining has already failed with CIF_FINAL_ERROR.
3804 (update_caller_keys): Call want_inline_small_function_p before
3806 (update_callee_keys): Likewise.
3808 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3810 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
3812 (rs6000_quadword_masked_address_p): Likewise.
3813 (quad_aligned_load_p): Likewise.
3814 (quad_aligned_store_p): Likewise.
3815 (const_load_sequence_p): Add comment to describe the outer-most loop.
3816 (mimic_memory_attributes_and_flags): New function.
3817 (rs6000_gen_stvx): Likewise.
3818 (replace_swapped_aligned_store): Likewise.
3819 (rs6000_gen_lvx): Likewise.
3820 (replace_swapped_aligned_load): Likewise.
3821 (replace_swapped_load_constant): Capitalize argument name in
3822 comment describing this function.
3823 (rs6000_analyze_swaps): Add a third pass to search for vector loads
3824 and stores that access quad-word aligned addresses and replace
3825 with stvx or lvx instructions when appropriate.
3826 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
3827 New function prototype.
3828 (rs6000_quadword_masked_address_p): Likewise.
3829 (rs6000_gen_lvx): Likewise.
3830 (rs6000_gen_stvx): Likewise.
3831 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
3832 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
3833 when memory address is aligned.
3834 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
3835 this split to select lvx instruction when memory address is aligned.
3836 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
3837 instruction when memory address is aligned.
3838 (*vsx_le_perm_load_v16qi): Likewise.
3839 (four unnamed splitters): Modify to select the stvx instruction
3840 when memory is aligned.
3842 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
3844 * predict.c (determine_unlikely_bbs): Handle correctly BBs
3845 which appears in the queue multiple times.
3847 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3848 Alan Hayward <alan.hayward@arm.com>
3849 David Sherwood <david.sherwood@arm.com>
3851 * tree-vectorizer.h (vec_lower_bound): New structure.
3852 (_loop_vec_info): Add check_nonzero and lower_bounds.
3853 (LOOP_VINFO_CHECK_NONZERO): New macro.
3854 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
3855 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
3856 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
3857 fields. Make seg_len the distance travelled, not including the
3859 (dr_direction_indicator): Declare.
3860 (dr_zero_step_indicator): Likewise.
3861 (dr_known_forward_stride_p): Likewise.
3862 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
3864 (runtime_alias_check_p): Allow runtime alias checks with
3866 (operator ==): Compare access_size and align.
3867 (prune_runtime_alias_test_list): Rework for new distinction between
3868 the access_size and seg_len.
3869 (create_intersect_range_checks_index): Likewise. Cope with polynomial
3871 (get_segment_min_max): New function.
3872 (create_intersect_range_checks): Use it.
3873 (dr_step_indicator): New function.
3874 (dr_direction_indicator): Likewise.
3875 (dr_zero_step_indicator): Likewise.
3876 (dr_known_forward_stride_p): Likewise.
3877 * tree-loop-distribution.c (data_ref_segment_size): Return
3878 DR_STEP * (niters - 1).
3879 (compute_alias_check_pairs): Update call to the dr_with_seg_len
3881 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
3882 (vect_preserves_scalar_order_p): New function, split out from...
3883 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
3884 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
3885 (vect_vfa_access_size): New function.
3886 (vect_vfa_align): Likewise.
3887 (vect_compile_time_alias): Take access_size_a and access_b arguments.
3888 (dump_lower_bound): New function.
3889 (vect_check_lower_bound): Likewise.
3890 (vect_small_gap_p): Likewise.
3891 (vectorizable_with_step_bound_p): Likewise.
3892 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
3893 depencies if the vectorization factor is 1. Convert the checks
3894 for nonzero steps into checks on the bounds of DR_STEP. Try using
3895 a bunds check for variable steps if the minimum required step is
3896 relatively small. Update calls to the dr_with_seg_len
3897 constructor and to vect_compile_time_alias.
3898 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
3900 (vect_loop_versioning): Call it.
3901 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
3903 (vect_estimate_min_profitable_iters): Account for any bounds checks.
3905 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3906 Alan Hayward <alan.hayward@arm.com>
3907 David Sherwood <david.sherwood@arm.com>
3909 * doc/sourcebuild.texi (vect_scatter_store): Document.
3910 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
3912 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
3914 * genopinit.c (main): Add supports_vec_scatter_store and
3915 supports_vec_scatter_store_cached to target_optabs.
3916 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
3917 IFN_MASK_SCATTER_STORE.
3918 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
3920 * internal-fn.h (internal_store_fn_p): Declare.
3921 (internal_fn_stored_value_index): Likewise.
3922 * internal-fn.c (scatter_store_direct): New macro.
3923 (expand_scatter_store_optab_fn): New function.
3924 (direct_scatter_store_optab_supported_p): New macro.
3925 (internal_store_fn_p): New function.
3926 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
3927 IFN_MASK_SCATTER_STORE.
3928 (internal_fn_mask_index): Likewise.
3929 (internal_fn_stored_value_index): New function.
3930 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
3932 * optabs-query.h (supports_vec_scatter_store_p): Declare.
3933 * optabs-query.c (supports_vec_scatter_store_p): New function.
3934 * tree-vectorizer.h (vect_get_store_rhs): Declare.
3935 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
3936 true for scatter stores.
3937 (vect_gather_scatter_fn_p): Handle scatter stores too.
3938 (vect_check_gather_scatter): Consider using scatter stores if
3939 supports_vec_scatter_store_p.
3940 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
3942 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
3943 internal_fn_stored_value_index.
3944 (check_load_store_masking): Handle scatter stores too.
3945 (vect_get_store_rhs): Make public.
3946 (vectorizable_call): Use internal_store_fn_p.
3947 (vectorizable_store): Handle scatter store internal functions.
3948 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
3949 when deciding whether the end of the group has been reached.
3950 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
3951 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
3952 (mask_scatter_store<mode>): New insns.
3954 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3955 Alan Hayward <alan.hayward@arm.com>
3956 David Sherwood <david.sherwood@arm.com>
3958 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
3959 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
3960 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
3962 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
3963 Use vect_truncate_gather_scatter_offset if we can't treat the
3964 operation as a normal gather load or scatter store.
3965 (get_group_load_store_type): Take the gather_scatter_info
3966 as argument. Try using a gather load or scatter store for
3967 single-element groups.
3968 (get_load_store_type): Update calls to get_group_load_store_type
3969 and vect_use_strided_gather_scatters_p.
3971 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3972 Alan Hayward <alan.hayward@arm.com>
3973 David Sherwood <david.sherwood@arm.com>
3975 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
3976 optional tree argument.
3977 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
3979 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
3980 but continue to use the current value as a fallback.
3981 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
3982 to compare the updates.
3983 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
3984 (get_load_store_type): Use it when handling a strided access.
3985 (vect_get_strided_load_store_ops): New function.
3986 (vect_get_data_ptr_increment): Likewise.
3987 (vectorizable_load): Handle strided gather loads. Always pass
3988 a step to vect_create_data_ref_ptr and bump_vector_ptr.
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 * doc/md.texi (gather_load@var{m}): Document.
3995 (mask_gather_load@var{m}): Likewise.
3996 * genopinit.c (main): Add supports_vec_gather_load and
3997 supports_vec_gather_load_cached to target_optabs.
3998 * optabs-tree.c (init_tree_optimization_optabs): Use
3999 ggc_cleared_alloc to allocate target_optabs.
4000 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
4001 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
4003 * internal-fn.h (internal_load_fn_p): Declare.
4004 (internal_gather_scatter_fn_p): Likewise.
4005 (internal_fn_mask_index): Likewise.
4006 (internal_gather_scatter_fn_supported_p): Likewise.
4007 * internal-fn.c (gather_load_direct): New macro.
4008 (expand_gather_load_optab_fn): New function.
4009 (direct_gather_load_optab_supported_p): New macro.
4010 (direct_internal_fn_optab): New function.
4011 (internal_load_fn_p): Likewise.
4012 (internal_gather_scatter_fn_p): Likewise.
4013 (internal_fn_mask_index): Likewise.
4014 (internal_gather_scatter_fn_supported_p): Likewise.
4015 * optabs-query.c (supports_at_least_one_mode_p): New function.
4016 (supports_vec_gather_load_p): Likewise.
4017 * optabs-query.h (supports_vec_gather_load_p): Declare.
4018 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
4019 and memory_type field.
4020 (NUM_PATTERNS): Bump to 15.
4021 * tree-vect-data-refs.c: Include internal-fn.h.
4022 (vect_gather_scatter_fn_p): New function.
4023 (vect_describe_gather_scatter_call): Likewise.
4024 (vect_check_gather_scatter): Try using internal functions for
4025 gather loads. Recognize existing calls to a gather load function.
4026 (vect_analyze_data_refs): Consider using gather loads if
4027 supports_vec_gather_load_p.
4028 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
4029 (vect_get_gather_scatter_offset_type): Likewise.
4030 (vect_convert_mask_for_vectype): Likewise.
4031 (vect_add_conversion_to_patterm): Likewise.
4032 (vect_try_gather_scatter_pattern): Likewise.
4033 (vect_recog_gather_scatter_pattern): New pattern recognizer.
4034 (vect_vect_recog_func_ptrs): Add it.
4035 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4036 internal_fn_mask_index and internal_gather_scatter_fn_p.
4037 (check_load_store_masking): Take the gather_scatter_info as an
4038 argument and handle gather loads.
4039 (vect_get_gather_scatter_ops): New function.
4040 (vectorizable_call): Check internal_load_fn_p.
4041 (vectorizable_load): Likewise. Handle gather load internal
4043 (vectorizable_store): Update call to check_load_store_masking.
4044 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
4045 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
4046 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
4047 (aarch64_gather_scale_operand_d): New predicates.
4048 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
4049 (mask_gather_load<mode>): New insns.
4051 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4052 Alan Hayward <alan.hayward@arm.com>
4053 David Sherwood <david.sherwood@arm.com>
4055 * optabs.def (fold_left_plus_optab): New optab.
4056 * doc/md.texi (fold_left_plus_@var{m}): Document.
4057 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
4058 * internal-fn.c (fold_left_direct): Define.
4059 (expand_fold_left_optab_fn): Likewise.
4060 (direct_fold_left_optab_supported_p): Likewise.
4061 * fold-const-call.c (fold_const_fold_left): New function.
4062 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
4063 * tree-parloops.c (valid_reduction_p): New function.
4064 (gather_scalar_reductions): Use it.
4065 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
4066 (vect_finish_replace_stmt): Declare.
4067 * tree-vect-loop.c (fold_left_reduction_fn): New function.
4068 (needs_fold_left_reduction_p): New function, split out from...
4069 (vect_is_simple_reduction): ...here. Accept reductions that
4070 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
4071 (vect_force_simple_reduction): Also store the reduction type in
4072 the assignment's STMT_VINFO_REDUC_TYPE.
4073 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
4074 (merge_with_identity): New function.
4075 (vect_expand_fold_left): Likewise.
4076 (vectorize_fold_left_reduction): Likewise.
4077 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
4078 scalar phi in place for it. Check for target support and reject
4079 cases that would reassociate the operation. Defer the transform
4080 phase to vectorize_fold_left_reduction.
4081 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
4082 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
4083 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
4085 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4087 * tree-if-conv.c (predicate_mem_writes): Remove redundant
4088 call to ifc_temp_var.
4090 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4091 Alan Hayward <alan.hayward@arm.com>
4092 David Sherwood <david.sherwood@arm.com>
4094 * target.def (legitimize_address_displacement): Take the original
4095 offset as a poly_int.
4096 * targhooks.h (default_legitimize_address_displacement): Update
4098 * targhooks.c (default_legitimize_address_displacement): Likewise.
4099 * doc/tm.texi: Regenerate.
4100 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
4101 as an argument, moving assert of ad->disp == ad->disp_term to...
4102 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
4103 Try calling targetm.legitimize_address_displacement before expanding
4104 the address rather than afterwards, and adjust for the new interface.
4105 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
4106 Match the new hook interface. Handle SVE addresses.
4107 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
4110 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4112 * Makefile.in (OBJS): Add early-remat.o.
4113 * target.def (select_early_remat_modes): New hook.
4114 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
4115 * doc/tm.texi: Regenerate.
4116 * targhooks.h (default_select_early_remat_modes): Declare.
4117 * targhooks.c (default_select_early_remat_modes): New function.
4118 * timevar.def (TV_EARLY_REMAT): New timevar.
4119 * passes.def (pass_early_remat): New pass.
4120 * tree-pass.h (make_pass_early_remat): Declare.
4121 * early-remat.c: New file.
4122 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
4124 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
4126 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4127 Alan Hayward <alan.hayward@arm.com>
4128 David Sherwood <david.sherwood@arm.com>
4130 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
4131 vfm1 with a bound_epilog parameter.
4132 (vect_do_peeling): Update calls accordingly, and move the prologue
4133 call earlier in the function. Treat the base bound_epilog as 0 for
4134 fully-masked loops and retain vf - 1 for other loops. Add 1 to
4135 this base when peeling for gaps.
4136 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
4137 with fully-masked loops.
4138 (vect_estimate_min_profitable_iters): Handle the single peeled
4139 iteration in that case.
4141 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4142 Alan Hayward <alan.hayward@arm.com>
4143 David Sherwood <david.sherwood@arm.com>
4145 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
4146 single-element interleaving even if the size is not a power of 2.
4147 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
4148 accesses for single-element interleaving if the group size is
4151 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4152 Alan Hayward <alan.hayward@arm.com>
4153 David Sherwood <david.sherwood@arm.com>
4155 * doc/md.texi (fold_extract_last_@var{m}): Document.
4156 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
4157 * optabs.def (fold_extract_last_optab): New optab.
4158 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
4159 * internal-fn.c (fold_extract_direct): New macro.
4160 (expand_fold_extract_optab_fn): Likewise.
4161 (direct_fold_extract_optab_supported_p): Likewise.
4162 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
4163 * tree-vect-loop.c (vect_model_reduction_cost): Handle
4164 EXTRACT_LAST_REDUCTION.
4165 (get_initial_def_for_reduction): Do not create an initial vector
4166 for EXTRACT_LAST_REDUCTION reductions.
4167 (vectorizable_reduction): Leave the scalar phi in place for
4168 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
4169 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
4170 epilogue code for EXTRACT_LAST_REDUCTION and defer the
4171 transform phase to vectorizable_condition.
4172 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
4174 (vect_finish_stmt_generation): ...here.
4175 (vect_finish_replace_stmt): New function.
4176 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
4177 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
4179 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
4181 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4182 Alan Hayward <alan.hayward@arm.com>
4183 David Sherwood <david.sherwood@arm.com>
4185 * doc/md.texi (extract_last_@var{m}): Document.
4186 * optabs.def (extract_last_optab): New optab.
4187 * internal-fn.def (EXTRACT_LAST): New internal function.
4188 * internal-fn.c (cond_unary_direct): New macro.
4189 (expand_cond_unary_optab_fn): Likewise.
4190 (direct_cond_unary_optab_supported_p): Likewise.
4191 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
4192 loops using EXTRACT_LAST.
4193 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
4194 (extract_last_<mode>): ...this optab.
4195 (vec_extract<mode><Vel>): Update accordingly.
4197 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4198 Alan Hayward <alan.hayward@arm.com>
4199 David Sherwood <david.sherwood@arm.com>
4201 * target.def (empty_mask_is_expensive): New hook.
4202 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
4203 * doc/tm.texi: Regenerate.
4204 * targhooks.h (default_empty_mask_is_expensive): Declare.
4205 * targhooks.c (default_empty_mask_is_expensive): New function.
4206 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
4207 if the target says that empty masks are expensive.
4208 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
4210 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
4212 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4213 Alan Hayward <alan.hayward@arm.com>
4214 David Sherwood <david.sherwood@arm.com>
4216 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
4217 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
4218 (vect_use_loop_mask_for_alignment_p): New function.
4219 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
4220 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
4221 niters_skip argument. Make sure that the first niters_skip elements
4222 of the first iteration are inactive.
4223 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
4224 Update call to vect_set_loop_masks_directly.
4225 (get_misalign_in_elems): New function, split out from...
4226 (vect_gen_prolog_loop_niters): ...here.
4227 (vect_update_init_of_dr): Take a code argument that specifies whether
4228 the adjustment should be added or subtracted.
4229 (vect_update_init_of_drs): Likewise.
4230 (vect_prepare_for_masked_peels): New function.
4231 (vect_do_peeling): Skip prologue peeling if we're using a mask
4232 instead. Update call to vect_update_inits_of_drs.
4233 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4235 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
4236 alignment. Do not include the number of peeled iterations in
4237 the minimum threshold in that case.
4238 (vectorizable_induction): Adjust the start value down by
4239 LOOP_VINFO_MASK_SKIP_NITERS iterations.
4240 (vect_transform_loop): Call vect_prepare_for_masked_peels.
4241 Take the number of skipped iterations into account when calculating
4243 * tree-vect-stmts.c (vect_gen_while_not): New function.
4245 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4246 Alan Hayward <alan.hayward@arm.com>
4247 David Sherwood <david.sherwood@arm.com>
4249 * doc/sourcebuild.texi (vect_fully_masked): Document.
4250 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
4252 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
4254 (vect_analyze_loop_2): ...here. Don't check the vectorization
4255 factor against the number of loop iterations if the loop is
4258 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4259 Alan Hayward <alan.hayward@arm.com>
4260 David Sherwood <david.sherwood@arm.com>
4262 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
4263 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
4264 (dump_groups): Update accordingly.
4265 (iv_use::mem_type): New member variable.
4266 (address_p): New function.
4267 (record_use): Add a mem_type argument and initialize the new
4269 (record_group_use): Add a mem_type argument. Use address_p.
4270 Remove obsolete null checks of base_object. Update call to record_use.
4271 (find_interesting_uses_op): Update call to record_group_use.
4272 (find_interesting_uses_cond): Likewise.
4273 (find_interesting_uses_address): Likewise.
4274 (get_mem_type_for_internal_fn): New function.
4275 (find_address_like_use): Likewise.
4276 (find_interesting_uses_stmt): Try find_address_like_use before
4277 calling find_interesting_uses_op.
4278 (addr_offset_valid_p): Use the iv mem_type field as the type
4279 of the addressed memory.
4280 (add_autoinc_candidates): Likewise.
4281 (get_address_cost): Likewise.
4282 (split_small_address_groups_p): Use address_p.
4283 (split_address_groups): Likewise.
4284 (add_iv_candidate_for_use): Likewise.
4285 (autoinc_possible_for_pair): Likewise.
4286 (rewrite_groups): Likewise.
4287 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
4288 (determine_group_iv_cost): Update after split of USE_ADDRESS.
4289 (get_alias_ptr_type_for_ptr_address): New function.
4290 (rewrite_use_address): Rewrite address uses in calls that were
4291 identified by find_address_like_use.
4293 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4294 Alan Hayward <alan.hayward@arm.com>
4295 David Sherwood <david.sherwood@arm.com>
4297 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
4299 * gimple-expr.h (is_gimple_addressable: Likewise.
4300 * gimple-expr.c (is_gimple_address): Likewise.
4301 * internal-fn.c (expand_call_mem_ref): New function.
4302 (expand_mask_load_optab_fn): Use it.
4303 (expand_mask_store_optab_fn): Likewise.
4305 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4306 Alan Hayward <alan.hayward@arm.com>
4307 David Sherwood <david.sherwood@arm.com>
4309 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
4310 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
4311 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
4312 (cond_umax@var{mode}): Document.
4313 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
4314 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
4315 (cond_umin_optab, cond_umax_optab): New optabs.
4316 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
4317 (COND_IOR, COND_XOR): New internal functions.
4318 * internal-fn.h (get_conditional_internal_fn): Declare.
4319 * internal-fn.c (cond_binary_direct): New macro.
4320 (expand_cond_binary_optab_fn): Likewise.
4321 (direct_cond_binary_optab_supported_p): Likewise.
4322 (get_conditional_internal_fn): New function.
4323 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
4324 Cope with reduction statements that are vectorized as calls rather
4326 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
4327 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
4328 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
4329 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4330 (UNSPEC_COND_EOR): New unspecs.
4331 (optab): Add mappings for them.
4332 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
4333 (sve_int_op, sve_fp_op): New int attributes.
4335 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4336 Alan Hayward <alan.hayward@arm.com>
4337 David Sherwood <david.sherwood@arm.com>
4339 * optabs.def (while_ult_optab): New optab.
4340 * doc/md.texi (while_ult@var{m}@var{n}): Document.
4341 * internal-fn.def (WHILE_ULT): New internal function.
4342 * internal-fn.h (direct_internal_fn_supported_p): New override
4343 that takes two types as argument.
4344 * internal-fn.c (while_direct): New macro.
4345 (expand_while_optab_fn): New function.
4346 (convert_optab_supported_p): Likewise.
4347 (direct_while_optab_supported_p): New macro.
4348 * wide-int.h (wi::udiv_ceil): New function.
4349 * tree-vectorizer.h (rgroup_masks): New structure.
4350 (vec_loop_masks): New typedef.
4351 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
4353 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
4354 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
4355 (vect_max_vf): New function.
4356 (slpeel_make_loop_iterate_ntimes): Delete.
4357 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
4358 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
4359 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
4360 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
4361 internal-fn.h, stor-layout.h and optabs-query.h.
4362 (vect_set_loop_mask): New function.
4363 (add_preheader_seq): Likewise.
4364 (add_header_seq): Likewise.
4365 (interleave_supported_p): Likewise.
4366 (vect_maybe_permute_loop_masks): Likewise.
4367 (vect_set_loop_masks_directly): Likewise.
4368 (vect_set_loop_condition_masked): Likewise.
4369 (vect_set_loop_condition_unmasked): New function, split out from
4370 slpeel_make_loop_iterate_ntimes.
4371 (slpeel_make_loop_iterate_ntimes): Rename to..
4372 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
4373 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
4374 (vect_do_peeling): Update call accordingly.
4375 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
4377 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4378 mask_compare_type, can_fully_mask_p and fully_masked_p.
4379 (release_vec_loop_masks): New function.
4380 (_loop_vec_info): Use it to free the loop masks.
4381 (can_produce_all_loop_masks_p): New function.
4382 (vect_get_max_nscalars_per_iter): Likewise.
4383 (vect_verify_full_masking): Likewise.
4384 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
4385 retries, and free the mask rgroups before retrying. Check loop-wide
4386 reasons for disallowing fully-masked loops. Make the final decision
4387 about whether use a fully-masked loop or not.
4388 (vect_estimate_min_profitable_iters): Do not assume that peeling
4389 for the number of iterations will be needed for fully-masked loops.
4390 (vectorizable_reduction): Disable fully-masked loops.
4391 (vectorizable_live_operation): Likewise.
4392 (vect_halve_mask_nunits): New function.
4393 (vect_double_mask_nunits): Likewise.
4394 (vect_record_loop_mask): Likewise.
4395 (vect_get_loop_mask): Likewise.
4396 (vect_transform_loop): Handle the case in which the final loop
4397 iteration might handle a partial vector. Call vect_set_loop_condition
4398 instead of slpeel_make_loop_iterate_ntimes.
4399 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
4400 (check_load_store_masking): New function.
4401 (prepare_load_store_mask): Likewise.
4402 (vectorizable_store): Handle fully-masked loops.
4403 (vectorizable_load): Likewise.
4404 (supportable_widening_operation): Use vect_halve_mask_nunits for
4406 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
4407 (vect_gen_while): New function.
4408 * config/aarch64/aarch64.md (umax<mode>3): New expander.
4409 (aarch64_uqdec<mode>): New insn.
4411 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4412 Alan Hayward <alan.hayward@arm.com>
4413 David Sherwood <david.sherwood@arm.com>
4415 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
4416 (reduc_xor_scal_optab): New optabs.
4417 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
4418 (reduc_xor_scal_@var{m}): Document.
4419 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
4420 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
4422 * fold-const-call.c (fold_const_call): Handle them.
4423 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
4424 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
4425 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
4426 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
4427 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
4428 (UNSPEC_XORV): New unspecs.
4429 (optab): Add entries for them.
4430 (BITWISEV): New int iterator.
4431 (bit_reduc_op): New int attributes.
4433 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4434 Alan Hayward <alan.hayward@arm.com>
4435 David Sherwood <david.sherwood@arm.com>
4437 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
4438 * internal-fn.def (VEC_SHL_INSERT): New internal function.
4439 * optabs.def (vec_shl_insert_optab): New optab.
4440 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
4441 (duplicate_and_interleave): Likewise.
4442 * tree-vect-loop.c: Include internal-fn.h.
4443 (neutral_op_for_slp_reduction): New function, split out from
4444 get_initial_defs_for_reduction.
4445 (get_initial_def_for_reduction): Handle option 2 for variable-length
4446 vectors by loading the neutral value into a vector and then shifting
4447 the initial value into element 0.
4448 (get_initial_defs_for_reduction): Replace the code argument with
4449 the neutral value calculated by neutral_op_for_slp_reduction.
4450 Use gimple_build_vector for constant-length vectors.
4451 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
4452 but the first group_size elements have a neutral value.
4453 Use duplicate_and_interleave otherwise.
4454 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
4455 Update call to get_initial_defs_for_reduction. Handle SLP
4456 reductions for variable-length vectors by creating one vector
4457 result for each scalar result, with the elements associated
4458 with other scalar results stubbed out with the neutral value.
4459 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
4460 Require IFN_VEC_SHL_INSERT for double reductions on
4461 variable-length vectors, or SLP reductions that have
4462 a neutral value. Require can_duplicate_and_interleave_p
4463 support for variable-length unchained SLP reductions if there
4464 is no neutral value, such as for MIN/MAX reductions. Also require
4465 the number of vector elements to be a multiple of the number of
4466 SLP statements when doing variable-length unchained SLP reductions.
4467 Update call to vect_create_epilog_for_reduction.
4468 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
4469 and remove initial values.
4470 (duplicate_and_interleave): Make public.
4471 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
4472 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
4474 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4475 Alan Hayward <alan.hayward@arm.com>
4476 David Sherwood <david.sherwood@arm.com>
4478 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
4479 (can_duplicate_and_interleave_p): New function.
4480 (vect_get_and_check_slp_defs): Take the vector of statements
4481 rather than just the current one. Remove excess parentheses.
4482 Restriction rejectinon of vect_constant_def and vect_external_def
4483 for variable-length vectors to boolean types, or types for which
4484 can_duplicate_and_interleave_p is false.
4485 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
4486 (duplicate_and_interleave): New function.
4487 (vect_get_constant_vectors): Use gimple_build_vector for
4488 constant-length vectors and suitable variable-length constant
4489 vectors. Use duplicate_and_interleave for other variable-length
4490 vectors. Don't defer the update when inserting new statements.
4492 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4493 Alan Hayward <alan.hayward@arm.com>
4494 David Sherwood <david.sherwood@arm.com>
4496 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
4497 min_profitable_iters doesn't go negative.
4499 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4500 Alan Hayward <alan.hayward@arm.com>
4501 David Sherwood <david.sherwood@arm.com>
4503 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
4504 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
4505 * optabs.def (vec_mask_load_lanes_optab): New optab.
4506 (vec_mask_store_lanes_optab): Likewise.
4507 * internal-fn.def (MASK_LOAD_LANES): New internal function.
4508 (MASK_STORE_LANES): Likewise.
4509 * internal-fn.c (mask_load_lanes_direct): New macro.
4510 (mask_store_lanes_direct): Likewise.
4511 (expand_mask_load_optab_fn): Handle masked operations.
4512 (expand_mask_load_lanes_optab_fn): New macro.
4513 (expand_mask_store_optab_fn): Handle masked operations.
4514 (expand_mask_store_lanes_optab_fn): New macro.
4515 (direct_mask_load_lanes_optab_supported_p): Likewise.
4516 (direct_mask_store_lanes_optab_supported_p): Likewise.
4517 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
4519 (vect_load_lanes_supported): Likewise.
4520 * tree-vect-data-refs.c (strip_conversion): New function.
4521 (can_group_stmts_p): Likewise.
4522 (vect_analyze_data_ref_accesses): Use it instead of checking
4523 for a pair of assignments.
4524 (vect_store_lanes_supported): Take a masked_p parameter.
4525 (vect_load_lanes_supported): Likewise.
4526 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
4527 vect_store_lanes_supported and vect_load_lanes_supported.
4528 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4529 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
4530 parameter. Don't allow gaps for masked accesses.
4531 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
4532 and vect_load_lanes_supported.
4533 (get_load_store_type): Take a masked_p parameter and update
4534 call to get_group_load_store_type.
4535 (vectorizable_store): Update call to get_load_store_type.
4536 Handle IFN_MASK_STORE_LANES.
4537 (vectorizable_load): Update call to get_load_store_type.
4538 Handle IFN_MASK_LOAD_LANES.
4540 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4541 Alan Hayward <alan.hayward@arm.com>
4542 David Sherwood <david.sherwood@arm.com>
4544 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
4546 * config/aarch64/aarch64-protos.h
4547 (aarch64_sve_struct_memory_operand_p): Declare.
4548 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
4549 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
4550 (VPRED, vpred): Handle SVE structure modes.
4551 * config/aarch64/constraints.md (Utx): New constraint.
4552 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
4553 (aarch64_sve_struct_nonimmediate_operand): New predicates.
4554 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
4555 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
4556 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
4557 structure modes. Split into pieces after RA.
4558 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
4559 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
4561 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
4562 SVE structure modes.
4563 (aarch64_classify_address): Likewise.
4564 (sizetochar): Move earlier in file.
4565 (aarch64_print_operand): Handle SVE register lists.
4566 (aarch64_array_mode): New function.
4567 (aarch64_sve_struct_memory_operand_p): Likewise.
4568 (TARGET_ARRAY_MODE): Redefine.
4570 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4571 Alan Hayward <alan.hayward@arm.com>
4572 David Sherwood <david.sherwood@arm.com>
4574 * target.def (array_mode): New target hook.
4575 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
4576 * doc/tm.texi: Regenerate.
4577 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
4578 * hooks.c (hook_optmode_mode_uhwi_none): New function.
4579 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
4581 * stor-layout.c (mode_for_array): Likewise. Support polynomial
4584 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4585 Alan Hayward <alan.hayward@arm.com>
4586 David Sherwood <david.sherwood@arm.com>
4588 * fold-const.c (fold_binary_loc): Check the argument types
4589 rather than the result type when testing for a vector operation.
4591 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4593 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
4594 * doc/tm.texi: Regenerate.
4596 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4597 Alan Hayward <alan.hayward@arm.com>
4598 David Sherwood <david.sherwood@arm.com>
4600 * doc/invoke.texi (-msve-vector-bits=): Document new option.
4601 (sve): Document new AArch64 extension.
4602 * doc/md.texi (w): Extend the description of the AArch64
4603 constraint to include SVE vectors.
4604 (Upl, Upa): Document new AArch64 predicate constraints.
4605 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
4607 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
4608 (msve-vector-bits=): New option.
4609 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
4610 SVE when these are disabled.
4611 (sve): New extension.
4612 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
4613 modes. Adjust their number of units based on aarch64_sve_vg.
4614 (MAX_BITSIZE_MODE_ANY_MODE): Define.
4615 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
4616 aarch64_addr_query_type.
4617 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
4618 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
4619 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
4620 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
4621 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
4622 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
4623 (aarch64_simd_imm_zero_p): Delete.
4624 (aarch64_check_zero_based_sve_index_immediate): Declare.
4625 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4626 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4627 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4628 (aarch64_sve_float_mul_immediate_p): Likewise.
4629 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4631 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
4632 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
4633 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
4634 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
4635 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
4636 (aarch64_regmode_natural_size): Likewise.
4637 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
4638 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
4640 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
4641 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
4642 for VG and the SVE predicate registers.
4643 (V_ALIASES): Add a "z"-prefixed alias.
4644 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
4645 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
4646 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
4647 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
4648 (REG_CLASS_NAMES): Add entries for them.
4649 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
4650 and the predicate registers.
4651 (aarch64_sve_vg): Declare.
4652 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
4653 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
4654 (REGMODE_NATURAL_SIZE): Define.
4655 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
4657 * config/aarch64/aarch64.c: Include cfgrtl.h.
4658 (simd_immediate_info): Add a constructor for series vectors,
4659 and an associated step field.
4660 (aarch64_sve_vg): New variable.
4661 (aarch64_dbx_register_number): Handle VG and the predicate registers.
4662 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
4663 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
4664 (VEC_ANY_DATA, VEC_STRUCT): New constants.
4665 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
4666 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
4667 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
4668 (aarch64_get_mask_mode): New functions.
4669 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
4670 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
4671 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
4672 predicate modes and predicate registers. Explicitly restrict
4673 GPRs to modes of 16 bytes or smaller. Only allow FP registers
4674 to store a vector mode if it is recognized by
4675 aarch64_classify_vector_mode.
4676 (aarch64_regmode_natural_size): New function.
4677 (aarch64_hard_regno_caller_save_mode): Return the original mode
4679 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
4680 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
4681 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
4682 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
4684 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
4685 does not overlap dest if the function is frame-related. Handle
4687 (aarch64_split_add_offset): New function.
4688 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
4689 them aarch64_add_offset.
4690 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
4691 and update call to aarch64_sub_sp.
4692 (aarch64_add_cfa_expression): New function.
4693 (aarch64_expand_prologue): Pass extra temporary registers to the
4694 functions above. Handle the case in which we need to emit new
4695 DW_CFA_expressions for registers that were originally saved
4696 relative to the stack pointer, but now have to be expressed
4697 relative to the frame pointer.
4698 (aarch64_output_mi_thunk): Pass extra temporary registers to the
4700 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
4701 IP0 and IP1 values for SVE frames.
4702 (aarch64_expand_vec_series): New function.
4703 (aarch64_expand_sve_widened_duplicate): Likewise.
4704 (aarch64_expand_sve_const_vector): Likewise.
4705 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
4706 Handle SVE constants. Use emit_move_insn to move a force_const_mem
4707 into the register, rather than emitting a SET directly.
4708 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
4709 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
4710 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
4711 (offset_9bit_signed_scaled_p): New functions.
4712 (aarch64_replicate_bitmask_imm): New function.
4713 (aarch64_bitmask_imm): Use it.
4714 (aarch64_cannot_force_const_mem): Reject expressions involving
4715 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
4716 (aarch64_classify_index): Handle SVE indices, by requiring
4717 a plain register index with a scale that matches the element size.
4718 (aarch64_classify_address): Handle SVE addresses. Assert that
4719 the mode of the address is VOIDmode or an integer mode.
4720 Update call to aarch64_classify_symbol.
4721 (aarch64_classify_symbolic_expression): Update call to
4722 aarch64_classify_symbol.
4723 (aarch64_const_vec_all_in_range_p): New function.
4724 (aarch64_print_vector_float_operand): Likewise.
4725 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
4726 "vN" for FP registers with SVE modes. Handle (const ...) vectors
4727 and the FP immediates 1.0 and 0.5.
4728 (aarch64_print_address_internal): Handle SVE addresses.
4729 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
4730 (aarch64_regno_regclass): Handle predicate registers.
4731 (aarch64_secondary_reload): Handle big-endian reloads of SVE
4733 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
4734 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
4735 (aarch64_convert_sve_vector_bits): New function.
4736 (aarch64_override_options): Use it to handle -msve-vector-bits=.
4737 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4739 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
4740 Handle SVE vector and predicate modes. Accept VL-based constants
4741 that need only one temporary register, and VL offsets that require
4742 no temporary registers.
4743 (aarch64_conditional_register_usage): Mark the predicate registers
4744 as fixed if SVE isn't available.
4745 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
4746 Return true for SVE vector and predicate modes.
4747 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
4748 rather than an unsigned int. Handle SVE modes.
4749 (aarch64_preferred_simd_mode): Update call accordingly. Handle
4751 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
4753 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4754 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4755 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4756 (aarch64_sve_float_mul_immediate_p): New functions.
4757 (aarch64_sve_valid_immediate): New function.
4758 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
4759 Explicitly reject structure modes. Check for INDEX constants.
4760 Handle PTRUE and PFALSE constants.
4761 (aarch64_check_zero_based_sve_index_immediate): New function.
4762 (aarch64_simd_imm_zero_p): Delete.
4763 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
4764 vector modes. Accept constants in the range of CNT[BHWD].
4765 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
4766 ask for an Advanced SIMD mode.
4767 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
4768 (aarch64_simd_vector_alignment): Handle SVE predicates.
4769 (aarch64_vectorize_preferred_vector_alignment): New function.
4770 (aarch64_simd_vector_alignment_reachable): Use it instead of
4772 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
4773 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
4775 (MAX_VECT_LEN): Delete.
4776 (expand_vec_perm_d): Add a vec_flags field.
4777 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
4778 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
4779 (aarch64_evpc_ext): Don't apply a big-endian lane correction
4781 (aarch64_evpc_rev): Rename to...
4782 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
4783 (aarch64_evpc_rev_global): New function.
4784 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
4785 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
4787 (aarch64_evpc_sve_tbl): New function.
4788 (aarch64_expand_vec_perm_const_1): Update after rename of
4789 aarch64_evpc_rev. Handle SVE permutes too, trying
4790 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
4791 than aarch64_evpc_tbl.
4792 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
4793 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
4794 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
4795 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
4796 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
4797 (aarch64_expand_sve_vcond): New functions.
4798 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
4799 of aarch64_vector_mode_p.
4800 (aarch64_dwarf_poly_indeterminate_value): New function.
4801 (aarch64_compute_pressure_classes): Likewise.
4802 (aarch64_can_change_mode_class): Likewise.
4803 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
4804 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
4805 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
4806 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
4807 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
4808 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
4809 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
4810 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
4812 (Dn, Dl, Dr): Accept const as well as const_vector.
4813 (Dz): Likewise. Compare against CONST0_RTX.
4814 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
4815 of "vector" where appropriate.
4816 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
4817 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
4818 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
4819 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
4820 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
4821 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
4822 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
4823 (v_int_equiv): Extend to SVE modes.
4824 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
4826 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
4827 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
4828 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
4829 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
4830 (SVE_COND_FP_CMP): New int iterators.
4831 (perm_hilo): Handle the new unpack unspecs.
4832 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
4834 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
4835 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
4836 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
4837 (aarch64_equality_operator, aarch64_constant_vector_operand)
4838 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
4839 (aarch64_sve_nonimmediate_operand): Likewise.
4840 (aarch64_sve_general_operand): Likewise.
4841 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
4842 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
4843 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
4844 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
4845 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
4846 (aarch64_sve_float_arith_immediate): Likewise.
4847 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
4848 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
4849 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
4850 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
4851 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
4852 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
4853 (aarch64_sve_float_arith_operand): Likewise.
4854 (aarch64_sve_float_arith_with_sub_operand): Likewise.
4855 (aarch64_sve_float_mul_operand): Likewise.
4856 (aarch64_sve_vec_perm_operand): Likewise.
4857 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
4858 (aarch64_mov_operand): Accept const_poly_int and const_vector.
4859 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
4860 as well as const_vector.
4861 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
4862 in file. Use CONST0_RTX and CONSTM1_RTX.
4863 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
4864 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
4865 Use aarch64_simd_imm_zero.
4866 * config/aarch64/aarch64-sve.md: New file.
4867 * config/aarch64/aarch64.md: Include it.
4868 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
4869 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
4870 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
4871 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
4872 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
4873 (sve): New attribute.
4874 (enabled): Disable instructions with the sve attribute unless
4876 (movqi, movhi): Pass CONST_POLY_INT operaneds through
4877 aarch64_expand_mov_immediate.
4878 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
4879 CNT[BHSD] immediates.
4880 (movti): Split CONST_POLY_INT moves into two halves.
4881 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
4882 Split additions that need a temporary here if the destination
4883 is the stack pointer.
4884 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
4885 (*add<mode>3_poly_1): New instruction.
4886 (set_clobber_cc): New expander.
4888 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4890 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
4891 parameter and use it instead of GET_MODE_SIZE (innermode). Use
4892 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
4893 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
4894 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
4895 Change innermode from fixed_mode_size to machine_mode.
4896 (simplify_subreg): Update call accordingly. Handle a constant-sized
4897 subreg of a variable-length CONST_VECTOR.
4899 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4900 Alan Hayward <alan.hayward@arm.com>
4901 David Sherwood <david.sherwood@arm.com>
4903 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
4904 (add_offset_to_base): New function, split out from...
4905 (create_mem_ref): ...here. When handling a scale other than 1,
4906 check first whether the address is valid without the offset.
4907 Add it into the base if so, leaving the index and scale as-is.
4909 2018-01-12 Jakub Jelinek <jakub@redhat.com>
4912 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
4913 fold_for_warn before checking if arg2 is INTEGER_CST.
4915 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
4917 * config/rs6000/predicates.md (load_multiple_operation): Delete.
4918 (store_multiple_operation): Delete.
4919 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
4920 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
4921 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
4922 guarded by TARGET_STRING.
4923 (rs6000_output_load_multiple): Delete.
4924 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
4925 OPTION_MASK_STRING / TARGET_STRING handling.
4926 (print_operand) <'N', 'O'>: Add comment that these are unused now.
4927 (const rs6000_opt_masks) <"string">: Change mask to 0.
4928 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
4929 (MASK_STRING): Delete.
4930 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
4932 (load_multiple): Delete.
4939 (store_multiple): Delete.
4946 (movmemsi_8reg): Delete.
4947 (corresponding unnamed define_insn): Delete.
4948 (movmemsi_6reg): Delete.
4949 (corresponding unnamed define_insn): Delete.
4950 (movmemsi_4reg): Delete.
4951 (corresponding unnamed define_insn): Delete.
4952 (movmemsi_2reg): Delete.
4953 (corresponding unnamed define_insn): Delete.
4954 (movmemsi_1reg): Delete.
4955 (corresponding unnamed define_insn): Delete.
4956 * config/rs6000/rs6000.opt (mno-string): New.
4957 (mstring): Replace by deprecation warning stub.
4958 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
4960 2018-01-12 Jakub Jelinek <jakub@redhat.com>
4962 * regrename.c (regrename_do_replace): If replacing the same
4963 reg multiple times, try to reuse last created gen_raw_REG.
4966 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
4967 main to workaround a bug in GDB.
4969 2018-01-12 Tom de Vries <tom@codesourcery.com>
4972 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
4974 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
4976 PR rtl-optimization/80481
4977 * ira-color.c (get_cap_member): New function.
4978 (allocnos_conflict_by_live_ranges_p): Use it.
4979 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
4980 (setup_slot_coalesced_allocno_live_ranges): Ditto.
4982 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
4985 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
4986 (*saddl_se_1): Ditto.
4988 (*ssubl_se_1): Ditto.
4990 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
4992 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
4993 rather than wi::to_widest for DR_INITs.
4994 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
4995 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
4996 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
4998 (vect_analyze_group_access_1): Note that here.
5000 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5002 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
5003 polynomial type sizes.
5005 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5007 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
5008 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
5009 (gimple_add_tmp_var): Likewise.
5011 2018-01-12 Martin Liska <mliska@suse.cz>
5013 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
5014 (gimple_alloc_sizes): Likewise.
5015 (dump_gimple_statistics): Use PRIu64 in printf format.
5016 * gimple.h: Change uint64_t to int.
5018 2018-01-12 Martin Liska <mliska@suse.cz>
5020 * tree-core.h: Use uint64_t instead of int.
5021 * tree.c (tree_node_counts): Likewise.
5022 (tree_node_sizes): Likewise.
5023 (dump_tree_statistics): Use PRIu64 in printf format.
5025 2018-01-12 Martin Liska <mliska@suse.cz>
5027 * Makefile.in: As qsort_chk is implemented in vec.c, add
5028 vec.o to linkage of gencfn-macros.
5029 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
5030 passing the info to record_node_allocation_statistics.
5031 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
5033 * ggc-common.c (struct ggc_usage): Add operator== and use
5034 it in operator< and compare function.
5035 * mem-stats.h (struct mem_usage): Likewise.
5036 * vec.c (struct vec_usage): Remove operator< and compare
5037 function. Can be simply inherited.
5039 2018-01-12 Martin Jambor <mjambor@suse.cz>
5042 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
5043 * tree-ssa-math-opts.c: Include domwalk.h.
5044 (convert_mult_to_fma_1): New function.
5045 (fma_transformation_info): New type.
5046 (fma_deferring_state): Likewise.
5047 (cancel_fma_deferring): New function.
5048 (result_of_phi): Likewise.
5049 (last_fma_candidate_feeds_initial_phi): Likewise.
5050 (convert_mult_to_fma): Added deferring logic, split actual
5051 transformation to convert_mult_to_fma_1.
5052 (math_opts_dom_walker): New type.
5053 (math_opts_dom_walker::after_dom_children): New method, body moved
5054 here from pass_optimize_widening_mul::execute, added deferring logic
5056 (pass_optimize_widening_mul::execute): Moved most of code to
5057 math_opts_dom_walker::after_dom_children.
5058 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
5059 * config/i386/i386.c (ix86_option_override_internal): Added
5060 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
5062 2018-01-12 Richard Biener <rguenther@suse.de>
5065 * dwarf2out.c (gen_variable_die): Do not reset old_die for
5066 inline instance vars.
5068 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
5071 * config/rx/rx.c (rx_is_restricted_memory_address):
5074 2018-01-12 Richard Biener <rguenther@suse.de>
5076 PR tree-optimization/80846
5077 * target.def (split_reduction): New target hook.
5078 * targhooks.c (default_split_reduction): New function.
5079 * targhooks.h (default_split_reduction): Declare.
5080 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
5081 target requests first reduce vectors by combining low and high
5083 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
5084 (get_vectype_for_scalar_type_and_size): Export.
5085 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
5086 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
5087 * doc/tm.texi: Regenerate.
5088 * config/i386/i386.c (ix86_split_reduction): Implement
5089 TARGET_VECTORIZE_SPLIT_REDUCTION.
5091 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5094 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
5095 in PIC mode except for TARGET_VXWORKS_RTP.
5096 * config/sparc/sparc.c: Include cfgrtl.h.
5097 (TARGET_INIT_PIC_REG): Define.
5098 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
5099 (sparc_pic_register_p): New predicate.
5100 (sparc_legitimate_address_p): Use it.
5101 (sparc_legitimize_pic_address): Likewise.
5102 (sparc_delegitimize_address): Likewise.
5103 (sparc_mode_dependent_address_p): Likewise.
5104 (gen_load_pcrel_sym): Remove 4th parameter.
5105 (load_got_register): Adjust call to above. Remove obsolete stuff.
5106 (sparc_expand_prologue): Do not call load_got_register here.
5107 (sparc_flat_expand_prologue): Likewise.
5108 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
5109 (sparc_use_pseudo_pic_reg): New function.
5110 (sparc_init_pic_reg): Likewise.
5111 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5112 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
5114 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
5116 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
5117 Add item for branch_cost.
5119 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5121 PR rtl-optimization/83565
5122 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
5123 not extend the result to a larger mode for rotate operations.
5124 (num_sign_bit_copies1): Likewise.
5126 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5129 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
5131 Use values-Xc.o for -pedantic.
5132 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
5134 2018-01-12 Martin Liska <mliska@suse.cz>
5137 * ipa-devirt.c (final_warning_record::grow_type_warnings):
5139 (possible_polymorphic_call_targets): Use it.
5140 (ipa_devirt): Likewise.
5142 2018-01-12 Martin Liska <mliska@suse.cz>
5144 * profile-count.h (enum profile_quality): Use 0 as invalid
5145 enum value of profile_quality.
5147 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
5149 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
5150 -mext-string options.
5152 2018-01-12 Richard Biener <rguenther@suse.de>
5154 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
5155 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
5156 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
5158 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
5160 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5162 * configure.ac (--with-long-double-format): Add support for the
5163 configuration option to change the default long double format on
5165 * config.gcc (powerpc*-linux*-*): Likewise.
5166 * configure: Regenerate.
5167 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
5168 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
5169 used without modification.
5171 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5173 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
5174 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
5175 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
5176 MISC_BUILTIN_SPEC_BARRIER.
5177 (rs6000_init_builtins): Likewise.
5178 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
5180 (speculation_barrier): New define_insn.
5181 * doc/extend.texi: Document __builtin_speculation_barrier.
5183 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5186 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
5187 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
5188 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
5190 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
5191 integral modes instead of "ss" and "sd".
5192 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
5193 vectors with 32-bit and 64-bit elements.
5194 (vecdupssescalarmodesuffix): New mode attribute.
5195 (vec_dup<mode>): Use it.
5197 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
5200 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
5201 frame if argument is passed on stack.
5203 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5206 * ree.c (combine_reaching_defs): Optimize also
5207 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
5208 reg2=any_extend(exp); reg1=reg2;, formatting fix.
5210 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5213 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
5215 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5218 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
5219 after they are computed.
5221 2018-01-11 Bin Cheng <bin.cheng@arm.com>
5223 PR tree-optimization/83695
5224 * gimple-loop-linterchange.cc
5225 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
5226 reset cached scev information after interchange.
5227 (pass_linterchange::execute): Remove call to scev_reset_htab.
5229 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5231 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
5232 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
5233 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
5234 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
5235 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
5236 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
5237 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
5238 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
5239 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
5240 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
5241 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
5242 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
5243 (V_lane_reg): Likewise.
5244 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
5246 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
5247 (vfmal_lane_low<mode>_intrinsic,
5248 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
5249 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
5250 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
5251 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
5252 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
5253 vfmsl_lane_high<mode>_intrinsic): New define_insns.
5255 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5257 * config/arm/arm-cpus.in (fp16fml): New feature.
5258 (ALL_SIMD): Add fp16fml.
5259 (armv8.2-a): Add fp16fml as an option.
5260 (armv8.3-a): Likewise.
5261 (armv8.4-a): Add fp16fml as part of fp16.
5262 * config/arm/arm.h (TARGET_FP16FML): Define.
5263 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
5265 * config/arm/arm-modes.def (V2HF): Define.
5266 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
5267 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
5268 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
5269 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
5270 vfmsl_low, vfmsl_high): New set of builtins.
5271 * config/arm/iterators.md (PLUSMINUS): New code iterator.
5272 (vfml_op): New code attribute.
5273 (VFMLHALVES): New int iterator.
5274 (VFML, VFMLSEL): New mode attributes.
5275 (V_reg): Define mapping for V2HF.
5276 (V_hi, V_lo): New mode attributes.
5277 (VF_constraint): Likewise.
5278 (vfml_half, vfml_half_selector): New int attributes.
5279 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
5281 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
5282 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
5284 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
5285 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
5286 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
5287 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
5289 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
5290 Document new effective target and option set.
5292 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5294 * config/arm/arm-cpus.in (armv8_4): New feature.
5295 (ARMv8_4a): New fgroup.
5296 (armv8.4-a): New arch.
5297 * config/arm/arm-tables.opt: Regenerate.
5298 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
5299 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
5300 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
5301 Add matching rules for -march=armv8.4-a and extensions.
5302 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
5304 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
5307 * config/rx/rx.md (BW): New mode attribute.
5308 (sync_lock_test_and_setsi): Add mode suffix to insn output.
5310 2018-01-11 Richard Biener <rguenther@suse.de>
5312 PR tree-optimization/83435
5313 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
5314 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
5315 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
5317 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5318 Alan Hayward <alan.hayward@arm.com>
5319 David Sherwood <david.sherwood@arm.com>
5321 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
5323 (aarch64_classify_address): Initialize it. Track polynomial offsets.
5324 (aarch64_print_address_internal): Use it to check for a zero offset.
5326 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5327 Alan Hayward <alan.hayward@arm.com>
5328 David Sherwood <david.sherwood@arm.com>
5330 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
5331 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
5332 Return a poly_int64 rather than a HOST_WIDE_INT.
5333 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
5334 rather than a HOST_WIDE_INT.
5335 * config/aarch64/aarch64.h (aarch64_frame): Protect with
5336 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
5337 hard_fp_offset, frame_size, initial_adjust, callee_offset and
5338 final_offset from HOST_WIDE_INT to poly_int64.
5339 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
5340 to_constant when getting the number of units in an Advanced SIMD
5342 (aarch64_builtin_vectorized_function): Check for a constant number
5344 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
5346 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
5347 attribute instead of GET_MODE_NUNITS.
5348 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
5349 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
5350 GET_MODE_SIZE for fixed-size registers.
5351 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
5352 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
5353 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
5354 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
5355 (aarch64_print_operand, aarch64_print_address_internal)
5356 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
5357 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
5358 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
5359 Handle polynomial GET_MODE_SIZE.
5360 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
5361 wider than SImode without modification.
5362 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
5363 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
5364 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
5365 passing and returning SVE modes.
5366 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
5367 rather than GEN_INT.
5368 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
5369 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
5370 (aarch64_allocate_and_probe_stack_space): Likewise.
5371 (aarch64_layout_frame): Cope with polynomial offsets.
5372 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
5373 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
5375 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
5376 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
5377 poly_int64 rather than a HOST_WIDE_INT.
5378 (aarch64_get_separate_components, aarch64_process_components)
5379 (aarch64_expand_prologue, aarch64_expand_epilogue)
5380 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
5381 (aarch64_anchor_offset): New function, split out from...
5382 (aarch64_legitimize_address): ...here.
5383 (aarch64_builtin_vectorization_cost): Handle polynomial
5384 TYPE_VECTOR_SUBPARTS.
5385 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
5387 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
5388 number of elements from the PARALLEL rather than the mode.
5389 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
5390 rather than GET_MODE_BITSIZE.
5391 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
5392 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
5393 (aarch64_expand_vec_perm_const_1): Handle polynomial
5394 d->perm.length () and d->perm elements.
5395 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
5396 Apply to_constant to d->perm elements.
5397 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
5398 polynomial CONST_VECTOR_NUNITS.
5399 (aarch64_move_pointer): Take amount as a poly_int64 rather
5401 (aarch64_progress_pointer): Avoid temporary variable.
5402 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
5403 the mode attribute instead of GET_MODE.
5405 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5406 Alan Hayward <alan.hayward@arm.com>
5407 David Sherwood <david.sherwood@arm.com>
5409 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
5410 x exists before using it.
5411 (aarch64_add_constant_internal): Rename to...
5412 (aarch64_add_offset_1): ...this. Replace regnum with separate
5413 src and dest rtxes. Handle the case in which they're different,
5414 including when the offset is zero. Replace scratchreg with an rtx.
5415 Use 2 additions if there is no spare register into which we can
5416 move a 16-bit constant.
5417 (aarch64_add_constant): Delete.
5418 (aarch64_add_offset): Replace reg with separate src and dest
5419 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
5420 Use aarch64_add_offset_1.
5421 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
5422 an rtx rather than an int. Take the delta as a poly_int64
5423 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
5424 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
5425 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
5426 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
5427 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
5429 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
5430 aarch64_add_constant.
5432 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5434 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
5435 Use scalar_float_mode.
5437 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5439 * config/aarch64/aarch64-simd.md
5440 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
5441 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
5442 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
5443 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
5444 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
5445 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
5446 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
5447 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
5448 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
5449 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
5451 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5454 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
5455 targ_options->x_arm_arch_string is non NULL.
5457 2018-01-11 Tamar Christina <tamar.christina@arm.com>
5459 * config/aarch64/aarch64.h
5460 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
5462 2018-01-11 Sudakshina Das <sudi.das@arm.com>
5465 * expmed.c (emit_store_flag_force): Swap if const op0
5466 and change VOIDmode to mode of op0.
5468 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5470 PR rtl-optimization/83761
5471 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
5472 than bytes to mode_for_size.
5474 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5477 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
5478 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
5481 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5484 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
5485 when in layout mode.
5486 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
5487 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
5490 2018-01-10 Michael Collison <michael.collison@arm.com>
5492 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
5493 * config/aarch64/aarch64-option-extension.def: Add
5494 AARCH64_OPT_EXTENSION of 'fp16fml'.
5495 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5496 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
5497 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
5498 * config/aarch64/constraints.md (Ui7): New constraint.
5499 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
5500 (VFMLA_SEL_W): Ditto.
5503 (VFMLA16_LOW): New int iterator.
5504 (VFMLA16_HIGH): Ditto.
5505 (UNSPEC_FMLAL): New unspec.
5506 (UNSPEC_FMLSL): Ditto.
5507 (UNSPEC_FMLAL2): Ditto.
5508 (UNSPEC_FMLSL2): Ditto.
5509 (f16mac): New code attribute.
5510 * config/aarch64/aarch64-simd-builtins.def
5511 (aarch64_fmlal_lowv2sf): Ditto.
5512 (aarch64_fmlsl_lowv2sf): Ditto.
5513 (aarch64_fmlalq_lowv4sf): Ditto.
5514 (aarch64_fmlslq_lowv4sf): Ditto.
5515 (aarch64_fmlal_highv2sf): Ditto.
5516 (aarch64_fmlsl_highv2sf): Ditto.
5517 (aarch64_fmlalq_highv4sf): Ditto.
5518 (aarch64_fmlslq_highv4sf): Ditto.
5519 (aarch64_fmlal_lane_lowv2sf): Ditto.
5520 (aarch64_fmlsl_lane_lowv2sf): Ditto.
5521 (aarch64_fmlal_laneq_lowv2sf): Ditto.
5522 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
5523 (aarch64_fmlalq_lane_lowv4sf): Ditto.
5524 (aarch64_fmlsl_lane_lowv4sf): Ditto.
5525 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
5526 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
5527 (aarch64_fmlal_lane_highv2sf): Ditto.
5528 (aarch64_fmlsl_lane_highv2sf): Ditto.
5529 (aarch64_fmlal_laneq_highv2sf): Ditto.
5530 (aarch64_fmlsl_laneq_highv2sf): Ditto.
5531 (aarch64_fmlalq_lane_highv4sf): Ditto.
5532 (aarch64_fmlsl_lane_highv4sf): Ditto.
5533 (aarch64_fmlalq_laneq_highv4sf): Ditto.
5534 (aarch64_fmlsl_laneq_highv4sf): Ditto.
5535 * config/aarch64/aarch64-simd.md:
5536 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
5537 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5538 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
5539 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5540 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
5541 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
5542 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
5543 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
5544 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
5545 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
5546 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
5547 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
5548 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
5549 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
5550 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
5551 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
5552 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
5553 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
5554 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
5555 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
5556 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
5557 (vfmlsl_low_u32): Ditto.
5558 (vfmlalq_low_u32): Ditto.
5559 (vfmlslq_low_u32): Ditto.
5560 (vfmlal_high_u32): Ditto.
5561 (vfmlsl_high_u32): Ditto.
5562 (vfmlalq_high_u32): Ditto.
5563 (vfmlslq_high_u32): Ditto.
5564 (vfmlal_lane_low_u32): Ditto.
5565 (vfmlsl_lane_low_u32): Ditto.
5566 (vfmlal_laneq_low_u32): Ditto.
5567 (vfmlsl_laneq_low_u32): Ditto.
5568 (vfmlalq_lane_low_u32): Ditto.
5569 (vfmlslq_lane_low_u32): Ditto.
5570 (vfmlalq_laneq_low_u32): Ditto.
5571 (vfmlslq_laneq_low_u32): Ditto.
5572 (vfmlal_lane_high_u32): Ditto.
5573 (vfmlsl_lane_high_u32): Ditto.
5574 (vfmlal_laneq_high_u32): Ditto.
5575 (vfmlsl_laneq_high_u32): Ditto.
5576 (vfmlalq_lane_high_u32): Ditto.
5577 (vfmlslq_lane_high_u32): Ditto.
5578 (vfmlalq_laneq_high_u32): Ditto.
5579 (vfmlslq_laneq_high_u32): Ditto.
5580 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
5581 (AARCH64_FL_FOR_ARCH8_4): New.
5582 (AARCH64_ISA_F16FML): New ISA flag.
5583 (TARGET_F16FML): New feature flag for fp16fml.
5584 (doc/invoke.texi): Document new fp16fml option.
5586 2018-01-10 Michael Collison <michael.collison@arm.com>
5588 * config/aarch64/aarch64-builtins.c:
5589 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
5590 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5591 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
5592 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
5593 (AARCH64_ISA_SHA3): New ISA flag.
5594 (TARGET_SHA3): New feature flag for sha3.
5595 * config/aarch64/iterators.md (sha512_op): New int attribute.
5596 (CRYPTO_SHA512): New int iterator.
5597 (UNSPEC_SHA512H): New unspec.
5598 (UNSPEC_SHA512H2): Ditto.
5599 (UNSPEC_SHA512SU0): Ditto.
5600 (UNSPEC_SHA512SU1): Ditto.
5601 * config/aarch64/aarch64-simd-builtins.def
5602 (aarch64_crypto_sha512hqv2di): New builtin.
5603 (aarch64_crypto_sha512h2qv2di): Ditto.
5604 (aarch64_crypto_sha512su0qv2di): Ditto.
5605 (aarch64_crypto_sha512su1qv2di): Ditto.
5606 (aarch64_eor3qv8hi): Ditto.
5607 (aarch64_rax1qv2di): Ditto.
5608 (aarch64_xarqv2di): Ditto.
5609 (aarch64_bcaxqv8hi): Ditto.
5610 * config/aarch64/aarch64-simd.md:
5611 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
5612 (aarch64_crypto_sha512su0qv2di): Ditto.
5613 (aarch64_crypto_sha512su1qv2di): Ditto.
5614 (aarch64_eor3qv8hi): Ditto.
5615 (aarch64_rax1qv2di): Ditto.
5616 (aarch64_xarqv2di): Ditto.
5617 (aarch64_bcaxqv8hi): Ditto.
5618 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
5619 (vsha512h2q_u64): Ditto.
5620 (vsha512su0q_u64): Ditto.
5621 (vsha512su1q_u64): Ditto.
5622 (veor3q_u16): Ditto.
5623 (vrax1q_u64): Ditto.
5625 (vbcaxq_u16): Ditto.
5626 * config/arm/types.md (crypto_sha512): New type attribute.
5627 (crypto_sha3): Ditto.
5628 (doc/invoke.texi): Document new sha3 option.
5630 2018-01-10 Michael Collison <michael.collison@arm.com>
5632 * config/aarch64/aarch64-builtins.c:
5633 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
5634 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5635 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
5636 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
5637 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
5638 (AARCH64_ISA_SM4): New ISA flag.
5639 (TARGET_SM4): New feature flag for sm4.
5640 * config/aarch64/aarch64-simd-builtins.def
5641 (aarch64_sm3ss1qv4si): Ditto.
5642 (aarch64_sm3tt1aq4si): Ditto.
5643 (aarch64_sm3tt1bq4si): Ditto.
5644 (aarch64_sm3tt2aq4si): Ditto.
5645 (aarch64_sm3tt2bq4si): Ditto.
5646 (aarch64_sm3partw1qv4si): Ditto.
5647 (aarch64_sm3partw2qv4si): Ditto.
5648 (aarch64_sm4eqv4si): Ditto.
5649 (aarch64_sm4ekeyqv4si): Ditto.
5650 * config/aarch64/aarch64-simd.md:
5651 (aarch64_sm3ss1qv4si): Ditto.
5652 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
5653 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
5654 (aarch64_sm4eqv4si): Ditto.
5655 (aarch64_sm4ekeyqv4si): Ditto.
5656 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
5657 (sm3part_op): Ditto.
5658 (CRYPTO_SM3TT): Ditto.
5659 (CRYPTO_SM3PART): Ditto.
5660 (UNSPEC_SM3SS1): New unspec.
5661 (UNSPEC_SM3TT1A): Ditto.
5662 (UNSPEC_SM3TT1B): Ditto.
5663 (UNSPEC_SM3TT2A): Ditto.
5664 (UNSPEC_SM3TT2B): Ditto.
5665 (UNSPEC_SM3PARTW1): Ditto.
5666 (UNSPEC_SM3PARTW2): Ditto.
5667 (UNSPEC_SM4E): Ditto.
5668 (UNSPEC_SM4EKEY): Ditto.
5669 * config/aarch64/constraints.md (Ui2): New constraint.
5670 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
5671 * config/arm/types.md (crypto_sm3): New type attribute.
5672 (crypto_sm4): Ditto.
5673 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
5674 (vsm3tt1aq_u32): Ditto.
5675 (vsm3tt1bq_u32): Ditto.
5676 (vsm3tt2aq_u32): Ditto.
5677 (vsm3tt2bq_u32): Ditto.
5678 (vsm3partw1q_u32): Ditto.
5679 (vsm3partw2q_u32): Ditto.
5680 (vsm4eq_u32): Ditto.
5681 (vsm4ekeyq_u32): Ditto.
5682 (doc/invoke.texi): Document new sm4 option.
5684 2018-01-10 Michael Collison <michael.collison@arm.com>
5686 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
5687 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
5688 (AARCH64_FL_FOR_ARCH8_4): New.
5689 (AARCH64_FL_V8_4): New flag.
5690 (doc/invoke.texi): Document new armv8.4-a option.
5692 2018-01-10 Michael Collison <michael.collison@arm.com>
5694 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5695 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
5696 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
5697 * config/aarch64/aarch64-option-extension.def: Add
5698 AARCH64_OPT_EXTENSION of 'sha2'.
5699 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
5700 (crypto): Disable sha2 and aes if crypto disabled.
5701 (crypto): Enable aes and sha2 if enabled.
5702 (simd): Disable sha2 and aes if simd disabled.
5703 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
5705 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
5706 (TARGET_SHA2): New feature flag for sha2.
5707 (TARGET_AES): New feature flag for aes.
5708 * config/aarch64/aarch64-simd.md:
5709 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
5710 conditional on TARGET_AES.
5711 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
5712 (aarch64_crypto_sha1hsi): Make pattern conditional
5714 (aarch64_crypto_sha1hv4si): Ditto.
5715 (aarch64_be_crypto_sha1hv4si): Ditto.
5716 (aarch64_crypto_sha1su1v4si): Ditto.
5717 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
5718 (aarch64_crypto_sha1su0v4si): Ditto.
5719 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
5720 (aarch64_crypto_sha256su0v4si): Ditto.
5721 (aarch64_crypto_sha256su1v4si): Ditto.
5722 (doc/invoke.texi): Document new aes and sha2 options.
5724 2018-01-10 Martin Sebor <msebor@redhat.com>
5726 PR tree-optimization/83781
5727 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
5730 2018-01-11 Martin Sebor <msebor@gmail.com>
5731 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5733 PR tree-optimization/83501
5734 PR tree-optimization/81703
5736 * tree-ssa-strlen.c (get_string_cst): Rename...
5737 (get_string_len): ...to this. Handle global constants.
5738 (handle_char_store): Adjust.
5740 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
5741 Jim Wilson <jimw@sifive.com>
5743 * config/riscv/riscv-protos.h (riscv_output_return): New.
5744 * config/riscv/riscv.c (struct machine_function): New naked_p field.
5745 (riscv_attribute_table, riscv_output_return),
5746 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
5747 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
5748 (riscv_compute_frame_info): Only compute frame->mask if not a naked
5750 (riscv_expand_prologue): Add early return for naked function.
5751 (riscv_expand_epilogue): Likewise.
5752 (riscv_function_ok_for_sibcall): Return false for naked function.
5753 (riscv_set_current_function): New.
5754 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
5755 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
5756 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
5757 * doc/extend.texi (RISC-V Function Attributes): New.
5759 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
5761 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
5762 check for 128-bit long double before checking TCmode.
5763 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
5764 128-bit long doubles before checking TFmode or TCmode.
5765 (FLOAT128_IBM_P): Likewise.
5767 2018-01-10 Martin Sebor <msebor@redhat.com>
5769 PR tree-optimization/83671
5770 * builtins.c (c_strlen): Unconditionally return zero for the empty
5772 Use -Warray-bounds for warnings.
5773 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
5774 for non-constant array indices with COMPONENT_REF, arrays of
5775 arrays, and pointers to arrays.
5776 (gimple_fold_builtin_strlen): Determine and set length range for
5777 non-constant character arrays.
5779 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
5782 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
5785 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
5787 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
5789 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5792 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
5793 VECTOR_MEM_ALTIVEC_OR_VSX_P.
5794 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
5795 indexed_or_indirect_operand predicate.
5796 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
5797 (*vsx_le_perm_load_v8hi): Likewise.
5798 (*vsx_le_perm_load_v16qi): Likewise.
5799 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
5800 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
5801 (*vsx_le_perm_store_v8hi): Likewise.
5802 (*vsx_le_perm_store_v16qi): Likewise.
5803 (eight unnamed splitters): Likewise.
5805 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5807 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
5808 * config/rs6000/emmintrin.h: Likewise.
5809 * config/rs6000/mmintrin.h: Likewise.
5810 * config/rs6000/xmmintrin.h: Likewise.
5812 2018-01-10 David Malcolm <dmalcolm@redhat.com>
5815 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
5817 * tree.c (tree_nop_conversion): Return true for location wrapper
5819 (maybe_wrap_with_location): New function.
5820 (selftest::check_strip_nops): New function.
5821 (selftest::test_location_wrappers): New function.
5822 (selftest::tree_c_tests): Call it.
5823 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
5824 (maybe_wrap_with_location): New decl.
5825 (EXPR_LOCATION_WRAPPER_P): New macro.
5826 (location_wrapper_p): New inline function.
5827 (tree_strip_any_location_wrapper): New inline function.
5829 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
5832 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
5833 stack_realign_offset for the largest alignment of stack slot
5835 (ix86_find_max_used_stack_alignment): New function.
5836 (ix86_finalize_stack_frame_flags): Use it. Set
5837 max_used_stack_alignment if we don't realign stack.
5838 * config/i386/i386.h (machine_function): Add
5839 max_used_stack_alignment.
5841 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
5843 * config/arm/arm.opt (-mbranch-cost): New option.
5844 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
5847 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
5850 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
5851 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
5853 2018-01-10 Richard Biener <rguenther@suse.de>
5856 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
5857 early out so it also covers the case where we have a non-NULL
5860 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
5862 PR tree-optimization/83753
5863 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
5864 for non-strided grouped accesses if the number of elements is 1.
5866 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5869 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
5870 * i386.h (TARGET_USE_GATHER): Define.
5871 * x86-tune.def (X86_TUNE_USE_GATHER): New.
5873 2018-01-10 Martin Liska <mliska@suse.cz>
5876 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
5877 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
5879 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
5880 CLEANUP_NO_PARTITIONING is not set.
5882 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
5884 * doc/rtl.texi: Remove documentation of (const ...) wrappers
5885 for vectors, as a partial revert of r254296.
5886 * rtl.h (const_vec_p): Delete.
5887 (const_vec_duplicate_p): Don't test for vector CONSTs.
5888 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
5889 * expmed.c (make_tree): Likewise.
5892 * common.md (E, F): Use CONSTANT_P instead of checking for
5894 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
5895 checking for CONST_VECTOR.
5897 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5900 * predict.c (force_edge_cold): Handle in more sane way edges
5903 2018-01-09 Carl Love <cel@us.ibm.com>
5905 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
5907 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
5908 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
5909 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
5910 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
5911 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
5912 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
5913 * config/rs6000/rs6000-protos.h: Add extern defition for
5914 rs6000_generate_float2_double_code.
5915 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
5917 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
5918 (float2_v2df): Add define_expand.
5920 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
5923 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
5924 op_mode in the force_to_mode call.
5926 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5928 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
5929 instead of checking each element individually.
5930 (aarch64_evpc_uzp): Likewise.
5931 (aarch64_evpc_zip): Likewise.
5932 (aarch64_evpc_ext): Likewise.
5933 (aarch64_evpc_rev): Likewise.
5934 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
5935 instead of checking each element individually. Return true without
5937 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
5938 whether all selected elements come from the same input, instead of
5939 checking each element individually. Remove calls to gen_rtx_REG,
5940 start_sequence and end_sequence and instead assert that no rtl is
5943 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5945 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
5946 order of HIGH and CONST checks.
5948 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5950 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
5951 if the destination isn't an SSA_NAME.
5953 2018-01-09 Richard Biener <rguenther@suse.de>
5955 PR tree-optimization/83668
5956 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
5958 (canonicalize_loop_form): ... here, renamed from ...
5959 (canonicalize_loop_closed_ssa_form): ... this and amended to
5960 swap successor edges for loop exit blocks to make us use
5961 the RPO order we need for initial schedule generation.
5963 2018-01-09 Joseph Myers <joseph@codesourcery.com>
5965 PR tree-optimization/64811
5966 * match.pd: When optimizing comparisons with Inf, avoid
5967 introducing or losing exceptions from comparisons with NaN.
5969 2018-01-09 Martin Liska <mliska@suse.cz>
5972 * asan.c (shadow_mem_size): Add gcc_assert.
5974 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
5976 Don't save registers in main().
5979 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
5980 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
5981 * config/avr/avr.c (avr_set_current_function): Don't error if
5982 naked, OS_task or OS_main are specified at the same time.
5983 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
5985 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
5987 * common/config/avr/avr-common.c (avr_option_optimization_table):
5988 Switch on -mmain-is-OS_task for optimizing compilations.
5990 2018-01-09 Richard Biener <rguenther@suse.de>
5992 PR tree-optimization/83572
5993 * graphite.c: Include cfganal.h.
5994 (graphite_transform_loops): Connect infinite loops to exit
5995 and remove fake edges at the end.
5997 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5999 * ipa-inline.c (edge_badness): Revert accidental checkin.
6001 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6004 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
6005 symbols; not inline clones.
6007 2018-01-09 Jakub Jelinek <jakub@redhat.com>
6010 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
6011 hard registers. Formatting fixes.
6013 PR preprocessor/83722
6014 * gcc.c (try_generate_repro): Pass
6015 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
6016 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
6019 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
6020 Kito Cheng <kito.cheng@gmail.com>
6022 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
6023 (riscv_leaf_function_p): Delete.
6024 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
6026 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6028 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
6030 (do_ifelse): New function.
6031 (do_isel): New function.
6032 (do_sub3): New function.
6033 (do_add3): New function.
6034 (do_load_mask_compare): New function.
6035 (do_overlap_load_compare): New function.
6036 (expand_compare_loop): New function.
6037 (expand_block_compare): Call expand_compare_loop() when appropriate.
6038 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
6040 (-mblock-compare-inline-loop-limit): New option.
6042 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6045 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
6046 Reverse order of second and third operands in first alternative.
6047 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
6048 of first and second elements in UNSPEC_VPERMR vector.
6049 (altivec_expand_vec_perm_le): Likewise.
6051 2018-01-08 Jeff Law <law@redhat.com>
6053 PR rtl-optimizatin/81308
6054 * tree-switch-conversion.c (cfg_altered): New file scoped static.
6055 (process_switch): If group_case_labels makes a change, then set
6057 (pass_convert_switch::execute): If a switch is converted, then
6058 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
6060 PR rtl-optimization/81308
6061 * recog.c (split_all_insns): Conditionally cleanup the CFG after
6064 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
6066 PR target/83663 - Revert r255946
6067 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
6068 generation for cases where splatting a value is not useful.
6069 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
6070 across a vec_duplicate and a paradoxical subreg forming a vector
6071 mode to a vec_concat.
6073 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6075 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
6076 -march=armv8.3-a variants.
6077 * config/arm/t-multilib: Likewise.
6078 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
6080 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6082 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
6084 (cceq_ior_compare_complement): Give it a name so I can use it, and
6085 change boolean_or_operator predicate to boolean_operator so it can
6086 be used to generate a crand.
6087 (eqne): New code iterator.
6088 (bd/bd_neg): New code_attrs.
6089 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
6090 a single define_insn.
6091 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
6092 decrement (bdnzt/bdnzf/bdzt/bdzf).
6093 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
6094 with the new names of the branch decrement patterns, and added the
6095 names of the branch decrement conditional patterns.
6097 2018-01-08 Richard Biener <rguenther@suse.de>
6099 PR tree-optimization/83563
6100 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
6103 2018-01-08 Richard Biener <rguenther@suse.de>
6106 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
6108 2018-01-08 Richard Biener <rguenther@suse.de>
6110 PR tree-optimization/83685
6111 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
6112 references to abnormals.
6114 2018-01-08 Richard Biener <rguenther@suse.de>
6117 * dwarf2out.c (output_indirect_strings): Handle empty
6118 skeleton_debug_str_hash.
6119 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
6121 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6123 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
6124 (emit_store_direct): Likewise.
6125 (arc_trampoline_adjust_address): Likewise.
6126 (arc_asm_trampoline_template): New function.
6127 (arc_initialize_trampoline): Use asm_trampoline_template.
6128 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
6129 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
6130 * config/arc/arc.md (flush_icache): Delete pattern.
6132 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6134 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
6135 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
6138 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6141 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
6142 by not USED_FOR_TARGET.
6143 (make_pass_resolve_sw_modes): Likewise.
6145 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6147 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
6150 2018-01-08 Richard Biener <rguenther@suse.de>
6153 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
6155 2018-01-08 Richard Biener <rguenther@suse.de>
6158 * match.pd ((t * 2) / 2) -> t): Add missing :c.
6160 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
6163 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
6164 basic blocks with a small number of successors.
6165 (convert_control_dep_chain_into_preds): Improve handling of
6167 (dump_predicates): Split apart into...
6168 (dump_pred_chain): ...here...
6169 (dump_pred_info): ...and here.
6170 (can_one_predicate_be_invalidated_p): Add debugging printfs.
6171 (can_chain_union_be_invalidated_p): Improve check for invalidation
6173 (uninit_uses_cannot_happen): Avoid unnecessary if
6174 convert_control_dep_chain_into_preds yielded nothing.
6176 2018-01-06 Martin Sebor <msebor@redhat.com>
6178 PR tree-optimization/83640
6179 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
6180 subtracting negative offset from size.
6181 (builtin_access::overlap): Adjust offset bounds of the access to fall
6182 within the size of the object if possible.
6184 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
6186 PR rtl-optimization/83699
6187 * expmed.c (extract_bit_field_1): Restrict the vector usage of
6188 extract_bit_field_as_subreg to cases in which the extracted
6189 value is also a vector.
6191 * lra-constraints.c (process_alt_operands): Test for the equivalence
6192 substitutions when detecting a possible reload cycle.
6194 2018-01-06 Jakub Jelinek <jakub@redhat.com>
6197 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
6198 by default if flag_selective_schedling{,2}. Formatting fixes.
6200 PR rtl-optimization/83682
6201 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
6202 if it has non-VECTOR_MODE element mode.
6203 (vec_duplicate_p): Likewise.
6206 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
6207 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
6209 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6212 * config/i386/i386-builtin.def
6213 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
6214 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
6215 Require also OPTION_MASK_ISA_AVX512F in addition to
6216 OPTION_MASK_ISA_GFNI.
6217 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
6218 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
6219 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
6220 to OPTION_MASK_ISA_GFNI.
6221 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
6222 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
6223 OPTION_MASK_ISA_AVX512BW.
6224 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
6225 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
6226 addition to OPTION_MASK_ISA_GFNI.
6227 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
6228 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
6229 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
6230 to OPTION_MASK_ISA_GFNI.
6231 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
6232 a requirement for all ISAs rather than any of them with a few
6234 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
6236 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
6237 bitmasks to be enabled with 3 exceptions, instead of requiring any
6238 enabled ISA with lots of exceptions.
6239 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
6240 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
6241 Change avx512bw in isa attribute to avx512f.
6242 * config/i386/sgxintrin.h: Add license boilerplate.
6243 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
6244 to __AVX512F__ and __AVX512VL to __AVX512VL__.
6245 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
6246 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
6248 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
6249 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
6250 temporarily sse2 rather than sse if not enabled already.
6253 * config/i386/sse.md (VI248_VLBW): Rename to ...
6254 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
6255 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
6256 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
6257 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
6258 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
6259 mode iterator instead of VI248_VLBW.
6261 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
6263 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
6264 (record_modified): Skip clobbers; add debug output.
6265 (param_change_prob): Use sreal frequencies.
6267 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6269 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
6270 punt for user-aligned variables.
6272 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6274 * tree-chrec.c (chrec_contains_symbols): Return true for
6277 2018-01-05 Sudakshina Das <sudi.das@arm.com>
6280 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
6281 of (x|y) == x for BICS pattern.
6283 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6285 PR tree-optimization/83605
6286 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
6287 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
6290 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
6292 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
6293 * config/epiphany/rtems.h: New file.
6295 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6296 Uros Bizjak <ubizjak@gmail.com>
6299 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
6300 QIreg_operand instead of register_operand predicate.
6301 * config/i386/i386.c (ix86_rop_should_change_byte_p,
6302 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
6303 comments instead of -fmitigate[-_]rop.
6305 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6308 * cgraphunit.c (symbol_table::compile): Switch to text_section
6309 before calling assembly_start debug hook.
6310 * run-rtl-passes.c (run_rtl_passes): Likewise.
6313 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6315 * tree-vrp.c (extract_range_from_binary_expr_1): Check
6316 range_int_cst_p rather than !symbolic_range_p before calling
6317 extract_range_from_multiplicative_op_1.
6319 2018-01-04 Jeff Law <law@redhat.com>
6321 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
6322 redundant test in assertion.
6324 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6326 * doc/rtl.texi: Document machine_mode wrapper classes.
6328 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6330 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
6333 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6335 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
6336 the VEC_PERM_EXPR fold to fail.
6338 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6341 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
6342 to switched_sections.
6344 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6347 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
6350 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
6353 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
6354 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
6356 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6359 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
6360 is BLKmode and bitpos not zero or mode change is needed.
6362 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6365 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
6368 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
6371 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
6372 instead of MULT rtx. Update all corresponding splitters.
6374 (*ssub<modesuffix>): Ditto.
6376 (*cmp_sadd_di): Update split patterns.
6377 (*cmp_sadd_si): Ditto.
6378 (*cmp_sadd_sidi): Ditto.
6379 (*cmp_ssub_di): Ditto.
6380 (*cmp_ssub_si): Ditto.
6381 (*cmp_ssub_sidi): Ditto.
6382 * config/alpha/predicates.md (const23_operand): New predicate.
6383 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
6384 Look for ASHIFT, not MULT inner operand.
6385 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
6387 2018-01-04 Martin Liska <mliska@suse.cz>
6389 PR gcov-profile/83669
6390 * gcov.c (output_intermediate_file): Add version to intermediate
6392 * doc/gcov.texi: Document new field 'version' in intermediate
6393 file format. Fix location of '-k' option of gcov command.
6395 2018-01-04 Martin Liska <mliska@suse.cz>
6398 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
6400 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6402 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
6404 2018-01-03 Martin Sebor <msebor@redhat.com>
6406 PR tree-optimization/83655
6407 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
6408 checking calls with invalid arguments.
6410 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6412 * tree-vect-stmts.c (vect_get_store_rhs): New function.
6413 (vectorizable_mask_load_store): Delete.
6414 (vectorizable_call): Return false for masked loads and stores.
6415 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
6416 instead of gimple_assign_rhs1.
6417 (vectorizable_load): Handle IFN_MASK_LOAD.
6418 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
6420 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6422 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
6424 (vectorizable_mask_load_store): ...here.
6425 (vectorizable_load): ...and here.
6427 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6429 * tree-vect-stmts.c (vect_build_all_ones_mask)
6430 (vect_build_zero_merge_argument): New functions, split out from...
6431 (vectorizable_load): ...here.
6433 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6435 * tree-vect-stmts.c (vect_check_store_rhs): New function,
6437 (vectorizable_mask_load_store): ...here.
6438 (vectorizable_store): ...and here.
6440 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6442 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
6444 (vectorizable_mask_load_store): ...here.
6446 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6448 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
6449 (vect_model_store_cost): Take a vec_load_store_type instead of a
6451 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
6452 (vect_model_store_cost): Take a vec_load_store_type instead of a
6454 (vectorizable_mask_load_store): Update accordingly.
6455 (vectorizable_store): Likewise.
6456 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
6458 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6460 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
6461 IFN_MASK_LOAD calls here rather than...
6462 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
6464 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6465 Alan Hayward <alan.hayward@arm.com>
6466 David Sherwood <david.sherwood@arm.com>
6468 * expmed.c (extract_bit_field_1): For vector extracts,
6469 fall back to extract_bit_field_as_subreg if vec_extract
6472 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6473 Alan Hayward <alan.hayward@arm.com>
6474 David Sherwood <david.sherwood@arm.com>
6476 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
6477 they are variable or constant sized.
6478 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
6479 slots for constant-sized data.
6481 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6482 Alan Hayward <alan.hayward@arm.com>
6483 David Sherwood <david.sherwood@arm.com>
6485 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
6486 handling COND_EXPRs with boolean comparisons, try to find a better
6487 basis for the mask type than the boolean itself.
6489 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6491 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
6492 is calculated and how it can be overridden.
6493 * genmodes.c (max_bitsize_mode_any_mode): New variable.
6494 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
6496 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
6499 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6500 Alan Hayward <alan.hayward@arm.com>
6501 David Sherwood <david.sherwood@arm.com>
6503 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
6504 Remove the mode argument.
6505 (aarch64_simd_valid_immediate): Remove the mode and inverse
6507 * config/aarch64/iterators.md (bitsize): New iterator.
6508 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
6509 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
6510 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
6511 aarch64_simd_valid_immediate.
6512 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
6513 (aarch64_reg_or_bic_imm): Likewise.
6514 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
6515 with an insn_type enum and msl with a modifier_type enum.
6516 Replace element_width with a scalar_mode. Change the shift
6517 to unsigned int. Add constructors for scalar_float_mode and
6518 scalar_int_mode elements.
6519 (aarch64_vect_float_const_representable_p): Delete.
6520 (aarch64_can_const_movi_rtx_p)
6521 (aarch64_simd_scalar_immediate_valid_for_move)
6522 (aarch64_simd_make_constant): Update call to
6523 aarch64_simd_valid_immediate.
6524 (aarch64_advsimd_valid_immediate_hs): New function.
6525 (aarch64_advsimd_valid_immediate): Likewise.
6526 (aarch64_simd_valid_immediate): Remove mode and inverse
6527 arguments. Rewrite to use the above. Use const_vec_duplicate_p
6528 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
6529 and aarch64_float_const_representable_p on the result.
6530 (aarch64_output_simd_mov_immediate): Remove mode argument.
6531 Update call to aarch64_simd_valid_immediate and use of
6532 simd_immediate_info.
6533 (aarch64_output_scalar_simd_mov_immediate): Update call
6536 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6537 Alan Hayward <alan.hayward@arm.com>
6538 David Sherwood <david.sherwood@arm.com>
6540 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
6541 (mode_nunits): Likewise CONST_MODE_NUNITS.
6542 * machmode.def (ADJUST_NUNITS): Document.
6543 * genmodes.c (mode_data::need_nunits_adj): New field.
6544 (blank_mode): Update accordingly.
6545 (adj_nunits): New variable.
6546 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
6548 (emit_mode_size_inline): Set need_bytesize_adj for all modes
6549 listed in adj_nunits.
6550 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
6551 listed in adj_nunits. Don't emit case statements for such modes.
6552 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
6553 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
6554 nothing if adj_nunits is nonnull.
6555 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
6556 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
6557 (emit_mode_fbit): Update use of print_maybe_const_decl.
6558 (emit_move_size): Likewise. Treat the array as non-const
6560 (emit_mode_adjustments): Handle adj_nunits.
6562 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6564 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
6565 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
6566 (VECTOR_MODES): Use it.
6567 (make_vector_modes): Take the prefix as an argument.
6569 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6570 Alan Hayward <alan.hayward@arm.com>
6571 David Sherwood <david.sherwood@arm.com>
6573 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
6574 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
6575 for MODE_VECTOR_BOOL.
6576 * machmode.def (VECTOR_BOOL_MODE): Document.
6577 * genmodes.c (VECTOR_BOOL_MODE): New macro.
6578 (make_vector_bool_mode): New function.
6579 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
6581 * lto-streamer-in.c (lto_input_mode_table): Likewise.
6582 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
6584 * stor-layout.c (int_mode_for_mode): Likewise.
6585 * tree.c (build_vector_type_for_mode): Likewise.
6586 * varasm.c (output_constant_pool_2): Likewise.
6587 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
6588 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
6589 for MODE_VECTOR_BOOL.
6590 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
6591 of mode class checks.
6592 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
6593 instead of a list of mode class checks.
6594 (expand_vector_scalar_condition): Likewise.
6595 (type_for_widest_vector_mode): Handle BImode as an inner mode.
6597 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6598 Alan Hayward <alan.hayward@arm.com>
6599 David Sherwood <david.sherwood@arm.com>
6601 * machmode.h (mode_size): Change from unsigned short to
6603 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
6604 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6605 or if measurement_type is not polynomial.
6606 (fixed_size_mode::includes_p): Check for constant-sized modes.
6607 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
6608 return a poly_uint16 rather than an unsigned short.
6609 (emit_mode_size): Change the type of mode_size from unsigned short
6610 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
6611 (emit_mode_adjustments): Cope with polynomial vector sizes.
6612 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6614 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6616 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
6617 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
6618 * caller-save.c (setup_save_areas): Likewise.
6619 (replace_reg_with_saved_mem): Likewise.
6620 * calls.c (emit_library_call_value_1): Likewise.
6621 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
6622 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
6623 (gen_lowpart_for_combine): Likewise.
6624 * convert.c (convert_to_integer_1): Likewise.
6625 * cse.c (equiv_constant, cse_insn): Likewise.
6626 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
6627 (cselib_subst_to_values): Likewise.
6628 * dce.c (word_dce_process_block): Likewise.
6629 * df-problems.c (df_word_lr_mark_ref): Likewise.
6630 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
6631 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
6632 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
6633 (rtl_for_decl_location): Likewise.
6634 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
6635 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
6636 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
6637 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
6638 (expand_expr_real_1): Likewise.
6639 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
6640 (pad_below): Likewise.
6641 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6642 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
6643 * ira.c (get_subreg_tracking_sizes): Likewise.
6644 * ira-build.c (ira_create_allocno_objects): Likewise.
6645 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
6646 (ira_sort_regnos_for_alter_reg): Likewise.
6647 * ira-costs.c (record_operand_costs): Likewise.
6648 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
6649 (resolve_simple_move): Likewise.
6650 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
6651 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
6652 (lra_constraints): Likewise.
6653 (CONST_POOL_OK_P): Reject variable-sized modes.
6654 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
6655 (add_pseudo_to_slot, lra_spill): Likewise.
6656 * omp-low.c (omp_clause_aligned_alignment): Likewise.
6657 * optabs-query.c (get_best_extraction_insn): Likewise.
6658 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6659 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
6660 (expand_mult_highpart, valid_multiword_target_p): Likewise.
6661 * recog.c (offsettable_address_addr_space_p): Likewise.
6662 * regcprop.c (maybe_mode_change): Likewise.
6663 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
6664 * regrename.c (build_def_use): Likewise.
6665 * regstat.c (dump_reg_info): Likewise.
6666 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
6667 (find_reloads, find_reloads_subreg_address): Likewise.
6668 * reload1.c (eliminate_regs_1): Likewise.
6669 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
6670 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
6671 (simplify_binary_operation_1, simplify_subreg): Likewise.
6672 * targhooks.c (default_function_arg_padding): Likewise.
6673 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
6674 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
6675 (verify_gimple_assign_ternary): Likewise.
6676 * tree-inline.c (estimate_move_cost): Likewise.
6677 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6678 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
6679 (get_address_cost_ainc): Likewise.
6680 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
6681 (vect_supportable_dr_alignment): Likewise.
6682 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6683 (vectorizable_reduction): Likewise.
6684 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
6685 (vectorizable_operation, vectorizable_load): Likewise.
6686 * tree.c (build_same_sized_truth_vector_type): Likewise.
6687 * valtrack.c (cleanup_auto_inc_dec): Likewise.
6688 * var-tracking.c (emit_note_insn_var_location): Likewise.
6689 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
6690 (ADDR_VEC_ALIGN): Likewise.
6692 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6693 Alan Hayward <alan.hayward@arm.com>
6694 David Sherwood <david.sherwood@arm.com>
6696 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
6698 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6699 or if measurement_type is polynomial.
6700 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
6701 * combine.c (make_extraction): Likewise.
6702 * dse.c (find_shift_sequence): Likewise.
6703 * dwarf2out.c (mem_loc_descriptor): Likewise.
6704 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
6705 (extract_bit_field, extract_low_bits): Likewise.
6706 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
6707 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
6708 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
6709 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
6710 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6711 * reload.c (find_reloads): Likewise.
6712 * reload1.c (alter_reg): Likewise.
6713 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
6714 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
6715 * tree-if-conv.c (predicate_mem_writes): Likewise.
6716 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
6717 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
6718 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
6719 * valtrack.c (dead_debug_insert_temp): Likewise.
6720 * varasm.c (mergeable_constant_section): Likewise.
6721 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
6723 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6724 Alan Hayward <alan.hayward@arm.com>
6725 David Sherwood <david.sherwood@arm.com>
6727 * expr.c (expand_assignment): Cope with polynomial mode sizes
6728 when assigning to a CONCAT.
6730 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6731 Alan Hayward <alan.hayward@arm.com>
6732 David Sherwood <david.sherwood@arm.com>
6734 * machmode.h (mode_precision): Change from unsigned short to
6736 (mode_to_precision): Return a poly_uint16 rather than an unsigned
6738 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
6739 or if measurement_type is not polynomial.
6740 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
6741 in which the mode is already known to be a scalar_int_mode.
6742 * genmodes.c (emit_mode_precision): Change the type of mode_precision
6743 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
6745 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6746 for GET_MODE_PRECISION.
6747 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6748 for GET_MODE_PRECISION.
6749 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
6751 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
6752 (expand_field_assignment, make_extraction): Likewise.
6753 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
6754 (get_last_value): Likewise.
6755 * convert.c (convert_to_integer_1): Likewise.
6756 * cse.c (cse_insn): Likewise.
6757 * expr.c (expand_expr_real_1): Likewise.
6758 * lra-constraints.c (simplify_operand_subreg): Likewise.
6759 * optabs-query.c (can_atomic_load_p): Likewise.
6760 * optabs.c (expand_atomic_load): Likewise.
6761 (expand_atomic_store): Likewise.
6762 * ree.c (combine_reaching_defs): Likewise.
6763 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
6764 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
6765 * tree.h (type_has_mode_precision_p): Likewise.
6766 * ubsan.c (instrument_si_overflow): Likewise.
6768 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6769 Alan Hayward <alan.hayward@arm.com>
6770 David Sherwood <david.sherwood@arm.com>
6772 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
6773 polynomial numbers of units.
6774 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
6775 (valid_vector_subparts_p): New function.
6776 (build_vector_type): Remove temporary shim and take the number
6777 of units as a poly_uint64 rather than an int.
6778 (build_opaque_vector_type): Take the number of units as a
6779 poly_uint64 rather than an int.
6780 * tree.c (build_vector_from_ctor): Handle polynomial
6781 TYPE_VECTOR_SUBPARTS.
6782 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
6783 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
6784 (build_vector_from_val): If the number of units is variable,
6785 use build_vec_duplicate_cst for constant operands and
6786 VEC_DUPLICATE_EXPR otherwise.
6787 (make_vector_type): Remove temporary is_constant ().
6788 (build_vector_type, build_opaque_vector_type): Take the number of
6789 units as a poly_uint64 rather than an int.
6790 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
6792 * cfgexpand.c (expand_debug_expr): Likewise.
6793 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
6794 (store_constructor, expand_expr_real_1): Likewise.
6795 (const_scalar_mask_from_tree): Likewise.
6796 * fold-const-call.c (fold_const_reduction): Likewise.
6797 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
6798 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
6799 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
6800 (fold_relational_const): Likewise.
6801 (native_interpret_vector): Likewise. Change the size from an
6802 int to an unsigned int.
6803 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
6804 TYPE_VECTOR_SUBPARTS.
6805 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
6806 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
6807 duplicating a non-constant operand into a variable-length vector.
6808 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
6809 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
6810 * ipa-icf.c (sem_variable::equals): Likewise.
6811 * match.pd: Likewise.
6812 * omp-simd-clone.c (simd_clone_subparts): Likewise.
6813 * print-tree.c (print_node): Likewise.
6814 * stor-layout.c (layout_type): Likewise.
6815 * targhooks.c (default_builtin_vectorization_cost): Likewise.
6816 * tree-cfg.c (verify_gimple_comparison): Likewise.
6817 (verify_gimple_assign_binary): Likewise.
6818 (verify_gimple_assign_ternary): Likewise.
6819 (verify_gimple_assign_single): Likewise.
6820 * tree-pretty-print.c (dump_generic_node): Likewise.
6821 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6822 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
6823 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
6824 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
6825 (vect_shift_permute_load_chain): Likewise.
6826 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
6827 (expand_vector_condition, optimize_vector_constructor): Likewise.
6828 (lower_vec_perm, get_compute_type): Likewise.
6829 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6830 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
6831 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
6832 (vect_recog_mask_conversion_pattern): Likewise.
6833 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
6834 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
6835 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
6836 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
6837 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
6838 (vectorizable_shift, vectorizable_operation, vectorizable_store)
6839 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
6840 (supportable_widening_operation): Likewise.
6841 (supportable_narrowing_operation): Likewise.
6842 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
6844 * varasm.c (output_constant): Likewise.
6846 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6847 Alan Hayward <alan.hayward@arm.com>
6848 David Sherwood <david.sherwood@arm.com>
6850 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
6851 so that both the length == 3 and length != 3 cases set up their
6852 own permute vectors. Add comments explaining why we know the
6853 number of elements is constant.
6854 (vect_permute_load_chain): Likewise.
6856 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6857 Alan Hayward <alan.hayward@arm.com>
6858 David Sherwood <david.sherwood@arm.com>
6860 * machmode.h (mode_nunits): Change from unsigned char to
6862 (ONLY_FIXED_SIZE_MODES): New macro.
6863 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
6864 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
6865 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
6867 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
6868 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
6869 or if measurement_type is not polynomial.
6870 * genmodes.c (ZERO_COEFFS): New macro.
6871 (emit_mode_nunits_inline): Make mode_nunits_inline return a
6873 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
6874 Use ZERO_COEFFS when emitting initializers.
6875 * data-streamer.h (bp_pack_poly_value): New function.
6876 (bp_unpack_poly_value): Likewise.
6877 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6878 for GET_MODE_NUNITS.
6879 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6880 for GET_MODE_NUNITS.
6881 * tree.c (make_vector_type): Remove temporary shim and make
6882 the real function take the number of units as a poly_uint64
6884 (build_vector_type_for_mode): Handle polynomial nunits.
6885 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
6886 * emit-rtl.c (const_vec_series_p_1): Likewise.
6887 (gen_rtx_CONST_VECTOR): Likewise.
6888 * fold-const.c (test_vec_duplicate_folding): Likewise.
6889 * genrecog.c (validate_pattern): Likewise.
6890 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
6891 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6892 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
6893 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
6894 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
6895 * rtlanal.c (subreg_get_info): Likewise.
6896 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
6897 (vect_grouped_load_supported): Likewise.
6898 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
6899 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
6900 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
6901 (simplify_const_unary_operation, simplify_binary_operation_1)
6902 (simplify_const_binary_operation, simplify_ternary_operation)
6903 (test_vector_ops_duplicate, test_vector_ops): Likewise.
6904 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
6905 instead of CONST_VECTOR_NUNITS.
6906 * varasm.c (output_constant_pool_2): Likewise.
6907 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
6908 explicit-encoded elements in the XVEC for variable-length vectors.
6910 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6912 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
6914 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6915 Alan Hayward <alan.hayward@arm.com>
6916 David Sherwood <david.sherwood@arm.com>
6918 * coretypes.h (fixed_size_mode): Declare.
6919 (fixed_size_mode_pod): New typedef.
6920 * builtins.h (target_builtins::x_apply_args_mode)
6921 (target_builtins::x_apply_result_mode): Change type to
6922 fixed_size_mode_pod.
6923 * builtins.c (apply_args_size, apply_result_size, result_vector)
6924 (expand_builtin_apply_args_1, expand_builtin_apply)
6925 (expand_builtin_return): Update accordingly.
6927 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6929 * cse.c (hash_rtx_cb): Hash only the encoded elements.
6930 * cselib.c (cselib_hash_rtx): Likewise.
6931 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
6932 CONST_VECTOR encoding.
6934 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6935 Jeff Law <law@redhat.com>
6938 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
6939 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
6940 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
6941 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
6944 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
6945 explicitly probe *sp in a noreturn function if there were any callee
6946 register saves or frame pointer is needed.
6948 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6951 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
6952 BLKmode for ternary, binary or unary expressions.
6955 * var-tracking.c (delete_vta_debug_insn): New inline function.
6956 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
6957 insns from get_insns () to NULL instead of each bb separately.
6958 Use delete_vta_debug_insn. No longer static.
6959 (vt_debug_insns_local, variable_tracking_main_1): Adjust
6960 delete_vta_debug_insns callers.
6961 * rtl.h (delete_vta_debug_insns): Declare.
6962 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
6963 instead of variable_tracking_main.
6965 2018-01-03 Martin Sebor <msebor@redhat.com>
6967 PR tree-optimization/83603
6968 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
6969 arguments past the endof the argument list in functions declared
6970 without a prototype.
6971 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
6972 Avoid checking when arguments are null.
6974 2018-01-03 Martin Sebor <msebor@redhat.com>
6977 * doc/extend.texi (attribute const): Fix a typo.
6978 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
6979 issuing -Wsuggest-attribute for void functions.
6981 2018-01-03 Martin Sebor <msebor@redhat.com>
6983 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
6984 offset_int::from instead of wide_int::to_shwi.
6985 (maybe_diag_overlap): Remove assertion.
6986 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
6987 * gimple-ssa-sprintf.c (format_directive): Same.
6988 (parse_directive): Same.
6989 (sprintf_dom_walker::compute_format_length): Same.
6990 (try_substitute_return_value): Same.
6992 2018-01-03 Jeff Law <law@redhat.com>
6995 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
6996 non-constant residual for zero at runtime and avoid probing in
6997 that case. Reorganize code for trailing problem to mirror handling
7000 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7002 PR tree-optimization/83501
7003 * tree-ssa-strlen.c (get_string_cst): New.
7004 (handle_char_store): Call get_string_cst.
7006 2018-01-03 Martin Liska <mliska@suse.cz>
7008 PR tree-optimization/83593
7009 * tree-ssa-strlen.c: Include tree-cfg.h.
7010 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
7011 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
7012 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
7014 (strlen_dom_walker::before_dom_children): Call
7015 gimple_purge_dead_eh_edges. Dump tranformation with details
7017 (strlen_dom_walker::before_dom_children): Update call by adding
7018 new argument cleanup_eh.
7019 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
7021 2018-01-03 Martin Liska <mliska@suse.cz>
7024 * cif-code.def (VARIADIC_THUNK): New enum value.
7025 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
7028 2018-01-03 Jan Beulich <jbeulich@suse.com>
7030 * sse.md (mov<mode>_internal): Tighten condition for when to use
7031 vmovdqu<ssescalarsize> for TI and OI modes.
7033 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7035 Update copyright years.
7037 2018-01-03 Martin Liska <mliska@suse.cz>
7040 * ipa-visibility.c (function_and_variable_visibility): Skip
7041 functions with noipa attribure.
7043 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7045 * gcc.c (process_command): Update copyright notice dates.
7046 * gcov-dump.c (print_version): Ditto.
7047 * gcov.c (print_version): Ditto.
7048 * gcov-tool.c (print_version): Ditto.
7049 * gengtype.c (create_file): Ditto.
7050 * doc/cpp.texi: Bump @copying's copyright year.
7051 * doc/cppinternals.texi: Ditto.
7052 * doc/gcc.texi: Ditto.
7053 * doc/gccint.texi: Ditto.
7054 * doc/gcov.texi: Ditto.
7055 * doc/install.texi: Ditto.
7056 * doc/invoke.texi: Ditto.
7058 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7060 * vector-builder.h (vector_builder::m_full_nelts): Change from
7061 unsigned int to poly_uint64.
7062 (vector_builder::full_nelts): Update prototype accordingly.
7063 (vector_builder::new_vector): Likewise.
7064 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
7065 (vector_builder::operator ==): Likewise.
7066 (vector_builder::finalize): Likewise.
7067 * int-vector-builder.h (int_vector_builder::int_vector_builder):
7068 Take the number of elements as a poly_uint64 rather than an
7070 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
7071 from unsigned int to poly_uint64.
7072 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
7073 (vec_perm_indices::new_vector): Likewise.
7074 (vec_perm_indices::length): Likewise.
7075 (vec_perm_indices::nelts_per_input): Likewise.
7076 (vec_perm_indices::input_nelts): Likewise.
7077 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
7078 number of elements per input as a poly_uint64 rather than an
7079 unsigned int. Use the original encoding for variable-length
7080 vectors, rather than clamping each individual element.
7081 For the second and subsequent elements in each pattern,
7082 clamp the step and base before clamping their sum.
7083 (vec_perm_indices::series_p): Handle polynomial element counts.
7084 (vec_perm_indices::all_in_range_p): Likewise.
7085 (vec_perm_indices_to_tree): Likewise.
7086 (vec_perm_indices_to_rtx): Likewise.
7087 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
7088 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
7089 (tree_vector_builder::new_binary_operation): Handle polynomial
7090 element counts. Return false if we need to know the number
7091 of elements at compile time.
7092 * fold-const.c (fold_vec_perm): Punt if the number of elements
7093 isn't known at compile time.
7095 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7097 * vec-perm-indices.h (vec_perm_builder): Change element type
7098 from HOST_WIDE_INT to poly_int64.
7099 (vec_perm_indices::element_type): Update accordingly.
7100 (vec_perm_indices::clamp): Handle polynomial element_types.
7101 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7102 (vec_perm_indices::all_in_range_p): Likewise.
7103 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
7105 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
7106 polynomial vec_perm_indices element types.
7107 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
7108 * fold-const.c (fold_vec_perm): Likewise.
7109 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
7110 * tree-vect-generic.c (lower_vec_perm): Likewise.
7111 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7112 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
7113 element type to HOST_WIDE_INT.
7115 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7116 Alan Hayward <alan.hayward@arm.com>
7117 David Sherwood <david.sherwood@arm.com>
7119 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
7120 rather than an int. Use plus_constant.
7121 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
7122 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
7124 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7125 Alan Hayward <alan.hayward@arm.com>
7126 David Sherwood <david.sherwood@arm.com>
7128 * calls.c (emit_call_1, expand_call): Change struct_value_size from
7129 a HOST_WIDE_INT to a poly_int64.
7131 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7132 Alan Hayward <alan.hayward@arm.com>
7133 David Sherwood <david.sherwood@arm.com>
7135 * calls.c (load_register_parameters): Cope with polynomial
7136 mode sizes. Require a constant size for BLKmode parameters
7137 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
7138 forces a parameter to be padded at the lsb end in order to
7139 fill a complete number of words, require the parameter size
7140 to be ordered wrt UNITS_PER_WORD.
7142 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7143 Alan Hayward <alan.hayward@arm.com>
7144 David Sherwood <david.sherwood@arm.com>
7146 * reload1.c (spill_stack_slot_width): Change element type
7147 from unsigned int to poly_uint64_pod.
7148 (alter_reg): Treat mode sizes as polynomial.
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 * reload.c (complex_word_subreg_p): New function.
7155 (reload_inner_reg_of_subreg, push_reload): Use it.
7157 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7158 Alan Hayward <alan.hayward@arm.com>
7159 David Sherwood <david.sherwood@arm.com>
7161 * lra-constraints.c (process_alt_operands): Reject matched
7162 operands whose sizes aren't ordered.
7163 (match_reload): Refer to this check here.
7165 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7166 Alan Hayward <alan.hayward@arm.com>
7167 David Sherwood <david.sherwood@arm.com>
7169 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
7170 that the mode size is in the set {1, 2, 4, 8, 16}.
7172 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7173 Alan Hayward <alan.hayward@arm.com>
7174 David Sherwood <david.sherwood@arm.com>
7176 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
7177 Use plus_constant instead of gen_rtx_PLUS.
7179 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7180 Alan Hayward <alan.hayward@arm.com>
7181 David Sherwood <david.sherwood@arm.com>
7183 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
7184 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
7185 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
7186 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
7187 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
7188 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
7189 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
7190 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
7191 * config/i386/i386.c (ix86_push_rounding): ...this new function.
7192 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
7194 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
7195 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
7196 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
7197 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
7198 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
7199 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
7200 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
7201 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
7202 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
7203 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
7205 * expr.c (emit_move_resolve_push): Treat the input and result
7206 of PUSH_ROUNDING as a poly_int64.
7207 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
7208 (emit_push_insn): Likewise.
7209 * lra-eliminations.c (mark_not_eliminable): Likewise.
7210 * recog.c (push_operand): Likewise.
7211 * reload1.c (elimination_effects): Likewise.
7212 * rtlanal.c (nonzero_bits1): Likewise.
7213 * calls.c (store_one_arg): Likewise. Require the padding to be
7214 known at compile time.
7216 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7217 Alan Hayward <alan.hayward@arm.com>
7218 David Sherwood <david.sherwood@arm.com>
7220 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
7221 Use plus_constant instead of gen_rtx_PLUS.
7223 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7224 Alan Hayward <alan.hayward@arm.com>
7225 David Sherwood <david.sherwood@arm.com>
7227 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
7230 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7231 Alan Hayward <alan.hayward@arm.com>
7232 David Sherwood <david.sherwood@arm.com>
7234 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
7235 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
7236 via stack temporaries. Treat the mode size as polynomial too.
7238 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7239 Alan Hayward <alan.hayward@arm.com>
7240 David Sherwood <david.sherwood@arm.com>
7242 * expr.c (expand_expr_real_2): When handling conversions involving
7243 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
7244 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
7245 as a poly_uint64 too.
7247 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7248 Alan Hayward <alan.hayward@arm.com>
7249 David Sherwood <david.sherwood@arm.com>
7251 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
7253 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7254 Alan Hayward <alan.hayward@arm.com>
7255 David Sherwood <david.sherwood@arm.com>
7257 * combine.c (can_change_dest_mode): Handle polynomial
7258 REGMODE_NATURAL_SIZE.
7259 * expmed.c (store_bit_field_1): Likewise.
7260 * expr.c (store_constructor): Likewise.
7261 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
7262 and polynomial REGMODE_NATURAL_SIZE.
7263 (gen_lowpart_common): Likewise.
7264 * reginfo.c (record_subregs_of_mode): Likewise.
7265 * rtlanal.c (read_modify_subreg_p): Likewise.
7267 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7268 Alan Hayward <alan.hayward@arm.com>
7269 David Sherwood <david.sherwood@arm.com>
7271 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
7272 numbers of elements.
7274 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7275 Alan Hayward <alan.hayward@arm.com>
7276 David Sherwood <david.sherwood@arm.com>
7278 * match.pd: Cope with polynomial numbers of vector elements.
7280 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7281 Alan Hayward <alan.hayward@arm.com>
7282 David Sherwood <david.sherwood@arm.com>
7284 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
7285 in a POINTER_PLUS_EXPR.
7287 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7288 Alan Hayward <alan.hayward@arm.com>
7289 David Sherwood <david.sherwood@arm.com>
7291 * omp-simd-clone.c (simd_clone_subparts): New function.
7292 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
7293 (ipa_simd_modify_function_body): Likewise.
7295 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7296 Alan Hayward <alan.hayward@arm.com>
7297 David Sherwood <david.sherwood@arm.com>
7299 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
7300 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
7301 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
7302 (expand_vector_condition, vector_element): Likewise.
7303 (subparts_gt): New function.
7304 (get_compute_type): Use subparts_gt.
7305 (count_type_subparts): Delete.
7306 (expand_vector_operations_1): Use subparts_gt instead of
7307 count_type_subparts.
7309 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7310 Alan Hayward <alan.hayward@arm.com>
7311 David Sherwood <david.sherwood@arm.com>
7313 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
7314 (vect_compile_time_alias): ...this new function. Do the calculation
7315 on poly_ints rather than trees.
7316 (vect_prune_runtime_alias_test_list): Update call accordingly.
7318 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7319 Alan Hayward <alan.hayward@arm.com>
7320 David Sherwood <david.sherwood@arm.com>
7322 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
7324 (vect_schedule_slp_instance): Likewise.
7326 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7327 Alan Hayward <alan.hayward@arm.com>
7328 David Sherwood <david.sherwood@arm.com>
7330 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
7331 constant and extern definitions for variable-length vectors.
7332 (vect_get_constant_vectors): Note that the number of units
7333 is known to be constant.
7335 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7336 Alan Hayward <alan.hayward@arm.com>
7337 David Sherwood <david.sherwood@arm.com>
7339 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
7340 of units as polynomial. Choose between WIDE and NARROW based
7343 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7344 Alan Hayward <alan.hayward@arm.com>
7345 David Sherwood <david.sherwood@arm.com>
7347 * tree-vect-stmts.c (simd_clone_subparts): New function.
7348 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
7350 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7351 Alan Hayward <alan.hayward@arm.com>
7352 David Sherwood <david.sherwood@arm.com>
7354 * tree-vect-stmts.c (vectorizable_call): Treat the number of
7355 vectors as polynomial. Use build_index_vector for
7358 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7359 Alan Hayward <alan.hayward@arm.com>
7360 David Sherwood <david.sherwood@arm.com>
7362 * tree-vect-stmts.c (get_load_store_type): Treat the number of
7363 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
7364 for variable-length vectors.
7365 (vectorizable_mask_load_store): Treat the number of units as
7366 polynomial, asserting that it is constant if the condition has
7367 already been enforced.
7368 (vectorizable_store, vectorizable_load): Likewise.
7370 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7371 Alan Hayward <alan.hayward@arm.com>
7372 David Sherwood <david.sherwood@arm.com>
7374 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
7375 of units as polynomial. Punt if we can't tell at compile time
7376 which vector contains the final result.
7378 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7379 Alan Hayward <alan.hayward@arm.com>
7380 David Sherwood <david.sherwood@arm.com>
7382 * tree-vect-loop.c (vectorizable_induction): Treat the number
7383 of units as polynomial. Punt on SLP inductions. Use an integer
7384 VEC_SERIES_EXPR for variable-length integer reductions. Use a
7385 cast of such a series for variable-length floating-point
7388 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7389 Alan Hayward <alan.hayward@arm.com>
7390 David Sherwood <david.sherwood@arm.com>
7392 * tree.h (build_index_vector): Declare.
7393 * tree.c (build_index_vector): New function.
7394 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
7395 of units as polynomial, forcibly converting it to a constant if
7396 vectorizable_reduction has already enforced the condition.
7397 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
7398 to create a {1,2,3,...} vector.
7399 (vectorizable_reduction): Treat the number of units as polynomial.
7400 Choose vectype_in based on the largest scalar element size rather
7401 than the smallest number of units. Enforce the restrictions
7404 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7405 Alan Hayward <alan.hayward@arm.com>
7406 David Sherwood <david.sherwood@arm.com>
7408 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
7409 number of units as polynomial.
7411 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7412 Alan Hayward <alan.hayward@arm.com>
7413 David Sherwood <david.sherwood@arm.com>
7415 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
7416 * target.def (autovectorize_vector_sizes): Return the vector sizes
7417 by pointer, using vector_sizes rather than a bitmask.
7418 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
7419 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
7420 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
7422 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
7423 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
7424 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
7425 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
7426 * omp-general.c (omp_max_vf): Likewise.
7427 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7428 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
7429 * tree-vect-loop.c (vect_analyze_loop): Likewise.
7430 * tree-vect-slp.c (vect_slp_bb): Likewise.
7431 * doc/tm.texi: Regenerate.
7432 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
7434 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
7435 the vector size as a poly_uint64 rather than an unsigned int.
7436 (current_vector_size): Change from an unsigned int to a poly_uint64.
7437 (get_vectype_for_scalar_type): Update accordingly.
7438 * tree.h (build_truth_vector_type): Take the size and number of
7439 units as a poly_uint64 rather than an unsigned int.
7440 (build_vector_type): Add a temporary overload that takes
7441 the number of units as a poly_uint64 rather than an unsigned int.
7442 * tree.c (make_vector_type): Likewise.
7443 (build_truth_vector_type): Take the number of units as a poly_uint64
7444 rather than an unsigned int.
7446 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7447 Alan Hayward <alan.hayward@arm.com>
7448 David Sherwood <david.sherwood@arm.com>
7450 * target.def (get_mask_mode): Take the number of units and length
7451 as poly_uint64s rather than unsigned ints.
7452 * targhooks.h (default_get_mask_mode): Update accordingly.
7453 * targhooks.c (default_get_mask_mode): Likewise.
7454 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
7455 * doc/tm.texi: Regenerate.
7457 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7458 Alan Hayward <alan.hayward@arm.com>
7459 David Sherwood <david.sherwood@arm.com>
7461 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
7462 * omp-general.c (omp_max_vf): Likewise.
7463 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
7464 (expand_omp_simd): Handle polynomial safelen.
7465 * omp-low.c (omplow_simd_context): Add a default constructor.
7466 (omplow_simd_context::max_vf): Change from int to poly_uint64.
7467 (lower_rec_simd_input_clauses): Update accordingly.
7468 (lower_rec_input_clauses): Likewise.
7470 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7471 Alan Hayward <alan.hayward@arm.com>
7472 David Sherwood <david.sherwood@arm.com>
7474 * tree-vectorizer.h (vect_nunits_for_cost): New function.
7475 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
7476 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
7477 (vect_analyze_slp_cost): Likewise.
7478 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
7479 (vect_model_load_cost): Likewise.
7481 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7482 Alan Hayward <alan.hayward@arm.com>
7483 David Sherwood <david.sherwood@arm.com>
7485 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
7486 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
7487 from an unsigned int * to a poly_uint64_pod *.
7488 (calculate_unrolling_factor): New function.
7489 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
7491 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7492 Alan Hayward <alan.hayward@arm.com>
7493 David Sherwood <david.sherwood@arm.com>
7495 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
7496 from an unsigned int to a poly_uint64.
7497 (_loop_vec_info::slp_unrolling_factor): Likewise.
7498 (_loop_vec_info::vectorization_factor): Change from an int
7500 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
7501 (vect_get_num_vectors): New function.
7502 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
7503 (vect_get_num_copies): Use vect_get_num_vectors.
7504 (vect_analyze_data_ref_dependences): Change max_vf from an int *
7505 to an unsigned int *.
7506 (vect_analyze_data_refs): Change min_vf from an int * to a
7508 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7509 than an unsigned HOST_WIDE_INT.
7510 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7511 (vect_analyze_data_ref_dependence): Change max_vf from an int *
7512 to an unsigned int *.
7513 (vect_analyze_data_ref_dependences): Likewise.
7514 (vect_compute_data_ref_alignment): Handle polynomial vf.
7515 (vect_enhance_data_refs_alignment): Likewise.
7516 (vect_prune_runtime_alias_test_list): Likewise.
7517 (vect_shift_permute_load_chain): Likewise.
7518 (vect_supportable_dr_alignment): Likewise.
7519 (dependence_distance_ge_vf): Take the vectorization factor as a
7520 poly_uint64 rather than an unsigned HOST_WIDE_INT.
7521 (vect_analyze_data_refs): Change min_vf from an int * to a
7523 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
7524 vfm1 as a poly_uint64 rather than an int. Make the same change
7525 for the returned bound_scalar.
7526 (vect_gen_vector_loop_niters): Handle polynomial vf.
7527 (vect_do_peeling): Likewise. Update call to
7528 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
7529 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
7531 * tree-vect-loop.c (vect_determine_vectorization_factor)
7532 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
7533 (vect_get_known_peeling_cost): Likewise.
7534 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
7535 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
7536 (vect_transform_loop): Likewise. Use the lowest possible VF when
7537 updating the upper bounds of the loop.
7538 (vect_min_worthwhile_factor): Make static. Return an unsigned int
7540 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
7541 polynomial unroll factors.
7542 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
7543 (vect_make_slp_decision): Likewise.
7544 (vect_supported_load_permutation_p): Likewise, and polynomial
7546 (vect_analyze_slp_cost): Handle polynomial vf.
7547 (vect_slp_analyze_node_operations): Likewise.
7548 (vect_slp_analyze_bb_1): Likewise.
7549 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7550 than an unsigned HOST_WIDE_INT.
7551 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
7552 (vectorizable_load): Handle polynomial vf.
7553 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
7555 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
7557 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7558 Alan Hayward <alan.hayward@arm.com>
7559 David Sherwood <david.sherwood@arm.com>
7561 * match.pd: Handle bit operations involving three constants
7562 and try to fold one pair.
7564 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7566 * tree-vect-loop-manip.c: Include gimple-fold.h.
7567 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
7568 niters_maybe_zero parameters. Handle other cases besides a step of 1.
7569 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
7570 Add a path that uses a step of VF instead of 1, but disable it
7572 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
7573 and niters_no_overflow parameters. Update calls to
7574 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
7575 Create a new SSA name if the latter choses to use a ste other
7576 than zero, and return it via niters_vector_mult_vf_var.
7577 * tree-vect-loop.c (vect_transform_loop): Update calls to
7578 vect_do_peeling, vect_gen_vector_loop_niters and
7579 slpeel_make_loop_iterate_ntimes.
7580 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
7581 (vect_gen_vector_loop_niters): Update declarations after above changes.
7583 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
7585 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
7586 128-bit round to integer instructions.
7587 (ceil<mode>2): Likewise.
7588 (btrunc<mode>2): Likewise.
7589 (round<mode>2): Likewise.
7591 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7593 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
7594 unaligned VSX load/store on P8/P9.
7595 (expand_block_clear): Allow the use of unaligned VSX
7596 load/store on P8/P9.
7598 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7600 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
7602 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
7603 swap associated with both a load and a store.
7605 2018-01-02 Andrew Waterman <andrew@sifive.com>
7607 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
7608 * config/riscv/riscv.md (clear_cache): Use it.
7610 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
7612 * web.c: Remove out-of-date comment.
7614 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7616 * expr.c (fixup_args_size_notes): Check that any existing
7617 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
7618 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
7619 (emit_single_push_insn): ...here.
7621 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7623 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
7624 (const_vector_encoded_nelts): New function.
7625 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
7626 (const_vector_int_elt, const_vector_elt): Declare.
7627 * emit-rtl.c (const_vector_int_elt_1): New function.
7628 (const_vector_elt): Likewise.
7629 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
7630 of CONST_VECTOR_ELT.
7632 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7634 * expr.c: Include rtx-vector-builder.h.
7635 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
7636 directly on the tree encoding.
7637 (const_vector_from_tree): Likewise.
7638 * optabs.c: Include rtx-vector-builder.h.
7639 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
7640 sequence of "u" values.
7641 * vec-perm-indices.c: Include rtx-vector-builder.h.
7642 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
7643 directly on the vec_perm_indices encoding.
7645 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7647 * doc/rtl.texi (const_vector): Describe new encoding scheme.
7648 * Makefile.in (OBJS): Add rtx-vector-builder.o.
7649 * rtx-vector-builder.h: New file.
7650 * rtx-vector-builder.c: Likewise.
7651 * rtl.h (rtx_def::u2): Add a const_vector field.
7652 (CONST_VECTOR_NPATTERNS): New macro.
7653 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
7654 (CONST_VECTOR_DUPLICATE_P): Likewise.
7655 (CONST_VECTOR_STEPPED_P): Likewise.
7656 (CONST_VECTOR_ENCODED_ELT): Likewise.
7657 (const_vec_duplicate_p): Check for a duplicated vector encoding.
7658 (unwrap_const_vec_duplicate): Likewise.
7659 (const_vec_series_p): Check for a non-duplicated vector encoding.
7660 Say that the function only returns true for integer vectors.
7661 * emit-rtl.c: Include rtx-vector-builder.h.
7662 (gen_const_vec_duplicate_1): Delete.
7663 (gen_const_vector): Call gen_const_vec_duplicate instead of
7664 gen_const_vec_duplicate_1.
7665 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
7666 (gen_const_vec_duplicate): Use rtx_vector_builder.
7667 (gen_const_vec_series): Likewise.
7668 (gen_rtx_CONST_VECTOR): Likewise.
7669 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
7670 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7671 Build a new vector rather than modifying a CONST_VECTOR in-place.
7672 (handle_special_swappables): Update call accordingly.
7673 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
7674 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7675 Build a new vector rather than modifying a CONST_VECTOR in-place.
7676 (handle_special_swappables): Update call accordingly.
7678 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7680 * simplify-rtx.c (simplify_const_binary_operation): Use
7681 CONST_VECTOR_ELT instead of XVECEXP.
7683 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7685 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
7686 the selector elements to be different from the data elements
7687 if the selector is a VECTOR_CST.
7688 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
7689 ssizetype for the selector.
7691 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7693 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
7694 before testing each element individually.
7695 * tree-vect-generic.c (lower_vec_perm): Likewise.
7697 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7699 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
7700 * selftest-run-tests.c (selftest::run_tests): Call it.
7701 * vector-builder.h (vector_builder::operator ==): New function.
7702 (vector_builder::operator !=): Likewise.
7703 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
7704 (vec_perm_indices::all_from_input_p): New function.
7705 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7706 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
7707 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
7708 instead of reading the VECTOR_CST directly. Detect whether both
7709 vector inputs are the same before constructing the vec_perm_indices,
7710 and update the number of inputs argument accordingly. Use the
7711 utility functions added above. Only construct sel2 if we need to.
7713 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7715 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
7716 the broadcast of the low byte.
7717 (expand_mult_highpart): Use an explicit encoding for the permutes.
7718 * optabs-query.c (can_mult_highpart_p): Likewise.
7719 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
7720 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7721 (vectorizable_bswap): Likewise.
7722 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
7723 explicit encoding for the power-of-2 permutes.
7724 (vect_permute_store_chain): Likewise.
7725 (vect_grouped_load_supported): Likewise.
7726 (vect_permute_load_chain): Likewise.
7728 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7730 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
7731 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
7732 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
7733 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7734 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
7735 (vect_gen_perm_mask_any): Likewise.
7737 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7739 * int-vector-builder.h: New file.
7740 * vec-perm-indices.h: Include int-vector-builder.h.
7741 (vec_perm_indices): Redefine as an int_vector_builder.
7742 (auto_vec_perm_indices): Delete.
7743 (vec_perm_builder): Redefine as a stand-alone class.
7744 (vec_perm_indices::vec_perm_indices): New function.
7745 (vec_perm_indices::clamp): Likewise.
7746 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
7747 (vec_perm_indices::new_vector): New function.
7748 (vec_perm_indices::new_expanded_vector): Update for new
7749 vec_perm_indices class.
7750 (vec_perm_indices::rotate_inputs): New function.
7751 (vec_perm_indices::all_in_range_p): Operate directly on the
7752 encoded form, without computing elided elements.
7753 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
7754 encoding. Update for new vec_perm_indices class.
7755 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
7756 the given vec_perm_builder.
7757 (expand_vec_perm_var): Update vec_perm_builder constructor.
7758 (expand_mult_highpart): Use vec_perm_builder instead of
7759 auto_vec_perm_indices.
7760 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
7761 vec_perm_indices instead of auto_vec_perm_indices. Use a single
7762 or double series encoding as appropriate.
7763 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
7764 vec_perm_indices instead of auto_vec_perm_indices.
7765 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7766 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7767 (vect_permute_store_chain): Likewise.
7768 (vect_grouped_load_supported): Likewise.
7769 (vect_permute_load_chain): Likewise.
7770 (vect_shift_permute_load_chain): Likewise.
7771 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
7772 (vect_transform_slp_perm_load): Likewise.
7773 (vect_schedule_slp_instance): Likewise.
7774 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7775 (vectorizable_mask_load_store): Likewise.
7776 (vectorizable_bswap): Likewise.
7777 (vectorizable_store): Likewise.
7778 (vectorizable_load): Likewise.
7779 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
7780 vec_perm_indices instead of auto_vec_perm_indices. Use
7781 tree_to_vec_perm_builder to read the vector from a tree.
7782 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
7783 vec_perm_builder instead of a vec_perm_indices.
7784 (have_whole_vector_shift): Use vec_perm_builder and
7785 vec_perm_indices instead of auto_vec_perm_indices. Leave the
7786 truncation to calc_vec_perm_mask_for_shift.
7787 (vect_create_epilog_for_reduction): Likewise.
7788 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
7789 from auto_vec_perm_indices to vec_perm_indices.
7790 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7791 instead of changing individual elements.
7792 (aarch64_vectorize_vec_perm_const): Use new_vector to install
7793 the vector in d.perm.
7794 * config/arm/arm.c (expand_vec_perm_d::perm): Change
7795 from auto_vec_perm_indices to vec_perm_indices.
7796 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7797 instead of changing individual elements.
7798 (arm_vectorize_vec_perm_const): Use new_vector to install
7799 the vector in d.perm.
7800 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
7801 Update vec_perm_builder constructor.
7802 (rs6000_expand_interleave): Likewise.
7803 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
7804 (rs6000_expand_interleave): Likewise.
7806 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7808 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
7809 to qimode could truncate the indices.
7810 * optabs.c (expand_vec_perm_var): Likewise.
7812 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7814 * Makefile.in (OBJS): Add vec-perm-indices.o.
7815 * vec-perm-indices.h: New file.
7816 * vec-perm-indices.c: Likewise.
7817 * target.h (vec_perm_indices): Replace with a forward class
7819 (auto_vec_perm_indices): Move to vec-perm-indices.h.
7820 * optabs.h: Include vec-perm-indices.h.
7821 (expand_vec_perm): Delete.
7822 (selector_fits_mode_p, expand_vec_perm_var): Declare.
7823 (expand_vec_perm_const): Declare.
7824 * target.def (vec_perm_const_ok): Replace with...
7825 (vec_perm_const): ...this new hook.
7826 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
7827 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
7828 * doc/tm.texi: Regenerate.
7829 * optabs.def (vec_perm_const): Delete.
7830 * doc/md.texi (vec_perm_const): Likewise.
7831 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
7832 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
7833 expand_vec_perm for constant permutation vectors. Assert that
7834 the mode of variable permutation vectors is the integer equivalent
7835 of the mode that is being permuted.
7836 * optabs-query.h (selector_fits_mode_p): Declare.
7837 * optabs-query.c: Include vec-perm-indices.h.
7838 (selector_fits_mode_p): New function.
7839 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
7840 is defined, instead of checking whether the vec_perm_const_optab
7841 exists. Use targetm.vectorize.vec_perm_const instead of
7842 targetm.vectorize.vec_perm_const_ok. Check whether the indices
7843 fit in the vector mode before using a variable permute.
7844 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
7845 vec_perm_indices instead of an rtx.
7846 (expand_vec_perm): Replace with...
7847 (expand_vec_perm_const): ...this new function. Take the selector
7848 as a vec_perm_indices rather than an rtx. Also take the mode of
7849 the selector. Update call to shift_amt_for_vec_perm_mask.
7850 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
7851 Use vec_perm_indices::new_expanded_vector to expand the original
7852 selector into bytes. Check whether the indices fit in the vector
7853 mode before using a variable permute.
7854 (expand_vec_perm_var): Make global.
7855 (expand_mult_highpart): Use expand_vec_perm_const.
7856 * fold-const.c: Includes vec-perm-indices.h.
7857 * tree-ssa-forwprop.c: Likewise.
7858 * tree-vect-data-refs.c: Likewise.
7859 * tree-vect-generic.c: Likewise.
7860 * tree-vect-loop.c: Likewise.
7861 * tree-vect-slp.c: Likewise.
7862 * tree-vect-stmts.c: Likewise.
7863 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
7865 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
7866 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
7867 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
7868 (aarch64_vectorize_vec_perm_const): ...this new function.
7869 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7870 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7871 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
7872 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
7873 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7874 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7875 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
7877 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
7878 check for NEON modes.
7879 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
7880 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
7881 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
7882 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
7884 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
7885 the old VEC_PERM_CONST conditions.
7886 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
7887 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
7888 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
7889 (ia64_vectorize_vec_perm_const_ok): Merge into...
7890 (ia64_vectorize_vec_perm_const): ...this new function.
7891 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
7892 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
7893 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
7894 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
7895 * config/mips/mips.c (mips_expand_vec_perm_const)
7896 (mips_vectorize_vec_perm_const_ok): Merge into...
7897 (mips_vectorize_vec_perm_const): ...this new function.
7898 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
7899 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
7900 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
7901 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
7902 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
7903 (rs6000_expand_vec_perm_const): Delete.
7904 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
7906 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7907 (altivec_expand_vec_perm_const_le): Take each operand individually.
7908 Operate on constant selectors rather than rtxes.
7909 (altivec_expand_vec_perm_const): Likewise. Update call to
7910 altivec_expand_vec_perm_const_le.
7911 (rs6000_expand_vec_perm_const): Delete.
7912 (rs6000_vectorize_vec_perm_const_ok): Delete.
7913 (rs6000_vectorize_vec_perm_const): New function.
7914 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
7915 an element count and rtx array.
7916 (rs6000_expand_extract_even): Update call accordingly.
7917 (rs6000_expand_interleave): Likewise.
7918 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
7919 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
7920 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
7921 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
7922 (rs6000_expand_vec_perm_const): Delete.
7923 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7924 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7925 (altivec_expand_vec_perm_const_le): Take each operand individually.
7926 Operate on constant selectors rather than rtxes.
7927 (altivec_expand_vec_perm_const): Likewise. Update call to
7928 altivec_expand_vec_perm_const_le.
7929 (rs6000_expand_vec_perm_const): Delete.
7930 (rs6000_vectorize_vec_perm_const_ok): Delete.
7931 (rs6000_vectorize_vec_perm_const): New function. Remove stray
7932 reference to the SPE evmerge intructions.
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/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
7938 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
7940 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7942 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7944 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
7945 vector mode and that that mode matches the mode of the data
7947 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
7948 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
7949 directly using expand_vec_perm_1 when forcing selectors into
7951 (expand_vec_perm_var): New function, split out from expand_vec_perm.
7953 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7955 * optabs-query.h (can_vec_perm_p): Delete.
7956 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
7957 * optabs-query.c (can_vec_perm_p): Split into...
7958 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
7959 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
7960 particular selector is valid.
7961 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7962 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7963 (vect_grouped_load_supported): Likewise.
7964 (vect_shift_permute_load_chain): Likewise.
7965 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
7966 (vect_transform_slp_perm_load): Likewise.
7967 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7968 (vectorizable_bswap): Likewise.
7969 (vect_gen_perm_mask_checked): Likewise.
7970 * fold-const.c (fold_ternary_loc): Likewise. Don't take
7971 implementations of variable permutation vectors into account
7972 when deciding which selector to use.
7973 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
7974 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
7975 with a false third argument.
7976 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
7977 to test whether the constant selector is valid and can_vec_perm_var_p
7978 to test whether a variable selector is valid.
7980 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7982 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
7983 * optabs-query.c (can_vec_perm_p): Likewise.
7984 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
7985 instead of vec_perm_indices.
7986 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
7987 (vect_gen_perm_mask_checked): Likewise,
7988 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
7989 (vect_gen_perm_mask_checked): Likewise,
7991 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7993 * optabs-query.h (qimode_for_vec_perm): Declare.
7994 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
7995 (qimode_for_vec_perm): ...this new function.
7996 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
7998 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8000 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
8001 does not have a conditional at the top.
8003 2018-01-02 Richard Biener <rguenther@suse.de>
8005 * ipa-inline.c (big_speedup_p): Fix expression.
8007 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8010 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
8013 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8017 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
8018 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
8019 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
8020 cond_taken_branch_cost 3->4.
8022 2018-01-01 Jakub Jelinek <jakub@redhat.com>
8024 PR tree-optimization/83581
8025 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
8026 TODO_cleanup_cfg if any changes have been made.
8029 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
8030 convert_modes if target mode has the right side, but different mode
8034 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
8035 last argument when extracting from CONCAT. If either from_real or
8036 from_imag is NULL, use expansion through memory. If result is not
8037 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
8038 the parts directly to inner mode, if even that fails, use expansion
8042 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
8043 check for bswap in mode rather than HImode and use that in expand_unop
8046 Copyright (C) 2018 Free Software Foundation, Inc.
8048 Copying and distribution of this file, with or without modification,
8049 are permitted in any medium without royalty provided the copyright
8050 notice and this notice are preserved.