1 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
6 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
8 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
11 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
13 2018-03-05 Richard Biener <rguenther@suse.de>
15 PR tree-optimization/84486
16 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
17 When inserting a __builtin_assume_aligned call set the LHS
18 SSA name alignment info accordingly.
20 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
22 PR tree-optimization/84114
23 * config/aarch64/aarch64.c (aarch64_reassociation_width)
24 Avoid reassociation of FLOAT_MODE addition.
26 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
28 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
29 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
30 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
31 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
32 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
33 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
34 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
36 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
37 __builtin_ia32_wbinvd): New builtins.
39 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
41 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
42 (ix86_valid_target_attribute_inner_p): Ditto.
43 (ix86_init_mmx_sse_builtins): Add special_args2.
44 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
45 TARGET_WBNOINVD_P): New.
46 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
47 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
48 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
49 * config/i386/immintrin.h (_wbinvd): New intrinsic.
50 * config/i386/pconfigintrin.h: New file.
51 * config/i386/wbnoinvdintrin.h: Ditto.
52 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
53 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
55 2018-03-05 Richard Biener <rguenther@suse.de>
57 PR tree-optimization/84670
58 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
60 (BB_VISITED_WITH_VISITED_SUCCS): New define.
61 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
62 (compute_antic_aux): Only assert the number of values in ANTIC_IN
63 doesn't grow if all successors (recursively) were visited at least
66 2018-03-05 Richard Biener <rguenther@suse.de>
68 PR tree-optimization/84650
69 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
70 if executed in the loop pipeline.
72 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
74 * doc/configfiles.texi (Configuration Files): Move info about
75 conditionalizing $target-protos.h to...
76 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
77 differs from $target-protos.h.
79 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
80 Chung-Ju Wu <jasonwucj@gmail.com>
82 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
83 * config/nds32/nds32-multiple.md (setmemsi): Define.
84 * config/nds32/nds32-memory-manipulation.c
85 (nds32_gen_dup_4_byte_to_word_value): New.
86 (emit_setmem_word_loop): New.
87 (emit_setmem_byte_loop): New.
88 (nds32_expand_setmem_loop): New.
89 (nds32_expand_setmem_loop_v3m): New.
90 (nds32_expand_setmem_unroll): New.
91 (nds32_expand_setmem): New.
93 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
94 Chung-Ju Wu <jasonwucj@gmail.com>
96 * config/nds32/nds32-memory-manipulation.c
97 (nds32_emit_load_store): New.
98 (nds32_emit_post_inc_load_store): New.
99 (nds32_emit_mem_move): New.
100 (nds32_emit_mem_move_block): New.
101 (nds32_expand_movmemsi_loop_unknown_size): New.
102 (nds32_expand_movmemsi_loop_known_size): New.
103 (nds32_expand_movmemsi_loop): New.
104 (nds32_expand_movmemsi_unroll): New.
105 (nds32_expand_movmemqi): Rename ...
106 (nds32_expand_movmemsi): ... to this.
107 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
108 (movmemsi): ... to this.
109 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
110 (nds32_expand_movmemsi): ... to this.
112 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
113 Monk Chiang <sh.chiang04@gmail.com>
114 Chung-Ju Wu <jasonwucj@gmail.com>
116 * config/nds32/nds32-protos.h
117 (nds32_expand_load_multiple): New arguments.
118 (nds32_expand_store_multiple): Ditto.
119 (nds32_valid_multiple_load_store): Rename ...
120 (nds32_valid_multiple_load_store_p): ... to this.
121 * config/nds32/nds32-memory-manipulation.c
122 (nds32_expand_load_multiple): Refine implementation.
123 (nds32_expand_store_multiple): Ditto.
124 * config/nds32/nds32-multiple.md
125 (load_multiple): Update nds32_expand_load_multiple interface.
126 (store_multiple): Update nds32_expand_store_multiple interface.
127 * config/nds32/nds32-predicates.c
128 (nds32_valid_multiple_load_store): Rename ...
129 (nds32_valid_multiple_load_store_p): ... to this and refine
131 * config/nds32/predicates.md
132 (nds32_load_multiple_and_update_address_operation): New predicate.
133 (nds32_store_multiple_and_update_address_operation): New predicate.
135 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
136 Chung-Ju Wu <jasonwucj@gmail.com>
138 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
139 (combo): New attribute.
140 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
142 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
144 * config/nds32/nds32.opt: Change -mcmodel= default value.
146 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
147 Monk Chiang <sh.chiang04@gmail.com>
148 Chung-Ju Wu <jasonwucj@gmail.com>
150 * config/nds32/constants.md (unspec_element): New enum.
151 * config/nds32/constraints.md (Umw): New constraint.
152 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
153 * config/nds32/nds32-intrinsic.md: Likewise.
154 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
155 (nds32_valid_smw_lwm_base_p): New.
156 (nds32_output_smw_single_word): New.
157 (nds32_output_lmw_single_word): New.
158 (nds32_expand_unaligned_load): New.
159 (nds32_expand_unaligned_store): New.
160 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
161 (nds32_output_smw_single_word): Declare.
162 (nds32_output_lmw_single_word): Declare.
163 (nds32_expand_unaligned_load): Declare.
164 (nds32_expand_unaligned_store): Declare.
165 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
166 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
167 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
168 NDS32_BUILTIN_UASTORE_DW.
169 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
172 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
173 Kito Cheng <kito.cheng@gmail.com>
174 Chung-Ju Wu <jasonwucj@gmail.com>
176 * config/nds32/nds32-intrinsic.c
177 (nds32_expand_builtin_null_ftype_reg): Delete.
178 (nds32_expand_builtin_reg_ftype_imm): Ditto.
179 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
180 (nds32_read_argument): New.
181 (nds32_legitimize_target): Ditto.
182 (nds32_legitimize_argument): Ditto.
183 (nds32_check_constant_argument): Ditto.
184 (nds32_expand_unop_builtin): Ditto.
185 (nds32_expand_unopimm_builtin): Ditto.
186 (nds32_expand_binop_builtin): Ditto.
187 (nds32_builtin_decl_impl): Ditto.
188 (builtin_description): Ditto.
189 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
190 (nds32_init_builtins_impl): Ditto.
191 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
192 (nds32_builtin_decl): New.
193 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
194 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
196 2018-03-02 Jeff Law <law@redhat.com>
198 * reorg.c (stop_search_p): Handle DEBUG_INSN.
199 (redundant_insn, fill_simple_delay_slots): Likewise.
200 (fill_slots_from_thread): Likewise.
201 * resource.c (mark_referenced_resources): Likewise.
202 (mark_set_resources, find_dead_or_set_registers): Likewise.
204 2018-03-02 Jakub Jelinek <jakub@redhat.com>
206 * substring-locations.h (format_warning_va): Formatting fix for
208 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
210 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
211 * substring-locations.c: Include intl.h.
212 (format_warning_va): Turned into small wrapper around
213 format_warning_n_va, renamed to ...
214 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
215 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
217 (format_warning_at_substring_n): New function.
218 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
219 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
220 format_warning_at_substring with just a shorter name instead of
221 const function pointer.
222 (fmtwarn_n): New function.
223 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
224 appropriate, get rid of all the fmtstr temporaries, move conditionals
225 with G_() wrapped string literals directly into fmtwarn arguments,
226 cast dir.len to (int), formatting fixes.
228 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
230 * doc/invoke.texi: Remove "Cilk Plus" references.
232 2018-03-02 Jakub Jelinek <jakub@redhat.com>
233 Richard Biener <rguenther@suse.de>
236 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
237 for error or warning attributes if CALL_FROM_THUNK_P is set.
240 2018-03-02 Jakub Jelinek <jakub@redhat.com>
243 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
244 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
247 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
248 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
250 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
251 instead of -1U in last predictors element's probability member.
253 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
256 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
257 arguments if they are comparable.
259 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
261 PR tree-optimization/84634
262 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
263 masks and masked_loop_p with a single loop_masks, making sure it's
264 null for bb vectorization.
266 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
268 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
269 (vect_analyze_data_ref_access): Use loop->safe_len rather than
270 loop->force_vectorize to check whether there is no alias.
272 2018-03-02 Jakub Jelinek <jakub@redhat.com>
275 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
277 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
279 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
280 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
281 instead of a loop around prev_real_insn.
282 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
286 * config/i386/i386.c (ix86_print_operand): Use conditional
287 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
290 2018-03-02 Richard Biener <rguenther@suse.de>
292 PR tree-optimization/84427
293 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
294 (bitmap_set_subtract_values): Rewrite to handle multiple
297 (prune_clobbered_mems): Likewise.
298 (phi_translate): Take edge instead of pred/phiblock.
299 (phi_translate_1): Likewise.
300 (phi_translate_set): Likewise. Insert all translated
301 exprs for a value into the set, keeping possibly multiple
302 expressions per value.
303 (compute_antic_aux): Adjust for phi_translate changes.
304 When intersecting union the expressions and prune those
305 not in the final value set, keeping possibly multiple
306 expressions per value. Do not use value-insertion
307 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
308 all expressions. Add verification that the value-sets
309 only shrink during iteration.
310 (compute_partial_antic_aux): Adjust for the phi_translate changes.
311 (do_pre_regular_insertion): Likewise.
312 (do_pre_partial_partial_insertion): Likewise.
314 2018-03-02 Richard Biener <rguenther@suse.de>
317 * config/darwin.c (saved_debug_info_level): New static global.
318 (darwin_asm_lto_start): Disable debug info generation for LTO out.
319 (darwin_asm_lto_end): Restore debug info generation settings.
321 2018-03-01 Martin Liska <mliska@suse.cz>
324 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
327 2018-03-01 Richard Biener <rguenther@suse.de>
330 * dwarf2out.c (gen_variable_die): Properly handle late VLA
331 type annotation with LTO when debug was disabled at compile-time.
333 2018-03-01 Matthew Fortune <mfortune@gmail.com>
335 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
337 (mips_final_postscan_insn): Likewise.
339 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
341 PR rtl-optimization/84528
342 * alias.c (init_alias_target): Add commentary.
343 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
344 a unique base value if the frame pointer is not eliminated
345 to the stack pointer.
347 2018-03-01 Tom de Vries <tom@codesourcery.com>
349 PR rtl-optimization/83327
350 * lra-int.h (hard_regs_spilled_into): Declare.
351 * lra.c (hard_regs_spilled_into): Define.
352 (init_reg_info): Init hard_regs_spilled_into.
353 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
354 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
355 (process_bb_lives): Handle hard_regs_spilled_into.
356 (lra_create_live_ranges_1): Before doing liveness propagation, clear
357 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
359 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
361 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
362 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
363 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
364 * config/rs6000/aix72.h: New file.
366 2018-02-28 Jakub Jelinek <jakub@redhat.com>
368 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
369 instead of warning_at with conditional singular and plural messages
373 * stor-layout.c (update_alignment_for_field): For
374 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
375 && !DECL_PACKED (field), do the alignment update, just use
376 only desired_align instead of MAX (type_align, desired_align)
378 (place_field): Don't do known_align < desired_align handling
379 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
380 is non-NULL, instead do it after rli->prev_field handling and
381 only if not within a bitfield word. For DECL_PACKED (field)
382 use type_align of BITS_PER_UNIT.
384 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
386 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
387 superfluous parentheses and trailing spaces.
389 2018-02-28 Richard Biener <rguenther@suse.de>
391 PR tree-optimization/84584
392 * graphite-scop-detection.c (scop_detection::add_scop): Discard
393 SCoPs with fake exit edge.
395 2018-02-28 Martin Liska <mliska@suse.cz>
398 * timevar.c (timer::print): Fix format to properly print 100%
401 2018-02-28 Richard Biener <rguenther@suse.de>
404 * genmatch.c (capture_info::walk_match): Do not mark
405 captured expressions without operands as expr_p given
406 they act more like predicates and should be subject to
407 "lost tail" side-effect preserving.
409 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
411 PR rtl-optimization/81611
412 * auto-inc-dec.c (attempt_change): Move dead note from
413 mem_insn if it's the next use of regno
414 (find_address): Take address use of reg holding
415 non-incremented value. Add parm to limit search to the named
417 (merge_in_block): Attempt to use a mem insn that is the next
418 use of the original regno.
420 2018-02-27 Martin Sebor <msebor@redhat.com>
423 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
424 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
426 2018-02-27 Martin Sebor <msebor@redhat.com>
429 * diagnostic-core.h (warning_n, error_n, inform_n): Change
430 n argument to unsigned HOST_WIDE_INT.
431 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
432 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
433 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
434 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
436 2018-02-27 Richard Biener <rguenther@suse.de>
438 PR tree-optimization/84512
439 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
440 Do not use the estimate returned from record_stmt_cost for
441 the scalar iteration cost but sum properly using add_stmt_cost.
443 2018-02-27 Richard Biener <rguenther@suse.de>
445 PR tree-optimization/84466
446 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
447 Adjust last change to less strictly validate use operands.
449 2018-02-27 Martin Liska <mliska@suse.cz>
451 PR gcov-profile/84548
452 * gcov.c (process_file): Allow partial overlap and consider it
453 also as group functions.
454 (output_lines): Properly calculate range of lines for a group.
456 2018-02-27 Martin Liska <mliska@suse.cz>
458 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
459 'ggc' suffixes. Change first column width.
460 (timer::print): Fix formatting of the column.
462 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
464 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
465 preserve inline entry blocks for the sake of debug inline
466 entry point markers alone.
467 (remove_unused_locals): Suggest in comments a better place to
468 force the preservation of inline entry blocks that are
469 otherwise unused, but do not preserve them.
471 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
473 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
475 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
478 * config/i386/constraints.md (Bs): Replace
479 ix86_indirect_branch_register with
480 TARGET_INDIRECT_BRANCH_REGISTER.
482 * config/i386/i386.md (indirect_jump): Likewise.
483 (tablejump): Likewise.
484 (*sibcall_memory): Likewise.
485 (*sibcall_value_memory): Likewise.
486 Peepholes of indirect call and jump via memory: Likewise.
487 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
488 (*sibcall_value_GOT_32): Likewise.
489 * config/i386/predicates.md (indirect_branch_operand): Likewise.
490 (GOT_memory_operand): Likewise.
491 (call_insn_operand): Likewise.
492 (sibcall_insn_operand): Likewise.
493 (GOT32_symbol_operand): Likewise.
494 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
496 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
498 PR rtl-optimization/83496
499 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
500 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
501 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
502 redundant insn, if any.
503 (relax_delay_slots): Likewise.
504 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
506 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
508 PR tree-optimization/83965
509 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
510 that grouped statements are part of a reduction chain. Return
511 true if the statement is not marked as a reduction itself but
513 (vect_recog_dot_prod_pattern): Don't check whether the statement
514 is part of a group here.
515 (vect_recog_sad_pattern): Likewise.
516 (vect_recog_widen_sum_pattern): Likewise.
518 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
521 * final.c (rest_of_clean_state): Also look for calls inside sequences.
523 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
526 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
528 (ix86_output_indirect_function_return): New prototype.
529 (ix86_split_simple_return_pop_internal): Likewise.
530 * config/i386/i386.c (indirect_return_via_cx): New.
531 (indirect_return_via_cx_bnd): Likewise.
532 (indirect_thunk_name): Handle return va CX_REG.
533 (output_indirect_thunk_function): Create alias for
534 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
535 (ix86_output_indirect_jmp): Remove the bool argument.
536 (ix86_output_indirect_function_return): New function.
537 (ix86_split_simple_return_pop_internal): Likewise.
538 * config/i386/i386.md (*indirect_jump): Don't pass false
539 to ix86_output_indirect_jmp.
540 (*tablejump_1): Likewise.
541 (simple_return_pop_internal): Change it to define_insn_and_split.
542 Call ix86_split_simple_return_pop_internal to split it for
544 (simple_return_indirect_internal): Call
545 ix86_output_indirect_function_return instead of
546 ix86_output_indirect_jmp.
548 2018-02-26 Jakub Jelinek <jakub@redhat.com>
551 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
552 memset and value initialization afterwards.
554 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
556 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
558 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
561 * common/config/aarch64/aarch64-common.c
562 (aarch_option_optimization_table[]): Switch
563 off fomit-frame-pointer
565 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
566 Chung-Ju Wu <jasonwucj@gmail.com>
568 * config/nds32/nds32-multiple.md (load_multiple): Disallow
570 (store_multiple): Ditto.
572 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
574 * config.gcc: Add --with-cpu support for nds32 target.
575 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
576 * config/nds32/nds32.opt: Add -mcpu= option.
578 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
580 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
581 isel=yes): Warn for these deprecated options.
583 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
585 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
586 ISA_2_5_MASKS_EMBEDDED.
588 2018-02-23 Jakub Jelinek <jakub@redhat.com>
590 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
591 p->max as pointers rather than using iterative_hash_expr.
593 2018-02-23 Carl Love <cel@us.ibm.com>
595 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
596 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
598 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
599 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
600 P8V_BUILTIN_VEC_VUNSIGNED2.
602 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
605 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
606 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
608 * lra-constraints.c (curr_insn_transform): Set up
609 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
611 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
612 * lra-eliminations.c (spill_pseudos): Ditto.
613 (process_insn_for_elimination): Ditto.
614 * lra-lives.c (reg_early_clobber_p): Use the new macros.
615 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
616 LRA_NON_CLOBBERED_ALT.
618 2018-02-22 Martin Sebor <msebor@redhat.com>
620 PR tree-optimization/84480
621 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
622 to maybe_diag_stxncpy_trunc. Call it.
623 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
624 from gimple_fold_builtin_strcpy. Print inlining stack.
625 (handle_builtin_stxncpy): Print inlining stack.
626 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
628 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
631 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
632 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
633 and -fcheck-pointer-bounds are used together.
634 (indirect_thunk_prefix): New enum.
635 (indirect_thunk_need_prefix): New function.
636 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
637 "_nt" instead of "_bnd" for NOTRACK prefix.
638 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
639 (output_indirect_thunk_function): Likewise.
641 (ix86_code_end): Update output_indirect_thunk_function calls.
642 (ix86_output_indirect_branch_via_reg): Replace
643 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
644 (ix86_output_indirect_branch_via_push): Likewise.
645 (ix86_output_function_return): Likewise.
646 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
647 incompatible with -fcf-protection=branch and
648 -fcheck-pointer-bounds.
650 2018-02-22 Steve Ellcey <sellcey@cavium.com>
653 * config/aarch64/aarch64.c (aarch64_print_address_internal):
654 Change gcc_assert call to output_operand_lossage.
656 2018-02-22 Steve Ellcey <sellcey@cavium.com>
658 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
660 2018-02-22 DJ Delorie <dj@redhat.com>
661 Sebastian Perta <sebastian.perta@renesas.com>
662 Oleg Endo <olegendo@gcc.gnu.org>
664 * config/rx/rx.c (rx_rtx_costs): New function.
665 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
667 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
669 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
671 2018-02-22 Martin Liska <mliska@suse.cz>
674 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
675 Add "native" as a possible value.
677 2018-02-22 Martin Liska <mliska@suse.cz>
680 * config/i386/i386.c (ix86_option_override_internal):
681 Add "native" as a possible value for -march and -mtune.
683 2018-02-22 Jakub Jelinek <jakub@redhat.com>
686 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
687 to all type variants.
689 PR tree-optimization/84503
690 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
691 width as info->bitpos + info->bitsize - start.
692 (merged_store_group::merge_overlapping): Simplify width computation.
693 (check_no_overlap): New function.
694 (imm_store_chain_info::try_coalesce_bswap): Compute expected
695 start + width and last_order of the group, fail if check_no_overlap
697 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
698 to group if check_no_overlap fails.
700 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
702 * config/rs6000/altivec.md: Delete contraint arguments to
703 define_expand, define_split, and define_peephole2, and in
704 define_insn_and_split if always unused.
705 * config/rs6000/darwin.md: Ditto.
706 * config/rs6000/dfp.md: Ditto.
707 * config/rs6000/rs6000.md: Ditto.
708 * config/rs6000/sync.md: Ditto.
709 * config/rs6000/vector.md: Ditto.
710 * config/rs6000/vsx.md: Ditto.
712 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
714 * config/rs6000/altivec.md: Write output control strings as braced
715 blocks instead of double-quoted strings.
716 * config/rs6000/darwin.md: Ditto.
717 * config/rs6000/rs6000.md: Ditto.
718 * config/rs6000/vector.md: Ditto.
719 * config/rs6000/vsx.md: Ditto.
721 2018-02-21 Jason Merrill <jason@redhat.com>
723 PR c++/84314 - ICE with templates and fastcall attribute.
724 * attribs.c (build_type_attribute_qual_variant): Remove assert.
726 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
728 * ipa-cp.c (determine_versionability): Fix comment typos.
730 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
733 * ipa-cp.c (determine_versionability): Do not version functions caling
736 2018-02-21 Martin Liska <mliska@suse.cz>
739 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
740 Add "native" as a possible value.
741 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
742 the macro when native cpu detection is available.
744 2018-02-21 Martin Liska <mliska@suse.cz>
747 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
748 Add "native" as a possible value.
749 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
750 when native cpu detection is available.
752 2018-02-21 Jakub Jelinek <jakub@redhat.com>
753 Martin Sebor <msebor@redhat.com>
755 PR tree-optimization/84478
756 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
758 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
759 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
760 support which is conservatively correct, for 2 only stay conservative
761 for maxlen. Formatting and comment capitalization fixes. Add STRICT
762 argument to the 2 argument get_range_strlen, adjust 6 arg
763 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
765 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
766 (gimple_fold_builtin_strlen): Pass true as last argument to
769 2018-02-20 Martin Sebor <msebor@redhat.com>
772 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
773 (builtin_memref::set_base_and_offset): Same. Handle inner references.
774 (builtin_memref::builtin_memref): Factor out parts into
775 set_base_and_offset and call it.
777 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
780 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
781 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
782 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
783 search at the associated MODE_INT.
785 2018-02-20 Jeff Law <law@redhat.com>
788 PR tree-optimization/81592
790 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
791 for range data rather than using global data.
792 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
793 range data rather than using global data.
794 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
795 pass it to children as needed.
796 (struct directive::fmtresult): Similarly.
797 (struct directive::set_width): Similarly.
798 (struct directive::set_precision): Similarly.
799 (format_integer, format_directive, parse_directive): Similarly.
800 (format_none): Accept unnamed vr_values parameter.
801 (format_percent, format_floating, format_character): Similarly.
802 (format_string, format_plain): Similarly.
803 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
804 the EVRP range analyzer for range data rather than using global data.
805 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
806 gimple-ssa-evrp-analyze.h
807 (class sprintf_dom_walker): Add after_dom_children member function.
808 Add evrp_range_analyzer member.
809 (sprintf_dom_walker::before_dom_children): Call into the EVRP
810 range analyzer as needed.
811 (sprintf_dom_walker::after_dom_children): New member function.
812 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
814 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
815 (evrp_range_analyzer::pop_to_marker): Likewise.
817 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
819 PR tree-optimization/84419
820 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
821 with the required type if its current type is compatible but
824 2018-02-20 Jakub Jelinek <jakub@redhat.com>
827 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
830 2018-02-20 Martin Liska <mliska@suse.cz>
833 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
834 possible values if we don't have a hint.
836 2018-02-20 Martin Liska <mliska@suse.cz>
840 * final.c (shorten_branches): Build align_tab array with one
842 * opts.c (finish_options): Add alignment option limit check.
843 (MAX_CODE_ALIGN): Likewise.
844 (MAX_CODE_ALIGN_VALUE): Likewise.
845 * doc/invoke.texi: Document maximum allowed option value for
846 all -falign-* options.
848 2018-02-19 Jakub Jelinek <jakub@redhat.com>
851 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
852 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
853 * var-tracking.c (emit_note_insn_var_location): Remove all references
854 to NOTE_INSN_CALL_ARG_LOCATION.
855 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
856 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
857 Use copy_rtx_if_shared.
858 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
859 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
860 (dwarf2out_var_location): Remove handling of
861 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
863 * final.c (final_scan_insn): Remove all references to
864 NOTE_INSN_CALL_ARG_LOCATION.
865 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
866 before dumping final insns.
867 * except.c (emit_note_eh_region_end): Remove all references to
868 NOTE_INSN_CALL_ARG_LOCATION.
869 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
870 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
871 * config/arc/arc.c (hwloop_optimize): Likewise.
872 * config/arm/arm.c (create_fix_barrier): Likewise.
873 * config/s390/s390.c (s390_chunkify_start): Likewise.
874 * config/sh/sh.c (find_barrier): Likewise.
875 * config/i386/i386.c (rest_of_insert_endbranch,
876 ix86_seh_fixup_eh_fallthru): Likewise.
877 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
878 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
879 * config/frv/frv.c (frv_function_prologue): Likewise.
880 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
882 (note_outside_basic_block_p): Remove all references to
883 NOTE_INSN_CALL_ARG_LOCATION.
884 * gengtype.c (adjust_field_rtx_def): Likewise.
885 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
887 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
888 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
891 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
894 PR tree-optimization/84452
895 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
896 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
899 2018-02-19 Martin Liska <mliska@suse.cz>
902 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
904 2018-02-19 Martin Liska <mliska@suse.cz>
905 Richard Sandiford <richard.sandiford@linaro.org>
907 PR tree-optimization/82491
908 * gimple-fold.c (get_base_constructor): Make earlier bail out
911 2018-02-19 Carl Love <cel@us.ibm.com>
913 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
914 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
916 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
919 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
921 * config/rl78/rl78.md (movdf): New define expand.
923 2018-02-19 Martin Liska <mliska@suse.cz>
926 * doc/invoke.texi: Fix typo.
927 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
929 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
931 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
932 handle rs6000_single_float and rs6000_double_float specially for
935 2018-02-16 Jeff Law <law@redhat.com>
937 * config/rx/rx.c (add_pop_cfi_notes): New function.;
940 2018-02-16 Jakub Jelinek <jakub@redhat.com>
943 * ipa-inline.c (inline_small_functions): Fix a typo.
945 2018-02-16 Nathan Sidwell <nathan@acm.org>
947 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
949 2018-02-16 Carl Love <cel@us.ibm.com>
951 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
952 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
953 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
954 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
955 expansion to P8V_BUILTIN_VEC_FLOAT2.
957 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
959 PR rtl-optimization/70023
960 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
961 src_regno into account.
963 2018-02-16 Carl Love <cel@us.ibm.com>
965 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
966 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
967 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
968 * config/rs6000/rs6000.c: Remove case statements for
969 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
970 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
971 and P9V_BUILTIN_VEC_VINSERT4B.
972 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
973 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
974 * config/rs6000/vsx.md:
975 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
978 2018-02-16 Carl Love <cel@us.ibm.com>
980 * config/rs6000/altivec.h: Add builtin names vec_extract4b
982 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
984 * config/rs6000/rs6000-c.c: Add the definitions for
985 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
986 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
987 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
988 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
989 definition for insert4b and define insn *insert3b_internal.
990 * doc/extend.texi: Add documentation for vec_extract4b.
992 2018-02-16 Nathan Sidwell <nathan@acm.org>
994 * doc/extend.texi (Backwards Compatibility): Mention friend
995 injection. Note for-scope is deprecated.
996 * doc/invoke.texi (-ffriend-injection): Deprecate.
998 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
1000 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
1001 that moved to I2, also allow destinations that are a paradoxical
1002 subreg (instead of a normal reg).
1004 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
1007 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
1010 2018-02-16 Richard Biener <rguenther@suse.de>
1012 PR tree-optimization/84037
1013 PR tree-optimization/84016
1015 * config/i386/i386.c (ix86_builtin_vectorization_cost):
1016 Adjust vec_construct for the fact we need additional higher latency
1017 128bit inserts for AVX256 and AVX512 vector builds.
1018 (ix86_add_stmt_cost): Scale vector construction cost for
1021 2018-02-16 Richard Biener <rguenther@suse.de>
1023 PR tree-optimization/84417
1024 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
1025 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
1026 (non_rewritable_lvalue_p): Likewise, use poly-ints.
1028 2018-02-16 Martin Liska <mliska@suse.cz>
1031 * internal-fn.def (ASAN_CHECK): Set proper flags.
1032 (ASAN_MARK): Likewise.
1034 2018-02-16 Julia Koval <julia.koval@intel.com>
1036 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
1037 from PTA_CANNONLAKE.
1039 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1042 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
1043 Use ++iter rather than iter++ for std::list iterators.
1044 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
1045 defer deleting them until all nodes in the forest are processed. Do
1046 free even leaf nodes. Change to_process into auto_vec.
1049 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
1050 * vec.h (vec_default_construct): Use memset instead of placement new
1051 if BROKEN_VALUE_INITIALIZATION is defined.
1052 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
1053 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
1056 PR rtl-optimization/83723
1057 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
1058 * lra.c (lra_substitute_pseudo): Likewise. If true, use
1059 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
1061 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
1063 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
1065 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
1067 PR rtl-optimization/81443
1068 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
1069 from inner REGs to paradoxical SUBREGs.
1071 2018-02-16 Richard Biener <rguenther@suse.de>
1073 PR tree-optimization/84399
1074 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1075 For operands we can analyze at their definition make sure we can
1076 analyze them at each use as well.
1078 2018-02-16 Richard Biener <rguenther@suse.de>
1080 PR tree-optimization/84190
1081 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
1082 volatile accesses if the decl isn't volatile.
1084 2018-02-15 Jason Merrill <jason@redhat.com>
1086 PR c++/84314 - ICE with templates and fastcall attribute.
1087 * attribs.c (build_type_attribute_qual_variant): Don't clobber
1088 TYPE_CANONICAL on an existing type.
1090 2018-02-15 Jakub Jelinek <jakub@redhat.com>
1092 PR tree-optimization/84383
1093 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
1094 dstoff nor call operand_equal_p if dstbase is NULL.
1096 PR tree-optimization/84334
1097 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
1098 also a CONSTANT_CLASS_P, punt.
1100 2018-02-14 Jim Wilson <jimw@sifive.com>
1102 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
1103 first SMALL_OPERAND check. New local min_second_step. Move assert
1104 to where locals are set. Add TARGET_RVC support.
1105 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
1107 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
1109 * doc/invoke.texi: Correct -Wformat-overflow code sample.
1111 2018-02-14 Martin Sebor <msebor@redhat.com>
1113 PR tree-optimization/83698
1114 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
1115 arrays constrain the offset range to their bounds.
1116 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
1117 (builtin_access::overlap): Avoid setting the size of overlap if it's
1119 (maybe_diag_overlap): Also consider arrays when deciding what values
1120 of offsets to include in diagnostics.
1122 2018-02-14 Martin Sebor <msebor@redhat.com>
1125 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
1126 that correspond to the kind of a declaration.
1128 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
1131 * config/pa/pa.md: Load address of PIC label using the linkage table
1132 if the label is nonlocal.
1134 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
1136 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
1137 warning message if user requests -maltivec=be.
1138 * doc/invoke.texi: Document deprecation of -maltivec=be.
1140 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
1143 * config/rs6000/rs6000-c.c: Update definitions for
1144 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
1145 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
1147 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1150 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
1151 add _get_ssp intrinsics. Remove argument from
1152 __builtin_ia32_rdssp[d|q].
1153 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
1154 * config/i386/i386-builtin.def: Remove argument from
1155 __builtin_ia32_rdssp[d|q].
1156 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
1157 ix86_expand_special_args_builtin for _rdssp[d|q].
1158 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
1159 Clear register before usage.
1160 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
1161 Add documentation for new _get_ssp and _inc_ssp intrinsics.
1163 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
1165 PR tree-optimization/84357
1166 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
1167 operand 1 of an ARRAY_REF too.
1169 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
1172 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
1173 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1175 (set_of_reg): New struct.
1176 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
1177 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
1178 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1180 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
1181 Split into bitclr, bitset, bitinvert patterns if appropriate.
1182 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
1183 use rx_fuse_in_memory_bitop.
1184 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
1185 to named insn, correct maximum insn length.
1187 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
1190 * machmode.def: Define a complex mode for PARTIAL_INT.
1191 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
1193 * doc/rtl.texi: Document CSPImode.
1194 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
1196 (msp430_hard_regno_nregs_with_padding): Likewise.
1198 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
1201 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
1203 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
1205 PR rtl-optimization/84169
1206 * combine.c (try_combine): New variable split_i2i3. Set it to true if
1207 we generated a parallel as new i3 and we split that to new i2 and i3
1208 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
1209 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
1210 those to i2, not i1. Partially rewrite this scan code.
1212 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1215 * stor-layout.c (place_field): For variable length fields, adjust
1216 offset_align afterwards not just based on the field's alignment,
1217 but also on the size.
1220 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
1221 of exps and logs in the use_exp2 case.
1223 2018-02-13 Jeff Law <law@redhat.com>
1225 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
1228 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
1231 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
1235 * common.opt (gas-loc-support, gas-locview-support): New.
1236 (ginline-points, ginternal-reset-location-views): New.
1237 * doc/invoke.texi: Document them. Use @itemx where intended.
1238 (gvariable-location-views): Adjust.
1239 * target.def (reset_location_view): New.
1240 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
1241 (TARGET_RESET_LOCATION_VIEW): New.
1242 * doc/tm.texi: Rebuilt.
1243 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
1244 (dwarf2out_default_as_locview_support): New.
1245 (output_asm_line_debug_info): Use option variables.
1246 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
1247 (output_loc_list): Likewise.
1248 (add_high_low_attributes): Check option variables.
1249 Don't output entry view attribute in strict mode.
1250 (gen_inlined_subroutine_die): Check option variables.
1251 (dwarf2out_inline_entry): Likewise.
1252 (init_sections_and_labels): Likewise.
1253 (dwarf2out_early_finish): Likewise.
1254 (maybe_reset_location_view): New, from...
1255 (dwarf2out_var_location): ... here. Call it.
1256 * debug.h (dwarf2out_default_as_loc_support): Declare.
1257 (dwarf2out_default_as_locview_support): Declare.
1258 * hooks.c (hook_int_rtx_insn_0): New.
1259 * hooks.h (hook_int_rtx_insn_0): Declare.
1260 * toplev.c (process_options): Take -gas-loc-support and
1261 -gas-locview-support from dwarf2out. Enable
1262 -gvariable-location-views by default only with locview
1263 assembler support. Enable -ginternal-reset-location-views by
1264 default only if the target defines the corresponding hook.
1265 Enable -ginline-points by default if location views are
1266 enabled; force it disabled if statement frontiers are
1268 * tree-inline.c (expand_call_inline): Check option variables.
1269 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1271 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
1273 PR tree-optimization/84321
1274 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
1275 handling. Also check whether the anti-range contains any values
1276 that satisfy the mask; switch to a VR_RANGE if not.
1278 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
1281 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
1283 2018-02-13 Martin Jambor <mjambor@suse.cz>
1286 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
1287 of call statements, also set location of a load to a temporary.
1289 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
1291 * config/rl78/rl78.c (add_vector_labels): New function.
1292 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
1293 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
1294 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
1295 which checks that no arguments are passed.
1296 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
1297 * doc/extend.texi: Documentation for the new attribute.
1299 2018-02-13 Andreas Schwab <schwab@suse.de>
1301 * config/riscv/linux.h (CPP_SPEC): Define.
1303 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1306 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
1307 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
1308 OPTION_MASK_ISA_AES as first argument to def_builtin_const
1309 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
1310 instead of OPTION_MASK_ISA_PCLMUL as first argument to
1311 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
1312 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
1313 temporarily for AES and PCLMUL builtins.
1315 PR tree-optimization/84339
1316 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
1317 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
1321 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
1322 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
1323 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
1325 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
1327 * omp-simd-clone.h: New file.
1328 * omp-simd-clone.c: Include omp-simd-clone.h.
1329 (expand_simd_clones): No longer static.
1330 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
1331 cgraph.h and omp-simd-clone.h.
1332 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
1333 (vect_recog_widen_shift_pattern): Formatting fix.
1334 (vect_pattern_recog_1): Don't check optab for calls.
1337 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
1338 operands[2] into a REG before using gen_lowpart on it.
1340 2018-02-12 Jeff Law <law@redhat.com>
1343 * config/sh/sh.c (find_barrier): Consider a sibling call
1346 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
1347 successfully back substituting a reg.
1349 2018-02-12 Richard Biener <rguenther@suse.de>
1351 PR tree-optimization/84037
1352 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
1353 parameter, move visited init to caller.
1354 (vect_slp_analyze_operations): Separate cost from validity
1355 check, initialize visited once for all instances.
1356 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
1358 * tree-vect-stmts.c (vect_model_simple_cost): Make early
1360 (vect_model_promotion_demotion_cost): Likewise.
1361 (vectorizable_bswap): Guard cost modeling with !slp_node
1362 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
1364 (vectorizable_call): Likewise.
1365 (vectorizable_conversion): Likewise.
1366 (vectorizable_assignment): Likewise.
1367 (vectorizable_shift): Likewise.
1368 (vectorizable_operation): Likewise.
1369 (vectorizable_store): Likewise.
1370 (vectorizable_load): Likewise.
1371 (vectorizable_condition): Likewise.
1372 (vectorizable_comparison): Likewise.
1374 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
1377 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
1378 (ASAN_MARK): Fix fnspec to account for return value, change pointer
1379 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
1381 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1384 * params.def (inline-min-speedup): Increase from 8 to 15.
1385 (max-inline-insns-auto): Decrease from 40 to 30.
1386 * ipa-split.c (consider_split): Add some buffer for function to
1387 be considered inlining candidate.
1388 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
1391 2018-02-12 Richard Biener <rguenther@suse.de>
1393 PR tree-optimization/84037
1394 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
1395 matched stmts if we cannot swap the non-matched ones.
1397 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1399 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
1400 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
1401 _mm_maskz_scalef_round_ss): New intrinsics.
1402 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
1403 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
1404 __builtin_ia32_scalefss_round): Remove.
1405 (__builtin_ia32_scalefsd_mask_round,
1406 __builtin_ia32_scalefss_mask_round): New intrinsics.
1407 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
1408 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
1409 ((match_operand:VF_128 2 "<round_nimm_predicate>"
1410 "<round_constraint>")): Changed to ...
1411 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
1412 "<round_scalar_constraint>")): ... this.
1413 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
1414 %0, %1, %2<round_op3>}"): Changed to ...
1415 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
1416 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
1417 %2<round_scalar_mask_op3>}"): ... this.
1418 * config/i386/subst.md (round_scalar_nimm_predicate): New.
1420 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1422 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
1423 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
1424 (_mm_maskz_sqrt_round_ss): New intrinsics.
1425 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
1426 (__builtin_ia32_sqrtsd_mask_round)
1427 (__builtin_ia32_sqrtss_mask_round): New builtins.
1428 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
1429 (__builtin_ia32_sqrtss_round): Remove.
1430 (__builtin_ia32_sqrtsd_mask_round)
1431 (__builtin_ia32_sqrtss_mask_round): New builtins.
1432 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
1433 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
1434 ((match_operand:VF_128 1 "vector_operand"
1435 "xBm,<round_constraint>")): Changed to ...
1436 ((match_operand:VF_128 1 "vector_operand"
1437 "xBm,<round_scalar_constraint>")): ... this.
1438 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
1439 %0, %2, %<iptr>1<round_op3>}): Changed to ...
1440 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
1441 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
1442 %<iptr>1<round_scalar_mask_op3>}): ... this.
1443 ((set_attr "prefix" "<round_prefix>")): Changed to ...
1444 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
1446 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
1449 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
1450 Cast vec_cmpeq result to correct type.
1451 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
1452 Cast vec_cmpgt result to correct type.
1454 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
1456 * final.c (final_scan_insn_1): Renamed from...
1457 (final_scan_insn): ... this. New wrapper, to recover
1458 seen from the outermost call in recursive ones.
1459 * config/sparc/sparc.c (output_return): Drop seen from call.
1460 (output_sibcall): Likewise.
1461 * config/visium/visium.c (output_branch): Likewise.
1463 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
1465 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
1468 2018-02-10 Alan Modra <amodra@gmail.com>
1471 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
1472 Specify LR as an input.
1474 2018-02-10 Jakub Jelinek <jakub@redhat.com>
1477 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
1478 remove_member_access_dummy_vars): New functions.
1479 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
1480 lower_omp_1, execute_lower_omp): Use them.
1482 PR rtl-optimization/84308
1483 * shrink-wrap.c (spread_components): Release todo vector.
1485 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
1487 PR rtl-optimization/57193
1488 * ira-color.c (struct allocno_color_data): Add member
1489 conflict_allocno_hard_prefs.
1490 (update_conflict_allocno_hard_prefs): New.
1491 (bucket_allocno_compare_func): Add a preference based on
1492 conflict_allocno_hard_prefs.
1493 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
1494 (color_allocnos): Remove a dead code. Initiate
1495 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
1497 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1500 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
1501 constraint from =wa to wa. Avoid a subreg on the output operand,
1502 instead use a pseudo and subreg it in a move.
1503 (p9_xxbrd_<mode>): Changed to ...
1504 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
1505 (p9_xxbrd_v2df): New expander.
1506 (p9_xxbrw_<mode>): Changed to ...
1507 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
1508 (p9_xxbrw_v4sf): New expander.
1510 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1512 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
1514 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
1517 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
1518 multiply in 32-bit mode.
1519 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
1520 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
1523 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1525 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
1526 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
1527 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
1528 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
1530 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
1533 * dwarf2out.c (is_trivial_indirect_ref): New function.
1534 (dwarf2out_late_global_decl): Do not generate a location
1535 attribute for variables that have a non-trivial DECL_VALUE_EXPR
1536 and that are not defined in the current unit.
1538 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1540 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
1541 instead of a libcall for UNORDERED.
1543 2018-02-09 Tamar Christina <tamar.christina@arm.com>
1546 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
1547 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
1549 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1552 * config/s390/s390.c (s390_set_current_function): Invoke
1553 s390_indirect_branch_settings also if fndecl didn't change.
1555 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1557 * config/rs6000/rs6000.md (blockage): Set length to zero.
1559 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1561 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
1563 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1566 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
1567 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
1571 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
1572 PARALLEL incoming that failed vt_get_decl_and_offset check.
1575 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
1576 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
1578 (get_variable_section): For decls in named .bss* sections pass true as
1579 second argument to bss_initializer_p.
1581 2018-02-09 Marek Polacek <polacek@redhat.com>
1582 Jakub Jelinek <jakub@redhat.com>
1585 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
1586 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
1587 Sync some changes from cxx_fold_indirect_ref.
1589 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1591 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
1593 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
1594 (BLOCK_INLINE_ENTRY_LABEL): New.
1595 (dwarf2out_var_location): Disregard inline entry markers.
1596 (inline_entry_data): New struct.
1597 (inline_entry_data_hasher): New hashtable type.
1598 (inline_entry_data_hasher::hash): New.
1599 (inline_entry_data_hasher::equal): New.
1600 (inline_entry_data_table): New variable.
1601 (add_high_low_attributes): Add DW_AT_entry_pc and
1602 DW_AT_GNU_entry_view attributes if a pending entry is found
1603 in inline_entry_data_table. Add old entry_pc attribute only
1604 if debug nonbinding markers are disabled.
1605 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
1606 markers are enabled.
1607 (block_within_block_p, dwarf2out_inline_entry): New.
1608 (dwarf2out_finish): Check that no entries remained in
1609 inline_entry_data_table.
1610 * final.c (reemit_insn_block_notes): Handle inline entry notes.
1611 (final_scan_insn, notice_source_line): Likewise.
1612 (rest_of_clean_state): Skip inline entry markers.
1613 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
1615 * gimple.c (gimple_build_debug_inline_entry): New.
1616 * gimple.h (enum gimple_debug_subcode): Add
1617 GIMPLE_DEBUG_INLINE_ENTRY.
1618 (gimple_build_debug_inline_entry): Declare.
1619 (gimple_debug_inline_entry_p): New.
1620 (gimple_debug_nonbind_marker_p): Adjust.
1621 * insn-notes.def (INLINE_ENTRY): New.
1622 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
1623 inline entry marker notes.
1624 (print_insn): Likewise.
1625 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
1626 (INSN_DEBUG_MARKER_KIND): Likewise.
1627 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
1628 * tree-inline.c (expand_call_inline): Build and insert
1629 debug_inline_entry stmt.
1630 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
1631 inline entry blocks early, if nonbind markers are enabled.
1632 (dump_scope_block): Dump fragment info.
1633 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
1634 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
1635 (gimple_build_debug_inline_entry): New.
1636 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
1637 Enable/disable inline entry points too.
1638 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
1639 (DEBUG_INSN): Describe inline entry markers.
1641 * common.opt (gvariable-location-views): New.
1642 (gvariable-location-views=incompat5): New.
1643 * config.in: Rebuilt.
1644 * configure: Rebuilt.
1645 * configure.ac: Test assembler for view support.
1646 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
1647 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
1648 * dwarf2out.c (var_loc_view): New typedef.
1649 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
1650 (dwarf2out_locviews_in_attribute): New.
1651 (dwarf2out_locviews_in_loclist): New.
1652 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
1653 (enum dw_line_info_opcode): Add LI_adv_address.
1654 (struct dw_line_info_table): Add view.
1655 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
1656 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
1657 (zero_view_p): New variable.
1658 (ZERO_VIEW_P): New macro.
1659 (output_asm_line_debug_info): New.
1660 (struct var_loc_node): Add view.
1661 (add_AT_view_list, AT_loc_list): New.
1662 (add_var_loc_to_decl): Add view param. Test it against last.
1663 (new_loc_list): Add view params. Record them.
1664 (AT_loc_list_ptr): Handle loc and view lists.
1665 (view_list_to_loc_list_val_node): New.
1666 (print_dw_val): Handle dw_val_class_view_list.
1667 (size_of_die): Likewise.
1668 (value_format): Likewise.
1669 (loc_list_has_views): New.
1670 (gen_llsym): Set vl_symbol too.
1671 (maybe_gen_llsym, skip_loc_list_entry): New.
1672 (dwarf2out_maybe_output_loclist_view_pair): New.
1673 (output_loc_list): Output view list or entries too.
1674 (output_view_list_offset): New.
1675 (output_die): Handle dw_val_class_view_list.
1676 (output_dwarf_version): New.
1677 (output_compilation_unit_header): Use it.
1678 (output_skeleton_debug_sections): Likewise.
1679 (output_rnglists, output_line_info): Likewise.
1680 (output_pubnames, output_aranges): Update version comments.
1681 (output_one_line_info_table): Output view numbers in asm comments.
1682 (dw_loc_list): Determine current endview, pass it to new_loc_list.
1683 Call maybe_gen_llsym.
1684 (loc_list_from_tree_1): Adjust.
1685 (add_AT_location_description): Create view list attribute if
1686 needed, check it's absent otherwise.
1687 (convert_cfa_to_fb_loc_list): Adjust.
1688 (maybe_emit_file): Call output_asm_line_debug_info for test.
1689 (dwarf2out_var_location): Reset views as needed. Precompute
1690 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
1691 attribute. Set view.
1692 (new_line_info_table): Reset next view.
1693 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
1694 (dwarf2out_source_line): Likewise. Output view resets and labels to
1695 the assembler, or select appropriate line info opcodes.
1696 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
1697 (optimize_string_length): Catch it. Adjust.
1698 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
1699 dw_val_class_view_list, and remove it if no longer needed.
1700 (hash_loc_list): Hash view numbers.
1701 (loc_list_hasher::equal): Compare them.
1702 (optimize_location_lists): Check whether a view list symbol is
1703 needed, and whether the locview attribute is present, and
1704 whether they match. Remove the locview attribute if no longer
1706 (index_location_lists): Call skip_loc_list_entry for test.
1707 (dwarf2out_finish): Call output_asm_line_debug_info for test.
1708 Use output_dwarf_version.
1709 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
1710 (struct dw_val_node): Add val_view_list.
1711 * final.c (SEEN_NEXT_VIEW): New.
1712 (set_next_view_needed): New.
1713 (clear_next_view_needed): New.
1714 (maybe_output_next_view): New.
1715 (final_start_function): Rename to...
1716 (final_start_function_1): ... this. Take pointer to FIRST,
1717 add SEEN parameter. Emit param bindings in the initial view.
1718 (final_start_function): Reintroduce SEEN-less interface.
1719 (final): Rename to...
1720 (final_1): ... this. Take SEEN parameter. Output final pending
1721 next view at the end.
1722 (final): Reintroduce seen-less interface.
1723 (final_scan_insn): Output pending next view before switching
1724 sections or ending a block. Mark the next view as needed when
1725 outputting variable locations. Notify debug backend of section
1726 changes, and of location view changes.
1727 (rest_of_handle_final): Adjust.
1728 * toplev.c (process_options): Autodetect value for debug variable
1729 location views option. Warn on incompat5 without -gdwarf-5.
1730 * doc/invoke.texi (gvariable-location-views): New.
1731 (gvariable-location-views=incompat5): New.
1732 (gno-variable-location-views): New.
1734 2018-02-08 David Malcolm <dmalcolm@redhat.com>
1736 PR tree-optimization/84136
1737 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
1738 that the result of find_edge is non-NULL.
1740 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1743 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
1744 storing integer register in SImode. Fix cost of 256 and 512
1745 byte aligned SSE register store.
1747 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1749 * config/i386/i386.c (ix86_multiplication_cost): Fix
1750 multiplication cost for TARGET_AVX512DQ.
1752 2018-02-08 Marek Polacek <polacek@redhat.com>
1754 PR tree-optimization/84238
1755 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
1758 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1760 PR tree-optimization/84265
1761 * tree-vect-stmts.c (vectorizable_store): Don't treat
1762 VMAT_CONTIGUOUS accesses as grouped.
1763 (vectorizable_load): Likewise.
1765 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1767 PR tree-optimization/81635
1768 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
1769 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
1770 (test_round_for_mask): New functions.
1771 (wide_int_cc_tests): Call test_round_for_mask.
1772 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
1773 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
1774 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
1775 range returned by get_range_info.
1777 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1780 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
1781 * symtab.c: Include builtins.h
1782 (symtab_node::output_to_lto_symbol_table_p): Move here
1783 from lto-streamer-out.c:output_symbol_p.
1784 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
1785 (output_symbol_p): Move all logic to symtab.c
1786 (produce_symtab): Update.
1788 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1790 * config/s390/s390-opts.h (enum indirect_branch): Define.
1791 * config/s390/s390-protos.h (s390_return_addr_from_memory)
1792 (s390_indirect_branch_via_thunk)
1793 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
1794 (enum s390_indirect_branch_type): Define.
1795 * config/s390/s390.c (struct s390_frame_layout, struct
1796 machine_function): Remove.
1797 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
1798 (indirect_branch_table_label_no, indirect_branch_table_name):
1800 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
1801 (enum s390_indirect_branch_option): Define.
1802 (s390_return_addr_from_memory): New function.
1803 (s390_handle_string_attribute): New function.
1804 (s390_attribute_table): Add new attribute handler.
1805 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
1806 (s390_indirect_branch_via_thunk): New function.
1807 (s390_indirect_branch_via_inline_thunk): New function.
1808 (s390_function_ok_for_sibcall): When jumping via thunk disallow
1809 sibling call optimization for non z10 compiles.
1810 (s390_emit_call): Force indirect branch target to be a single
1811 register. Add r1 clobber for non-z10 compiles.
1812 (s390_emit_epilogue): Emit return jump via return_use expander.
1813 (s390_reorg): Handle JUMP_INSNs as execute targets.
1814 (s390_option_override_internal): Perform validity checks for the
1815 new command line options.
1816 (s390_indirect_branch_attrvalue): New function.
1817 (s390_indirect_branch_settings): New function.
1818 (s390_set_current_function): Invoke s390_indirect_branch_settings.
1819 (s390_output_indirect_thunk_function): New function.
1820 (s390_code_end): Implement target hook.
1821 (s390_case_values_threshold): Implement target hook.
1822 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
1824 * config/s390/s390.h (struct s390_frame_layout)
1825 (struct machine_function): Move here from s390.c.
1826 (TARGET_INDIRECT_BRANCH_NOBP_RET)
1827 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
1828 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
1829 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
1830 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
1831 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
1832 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
1833 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
1834 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
1835 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
1836 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
1837 (mnemonic attribute): Add values which aren't recognized
1839 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
1840 pattern for branch conversion. Fix mnemonic attribute.
1841 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
1842 indirect branch via thunk if requested.
1843 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
1844 ("*indirect_jump"): Disable for branch conversion using out of
1846 ("indirect_jump_via_thunk<mode>_z10")
1847 ("indirect_jump_via_thunk<mode>")
1848 ("indirect_jump_via_inlinethunk<mode>_z10")
1849 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
1850 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
1851 ("casesi_jump_via_inlinethunk<mode>_z10")
1852 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
1853 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
1854 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
1855 ("*indirect2_jump"): Disable for branch conversion.
1856 ("casesi_jump"): Turn into expander and expand patterns for branch
1858 ("return_use"): New expander.
1859 ("*return"): Emit return via thunk and rename it to ...
1860 ("*return<mode>"): ... this one.
1861 * config/s390/s390.opt: Add new options and and enum for the
1864 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1866 * lra-constraints.c (match_reload): Unconditionally use
1867 gen_lowpart_SUBREG, rather than selecting between that
1868 and equivalent gen_rtx_SUBREG code.
1870 2018-02-08 Richard Biener <rguenther@suse.de>
1872 PR tree-optimization/84233
1873 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
1874 changed flag instead of boguously re-using phi_inserted.
1876 2018-02-08 Martin Jambor <mjambor@suse.cz>
1878 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
1879 static local variables.
1881 2018-02-08 Richard Biener <rguenther@suse.de>
1883 PR tree-optimization/84278
1884 * tree-vect-stmts.c (vectorizable_store): When looking for
1885 smaller vector types to perform grouped strided loads/stores
1886 make sure the mode is supported by the target.
1887 (vectorizable_load): Likewise.
1889 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1891 * config/aarch64/aarch64.c (aarch64_components_for_bb):
1892 Increase LDP/STP opportunities by adding adjacent callee-saves.
1894 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1896 PR rtl-optimization/84068
1897 PR rtl-optimization/83459
1898 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
1900 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
1902 PR tree-optimization/84224
1903 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
1904 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
1907 2018-02-07 Iain Sandoe <iain@codesourcery.com>
1910 * config/rs6000/altivec.md (*restore_world): Remove LR use.
1911 * config/rs6000/predicates.md (restore_world_operation): Adjust op
1912 count, remove one USE.
1914 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
1916 * doc/install.texi (Configuration): Document the
1917 --with-long-double-format={ibm,ieee} PowerPC configuration
1921 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
1922 Convert from define_expand to be define_insn_and_split. Rework
1923 float/double/_Float128 conversions to QI/HI/SImode to work with
1924 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
1925 conversions to QI/HImode types did a store and then a load to
1926 truncate the value. For conversions to VSX registers, don't split
1927 the insn, instead emit the code directly. Use the code iterator
1928 any_fix to combine signed and unsigned conversions.
1929 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
1930 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
1931 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
1932 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
1933 (fix_<mode>di2_hw): Likewise.
1934 (fixuns_<mode>di2_hw): Likewise.
1935 (fix_<mode>si2_hw): Likewise.
1936 (fixuns_<mode>si2_hw): Likewise.
1937 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
1938 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
1939 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
1940 fix<uns>_trunc<SFDF:mode>si2_p8.
1941 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
1943 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
1944 (fix<uns>_<mode>_mem): Likewise.
1945 (fctiw<u>z_<mode>_mem): Likewise.
1946 (fix<uns>_<mode>_mem): Likewise.
1947 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
1948 the register allocator from doing a direct move to the GPRs to do
1949 a store, and instead use the ISA 3.0 store byte/half-word from
1950 vector register instruction. For IEEE 128-bit floating point,
1951 also optimize stores of 32-bit ints.
1952 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
1954 2018-02-07 Alan Hayward <alan.hayward@arm.com>
1956 * genextract.c (push_pathstr_operand): New function to support
1958 (walk_rtx): Call push_pathstr_operand.
1959 (print_path): Support [a-zA-Z].
1961 2018-02-07 Richard Biener <rguenther@suse.de>
1963 PR tree-optimization/84037
1964 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
1965 (cse_and_gimplify_to_preheader): Declare.
1966 (vect_get_place_in_interleaving_chain): Likewise.
1967 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
1969 (_loop_vec_info::~_loop_vec_info): Delete it.
1970 (cse_and_gimplify_to_preheader): New function.
1971 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
1972 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
1973 (vectorizable_load): Likewise. For grouped stores always base
1974 the IV on the first element.
1975 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
1976 condition before gimplifying.
1978 2018-02-07 Jakub Jelinek <jakub@redhat.com>
1980 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
1981 *DIV_EXPR and *MOD_EXPR.
1983 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
1986 * config/i386/i386.c (ix86_option_override_internal): Mask out
1987 the CF_SET bit when checking -fcf-protection.
1989 2018-02-07 Tom de Vries <tom@codesourcery.com>
1992 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
1995 2018-02-07 Richard Biener <rguenther@suse.de>
1997 PR tree-optimization/84204
1998 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2001 PR tree-optimization/84205
2002 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
2003 special-case isl_ast_op_zdiv_r.
2005 PR tree-optimization/84223
2006 * graphite-scop-detection.c (gather_bbs::before_dom_children):
2007 Only add conditions from within the region.
2008 (gather_bbs::after_dom_children): Adjust.
2010 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
2013 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
2014 * config/avr/avr.md: Only post-reload split REG-REG moves if
2015 either register is REGERAL_REG_P.
2017 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2019 PR tree-optimization/84235
2020 * tree-ssa-scopedtables.c
2021 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
2022 if the subtraction is performed in floating point type where NaNs are
2023 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
2024 build 1. Formatting fix.
2026 2018-02-06 Jakub Jelinek <jakub@redhat.com>
2029 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
2030 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
2031 and skip it regardless of bb boundaries. Use CALL_P macro,
2032 don't test INSN_P (insn) together with CALL_P or JUMP_P check
2033 unnecessarily, formatting fix.
2035 2018-02-06 Michael Collison <michael.collison@arm.com>
2037 * config/arm/thumb2.md:
2038 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
2039 (*thumb_mov_notscc): Ditto.
2041 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2044 * config/rs6000/rs6000.md (su code attribute): Use "u" for
2045 unsigned_fix, not "s".
2047 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2049 * configure.ac (gcc_fn_eh_frame_ro): New function.
2050 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
2051 correct .eh_frame permissions.
2052 * configure: Regenerate.
2054 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
2056 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
2059 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2061 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2062 Display warning message for -mno-speculate-indirect-jumps.
2064 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
2066 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
2068 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
2070 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
2072 PR tree-optimization/84225
2073 * tree-eh.c (find_trapping_overflow): Only call
2074 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
2076 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2079 * config/i386/i386.c: Reimplement the check of possible options
2080 -mibt/-mshstk conbination. Change error messages.
2081 * doc/invoke.texi: Fix a typo: remove extra '='.
2083 2018-02-06 Marek Polacek <polacek@redhat.com>
2085 PR tree-optimization/84228
2086 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
2088 2018-02-06 Tamar Christina <tamar.christina@arm.com>
2091 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
2092 emitted arch directives.
2093 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
2094 __ARM_FEATURE_COPROC before changing architectures.
2096 2018-02-06 Richard Biener <rguenther@suse.de>
2098 * config/i386/i386.c (print_reg): Fix typo.
2099 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
2101 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
2103 * configure: Regenerate.
2105 2018-02-05 Martin Sebor <msebor@redhat.com>
2107 PR tree-optimization/83369
2108 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
2111 2018-02-05 Martin Liska <mliska@suse.cz>
2113 * doc/invoke.texi: Cherry-pick upstream r323995.
2115 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
2117 * ira.c (ira_init_register_move_cost): Adjust comment.
2119 2018-02-05 Martin Liska <mliska@suse.cz>
2121 PR gcov-profile/84137
2122 * doc/gcov.texi: Fix typo in documentation.
2124 2018-02-05 Martin Liska <mliska@suse.cz>
2126 PR gcov-profile/83879
2127 * doc/gcov.texi: Document necessity of --dynamic-list-data when
2128 using dlopen functionality.
2130 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
2132 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
2133 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
2134 _mm_maskz_range_ss, _mm_mask_range_round_ss,
2135 _mm_maskz_range_round_ss): New intrinsics.
2136 (__builtin_ia32_rangesd128_round)
2137 (__builtin_ia32_rangess128_round): Remove.
2138 (__builtin_ia32_rangesd128_mask_round,
2139 __builtin_ia32_rangess128_mask_round): New builtins.
2140 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
2141 __builtin_ia32_rangess128_round): Remove.
2142 (__builtin_ia32_rangesd128_mask_round,
2143 __builtin_ia32_rangess128_mask_round): New builtins.
2144 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
2145 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
2146 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
2147 "<round_saeonly_constraint>")): Changed to ...
2148 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
2149 "<round_saeonly_scalar_constraint>")): ... this.
2150 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
2151 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
2152 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
2153 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
2154 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
2156 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
2158 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
2160 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
2161 Remove all values except native, 8540 and 8548.
2163 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
2165 * config/i386/i386.c (ix86_output_function_return): Pass
2166 INVALID_REGNUM, instead of -1, as invalid register number to
2167 indirect_thunk_name and output_indirect_thunk.
2169 2018-02-02 Julia Koval <julia.koval@intel.com>
2171 * config.gcc: Add -march=icelake.
2172 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
2173 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
2174 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
2175 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
2176 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
2177 (processor_target_table): Add icelake.
2178 (ix86_option_override_internal): Handle new PTAs.
2179 (get_builtin_code_for_version): Handle icelake.
2180 (M_INTEL_COREI7_ICELAKE): New.
2181 (fold_builtin_cpu): Handle icelake.
2182 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
2183 * doc/invoke.texi: Add -march=icelake.
2185 2018-02-02 Julia Koval <julia.koval@intel.com>
2187 * config/i386/i386.c (ix86_option_override_internal): Change flags type
2188 to wide_int_bitmask.
2189 * wide-int-bitmask.h: New.
2191 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2194 * config/i386/i386.md: Replace Pmode with word_mode in
2195 builtin_setjmp_setup and builtin_longjmp to support x32.
2197 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
2201 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
2203 (rs6000_supported_cpu_names): New static variable.
2204 (linux_cpu_translation_table): Likewise.
2205 (elf_platform) <cpu>: Define new static variable and use it.
2206 Translate kernel AT_PLATFORM name to canonical name if needed.
2207 Error if platform name is unknown.
2209 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
2212 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
2214 2018-02-01 Jeff Law <law@redhat.com>
2217 * config/i386/i386.c (release_scratch_register_on_entry): Add new
2218 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
2219 the scratch if RELEASE_VIA_POP is false.
2220 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
2221 If we have to save a temporary register, decrement SIZE appropriately.
2222 Pass new arguments to release_scratch_register_on_entry.
2223 (ix86_adjust_stack_and_probe): Likewise.
2224 (ix86_emit_probe_stack_range): Pass new arguments to
2225 release_scratch_register_on_entry.
2227 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
2229 PR rtl-optimization/84157
2230 * combine.c (change_zero_ext): Use REG_P predicate in
2231 front of HARD_REGISTER_P predicate.
2233 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
2235 * config/avr/avr.c (avr_option_override): Move disabling of
2236 -fdelete-null-pointer-checks to...
2237 * common/config/avr/avr-common.c (avr_option_optimization_table):
2240 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2242 PR tree-optimization/81635
2243 * tree-data-ref.c (split_constant_offset_1): For types that
2244 wrap on overflow, try to use range info to prove that wrapping
2247 2018-02-01 Renlin Li <renlin.li@arm.com>
2250 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
2252 (aarch64_register_move_cost): Likewise.
2253 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
2255 (REG_CLASS_NAMES): Likewise.
2256 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
2257 TAILCALL_ADDR_REGS. Remove IP registers.
2258 * config/aarch64/aarch64.md (Ucs): Update register constraint.
2260 2018-02-01 Richard Biener <rguenther@suse.de>
2262 * domwalk.h (dom_walker::dom_walker): Add additional constructor
2263 for specifying RPO order and allow NULL for that.
2264 * domwalk.c (dom_walker::dom_walker): Likewise.
2265 (dom_walker::walk): Handle NULL RPO order.
2266 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
2268 (rewrite_update_dom_walker): Likewise.
2269 (mark_def_dom_walker): Likewise.
2271 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2273 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
2274 (aarch64_maybe_expand_sve_subreg_move): Declare.
2275 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
2276 * config/aarch64/predicates.md (aarch64_any_register_operand): New
2278 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
2279 that are semantically a reverse operation.
2280 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
2281 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
2282 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
2284 (aarch64_can_change_mode_class): For big-endian, forbid changes
2285 between two SVE modes if they have different element sizes.
2287 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2289 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
2290 the TImode handling for big-endian targets.
2292 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2294 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
2295 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
2297 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
2298 Remove BSWAP handing for big-endian targets and use the form of
2299 LD1RQ appropariate for the mode.
2301 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2303 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
2304 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
2307 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2310 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
2311 check for operands that need to go through aarch64_sve_reload_be.
2313 2018-02-01 Jakub Jelinek <jakub@redhat.com>
2315 PR tree-optimization/81661
2316 PR tree-optimization/84117
2317 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
2318 * tree-eh.c: Include gimplify.h.
2319 (find_trapping_overflow, replace_trapping_overflow,
2320 rewrite_to_non_trapping_overflow): New functions.
2321 * tree-vect-loop.c: Include tree-eh.h.
2322 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
2323 * tree-data-ref.c: Include tree-eh.h.
2324 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
2326 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
2328 PR rtl-optimization/84123
2329 * combine.c (change_zero_ext): Check if hard register satisfies
2330 can_change_dest_mode before calling gen_lowpart_SUBREG.
2332 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
2335 * ira.c (ira_init_register_move_cost): Remove assert.
2337 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2339 PR rtl-optimization/84071
2340 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
2341 * doc/tm.texi: Regenerate.
2343 2018-01-31 Richard Biener <rguenther@suse.de>
2345 PR tree-optimization/84132
2346 * tree-data-ref.c (analyze_miv_subscript): Properly
2347 check whether evolution_function_is_affine_multivariate_p
2348 before calling gcd_of_steps_may_divide_p.
2350 2018-01-31 Julia Koval <julia.koval@intel.com>
2353 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
2354 * config/i386/i386.md (rdpid_rex64) New.
2355 (rdpid): Make 32bit only.
2357 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
2360 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
2361 an IDENTIFIER_NODE for FUNCTION_TYPE's.
2363 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2366 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
2368 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
2370 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2372 PR rtl-optimization/84071
2373 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
2374 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
2376 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2378 * config/arc/arc.c (arc_handle_aux_attribute): New function.
2379 (arc_attribute_table): Add 'aux' attribute.
2380 (arc_in_small_data_p): Consider aux like variables.
2381 (arc_is_aux_reg_p): New function.
2382 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
2383 (arc_get_aux_arg): New function.
2384 (prepare_move_operands): Handle aux-register access.
2385 (arc_handle_aux_attribute): New function.
2386 * doc/extend.texi (ARC Variable attributes): Add subsection.
2388 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2390 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
2391 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
2392 (arc_attribute_table): Add 'uncached' attribute.
2393 (arc_print_operand): Print '.di' flag for uncached memory
2395 (arc_in_small_data_p): Do not consider for small data the uncached
2397 (arc_is_uncached_mem_p): New function.
2398 * config/arc/predicates.md (compact_store_memory_operand): Check
2399 for uncached memory accesses.
2400 (nonvol_nonimm_operand): Likewise.
2401 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
2403 2018-01-31 Jakub Jelinek <jakub@redhat.com>
2406 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
2407 falign-loops=): Add Optimization flag.
2409 2018-01-30 Jeff Law <law@redhat.com>
2412 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
2413 INT_REGISTERS_SAVED. Check it prior to calling
2414 get_scratch_register_on_entry.
2415 (ix86_adjust_stack_and_probe): Similarly.
2416 (ix86_emit_probe_stack_range): Similarly.
2417 (ix86_expand_prologue): Corresponding changes.
2419 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2422 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
2423 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
2425 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
2428 * lra-constraints.c (curr_insn_transform): Process AND in the
2431 2018-01-30 Jakub Jelinek <jakub@redhat.com>
2433 PR rtl-optimization/83986
2434 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
2435 dependence against last_pending_memory_flush in addition to
2438 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
2440 PR tree-optimization/81611
2441 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
2444 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2447 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
2450 2018-01-30 Richard Biener <rguenther@suse.de>
2451 Jakub Jelinek <jakub@redhat.com>
2453 PR tree-optimization/84111
2454 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
2455 inner loops added during recursion, as they don't have up-to-date
2458 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2461 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
2462 (can_inline_edge_by_limits_p): ... here.
2463 (can_early_inline_edge_p, check_callers,
2464 update_caller_keys, update_callee_keys, recursive_inlining,
2465 add_new_edges_to_heap, speculation_useful_p,
2466 inline_small_functions,
2467 inline_small_functions, flatten_function,
2468 inline_to_all_callers_1): Update.
2470 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2472 * profile-count.c (profile_count::combine_with_ipa_count): Handle
2475 2018-01-30 Richard Biener <rguenther@suse.de>
2477 PR tree-optimization/83008
2478 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
2479 invariant and constant vector uses in stmts when they need
2482 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2485 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
2486 * configure: Regenerate.
2488 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2490 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
2492 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
2493 Use gen_rtx_REG rather than gen_lowpart.
2495 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2497 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
2498 rather than 0 when creating partial subregs.
2500 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2502 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
2505 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
2508 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
2509 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
2510 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
2511 flags. This restores the settings used before the 2017-07-24.
2512 Turning off pre increment/decrement/modify allows IVOPTS to
2513 optimize DF/SF loops where the index is an int.
2515 2018-01-29 Richard Biener <rguenther@suse.de>
2516 Kelvin Nilsen <kelvin@gcc.gnu.org>
2519 * tree-vect-stmts.c (vectorizable_call): Don't call
2520 targetm.vectorize_builtin_md_vectorized_function if callee is
2523 2018-01-22 Carl Love <cel@us.ibm.com>
2525 * doc/extend.tex: Fix typo in second arg in
2526 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
2528 2018-01-29 Richard Biener <rguenther@suse.de>
2530 PR tree-optimization/84086
2531 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
2532 (flush_ssaname_freelist): When SSA names were released reset
2533 the SCEV hash table.
2535 2018-01-29 Richard Biener <rguenther@suse.de>
2537 PR tree-optimization/84057
2538 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
2539 removed paths when removing edges.
2541 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
2543 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
2544 -mfunction-return=@var{choice}.
2546 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
2549 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
2551 (layout::print_source_line): Likewise.
2552 (test_get_line_width_without_trailing_whitespace): Add test cases.
2554 2018-01-27 Jakub Jelinek <jakub@redhat.com>
2557 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
2560 2018-01-26 Jim Wilson <jimw@sifive.com>
2562 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
2564 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
2567 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2569 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
2570 and CMP + SUB-immediate -> SUBS.
2572 2018-01-26 Martin Sebor <msebor@redhat.com>
2574 PR tree-optimization/83896
2575 * tree-ssa-strlen.c (get_string_len): Rename...
2576 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
2577 Avoid assuming length is constant.
2578 (handle_char_store): Use HOST_WIDE_INT for string length.
2580 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
2583 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
2584 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
2586 2018-01-26 Richard Biener <rguenther@suse.de>
2588 PR rtl-optimization/84003
2589 * dse.c (record_store): Only record redundant stores when
2590 the earlier store aliases at least all accesses the later one does.
2592 2018-01-26 Jakub Jelinek <jakub@redhat.com>
2594 PR rtl-optimization/83985
2595 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
2596 REG_CFA_RESTORE insns.
2597 (delete_unmarked_insns): Don't ignore separate shrink wrapping
2598 REG_CFA_RESTORE insns here.
2601 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
2602 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
2604 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2606 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
2607 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
2608 (arc_init): Likewise.
2609 (arc_override_options): Likewise.
2610 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
2612 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
2614 * config/arc/arc.h (TARGET_DBNZ): Define.
2615 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
2616 properly set the tune attribute.
2617 (dbnz): Use TARGET_DBNZ guard.
2618 * config/arc/arc.opt (mtune): Add core3 option.
2620 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2622 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
2623 recognize new pic like addresses.
2624 (arc_delegitimize_address): Clean up.
2626 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2628 * config/arc/arc-arches.def: Option mrf16 valid for all
2630 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
2631 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
2632 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
2633 * config/arc/arc-tables.opt: Regenerate.
2634 * config/arc/arc.c (arc_conditional_register_usage): Handle
2635 reduced register file case.
2636 (arc_file_start): Set must have build attributes.
2637 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
2639 * config/arc/arc.opt (mrf16): Add new option.
2640 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
2641 * config/arc/genmultilib.awk: Handle new mrf16 option.
2642 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
2643 * config/arc/t-multilib: Regenerate.
2644 * doc/invoke.texi (ARC Options): Document mrf16 option.
2646 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2648 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
2649 * config/arc/arc.c (arc_handle_secure_attribute): New function.
2650 (arc_attribute_table): Add 'secure_call' attribute.
2651 (arc_print_operand): Print secure call operand.
2652 (arc_function_ok_for_sibcall): Don't optimize tail calls when
2654 (arc_is_secure_call_p): New function. * config/arc/arc.md
2655 (call_i): Add support for sjli instruction.
2656 (call_value_i): Likewise.
2657 * config/arc/constraints.md (Csc): New constraint.
2659 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2660 John Eric Martin <John.Martin@emmicro-us.com>
2662 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
2663 * config/arc/arc.c (_arc_jli_section): New struct.
2664 (arc_jli_section): New type.
2665 (rc_jli_sections): New static variable.
2666 (arc_handle_jli_attribute): New function.
2667 (arc_attribute_table): Add jli_always and jli_fixed attribute.
2668 (arc_file_end): New function.
2669 (TARGET_ASM_FILE_END): Define.
2670 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
2671 (arc_add_jli_section): New function.
2672 (jli_call_scan): Likewise.
2673 (arc_reorg): Call jli_call_scan.
2674 (arc_output_addsi): Remove 'S' from printing asm operand.
2675 (arc_is_jli_call_p): New function.
2676 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
2678 (movhi_insn): Likewise.
2679 (movsi_insn): Likewise.
2680 (movsi_set_cc_insn): Likewise.
2681 (loadqi_update): Likewise.
2682 (load_zeroextendqisi_update): Likewise.
2683 (load_signextendqisi_update): Likewise.
2684 (loadhi_update): Likewise.
2685 (load_zeroextendhisi_update): Likewise.
2686 (load_signextendhisi_update): Likewise.
2687 (loadsi_update): Likewise.
2688 (loadsf_update): Likewise.
2689 (movsicc_insn): Likewise.
2690 (bset_insn): Likewise.
2691 (bxor_insn): Likewise.
2692 (bclr_insn): Likewise.
2693 (bmsk_insn): Likewise.
2694 (bicsi3_insn): Likewise.
2695 (cmpsi_cc_c_insn): Likewise.
2696 (movsi_ne): Likewise.
2697 (movsi_cond_exec): Likewise.
2698 (clrsbsi2): Likewise.
2709 (call_i): Remove 'S' asm letter, add jli instruction.
2710 (call_value_i): Likewise.
2711 * config/arc/arc.op (mjli-always): New option.
2712 * config/arc/constraints.md (Cji): New constraint.
2713 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
2715 (subsf3_fpx): Likewise.
2716 (mulsf3_fpx): Likewise.
2717 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
2719 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
2721 * doc/invoke.texi (ARC): Document mjli-always option.
2723 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
2725 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
2726 avoid addition with 0 and use incw and decw where possible.
2728 2018-01-26 Richard Biener <rguenther@suse.de>
2730 PR tree-optimization/81082
2731 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
2732 association if it requires casting to unsigned.
2733 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
2734 from fold_plusminus_mult_expr to catch important cases late when
2735 range info is available.
2737 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2739 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
2740 * configure.ac (hidden_linkonce): New test.
2741 * configure: Regenerate.
2742 * config.in: Regenerate.
2744 2018-01-26 Julia Koval <julia.koval@intel.com>
2746 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
2747 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
2748 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
2749 _mm_mask_bitshuffle_epi64_mask): Fix type.
2750 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
2751 USI_FTYPE_V4DI_V4DI_USI): Remove.
2752 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
2753 __builtin_ia32_vpshufbitqmb256_mask,
2754 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
2755 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
2756 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
2758 2018-01-26 Alan Modra <amodra@gmail.com>
2761 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
2762 UNSPEC_VBPERMQ. Sort other unspecs.
2764 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
2766 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
2768 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
2771 * predict.c (drop_profile): Do not push/pop cfun; update also
2773 (handle_missing_profiles): Fix logic looking for zero profiles.
2775 2018-01-25 Jakub Jelinek <jakub@redhat.com>
2778 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
2779 on functions with #pragma omp declare simd or functions with simd
2781 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
2782 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
2783 Remove trailing \n from warning_at calls.
2785 2018-01-25 Tom de Vries <tom@codesourcery.com>
2788 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2789 for neutered workers.
2791 2018-01-24 Joseph Myers <joseph@codesourcery.com>
2794 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
2795 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
2797 2018-01-24 Jeff Law <law@redhat.com>
2800 * i386.c (get_probe_interval): Move to earlier point.
2801 (ix86_compute_frame_layout): If -fstack-clash-protection and
2802 the frame is larger than the probe interval, then use pushes
2803 to save registers rather than reg->mem moves.
2804 (ix86_expand_prologue): Remove conditional for int_registers_saved
2807 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
2810 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
2811 min/max for never referenced object.
2813 2018-01-24 Jakub Jelinek <jakub@redhat.com>
2816 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
2818 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
2819 attributes from DECL_ATTRIBUTES (decl) without affecting
2820 DECL_ATTRIBUTES (current_function_decl).
2821 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
2822 functions with non-NULL DECL_ABSTRACT_ORIGIN.
2824 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
2826 PR tree-optimization/83979
2827 * fold-const.c (fold_comparison): Use constant_boolean_node
2828 instead of boolean_{true,false}_node.
2830 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
2832 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
2835 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2837 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
2838 Simplify the clause that sets the length attribute.
2839 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
2840 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
2841 clause that sets the length attribute.
2842 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
2844 2018-01-24 Tom de Vries <tom@codesourcery.com>
2847 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
2848 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
2849 Add strict parameter.
2850 (prevent_branch_around_nothing): Insert dummy insn between branch to
2851 label and label with no ptx insn inbetween.
2852 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
2854 2018-01-24 Tom de Vries <tom@codesourcery.com>
2857 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2858 for neutered threads in warp.
2859 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
2861 2018-01-24 Richard Biener <rguenther@suse.de>
2863 PR tree-optimization/83176
2864 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
2867 2018-01-24 Richard Biener <rguenther@suse.de>
2869 PR tree-optimization/82819
2870 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
2871 code generating pluses that are no-ops in the target precision.
2873 2018-01-24 Richard Biener <rguenther@suse.de>
2876 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
2878 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2880 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
2881 to merge probabilities.
2882 * predict.c (probably_never_executed): Also mark as cold functions
2883 with global 0 profile and guessed local profile.
2884 * profile-count.c (profile_probability::combine_with_count): New
2886 * profile-count.h (profile_probability::operator*,
2887 profile_probability::operator*=, profile_probability::operator/,
2888 profile_probability::operator/=): Reduce precision to adjusted
2889 and set value to guessed on contradictory divisions.
2890 (profile_probability::combine_with_freq): Remove.
2891 (profile_probability::combine_wiht_count): Declare.
2892 (profile_count::force_nonzero):: Set to adjusted.
2893 (profile_count::probability_in):: Set quality to adjusted.
2894 * tree-ssa-tail-merge.c (replace_block_by): Use
2897 2018-01-23 Andrew Waterman <andrew@sifive.com>
2898 Jim Wilson <jimw@sifive.com>
2900 * config/riscv/riscv.c (riscv_stack_boundary): New.
2901 (riscv_option_override): Set riscv_stack_boundary. Handle
2902 riscv_preferred_stack_boundary_arg.
2903 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
2904 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
2905 (STACK_BOUNDARY): Set to riscv_stack_boundary.
2906 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
2907 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
2908 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
2910 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
2913 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
2914 of struct ix86_frame.
2915 (ix86_expand_epilogue): Likewise. Add a local variable for
2916 the reg_save_offset field in struct ix86_frame.
2918 2018-01-23 Bin Cheng <bin.cheng@arm.com>
2920 PR tree-optimization/82604
2921 * tree-loop-distribution.c (enum partition_kind): New enum item
2922 PKIND_PARTIAL_MEMSET.
2923 (partition_builtin_p): Support above new enum item.
2924 (generate_code_for_partition): Ditto.
2925 (compute_access_range): Differentiate cases that equality can be
2926 proven at all loops, the innermost loops or no loops.
2927 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
2928 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
2929 (finalize_partitions, distribute_loop): Don't fuse partition of
2930 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
2931 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
2934 2018-01-23 Martin Liska <mliska@suse.cz>
2936 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
2937 order to ignore the predictor.
2938 (PRED_POLYMORPHIC_CALL): Likewise.
2939 (PRED_RECURSIVE_CALL): Likewise.
2941 2018-01-23 Martin Liska <mliska@suse.cz>
2943 * tree-profile.c (tree_profiling): Print function header to
2944 aware reader which function we are working on.
2945 * value-prof.c (gimple_find_values_to_profile): Do not print
2946 not interesting value histograms.
2948 2018-01-23 Martin Liska <mliska@suse.cz>
2950 * profile-count.h (enum profile_quality): Add
2951 profile_uninitialized as the first value. Do not number values
2952 as they are zero based.
2953 (profile_count::verify): Update sanity check.
2954 (profile_probability::verify): Likewise.
2956 2018-01-23 Nathan Sidwell <nathan@acm.org>
2958 * doc/invoke.texi (ffor-scope): Deprecate.
2960 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2962 PR tree-optimization/83510
2963 * domwalk.c (set_all_edges_as_executable): New function.
2964 (dom_walker::dom_walker): Convert bool param
2965 "skip_unreachable_blocks" to enum reachability. Move setup of
2966 edge flags to set_all_edges_as_executable and only do it when
2967 reachability is REACHABLE_BLOCKS.
2968 * domwalk.h (enum dom_walker::reachability): New enum.
2969 (dom_walker::dom_walker): Convert bool param
2970 "skip_unreachable_blocks" to enum reachability.
2971 (set_all_edges_as_executable): New decl.
2972 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
2973 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
2975 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
2976 but converting true to REACHABLE_BLOCKS.
2977 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
2979 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
2980 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
2981 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
2983 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
2984 if check_all_array_refs will be called.
2986 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2988 * tree.c (selftest::test_location_wrappers): Add more test
2991 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2993 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
2994 (selftest::test_bit_in_range): Likewise.
2996 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
2999 * doc/sourcebuild.texi (vect_float): Say that the selector
3000 only describes the situation when -funsafe-math-optimizations is on.
3001 (vect_float_strict): Document.
3003 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3005 PR tree-optimization/83965
3006 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3007 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
3008 instead of checking only for a reduction.
3009 (vect_recog_widen_sum_pattern): Likewise.
3011 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3013 * predict.c (probably_never_executed): Only use precise profile info.
3014 (compute_function_frequency): Skip after inlining hack since we now
3015 have quality checking.
3017 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3019 * profile-count.h (profile_probability::very_unlikely,
3020 profile_probability::unlikely, profile_probability::even): Set
3021 precision to guessed.
3023 2018-01-23 Richard Biener <rguenther@suse.de>
3025 PR tree-optimization/83963
3026 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
3027 Properly terminate dominator walk when crossing the exit edge not
3028 when visiting its source block.
3030 2018-01-23 Jakub Jelinek <jakub@redhat.com>
3033 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
3034 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
3036 2018-01-22 Jakub Jelinek <jakub@redhat.com>
3038 PR tree-optimization/83957
3039 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
3040 semicolon after for body surrounded by braces.
3042 PR tree-optimization/83081
3043 * profile-count.h (profile_probability::split): New method.
3044 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
3045 Use profile_probability::split.
3046 (do_compare_rtx_and_jump): Fix adjustment of probabilities
3047 when splitting a single conditional jump into 2.
3049 2018-01-22 David Malcolm <dmalcolm@redhat.com>
3051 PR tree-optimization/69452
3052 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
3055 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3057 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
3058 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
3059 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
3061 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3063 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
3064 * config/rl78/rl78.md (movdi): New define_expand.
3065 * config/rl78/rl78.c (rl78_split_movdi): New function.
3067 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
3070 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
3072 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
3073 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
3074 128-bit to produce an UNSPEC move to get the double word with the
3075 signbit and then a shift directly to do signbit.
3076 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
3077 implementation with a new version that just does either a direct
3078 move or a regular move. Move memory interface to separate insns.
3079 Move insns so they are next to the expander.
3080 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
3081 with signbit move. Split big and little endian case.
3082 (signbit<mode>2_dm_mem_le): Likewise.
3083 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
3084 (signbit<mode>2_dm2): Likewise.
3086 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3088 * config/rl78/rl78.md (anddi3): New define_expand.
3090 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3092 * config/rl78/rl78.md (umindi3): New define_expand.
3094 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3096 * config/rl78/rl78.md (smindi3): New define_expand.
3098 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3100 * config/rl78/rl78.md (smaxdi3): New define_expand.
3102 2018-01-22 Carl Love <cel@us.ibm.com>
3104 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
3105 LVX_V1TI): Add macro expansion.
3106 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
3107 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
3108 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
3109 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
3110 Change check to determine if the instruction is a byte reversing
3111 entry. Fix typo in comment.
3112 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
3113 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
3114 Add def_builtin calls for new builtins.
3115 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
3116 Add define_insn expansion.
3118 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3120 * config/rl78/rl78.md (umaxdi3): New define_expand.
3122 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3124 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
3125 for non-QImode registers.
3127 2018-01-22 Richard Biener <rguenther@suse.de>
3129 PR tree-optimization/83963
3130 * graphite-scop-detection.c (scop_detection::get_sese): Delay
3131 including the loop exit block.
3132 (scop_detection::merge_sese): Likewise.
3133 (scop_detection::add_scop): Do it here instead.
3135 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3137 * doc/sourcebuild.texi (arm_softfloat): Document.
3139 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
3142 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
3143 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
3144 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
3146 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3147 David Edelsohn <dje.gcc@gmail.com>
3150 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3151 Change "crset eq" to "crset 2".
3152 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3153 (*call_indirect_aix<mode>_nospec): Likewise.
3154 (*call_value_indirect_aix<mode>_nospec): Likewise.
3155 (*call_indirect_elfv2<mode>_nospec): Likewise.
3156 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
3157 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
3158 change assembly output from . to $.
3159 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3160 (indirect_jump<mode>_nospec): Change assembly output from . to $.
3161 (*tablejump<mode>_internal1_nospec): Likewise.
3163 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
3166 * config/sh/sh_optimize_sett_clrt.cc:
3167 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
3169 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
3171 PR tree-optimization/83940
3172 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
3173 offset_dt to vect_constant_def rather than vect_unknown_def_type.
3174 (vect_check_load_store_mask): Add a mask_dt_out parameter and
3175 use it to pass back the definition type.
3176 (vect_check_store_rhs): Likewise rhs_dt_out.
3177 (vect_build_gather_load_calls): Add a mask_dt argument and use
3178 it instead of a call to vect_is_simple_use.
3179 (vectorizable_store): Update calls to vect_check_load_store_mask
3180 and vect_check_store_rhs. Use the dt returned by the latter instead
3181 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
3182 instead of calls to vect_is_simple_use. Pass the scalar rather
3183 than the vector operand to vect_is_simple_use when handling
3184 second and subsequent copies of an rhs value.
3185 (vectorizable_load): Update calls to vect_check_load_store_mask
3186 and vect_build_gather_load_calls. Use the cached mask_dt and
3187 gs_info.offset_dt instead of calls to vect_is_simple_use.
3189 2018-01-20 Jakub Jelinek <jakub@redhat.com>
3192 * tree-emutls.c: Include gimplify.h.
3193 (lower_emutls_2): New function.
3194 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
3195 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
3196 it before further processing.
3199 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
3200 UINTVAL (trueop1) instead of INTVAL (op1).
3202 2018-01-19 Jakub Jelinek <jakub@redhat.com>
3206 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
3207 INCOMING_FRAME_SP_OFFSET if not defined.
3208 (scan_trace): Add ENTRY argument. If true and
3209 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
3210 emit a note to adjust the CFA offset.
3211 (create_cfi_notes): Adjust scan_trace callers.
3212 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
3213 INCOMING_FRAME_SP_OFFSET in the CIE.
3214 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
3215 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
3217 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
3218 * doc/tm.texi: Regenerated.
3220 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3222 PR rtl-optimization/83147
3223 * lra-constraints.c (remove_inheritance_pseudos): Use
3224 lra_substitute_pseudo_within_insn.
3226 2018-01-19 Tom de Vries <tom@codesourcery.com>
3227 Cesar Philippidis <cesar@codesourcery.com>
3230 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
3232 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
3235 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
3236 spaces for function labels.
3238 2018-01-19 Martin Liska <mliska@suse.cz>
3240 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
3241 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
3242 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
3243 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
3244 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
3245 (PRED_CONST_RETURN): Change from 69 to 65.
3246 (PRED_NULL_RETURN): Change from 91 to 71.
3247 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
3248 (PRED_LOOP_GUARD): Change from 66 to 73.
3250 2018-01-19 Martin Liska <mliska@suse.cz>
3252 * predict.c (predict_insn_def): Add new assert.
3253 (struct branch_predictor): Change type to signed integer.
3254 (test_prediction_value_range): Amend test to cover
3256 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
3257 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
3258 (PRED_LOOP_ITERATIONS_MAX): Likewise.
3259 (PRED_LOOP_IV_COMPARE): Likewise.
3260 * predict.h (PROB_UNINITIALIZED): Define new constant.
3262 2018-01-19 Martin Liska <mliska@suse.cz>
3264 * predict.c (dump_prediction): Add new format for
3265 analyze_brprob.py script which is enabled with -details
3267 * profile-count.h (precise_p): New function.
3269 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3271 PR tree-optimization/83922
3272 * tree-vect-loop.c (vect_verify_full_masking): Return false if
3273 there are no statements that need masking.
3274 (vect_active_double_reduction_p): New function.
3275 (vect_analyze_loop_operations): Use it when handling phis that
3276 are not in the loop header.
3278 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3280 PR tree-optimization/83914
3281 * tree-vect-loop.c (vectorizable_induction): Don't convert
3282 init_expr or apply the peeling adjustment for inductions
3283 that are nested within the vectorized loop.
3285 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3287 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
3290 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3294 * function.h (gimplify_parameters): Add gimple_seq * argument.
3295 * function.c: Include gimple.h and options.h.
3296 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
3297 for the added local temporaries if needed.
3298 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
3299 if there are any parameter cleanups, wrap whole body into a
3300 try/finally with the cleanups.
3302 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
3305 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
3306 Use GET_MODE_CLASS for scalar floating point.
3308 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3312 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3313 Fix call of call_cgraph_insertion_hooks.
3315 2018-01-18 Martin Sebor <msebor@redhat.com>
3317 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
3319 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3322 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
3325 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
3328 * common.opt: (-ffile-prefix-map): New option.
3329 * opts.c (common_handle_option): Defer it.
3330 * opts-global.c (handle_common_deferred_options): Handle it.
3331 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
3332 * file-prefix-map.h: New file.
3333 (remap_debug_filename, add_debug_prefix_map): ...here.
3334 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
3335 * final.c (debug_prefix_map, add_debug_prefix_map
3336 remap_debug_filename): Move to...
3337 * file-prefix-map.c: New file.
3338 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
3339 generalize, get rid of alloca(), use strrchr() instead of strchr().
3340 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
3341 Implement in terms of add_prefix_map().
3342 (remap_macro_filename, remap_debug_filename): Implement in term of
3344 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
3345 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
3346 * dbxout.c: Include file-prefix-map.h.
3347 * varasm.c: Likewise.
3348 * vmsdbgout.c: Likewise.
3349 * xcoffout.c: Likewise.
3350 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
3351 * doc/cppopts.texi (-fmacro-prefix-map): Document.
3352 * doc/invoke.texi (-ffile-prefix-map): Document.
3353 (-fdebug-prefix-map): Update description.
3355 2018-01-18 Martin Liska <mliska@suse.cz>
3357 * config/i386/i386.c (indirect_thunk_name): Document that also
3359 (output_indirect_thunk): Document why both instructions
3360 (pause and lfence) are generated.
3362 2018-01-18 Richard Biener <rguenther@suse.de>
3364 PR tree-optimization/83887
3365 * graphite-scop-detection.c
3366 (scop_detection::get_nearest_dom_with_single_entry): Remove.
3367 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
3368 (scop_detection::merge_sese): Re-implement with a flood-fill
3369 algorithm that properly finds a SESE region if it exists.
3371 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3374 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
3375 pointer_diff optimizations use view_convert instead of convert.
3377 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3379 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3380 Generate different code for -mno-speculate-indirect-jumps.
3381 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3382 (*call_indirect_aix<mode>): Disable for
3383 -mno-speculate-indirect-jumps.
3384 (*call_indirect_aix<mode>_nospec): New define_insn.
3385 (*call_value_indirect_aix<mode>): Disable for
3386 -mno-speculate-indirect-jumps.
3387 (*call_value_indirect_aix<mode>_nospec): New define_insn.
3388 (*sibcall_nonlocal_sysv<mode>): Generate different code for
3389 -mno-speculate-indirect-jumps.
3390 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3392 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3394 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
3395 long double type, set the flags for noting the default long double
3396 type, even if we don't pass or return a long double type.
3398 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
3401 * ipa-inline.c (flatten_function): Do not overwrite final inlining
3404 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
3406 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
3407 support for merge[hl].
3408 (fold_mergehl_helper): New helper function.
3409 (tree-vector-builder.h): New #include for tree_vector_builder usage.
3410 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
3411 (altivec_vmrglw_direct): Add xxmrglw insn.
3413 2018-01-17 Andrew Waterman <andrew@sifive.com>
3415 * config/riscv/riscv.c (riscv_conditional_register_usage): If
3416 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
3418 2018-01-17 David Malcolm <dmalcolm@redhat.com>
3421 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
3422 call the lto_location_cache before reading the
3423 DECL_SOURCE_LOCATION of the types.
3425 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3426 Richard Sandiford <richard.sandiford@linaro.org>
3428 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
3429 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
3430 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
3431 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
3432 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
3434 * config/aarch64/constraints.md (aarch64_movti_operand):
3436 * config/aarch64/predicates.md (Uti): Add new constraint.
3438 2018-01-17 Carl Love <cel@us.ibm.com>
3440 * config/rs6000/vsx.md (define_expand xl_len_r,
3441 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
3442 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
3444 (define_expand, define_insn): Move the shift left from the
3445 define_insn to the define_expand for lxvl and stxvl instructions.
3446 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
3447 and XL_LEN_R definitions to PURE.
3449 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
3451 * config/i386/i386.c (indirect_thunk_name): Declare regno
3452 as unsigned int. Compare regno with INVALID_REGNUM.
3453 (output_indirect_thunk): Ditto.
3454 (output_indirect_thunk_function): Ditto.
3455 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
3456 in the call to output_indirect_thunk_function.
3458 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
3461 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
3462 rather than the size of inner_type to determine the stack slot size
3463 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
3465 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
3468 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
3471 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
3473 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
3474 endian Linux systems to optionally enable multilibs for selecting
3475 the long double type if the user configured an explicit type.
3476 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
3477 have no long double multilibs if not defined.
3478 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
3479 warn if the user used -mabi={ieee,ibm}longdouble and we built
3480 multilibs for long double.
3481 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
3482 appropriate multilib option.
3483 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
3485 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
3486 for building long double multilibs.
3487 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
3489 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
3491 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
3494 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
3496 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
3499 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
3502 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
3505 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3507 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
3508 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
3510 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3512 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
3513 different rtl trees depending on TARGET_64BIT.
3514 (rs6000_gen_lvx): Likewise.
3516 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3518 * config/visium/visium.md (nop): Tweak comment.
3519 (hazard_nop): Likewise.
3521 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3523 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
3524 -mspeculate-indirect-jumps.
3525 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
3526 for -mno-speculate-indirect-jumps.
3527 (*call_indirect_elfv2<mode>_nospec): New define_insn.
3528 (*call_value_indirect_elfv2<mode>): Disable for
3529 -mno-speculate-indirect-jumps.
3530 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
3531 (indirect_jump): Emit different RTL for
3532 -mno-speculate-indirect-jumps.
3533 (*indirect_jump<mode>): Disable for
3534 -mno-speculate-indirect-jumps.
3535 (*indirect_jump<mode>_nospec): New define_insn.
3536 (tablejump): Emit different RTL for
3537 -mno-speculate-indirect-jumps.
3538 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
3539 (tablejumpsi_nospec): New define_expand.
3540 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
3541 (tablejumpdi_nospec): New define_expand.
3542 (*tablejump<mode>_internal1): Disable for
3543 -mno-speculate-indirect-jumps.
3544 (*tablejump<mode>_internal1_nospec): New define_insn.
3545 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
3548 2018-01-16 Artyom Skrobov tyomitch@gmail.com
3550 * caller-save.c (insert_save): Drop unnecessary parameter. All
3553 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3554 Richard Biener <rguenth@suse.de>
3557 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
3558 return early, inline manually is_gimple_sizepos. Make sure if we
3559 call gimplify_expr we don't end up with a gimple constant.
3560 * tree.c (variably_modified_type_p): Don't return true for
3561 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
3562 * gimplify.h (is_gimple_sizepos): Remove.
3564 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3566 PR tree-optimization/83857
3567 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
3568 vectorizable_live_operation for pure SLP statements.
3569 (vectorizable_live_operation): Handle PHIs.
3571 2018-01-16 Richard Biener <rguenther@suse.de>
3573 PR tree-optimization/83867
3574 * tree-vect-stmts.c (vect_transform_stmt): Precompute
3575 nested_in_vect_loop_p since the scalar stmt may get invalidated.
3577 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3580 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
3581 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
3582 If off is not INTEGER_CST, issue a may not be aligned warning
3583 rather than isn't aligned. Use isn%'t rather than isn't.
3584 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
3586 <case MULT_EXPR>: Improve the case when bottom and one of the
3587 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
3588 operand, in that case check if the other operand is multiple of
3589 bottom divided by the INTEGER_CST operand.
3591 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3594 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
3595 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
3596 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
3597 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
3598 * config/pa/pa.c (pa_function_arg_advance): Likewise.
3599 (pa_function_arg, pa_arg_partial_bytes): Likewise.
3600 (pa_function_arg_size): New function.
3602 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3604 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
3605 in a separate statement.
3607 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3609 PR tree-optimization/83847
3610 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
3611 group gathers and scatters.
3613 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3615 PR rtl-optimization/86620
3616 * params.def (max-sched-ready-insns): Bump minimum value to 1.
3618 PR rtl-optimization/83213
3619 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
3620 to last if both are JUMP_INSNs.
3622 PR tree-optimization/83843
3623 * gimple-ssa-store-merging.c
3624 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
3625 store_immediate_info for bswap/nop orig_stores.
3627 2018-01-15 Andrew Waterman <andrew@sifive.com>
3629 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
3631 <UDIV>: Increase cost if !TARGET_DIV.
3633 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
3635 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
3636 (define_attr "cr_logical_3op"): New.
3637 (cceq_ior_compare): Adjust.
3638 (cceq_ior_compare_complement): Adjust.
3639 (*cceq_rev_compare): Adjust.
3640 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
3641 (is_cracked_insn): Adjust.
3642 (insn_must_be_first_in_group): Adjust.
3643 * config/rs6000/40x.md: Adjust.
3644 * config/rs6000/440.md: Adjust.
3645 * config/rs6000/476.md: Adjust.
3646 * config/rs6000/601.md: Adjust.
3647 * config/rs6000/603.md: Adjust.
3648 * config/rs6000/6xx.md: Adjust.
3649 * config/rs6000/7450.md: Adjust.
3650 * config/rs6000/7xx.md: Adjust.
3651 * config/rs6000/8540.md: Adjust.
3652 * config/rs6000/cell.md: Adjust.
3653 * config/rs6000/e300c2c3.md: Adjust.
3654 * config/rs6000/e500mc.md: Adjust.
3655 * config/rs6000/e500mc64.md: Adjust.
3656 * config/rs6000/e5500.md: Adjust.
3657 * config/rs6000/e6500.md: Adjust.
3658 * config/rs6000/mpc.md: Adjust.
3659 * config/rs6000/power4.md: Adjust.
3660 * config/rs6000/power5.md: Adjust.
3661 * config/rs6000/power6.md: Adjust.
3662 * config/rs6000/power7.md: Adjust.
3663 * config/rs6000/power8.md: Adjust.
3664 * config/rs6000/power9.md: Adjust.
3665 * config/rs6000/rs64.md: Adjust.
3666 * config/rs6000/titan.md: Adjust.
3668 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3670 * config/i386/predicates.md (indirect_branch_operand): Rewrite
3671 ix86_indirect_branch_register logic.
3673 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3675 * config/i386/constraints.md (Bs): Update
3676 ix86_indirect_branch_register check. Don't check
3677 ix86_indirect_branch_register with GOT_memory_operand.
3679 * config/i386/predicates.md (GOT_memory_operand): Don't check
3680 ix86_indirect_branch_register here.
3681 (GOT32_symbol_operand): Likewise.
3683 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3685 * config/i386/predicates.md (constant_call_address_operand):
3686 Rewrite ix86_indirect_branch_register logic.
3687 (sibcall_insn_operand): Likewise.
3689 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3691 * config/i386/constraints.md (Bs): Replace
3692 ix86_indirect_branch_thunk_register with
3693 ix86_indirect_branch_register.
3695 * config/i386/i386.md (indirect_jump): Likewise.
3696 (tablejump): Likewise.
3697 (*sibcall_memory): Likewise.
3698 (*sibcall_value_memory): Likewise.
3699 Peepholes of indirect call and jump via memory: Likewise.
3700 * config/i386/i386.opt: Likewise.
3701 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3702 (GOT_memory_operand): Likewise.
3703 (call_insn_operand): Likewise.
3704 (sibcall_insn_operand): Likewise.
3705 (GOT32_symbol_operand): Likewise.
3707 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3710 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
3711 type rather than type addr's type points to.
3712 (expand_omp_atomic_mutex): Likewise.
3713 (expand_omp_atomic): Likewise.
3715 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3718 * config/i386/i386.c (output_indirect_thunk_function): Use
3719 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
3720 for __x86_return_thunk.
3722 2018-01-15 Richard Biener <rguenther@suse.de>
3725 * expmed.c (extract_bit_field_1): Fix typo.
3727 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3730 * config/arm/iterators.md (VF): New mode iterator.
3731 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
3732 Remove integer-related logic from pattern.
3733 (neon_vabd<mode>_3): Likewise.
3735 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3738 * common.opt (fstrict-overflow): No longer an alias.
3739 (fwrapv-pointer): New option.
3740 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
3741 also for pointer types based on flag_wrapv_pointer.
3742 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
3743 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
3744 opts->x_flag_wrapv got set.
3745 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
3746 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
3747 POINTER_TYPE_OVERFLOW_UNDEFINED.
3748 * match.pd: Likewise in address comparison pattern.
3749 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
3751 2018-01-15 Richard Biener <rguenther@suse.de>
3754 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
3755 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
3756 Reset type names to their identifier if their TYPE_DECL doesn't
3757 have linkage (and thus is used for ODR and devirt).
3758 (save_debug_info_for_decl): Remove.
3759 (save_debug_info_for_type): Likewise.
3760 (add_tree_to_fld_list): Adjust.
3761 * tree-pretty-print.c (dump_generic_node): Make dumping of
3762 type names more robust.
3764 2018-01-15 Richard Biener <rguenther@suse.de>
3766 * BASE-VER: Bump to 8.0.1.
3768 2018-01-14 Martin Sebor <msebor@redhat.com>
3771 * builtins.c (check_access): Avoid warning when the no-warning bit
3774 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
3776 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
3777 * ira-color (allocno_hard_regs_compare): Likewise.
3779 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
3782 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
3783 Use .pushsection/.popsection.
3785 2018-01-14 Martin Sebor <msebor@redhat.com>
3788 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
3790 2018-01-14 Jakub Jelinek <jakub@redhat.com>
3792 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
3793 entry from extra_headers.
3794 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
3795 extra_headers, make the list bitwise identical to the i?86-*-* one.
3797 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3799 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
3800 -mcmodel=large with -mindirect-branch=thunk,
3801 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
3802 -mfunction-return=thunk-extern.
3803 * doc/invoke.texi: Document -mcmodel=large is incompatible with
3804 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
3805 -mfunction-return=thunk and -mfunction-return=thunk-extern.
3807 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3809 * config/i386/i386.c (print_reg): Print the name of the full
3810 integer register without '%'.
3811 (ix86_print_operand): Handle 'V'.
3812 * doc/extend.texi: Document 'V' modifier.
3814 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3816 * config/i386/constraints.md (Bs): Disallow memory operand for
3817 -mindirect-branch-register.
3819 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3820 (GOT_memory_operand): Likewise.
3821 (call_insn_operand): Likewise.
3822 (sibcall_insn_operand): Likewise.
3823 (GOT32_symbol_operand): Likewise.
3824 * config/i386/i386.md (indirect_jump): Call convert_memory_address
3825 for -mindirect-branch-register.
3826 (tablejump): Likewise.
3827 (*sibcall_memory): Likewise.
3828 (*sibcall_value_memory): Likewise.
3829 Disallow peepholes of indirect call and jump via memory for
3830 -mindirect-branch-register.
3831 (*call_pop): Replace m with Bw.
3832 (*call_value_pop): Likewise.
3833 (*sibcall_pop_memory): Replace m with Bs.
3834 * config/i386/i386.opt (mindirect-branch-register): New option.
3835 * doc/invoke.texi: Document -mindirect-branch-register option.
3837 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3839 * config/i386/i386-protos.h (ix86_output_function_return): New.
3840 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
3841 set function_return_type.
3842 (indirect_thunk_name): Add ret_p to indicate thunk for function
3844 (output_indirect_thunk_function): Pass false to
3845 indirect_thunk_name.
3846 (ix86_output_indirect_branch_via_reg): Likewise.
3847 (ix86_output_indirect_branch_via_push): Likewise.
3848 (output_indirect_thunk_function): Create alias for function
3849 return thunk if regno < 0.
3850 (ix86_output_function_return): New function.
3851 (ix86_handle_fndecl_attribute): Handle function_return.
3852 (ix86_attribute_table): Add function_return.
3853 * config/i386/i386.h (machine_function): Add
3854 function_return_type.
3855 * config/i386/i386.md (simple_return_internal): Use
3856 ix86_output_function_return.
3857 (simple_return_internal_long): Likewise.
3858 * config/i386/i386.opt (mfunction-return=): New option.
3859 (indirect_branch): Mention -mfunction-return=.
3860 * doc/extend.texi: Document function_return function attribute.
3861 * doc/invoke.texi: Document -mfunction-return= option.
3863 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3865 * config/i386/i386-opts.h (indirect_branch): New.
3866 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
3867 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
3868 with local indirect jump when converting indirect call and jump.
3869 (ix86_set_indirect_branch_type): New.
3870 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
3871 (indirectlabelno): New.
3872 (indirect_thunk_needed): Likewise.
3873 (indirect_thunk_bnd_needed): Likewise.
3874 (indirect_thunks_used): Likewise.
3875 (indirect_thunks_bnd_used): Likewise.
3876 (INDIRECT_LABEL): Likewise.
3877 (indirect_thunk_name): Likewise.
3878 (output_indirect_thunk): Likewise.
3879 (output_indirect_thunk_function): Likewise.
3880 (ix86_output_indirect_branch_via_reg): Likewise.
3881 (ix86_output_indirect_branch_via_push): Likewise.
3882 (ix86_output_indirect_branch): Likewise.
3883 (ix86_output_indirect_jmp): Likewise.
3884 (ix86_code_end): Call output_indirect_thunk_function if needed.
3885 (ix86_output_call_insn): Call ix86_output_indirect_branch if
3887 (ix86_handle_fndecl_attribute): Handle indirect_branch.
3888 (ix86_attribute_table): Add indirect_branch.
3889 * config/i386/i386.h (machine_function): Add indirect_branch_type
3890 and has_local_indirect_jump.
3891 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
3893 (tablejump): Likewise.
3894 (*indirect_jump): Use ix86_output_indirect_jmp.
3895 (*tablejump_1): Likewise.
3896 (simple_return_indirect_internal): Likewise.
3897 * config/i386/i386.opt (mindirect-branch=): New option.
3898 (indirect_branch): New.
3901 (thunk-inline): Likewise.
3902 (thunk-extern): Likewise.
3903 * doc/extend.texi: Document indirect_branch function attribute.
3904 * doc/invoke.texi: Document -mindirect-branch= option.
3906 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
3909 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
3911 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
3913 * ipa-inline.c (want_inline_small_function_p): Return false if
3914 inlining has already failed with CIF_FINAL_ERROR.
3915 (update_caller_keys): Call want_inline_small_function_p before
3917 (update_callee_keys): Likewise.
3919 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3921 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
3923 (rs6000_quadword_masked_address_p): Likewise.
3924 (quad_aligned_load_p): Likewise.
3925 (quad_aligned_store_p): Likewise.
3926 (const_load_sequence_p): Add comment to describe the outer-most loop.
3927 (mimic_memory_attributes_and_flags): New function.
3928 (rs6000_gen_stvx): Likewise.
3929 (replace_swapped_aligned_store): Likewise.
3930 (rs6000_gen_lvx): Likewise.
3931 (replace_swapped_aligned_load): Likewise.
3932 (replace_swapped_load_constant): Capitalize argument name in
3933 comment describing this function.
3934 (rs6000_analyze_swaps): Add a third pass to search for vector loads
3935 and stores that access quad-word aligned addresses and replace
3936 with stvx or lvx instructions when appropriate.
3937 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
3938 New function prototype.
3939 (rs6000_quadword_masked_address_p): Likewise.
3940 (rs6000_gen_lvx): Likewise.
3941 (rs6000_gen_stvx): Likewise.
3942 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
3943 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
3944 when memory address is aligned.
3945 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
3946 this split to select lvx instruction when memory address is aligned.
3947 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
3948 instruction when memory address is aligned.
3949 (*vsx_le_perm_load_v16qi): Likewise.
3950 (four unnamed splitters): Modify to select the stvx instruction
3951 when memory is aligned.
3953 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
3955 * predict.c (determine_unlikely_bbs): Handle correctly BBs
3956 which appears in the queue multiple times.
3958 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3959 Alan Hayward <alan.hayward@arm.com>
3960 David Sherwood <david.sherwood@arm.com>
3962 * tree-vectorizer.h (vec_lower_bound): New structure.
3963 (_loop_vec_info): Add check_nonzero and lower_bounds.
3964 (LOOP_VINFO_CHECK_NONZERO): New macro.
3965 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
3966 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
3967 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
3968 fields. Make seg_len the distance travelled, not including the
3970 (dr_direction_indicator): Declare.
3971 (dr_zero_step_indicator): Likewise.
3972 (dr_known_forward_stride_p): Likewise.
3973 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
3975 (runtime_alias_check_p): Allow runtime alias checks with
3977 (operator ==): Compare access_size and align.
3978 (prune_runtime_alias_test_list): Rework for new distinction between
3979 the access_size and seg_len.
3980 (create_intersect_range_checks_index): Likewise. Cope with polynomial
3982 (get_segment_min_max): New function.
3983 (create_intersect_range_checks): Use it.
3984 (dr_step_indicator): New function.
3985 (dr_direction_indicator): Likewise.
3986 (dr_zero_step_indicator): Likewise.
3987 (dr_known_forward_stride_p): Likewise.
3988 * tree-loop-distribution.c (data_ref_segment_size): Return
3989 DR_STEP * (niters - 1).
3990 (compute_alias_check_pairs): Update call to the dr_with_seg_len
3992 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
3993 (vect_preserves_scalar_order_p): New function, split out from...
3994 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
3995 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
3996 (vect_vfa_access_size): New function.
3997 (vect_vfa_align): Likewise.
3998 (vect_compile_time_alias): Take access_size_a and access_b arguments.
3999 (dump_lower_bound): New function.
4000 (vect_check_lower_bound): Likewise.
4001 (vect_small_gap_p): Likewise.
4002 (vectorizable_with_step_bound_p): Likewise.
4003 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
4004 depencies if the vectorization factor is 1. Convert the checks
4005 for nonzero steps into checks on the bounds of DR_STEP. Try using
4006 a bunds check for variable steps if the minimum required step is
4007 relatively small. Update calls to the dr_with_seg_len
4008 constructor and to vect_compile_time_alias.
4009 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
4011 (vect_loop_versioning): Call it.
4012 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
4014 (vect_estimate_min_profitable_iters): Account for any bounds checks.
4016 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4017 Alan Hayward <alan.hayward@arm.com>
4018 David Sherwood <david.sherwood@arm.com>
4020 * doc/sourcebuild.texi (vect_scatter_store): Document.
4021 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
4023 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
4025 * genopinit.c (main): Add supports_vec_scatter_store and
4026 supports_vec_scatter_store_cached to target_optabs.
4027 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
4028 IFN_MASK_SCATTER_STORE.
4029 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
4031 * internal-fn.h (internal_store_fn_p): Declare.
4032 (internal_fn_stored_value_index): Likewise.
4033 * internal-fn.c (scatter_store_direct): New macro.
4034 (expand_scatter_store_optab_fn): New function.
4035 (direct_scatter_store_optab_supported_p): New macro.
4036 (internal_store_fn_p): New function.
4037 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
4038 IFN_MASK_SCATTER_STORE.
4039 (internal_fn_mask_index): Likewise.
4040 (internal_fn_stored_value_index): New function.
4041 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
4043 * optabs-query.h (supports_vec_scatter_store_p): Declare.
4044 * optabs-query.c (supports_vec_scatter_store_p): New function.
4045 * tree-vectorizer.h (vect_get_store_rhs): Declare.
4046 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
4047 true for scatter stores.
4048 (vect_gather_scatter_fn_p): Handle scatter stores too.
4049 (vect_check_gather_scatter): Consider using scatter stores if
4050 supports_vec_scatter_store_p.
4051 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
4053 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4054 internal_fn_stored_value_index.
4055 (check_load_store_masking): Handle scatter stores too.
4056 (vect_get_store_rhs): Make public.
4057 (vectorizable_call): Use internal_store_fn_p.
4058 (vectorizable_store): Handle scatter store internal functions.
4059 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
4060 when deciding whether the end of the group has been reached.
4061 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
4062 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
4063 (mask_scatter_store<mode>): New insns.
4065 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4066 Alan Hayward <alan.hayward@arm.com>
4067 David Sherwood <david.sherwood@arm.com>
4069 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
4070 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
4071 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
4073 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
4074 Use vect_truncate_gather_scatter_offset if we can't treat the
4075 operation as a normal gather load or scatter store.
4076 (get_group_load_store_type): Take the gather_scatter_info
4077 as argument. Try using a gather load or scatter store for
4078 single-element groups.
4079 (get_load_store_type): Update calls to get_group_load_store_type
4080 and vect_use_strided_gather_scatters_p.
4082 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4083 Alan Hayward <alan.hayward@arm.com>
4084 David Sherwood <david.sherwood@arm.com>
4086 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
4087 optional tree argument.
4088 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
4090 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
4091 but continue to use the current value as a fallback.
4092 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
4093 to compare the updates.
4094 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
4095 (get_load_store_type): Use it when handling a strided access.
4096 (vect_get_strided_load_store_ops): New function.
4097 (vect_get_data_ptr_increment): Likewise.
4098 (vectorizable_load): Handle strided gather loads. Always pass
4099 a step to vect_create_data_ref_ptr and bump_vector_ptr.
4101 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4102 Alan Hayward <alan.hayward@arm.com>
4103 David Sherwood <david.sherwood@arm.com>
4105 * doc/md.texi (gather_load@var{m}): Document.
4106 (mask_gather_load@var{m}): Likewise.
4107 * genopinit.c (main): Add supports_vec_gather_load and
4108 supports_vec_gather_load_cached to target_optabs.
4109 * optabs-tree.c (init_tree_optimization_optabs): Use
4110 ggc_cleared_alloc to allocate target_optabs.
4111 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
4112 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
4114 * internal-fn.h (internal_load_fn_p): Declare.
4115 (internal_gather_scatter_fn_p): Likewise.
4116 (internal_fn_mask_index): Likewise.
4117 (internal_gather_scatter_fn_supported_p): Likewise.
4118 * internal-fn.c (gather_load_direct): New macro.
4119 (expand_gather_load_optab_fn): New function.
4120 (direct_gather_load_optab_supported_p): New macro.
4121 (direct_internal_fn_optab): New function.
4122 (internal_load_fn_p): Likewise.
4123 (internal_gather_scatter_fn_p): Likewise.
4124 (internal_fn_mask_index): Likewise.
4125 (internal_gather_scatter_fn_supported_p): Likewise.
4126 * optabs-query.c (supports_at_least_one_mode_p): New function.
4127 (supports_vec_gather_load_p): Likewise.
4128 * optabs-query.h (supports_vec_gather_load_p): Declare.
4129 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
4130 and memory_type field.
4131 (NUM_PATTERNS): Bump to 15.
4132 * tree-vect-data-refs.c: Include internal-fn.h.
4133 (vect_gather_scatter_fn_p): New function.
4134 (vect_describe_gather_scatter_call): Likewise.
4135 (vect_check_gather_scatter): Try using internal functions for
4136 gather loads. Recognize existing calls to a gather load function.
4137 (vect_analyze_data_refs): Consider using gather loads if
4138 supports_vec_gather_load_p.
4139 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
4140 (vect_get_gather_scatter_offset_type): Likewise.
4141 (vect_convert_mask_for_vectype): Likewise.
4142 (vect_add_conversion_to_patterm): Likewise.
4143 (vect_try_gather_scatter_pattern): Likewise.
4144 (vect_recog_gather_scatter_pattern): New pattern recognizer.
4145 (vect_vect_recog_func_ptrs): Add it.
4146 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4147 internal_fn_mask_index and internal_gather_scatter_fn_p.
4148 (check_load_store_masking): Take the gather_scatter_info as an
4149 argument and handle gather loads.
4150 (vect_get_gather_scatter_ops): New function.
4151 (vectorizable_call): Check internal_load_fn_p.
4152 (vectorizable_load): Likewise. Handle gather load internal
4154 (vectorizable_store): Update call to check_load_store_masking.
4155 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
4156 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
4157 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
4158 (aarch64_gather_scale_operand_d): New predicates.
4159 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
4160 (mask_gather_load<mode>): New insns.
4162 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4163 Alan Hayward <alan.hayward@arm.com>
4164 David Sherwood <david.sherwood@arm.com>
4166 * optabs.def (fold_left_plus_optab): New optab.
4167 * doc/md.texi (fold_left_plus_@var{m}): Document.
4168 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
4169 * internal-fn.c (fold_left_direct): Define.
4170 (expand_fold_left_optab_fn): Likewise.
4171 (direct_fold_left_optab_supported_p): Likewise.
4172 * fold-const-call.c (fold_const_fold_left): New function.
4173 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
4174 * tree-parloops.c (valid_reduction_p): New function.
4175 (gather_scalar_reductions): Use it.
4176 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
4177 (vect_finish_replace_stmt): Declare.
4178 * tree-vect-loop.c (fold_left_reduction_fn): New function.
4179 (needs_fold_left_reduction_p): New function, split out from...
4180 (vect_is_simple_reduction): ...here. Accept reductions that
4181 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
4182 (vect_force_simple_reduction): Also store the reduction type in
4183 the assignment's STMT_VINFO_REDUC_TYPE.
4184 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
4185 (merge_with_identity): New function.
4186 (vect_expand_fold_left): Likewise.
4187 (vectorize_fold_left_reduction): Likewise.
4188 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
4189 scalar phi in place for it. Check for target support and reject
4190 cases that would reassociate the operation. Defer the transform
4191 phase to vectorize_fold_left_reduction.
4192 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
4193 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
4194 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
4196 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4198 * tree-if-conv.c (predicate_mem_writes): Remove redundant
4199 call to ifc_temp_var.
4201 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4202 Alan Hayward <alan.hayward@arm.com>
4203 David Sherwood <david.sherwood@arm.com>
4205 * target.def (legitimize_address_displacement): Take the original
4206 offset as a poly_int.
4207 * targhooks.h (default_legitimize_address_displacement): Update
4209 * targhooks.c (default_legitimize_address_displacement): Likewise.
4210 * doc/tm.texi: Regenerate.
4211 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
4212 as an argument, moving assert of ad->disp == ad->disp_term to...
4213 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
4214 Try calling targetm.legitimize_address_displacement before expanding
4215 the address rather than afterwards, and adjust for the new interface.
4216 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
4217 Match the new hook interface. Handle SVE addresses.
4218 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
4221 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4223 * Makefile.in (OBJS): Add early-remat.o.
4224 * target.def (select_early_remat_modes): New hook.
4225 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
4226 * doc/tm.texi: Regenerate.
4227 * targhooks.h (default_select_early_remat_modes): Declare.
4228 * targhooks.c (default_select_early_remat_modes): New function.
4229 * timevar.def (TV_EARLY_REMAT): New timevar.
4230 * passes.def (pass_early_remat): New pass.
4231 * tree-pass.h (make_pass_early_remat): Declare.
4232 * early-remat.c: New file.
4233 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
4235 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
4237 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4238 Alan Hayward <alan.hayward@arm.com>
4239 David Sherwood <david.sherwood@arm.com>
4241 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
4242 vfm1 with a bound_epilog parameter.
4243 (vect_do_peeling): Update calls accordingly, and move the prologue
4244 call earlier in the function. Treat the base bound_epilog as 0 for
4245 fully-masked loops and retain vf - 1 for other loops. Add 1 to
4246 this base when peeling for gaps.
4247 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
4248 with fully-masked loops.
4249 (vect_estimate_min_profitable_iters): Handle the single peeled
4250 iteration in that case.
4252 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4253 Alan Hayward <alan.hayward@arm.com>
4254 David Sherwood <david.sherwood@arm.com>
4256 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
4257 single-element interleaving even if the size is not a power of 2.
4258 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
4259 accesses for single-element interleaving if the group size is
4262 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4263 Alan Hayward <alan.hayward@arm.com>
4264 David Sherwood <david.sherwood@arm.com>
4266 * doc/md.texi (fold_extract_last_@var{m}): Document.
4267 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
4268 * optabs.def (fold_extract_last_optab): New optab.
4269 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
4270 * internal-fn.c (fold_extract_direct): New macro.
4271 (expand_fold_extract_optab_fn): Likewise.
4272 (direct_fold_extract_optab_supported_p): Likewise.
4273 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
4274 * tree-vect-loop.c (vect_model_reduction_cost): Handle
4275 EXTRACT_LAST_REDUCTION.
4276 (get_initial_def_for_reduction): Do not create an initial vector
4277 for EXTRACT_LAST_REDUCTION reductions.
4278 (vectorizable_reduction): Leave the scalar phi in place for
4279 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
4280 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
4281 epilogue code for EXTRACT_LAST_REDUCTION and defer the
4282 transform phase to vectorizable_condition.
4283 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
4285 (vect_finish_stmt_generation): ...here.
4286 (vect_finish_replace_stmt): New function.
4287 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
4288 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
4290 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
4292 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4293 Alan Hayward <alan.hayward@arm.com>
4294 David Sherwood <david.sherwood@arm.com>
4296 * doc/md.texi (extract_last_@var{m}): Document.
4297 * optabs.def (extract_last_optab): New optab.
4298 * internal-fn.def (EXTRACT_LAST): New internal function.
4299 * internal-fn.c (cond_unary_direct): New macro.
4300 (expand_cond_unary_optab_fn): Likewise.
4301 (direct_cond_unary_optab_supported_p): Likewise.
4302 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
4303 loops using EXTRACT_LAST.
4304 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
4305 (extract_last_<mode>): ...this optab.
4306 (vec_extract<mode><Vel>): Update accordingly.
4308 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4309 Alan Hayward <alan.hayward@arm.com>
4310 David Sherwood <david.sherwood@arm.com>
4312 * target.def (empty_mask_is_expensive): New hook.
4313 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
4314 * doc/tm.texi: Regenerate.
4315 * targhooks.h (default_empty_mask_is_expensive): Declare.
4316 * targhooks.c (default_empty_mask_is_expensive): New function.
4317 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
4318 if the target says that empty masks are expensive.
4319 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
4321 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
4323 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4324 Alan Hayward <alan.hayward@arm.com>
4325 David Sherwood <david.sherwood@arm.com>
4327 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
4328 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
4329 (vect_use_loop_mask_for_alignment_p): New function.
4330 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
4331 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
4332 niters_skip argument. Make sure that the first niters_skip elements
4333 of the first iteration are inactive.
4334 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
4335 Update call to vect_set_loop_masks_directly.
4336 (get_misalign_in_elems): New function, split out from...
4337 (vect_gen_prolog_loop_niters): ...here.
4338 (vect_update_init_of_dr): Take a code argument that specifies whether
4339 the adjustment should be added or subtracted.
4340 (vect_update_init_of_drs): Likewise.
4341 (vect_prepare_for_masked_peels): New function.
4342 (vect_do_peeling): Skip prologue peeling if we're using a mask
4343 instead. Update call to vect_update_inits_of_drs.
4344 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4346 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
4347 alignment. Do not include the number of peeled iterations in
4348 the minimum threshold in that case.
4349 (vectorizable_induction): Adjust the start value down by
4350 LOOP_VINFO_MASK_SKIP_NITERS iterations.
4351 (vect_transform_loop): Call vect_prepare_for_masked_peels.
4352 Take the number of skipped iterations into account when calculating
4354 * tree-vect-stmts.c (vect_gen_while_not): New function.
4356 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4357 Alan Hayward <alan.hayward@arm.com>
4358 David Sherwood <david.sherwood@arm.com>
4360 * doc/sourcebuild.texi (vect_fully_masked): Document.
4361 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
4363 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
4365 (vect_analyze_loop_2): ...here. Don't check the vectorization
4366 factor against the number of loop iterations if the loop is
4369 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4370 Alan Hayward <alan.hayward@arm.com>
4371 David Sherwood <david.sherwood@arm.com>
4373 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
4374 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
4375 (dump_groups): Update accordingly.
4376 (iv_use::mem_type): New member variable.
4377 (address_p): New function.
4378 (record_use): Add a mem_type argument and initialize the new
4380 (record_group_use): Add a mem_type argument. Use address_p.
4381 Remove obsolete null checks of base_object. Update call to record_use.
4382 (find_interesting_uses_op): Update call to record_group_use.
4383 (find_interesting_uses_cond): Likewise.
4384 (find_interesting_uses_address): Likewise.
4385 (get_mem_type_for_internal_fn): New function.
4386 (find_address_like_use): Likewise.
4387 (find_interesting_uses_stmt): Try find_address_like_use before
4388 calling find_interesting_uses_op.
4389 (addr_offset_valid_p): Use the iv mem_type field as the type
4390 of the addressed memory.
4391 (add_autoinc_candidates): Likewise.
4392 (get_address_cost): Likewise.
4393 (split_small_address_groups_p): Use address_p.
4394 (split_address_groups): Likewise.
4395 (add_iv_candidate_for_use): Likewise.
4396 (autoinc_possible_for_pair): Likewise.
4397 (rewrite_groups): Likewise.
4398 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
4399 (determine_group_iv_cost): Update after split of USE_ADDRESS.
4400 (get_alias_ptr_type_for_ptr_address): New function.
4401 (rewrite_use_address): Rewrite address uses in calls that were
4402 identified by find_address_like_use.
4404 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4405 Alan Hayward <alan.hayward@arm.com>
4406 David Sherwood <david.sherwood@arm.com>
4408 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
4410 * gimple-expr.h (is_gimple_addressable: Likewise.
4411 * gimple-expr.c (is_gimple_address): Likewise.
4412 * internal-fn.c (expand_call_mem_ref): New function.
4413 (expand_mask_load_optab_fn): Use it.
4414 (expand_mask_store_optab_fn): Likewise.
4416 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4417 Alan Hayward <alan.hayward@arm.com>
4418 David Sherwood <david.sherwood@arm.com>
4420 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
4421 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
4422 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
4423 (cond_umax@var{mode}): Document.
4424 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
4425 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
4426 (cond_umin_optab, cond_umax_optab): New optabs.
4427 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
4428 (COND_IOR, COND_XOR): New internal functions.
4429 * internal-fn.h (get_conditional_internal_fn): Declare.
4430 * internal-fn.c (cond_binary_direct): New macro.
4431 (expand_cond_binary_optab_fn): Likewise.
4432 (direct_cond_binary_optab_supported_p): Likewise.
4433 (get_conditional_internal_fn): New function.
4434 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
4435 Cope with reduction statements that are vectorized as calls rather
4437 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
4438 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
4439 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
4440 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4441 (UNSPEC_COND_EOR): New unspecs.
4442 (optab): Add mappings for them.
4443 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
4444 (sve_int_op, sve_fp_op): New int attributes.
4446 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4447 Alan Hayward <alan.hayward@arm.com>
4448 David Sherwood <david.sherwood@arm.com>
4450 * optabs.def (while_ult_optab): New optab.
4451 * doc/md.texi (while_ult@var{m}@var{n}): Document.
4452 * internal-fn.def (WHILE_ULT): New internal function.
4453 * internal-fn.h (direct_internal_fn_supported_p): New override
4454 that takes two types as argument.
4455 * internal-fn.c (while_direct): New macro.
4456 (expand_while_optab_fn): New function.
4457 (convert_optab_supported_p): Likewise.
4458 (direct_while_optab_supported_p): New macro.
4459 * wide-int.h (wi::udiv_ceil): New function.
4460 * tree-vectorizer.h (rgroup_masks): New structure.
4461 (vec_loop_masks): New typedef.
4462 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
4464 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
4465 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
4466 (vect_max_vf): New function.
4467 (slpeel_make_loop_iterate_ntimes): Delete.
4468 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
4469 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
4470 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
4471 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
4472 internal-fn.h, stor-layout.h and optabs-query.h.
4473 (vect_set_loop_mask): New function.
4474 (add_preheader_seq): Likewise.
4475 (add_header_seq): Likewise.
4476 (interleave_supported_p): Likewise.
4477 (vect_maybe_permute_loop_masks): Likewise.
4478 (vect_set_loop_masks_directly): Likewise.
4479 (vect_set_loop_condition_masked): Likewise.
4480 (vect_set_loop_condition_unmasked): New function, split out from
4481 slpeel_make_loop_iterate_ntimes.
4482 (slpeel_make_loop_iterate_ntimes): Rename to..
4483 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
4484 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
4485 (vect_do_peeling): Update call accordingly.
4486 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
4488 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4489 mask_compare_type, can_fully_mask_p and fully_masked_p.
4490 (release_vec_loop_masks): New function.
4491 (_loop_vec_info): Use it to free the loop masks.
4492 (can_produce_all_loop_masks_p): New function.
4493 (vect_get_max_nscalars_per_iter): Likewise.
4494 (vect_verify_full_masking): Likewise.
4495 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
4496 retries, and free the mask rgroups before retrying. Check loop-wide
4497 reasons for disallowing fully-masked loops. Make the final decision
4498 about whether use a fully-masked loop or not.
4499 (vect_estimate_min_profitable_iters): Do not assume that peeling
4500 for the number of iterations will be needed for fully-masked loops.
4501 (vectorizable_reduction): Disable fully-masked loops.
4502 (vectorizable_live_operation): Likewise.
4503 (vect_halve_mask_nunits): New function.
4504 (vect_double_mask_nunits): Likewise.
4505 (vect_record_loop_mask): Likewise.
4506 (vect_get_loop_mask): Likewise.
4507 (vect_transform_loop): Handle the case in which the final loop
4508 iteration might handle a partial vector. Call vect_set_loop_condition
4509 instead of slpeel_make_loop_iterate_ntimes.
4510 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
4511 (check_load_store_masking): New function.
4512 (prepare_load_store_mask): Likewise.
4513 (vectorizable_store): Handle fully-masked loops.
4514 (vectorizable_load): Likewise.
4515 (supportable_widening_operation): Use vect_halve_mask_nunits for
4517 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
4518 (vect_gen_while): New function.
4519 * config/aarch64/aarch64.md (umax<mode>3): New expander.
4520 (aarch64_uqdec<mode>): New insn.
4522 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4523 Alan Hayward <alan.hayward@arm.com>
4524 David Sherwood <david.sherwood@arm.com>
4526 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
4527 (reduc_xor_scal_optab): New optabs.
4528 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
4529 (reduc_xor_scal_@var{m}): Document.
4530 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
4531 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
4533 * fold-const-call.c (fold_const_call): Handle them.
4534 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
4535 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
4536 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
4537 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
4538 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
4539 (UNSPEC_XORV): New unspecs.
4540 (optab): Add entries for them.
4541 (BITWISEV): New int iterator.
4542 (bit_reduc_op): New int attributes.
4544 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4545 Alan Hayward <alan.hayward@arm.com>
4546 David Sherwood <david.sherwood@arm.com>
4548 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
4549 * internal-fn.def (VEC_SHL_INSERT): New internal function.
4550 * optabs.def (vec_shl_insert_optab): New optab.
4551 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
4552 (duplicate_and_interleave): Likewise.
4553 * tree-vect-loop.c: Include internal-fn.h.
4554 (neutral_op_for_slp_reduction): New function, split out from
4555 get_initial_defs_for_reduction.
4556 (get_initial_def_for_reduction): Handle option 2 for variable-length
4557 vectors by loading the neutral value into a vector and then shifting
4558 the initial value into element 0.
4559 (get_initial_defs_for_reduction): Replace the code argument with
4560 the neutral value calculated by neutral_op_for_slp_reduction.
4561 Use gimple_build_vector for constant-length vectors.
4562 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
4563 but the first group_size elements have a neutral value.
4564 Use duplicate_and_interleave otherwise.
4565 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
4566 Update call to get_initial_defs_for_reduction. Handle SLP
4567 reductions for variable-length vectors by creating one vector
4568 result for each scalar result, with the elements associated
4569 with other scalar results stubbed out with the neutral value.
4570 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
4571 Require IFN_VEC_SHL_INSERT for double reductions on
4572 variable-length vectors, or SLP reductions that have
4573 a neutral value. Require can_duplicate_and_interleave_p
4574 support for variable-length unchained SLP reductions if there
4575 is no neutral value, such as for MIN/MAX reductions. Also require
4576 the number of vector elements to be a multiple of the number of
4577 SLP statements when doing variable-length unchained SLP reductions.
4578 Update call to vect_create_epilog_for_reduction.
4579 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
4580 and remove initial values.
4581 (duplicate_and_interleave): Make public.
4582 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
4583 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
4585 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4586 Alan Hayward <alan.hayward@arm.com>
4587 David Sherwood <david.sherwood@arm.com>
4589 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
4590 (can_duplicate_and_interleave_p): New function.
4591 (vect_get_and_check_slp_defs): Take the vector of statements
4592 rather than just the current one. Remove excess parentheses.
4593 Restriction rejectinon of vect_constant_def and vect_external_def
4594 for variable-length vectors to boolean types, or types for which
4595 can_duplicate_and_interleave_p is false.
4596 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
4597 (duplicate_and_interleave): New function.
4598 (vect_get_constant_vectors): Use gimple_build_vector for
4599 constant-length vectors and suitable variable-length constant
4600 vectors. Use duplicate_and_interleave for other variable-length
4601 vectors. Don't defer the update when inserting new statements.
4603 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4604 Alan Hayward <alan.hayward@arm.com>
4605 David Sherwood <david.sherwood@arm.com>
4607 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
4608 min_profitable_iters doesn't go negative.
4610 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4611 Alan Hayward <alan.hayward@arm.com>
4612 David Sherwood <david.sherwood@arm.com>
4614 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
4615 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
4616 * optabs.def (vec_mask_load_lanes_optab): New optab.
4617 (vec_mask_store_lanes_optab): Likewise.
4618 * internal-fn.def (MASK_LOAD_LANES): New internal function.
4619 (MASK_STORE_LANES): Likewise.
4620 * internal-fn.c (mask_load_lanes_direct): New macro.
4621 (mask_store_lanes_direct): Likewise.
4622 (expand_mask_load_optab_fn): Handle masked operations.
4623 (expand_mask_load_lanes_optab_fn): New macro.
4624 (expand_mask_store_optab_fn): Handle masked operations.
4625 (expand_mask_store_lanes_optab_fn): New macro.
4626 (direct_mask_load_lanes_optab_supported_p): Likewise.
4627 (direct_mask_store_lanes_optab_supported_p): Likewise.
4628 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
4630 (vect_load_lanes_supported): Likewise.
4631 * tree-vect-data-refs.c (strip_conversion): New function.
4632 (can_group_stmts_p): Likewise.
4633 (vect_analyze_data_ref_accesses): Use it instead of checking
4634 for a pair of assignments.
4635 (vect_store_lanes_supported): Take a masked_p parameter.
4636 (vect_load_lanes_supported): Likewise.
4637 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
4638 vect_store_lanes_supported and vect_load_lanes_supported.
4639 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4640 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
4641 parameter. Don't allow gaps for masked accesses.
4642 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
4643 and vect_load_lanes_supported.
4644 (get_load_store_type): Take a masked_p parameter and update
4645 call to get_group_load_store_type.
4646 (vectorizable_store): Update call to get_load_store_type.
4647 Handle IFN_MASK_STORE_LANES.
4648 (vectorizable_load): Update call to get_load_store_type.
4649 Handle IFN_MASK_LOAD_LANES.
4651 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4652 Alan Hayward <alan.hayward@arm.com>
4653 David Sherwood <david.sherwood@arm.com>
4655 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
4657 * config/aarch64/aarch64-protos.h
4658 (aarch64_sve_struct_memory_operand_p): Declare.
4659 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
4660 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
4661 (VPRED, vpred): Handle SVE structure modes.
4662 * config/aarch64/constraints.md (Utx): New constraint.
4663 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
4664 (aarch64_sve_struct_nonimmediate_operand): New predicates.
4665 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
4666 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
4667 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
4668 structure modes. Split into pieces after RA.
4669 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
4670 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
4672 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
4673 SVE structure modes.
4674 (aarch64_classify_address): Likewise.
4675 (sizetochar): Move earlier in file.
4676 (aarch64_print_operand): Handle SVE register lists.
4677 (aarch64_array_mode): New function.
4678 (aarch64_sve_struct_memory_operand_p): Likewise.
4679 (TARGET_ARRAY_MODE): Redefine.
4681 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4682 Alan Hayward <alan.hayward@arm.com>
4683 David Sherwood <david.sherwood@arm.com>
4685 * target.def (array_mode): New target hook.
4686 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
4687 * doc/tm.texi: Regenerate.
4688 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
4689 * hooks.c (hook_optmode_mode_uhwi_none): New function.
4690 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
4692 * stor-layout.c (mode_for_array): Likewise. Support polynomial
4695 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4696 Alan Hayward <alan.hayward@arm.com>
4697 David Sherwood <david.sherwood@arm.com>
4699 * fold-const.c (fold_binary_loc): Check the argument types
4700 rather than the result type when testing for a vector operation.
4702 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4704 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
4705 * doc/tm.texi: Regenerate.
4707 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4708 Alan Hayward <alan.hayward@arm.com>
4709 David Sherwood <david.sherwood@arm.com>
4711 * doc/invoke.texi (-msve-vector-bits=): Document new option.
4712 (sve): Document new AArch64 extension.
4713 * doc/md.texi (w): Extend the description of the AArch64
4714 constraint to include SVE vectors.
4715 (Upl, Upa): Document new AArch64 predicate constraints.
4716 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
4718 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
4719 (msve-vector-bits=): New option.
4720 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
4721 SVE when these are disabled.
4722 (sve): New extension.
4723 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
4724 modes. Adjust their number of units based on aarch64_sve_vg.
4725 (MAX_BITSIZE_MODE_ANY_MODE): Define.
4726 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
4727 aarch64_addr_query_type.
4728 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
4729 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
4730 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
4731 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
4732 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
4733 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
4734 (aarch64_simd_imm_zero_p): Delete.
4735 (aarch64_check_zero_based_sve_index_immediate): Declare.
4736 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4737 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4738 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4739 (aarch64_sve_float_mul_immediate_p): Likewise.
4740 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4742 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
4743 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
4744 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
4745 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
4746 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
4747 (aarch64_regmode_natural_size): Likewise.
4748 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
4749 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
4751 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
4752 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
4753 for VG and the SVE predicate registers.
4754 (V_ALIASES): Add a "z"-prefixed alias.
4755 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
4756 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
4757 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
4758 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
4759 (REG_CLASS_NAMES): Add entries for them.
4760 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
4761 and the predicate registers.
4762 (aarch64_sve_vg): Declare.
4763 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
4764 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
4765 (REGMODE_NATURAL_SIZE): Define.
4766 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
4768 * config/aarch64/aarch64.c: Include cfgrtl.h.
4769 (simd_immediate_info): Add a constructor for series vectors,
4770 and an associated step field.
4771 (aarch64_sve_vg): New variable.
4772 (aarch64_dbx_register_number): Handle VG and the predicate registers.
4773 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
4774 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
4775 (VEC_ANY_DATA, VEC_STRUCT): New constants.
4776 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
4777 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
4778 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
4779 (aarch64_get_mask_mode): New functions.
4780 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
4781 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
4782 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
4783 predicate modes and predicate registers. Explicitly restrict
4784 GPRs to modes of 16 bytes or smaller. Only allow FP registers
4785 to store a vector mode if it is recognized by
4786 aarch64_classify_vector_mode.
4787 (aarch64_regmode_natural_size): New function.
4788 (aarch64_hard_regno_caller_save_mode): Return the original mode
4790 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
4791 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
4792 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
4793 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
4795 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
4796 does not overlap dest if the function is frame-related. Handle
4798 (aarch64_split_add_offset): New function.
4799 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
4800 them aarch64_add_offset.
4801 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
4802 and update call to aarch64_sub_sp.
4803 (aarch64_add_cfa_expression): New function.
4804 (aarch64_expand_prologue): Pass extra temporary registers to the
4805 functions above. Handle the case in which we need to emit new
4806 DW_CFA_expressions for registers that were originally saved
4807 relative to the stack pointer, but now have to be expressed
4808 relative to the frame pointer.
4809 (aarch64_output_mi_thunk): Pass extra temporary registers to the
4811 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
4812 IP0 and IP1 values for SVE frames.
4813 (aarch64_expand_vec_series): New function.
4814 (aarch64_expand_sve_widened_duplicate): Likewise.
4815 (aarch64_expand_sve_const_vector): Likewise.
4816 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
4817 Handle SVE constants. Use emit_move_insn to move a force_const_mem
4818 into the register, rather than emitting a SET directly.
4819 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
4820 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
4821 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
4822 (offset_9bit_signed_scaled_p): New functions.
4823 (aarch64_replicate_bitmask_imm): New function.
4824 (aarch64_bitmask_imm): Use it.
4825 (aarch64_cannot_force_const_mem): Reject expressions involving
4826 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
4827 (aarch64_classify_index): Handle SVE indices, by requiring
4828 a plain register index with a scale that matches the element size.
4829 (aarch64_classify_address): Handle SVE addresses. Assert that
4830 the mode of the address is VOIDmode or an integer mode.
4831 Update call to aarch64_classify_symbol.
4832 (aarch64_classify_symbolic_expression): Update call to
4833 aarch64_classify_symbol.
4834 (aarch64_const_vec_all_in_range_p): New function.
4835 (aarch64_print_vector_float_operand): Likewise.
4836 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
4837 "vN" for FP registers with SVE modes. Handle (const ...) vectors
4838 and the FP immediates 1.0 and 0.5.
4839 (aarch64_print_address_internal): Handle SVE addresses.
4840 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
4841 (aarch64_regno_regclass): Handle predicate registers.
4842 (aarch64_secondary_reload): Handle big-endian reloads of SVE
4844 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
4845 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
4846 (aarch64_convert_sve_vector_bits): New function.
4847 (aarch64_override_options): Use it to handle -msve-vector-bits=.
4848 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4850 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
4851 Handle SVE vector and predicate modes. Accept VL-based constants
4852 that need only one temporary register, and VL offsets that require
4853 no temporary registers.
4854 (aarch64_conditional_register_usage): Mark the predicate registers
4855 as fixed if SVE isn't available.
4856 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
4857 Return true for SVE vector and predicate modes.
4858 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
4859 rather than an unsigned int. Handle SVE modes.
4860 (aarch64_preferred_simd_mode): Update call accordingly. Handle
4862 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
4864 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4865 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4866 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4867 (aarch64_sve_float_mul_immediate_p): New functions.
4868 (aarch64_sve_valid_immediate): New function.
4869 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
4870 Explicitly reject structure modes. Check for INDEX constants.
4871 Handle PTRUE and PFALSE constants.
4872 (aarch64_check_zero_based_sve_index_immediate): New function.
4873 (aarch64_simd_imm_zero_p): Delete.
4874 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
4875 vector modes. Accept constants in the range of CNT[BHWD].
4876 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
4877 ask for an Advanced SIMD mode.
4878 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
4879 (aarch64_simd_vector_alignment): Handle SVE predicates.
4880 (aarch64_vectorize_preferred_vector_alignment): New function.
4881 (aarch64_simd_vector_alignment_reachable): Use it instead of
4883 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
4884 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
4886 (MAX_VECT_LEN): Delete.
4887 (expand_vec_perm_d): Add a vec_flags field.
4888 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
4889 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
4890 (aarch64_evpc_ext): Don't apply a big-endian lane correction
4892 (aarch64_evpc_rev): Rename to...
4893 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
4894 (aarch64_evpc_rev_global): New function.
4895 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
4896 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
4898 (aarch64_evpc_sve_tbl): New function.
4899 (aarch64_expand_vec_perm_const_1): Update after rename of
4900 aarch64_evpc_rev. Handle SVE permutes too, trying
4901 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
4902 than aarch64_evpc_tbl.
4903 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
4904 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
4905 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
4906 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
4907 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
4908 (aarch64_expand_sve_vcond): New functions.
4909 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
4910 of aarch64_vector_mode_p.
4911 (aarch64_dwarf_poly_indeterminate_value): New function.
4912 (aarch64_compute_pressure_classes): Likewise.
4913 (aarch64_can_change_mode_class): Likewise.
4914 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
4915 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
4916 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
4917 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
4918 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
4919 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
4920 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
4921 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
4923 (Dn, Dl, Dr): Accept const as well as const_vector.
4924 (Dz): Likewise. Compare against CONST0_RTX.
4925 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
4926 of "vector" where appropriate.
4927 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
4928 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
4929 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
4930 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
4931 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
4932 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
4933 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
4934 (v_int_equiv): Extend to SVE modes.
4935 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
4937 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
4938 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
4939 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
4940 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
4941 (SVE_COND_FP_CMP): New int iterators.
4942 (perm_hilo): Handle the new unpack unspecs.
4943 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
4945 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
4946 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
4947 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
4948 (aarch64_equality_operator, aarch64_constant_vector_operand)
4949 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
4950 (aarch64_sve_nonimmediate_operand): Likewise.
4951 (aarch64_sve_general_operand): Likewise.
4952 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
4953 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
4954 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
4955 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
4956 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
4957 (aarch64_sve_float_arith_immediate): Likewise.
4958 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
4959 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
4960 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
4961 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
4962 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
4963 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
4964 (aarch64_sve_float_arith_operand): Likewise.
4965 (aarch64_sve_float_arith_with_sub_operand): Likewise.
4966 (aarch64_sve_float_mul_operand): Likewise.
4967 (aarch64_sve_vec_perm_operand): Likewise.
4968 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
4969 (aarch64_mov_operand): Accept const_poly_int and const_vector.
4970 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
4971 as well as const_vector.
4972 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
4973 in file. Use CONST0_RTX and CONSTM1_RTX.
4974 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
4975 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
4976 Use aarch64_simd_imm_zero.
4977 * config/aarch64/aarch64-sve.md: New file.
4978 * config/aarch64/aarch64.md: Include it.
4979 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
4980 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
4981 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
4982 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
4983 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
4984 (sve): New attribute.
4985 (enabled): Disable instructions with the sve attribute unless
4987 (movqi, movhi): Pass CONST_POLY_INT operaneds through
4988 aarch64_expand_mov_immediate.
4989 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
4990 CNT[BHSD] immediates.
4991 (movti): Split CONST_POLY_INT moves into two halves.
4992 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
4993 Split additions that need a temporary here if the destination
4994 is the stack pointer.
4995 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
4996 (*add<mode>3_poly_1): New instruction.
4997 (set_clobber_cc): New expander.
4999 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5001 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
5002 parameter and use it instead of GET_MODE_SIZE (innermode). Use
5003 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
5004 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
5005 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
5006 Change innermode from fixed_mode_size to machine_mode.
5007 (simplify_subreg): Update call accordingly. Handle a constant-sized
5008 subreg of a variable-length CONST_VECTOR.
5010 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5011 Alan Hayward <alan.hayward@arm.com>
5012 David Sherwood <david.sherwood@arm.com>
5014 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
5015 (add_offset_to_base): New function, split out from...
5016 (create_mem_ref): ...here. When handling a scale other than 1,
5017 check first whether the address is valid without the offset.
5018 Add it into the base if so, leaving the index and scale as-is.
5020 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5023 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
5024 fold_for_warn before checking if arg2 is INTEGER_CST.
5026 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
5028 * config/rs6000/predicates.md (load_multiple_operation): Delete.
5029 (store_multiple_operation): Delete.
5030 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
5031 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
5032 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
5033 guarded by TARGET_STRING.
5034 (rs6000_output_load_multiple): Delete.
5035 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
5036 OPTION_MASK_STRING / TARGET_STRING handling.
5037 (print_operand) <'N', 'O'>: Add comment that these are unused now.
5038 (const rs6000_opt_masks) <"string">: Change mask to 0.
5039 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
5040 (MASK_STRING): Delete.
5041 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
5043 (load_multiple): Delete.
5050 (store_multiple): Delete.
5057 (movmemsi_8reg): Delete.
5058 (corresponding unnamed define_insn): Delete.
5059 (movmemsi_6reg): Delete.
5060 (corresponding unnamed define_insn): Delete.
5061 (movmemsi_4reg): Delete.
5062 (corresponding unnamed define_insn): Delete.
5063 (movmemsi_2reg): Delete.
5064 (corresponding unnamed define_insn): Delete.
5065 (movmemsi_1reg): Delete.
5066 (corresponding unnamed define_insn): Delete.
5067 * config/rs6000/rs6000.opt (mno-string): New.
5068 (mstring): Replace by deprecation warning stub.
5069 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
5071 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5073 * regrename.c (regrename_do_replace): If replacing the same
5074 reg multiple times, try to reuse last created gen_raw_REG.
5077 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
5078 main to workaround a bug in GDB.
5080 2018-01-12 Tom de Vries <tom@codesourcery.com>
5083 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
5085 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
5087 PR rtl-optimization/80481
5088 * ira-color.c (get_cap_member): New function.
5089 (allocnos_conflict_by_live_ranges_p): Use it.
5090 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
5091 (setup_slot_coalesced_allocno_live_ranges): Ditto.
5093 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
5096 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
5097 (*saddl_se_1): Ditto.
5099 (*ssubl_se_1): Ditto.
5101 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5103 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
5104 rather than wi::to_widest for DR_INITs.
5105 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
5106 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
5107 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
5109 (vect_analyze_group_access_1): Note that here.
5111 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5113 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
5114 polynomial type sizes.
5116 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5118 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
5119 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
5120 (gimple_add_tmp_var): Likewise.
5122 2018-01-12 Martin Liska <mliska@suse.cz>
5124 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
5125 (gimple_alloc_sizes): Likewise.
5126 (dump_gimple_statistics): Use PRIu64 in printf format.
5127 * gimple.h: Change uint64_t to int.
5129 2018-01-12 Martin Liska <mliska@suse.cz>
5131 * tree-core.h: Use uint64_t instead of int.
5132 * tree.c (tree_node_counts): Likewise.
5133 (tree_node_sizes): Likewise.
5134 (dump_tree_statistics): Use PRIu64 in printf format.
5136 2018-01-12 Martin Liska <mliska@suse.cz>
5138 * Makefile.in: As qsort_chk is implemented in vec.c, add
5139 vec.o to linkage of gencfn-macros.
5140 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
5141 passing the info to record_node_allocation_statistics.
5142 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
5144 * ggc-common.c (struct ggc_usage): Add operator== and use
5145 it in operator< and compare function.
5146 * mem-stats.h (struct mem_usage): Likewise.
5147 * vec.c (struct vec_usage): Remove operator< and compare
5148 function. Can be simply inherited.
5150 2018-01-12 Martin Jambor <mjambor@suse.cz>
5153 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
5154 * tree-ssa-math-opts.c: Include domwalk.h.
5155 (convert_mult_to_fma_1): New function.
5156 (fma_transformation_info): New type.
5157 (fma_deferring_state): Likewise.
5158 (cancel_fma_deferring): New function.
5159 (result_of_phi): Likewise.
5160 (last_fma_candidate_feeds_initial_phi): Likewise.
5161 (convert_mult_to_fma): Added deferring logic, split actual
5162 transformation to convert_mult_to_fma_1.
5163 (math_opts_dom_walker): New type.
5164 (math_opts_dom_walker::after_dom_children): New method, body moved
5165 here from pass_optimize_widening_mul::execute, added deferring logic
5167 (pass_optimize_widening_mul::execute): Moved most of code to
5168 math_opts_dom_walker::after_dom_children.
5169 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
5170 * config/i386/i386.c (ix86_option_override_internal): Added
5171 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
5173 2018-01-12 Richard Biener <rguenther@suse.de>
5176 * dwarf2out.c (gen_variable_die): Do not reset old_die for
5177 inline instance vars.
5179 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
5182 * config/rx/rx.c (rx_is_restricted_memory_address):
5185 2018-01-12 Richard Biener <rguenther@suse.de>
5187 PR tree-optimization/80846
5188 * target.def (split_reduction): New target hook.
5189 * targhooks.c (default_split_reduction): New function.
5190 * targhooks.h (default_split_reduction): Declare.
5191 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
5192 target requests first reduce vectors by combining low and high
5194 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
5195 (get_vectype_for_scalar_type_and_size): Export.
5196 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
5197 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
5198 * doc/tm.texi: Regenerate.
5199 * config/i386/i386.c (ix86_split_reduction): Implement
5200 TARGET_VECTORIZE_SPLIT_REDUCTION.
5202 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5205 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
5206 in PIC mode except for TARGET_VXWORKS_RTP.
5207 * config/sparc/sparc.c: Include cfgrtl.h.
5208 (TARGET_INIT_PIC_REG): Define.
5209 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
5210 (sparc_pic_register_p): New predicate.
5211 (sparc_legitimate_address_p): Use it.
5212 (sparc_legitimize_pic_address): Likewise.
5213 (sparc_delegitimize_address): Likewise.
5214 (sparc_mode_dependent_address_p): Likewise.
5215 (gen_load_pcrel_sym): Remove 4th parameter.
5216 (load_got_register): Adjust call to above. Remove obsolete stuff.
5217 (sparc_expand_prologue): Do not call load_got_register here.
5218 (sparc_flat_expand_prologue): Likewise.
5219 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
5220 (sparc_use_pseudo_pic_reg): New function.
5221 (sparc_init_pic_reg): Likewise.
5222 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5223 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
5225 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
5227 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
5228 Add item for branch_cost.
5230 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5232 PR rtl-optimization/83565
5233 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
5234 not extend the result to a larger mode for rotate operations.
5235 (num_sign_bit_copies1): Likewise.
5237 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5240 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
5242 Use values-Xc.o for -pedantic.
5243 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
5245 2018-01-12 Martin Liska <mliska@suse.cz>
5248 * ipa-devirt.c (final_warning_record::grow_type_warnings):
5250 (possible_polymorphic_call_targets): Use it.
5251 (ipa_devirt): Likewise.
5253 2018-01-12 Martin Liska <mliska@suse.cz>
5255 * profile-count.h (enum profile_quality): Use 0 as invalid
5256 enum value of profile_quality.
5258 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
5260 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
5261 -mext-string options.
5263 2018-01-12 Richard Biener <rguenther@suse.de>
5265 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
5266 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
5267 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
5269 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
5271 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5273 * configure.ac (--with-long-double-format): Add support for the
5274 configuration option to change the default long double format on
5276 * config.gcc (powerpc*-linux*-*): Likewise.
5277 * configure: Regenerate.
5278 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
5279 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
5280 used without modification.
5282 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5284 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
5285 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
5286 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
5287 MISC_BUILTIN_SPEC_BARRIER.
5288 (rs6000_init_builtins): Likewise.
5289 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
5291 (speculation_barrier): New define_insn.
5292 * doc/extend.texi: Document __builtin_speculation_barrier.
5294 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5297 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
5298 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
5299 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
5301 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
5302 integral modes instead of "ss" and "sd".
5303 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
5304 vectors with 32-bit and 64-bit elements.
5305 (vecdupssescalarmodesuffix): New mode attribute.
5306 (vec_dup<mode>): Use it.
5308 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
5311 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
5312 frame if argument is passed on stack.
5314 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5317 * ree.c (combine_reaching_defs): Optimize also
5318 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
5319 reg2=any_extend(exp); reg1=reg2;, formatting fix.
5321 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5324 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
5326 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5329 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
5330 after they are computed.
5332 2018-01-11 Bin Cheng <bin.cheng@arm.com>
5334 PR tree-optimization/83695
5335 * gimple-loop-linterchange.cc
5336 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
5337 reset cached scev information after interchange.
5338 (pass_linterchange::execute): Remove call to scev_reset_htab.
5340 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5342 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
5343 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
5344 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
5345 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
5346 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
5347 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
5348 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
5349 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
5350 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
5351 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
5352 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
5353 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
5354 (V_lane_reg): Likewise.
5355 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
5357 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
5358 (vfmal_lane_low<mode>_intrinsic,
5359 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
5360 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
5361 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
5362 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
5363 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
5364 vfmsl_lane_high<mode>_intrinsic): New define_insns.
5366 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5368 * config/arm/arm-cpus.in (fp16fml): New feature.
5369 (ALL_SIMD): Add fp16fml.
5370 (armv8.2-a): Add fp16fml as an option.
5371 (armv8.3-a): Likewise.
5372 (armv8.4-a): Add fp16fml as part of fp16.
5373 * config/arm/arm.h (TARGET_FP16FML): Define.
5374 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
5376 * config/arm/arm-modes.def (V2HF): Define.
5377 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
5378 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
5379 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
5380 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
5381 vfmsl_low, vfmsl_high): New set of builtins.
5382 * config/arm/iterators.md (PLUSMINUS): New code iterator.
5383 (vfml_op): New code attribute.
5384 (VFMLHALVES): New int iterator.
5385 (VFML, VFMLSEL): New mode attributes.
5386 (V_reg): Define mapping for V2HF.
5387 (V_hi, V_lo): New mode attributes.
5388 (VF_constraint): Likewise.
5389 (vfml_half, vfml_half_selector): New int attributes.
5390 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
5392 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
5393 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
5395 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
5396 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
5397 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
5398 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
5400 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
5401 Document new effective target and option set.
5403 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5405 * config/arm/arm-cpus.in (armv8_4): New feature.
5406 (ARMv8_4a): New fgroup.
5407 (armv8.4-a): New arch.
5408 * config/arm/arm-tables.opt: Regenerate.
5409 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
5410 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
5411 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
5412 Add matching rules for -march=armv8.4-a and extensions.
5413 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
5415 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
5418 * config/rx/rx.md (BW): New mode attribute.
5419 (sync_lock_test_and_setsi): Add mode suffix to insn output.
5421 2018-01-11 Richard Biener <rguenther@suse.de>
5423 PR tree-optimization/83435
5424 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
5425 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
5426 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
5428 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5429 Alan Hayward <alan.hayward@arm.com>
5430 David Sherwood <david.sherwood@arm.com>
5432 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
5434 (aarch64_classify_address): Initialize it. Track polynomial offsets.
5435 (aarch64_print_address_internal): Use it to check for a zero offset.
5437 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5438 Alan Hayward <alan.hayward@arm.com>
5439 David Sherwood <david.sherwood@arm.com>
5441 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
5442 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
5443 Return a poly_int64 rather than a HOST_WIDE_INT.
5444 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
5445 rather than a HOST_WIDE_INT.
5446 * config/aarch64/aarch64.h (aarch64_frame): Protect with
5447 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
5448 hard_fp_offset, frame_size, initial_adjust, callee_offset and
5449 final_offset from HOST_WIDE_INT to poly_int64.
5450 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
5451 to_constant when getting the number of units in an Advanced SIMD
5453 (aarch64_builtin_vectorized_function): Check for a constant number
5455 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
5457 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
5458 attribute instead of GET_MODE_NUNITS.
5459 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
5460 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
5461 GET_MODE_SIZE for fixed-size registers.
5462 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
5463 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
5464 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
5465 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
5466 (aarch64_print_operand, aarch64_print_address_internal)
5467 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
5468 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
5469 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
5470 Handle polynomial GET_MODE_SIZE.
5471 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
5472 wider than SImode without modification.
5473 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
5474 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
5475 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
5476 passing and returning SVE modes.
5477 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
5478 rather than GEN_INT.
5479 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
5480 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
5481 (aarch64_allocate_and_probe_stack_space): Likewise.
5482 (aarch64_layout_frame): Cope with polynomial offsets.
5483 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
5484 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
5486 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
5487 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
5488 poly_int64 rather than a HOST_WIDE_INT.
5489 (aarch64_get_separate_components, aarch64_process_components)
5490 (aarch64_expand_prologue, aarch64_expand_epilogue)
5491 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
5492 (aarch64_anchor_offset): New function, split out from...
5493 (aarch64_legitimize_address): ...here.
5494 (aarch64_builtin_vectorization_cost): Handle polynomial
5495 TYPE_VECTOR_SUBPARTS.
5496 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
5498 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
5499 number of elements from the PARALLEL rather than the mode.
5500 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
5501 rather than GET_MODE_BITSIZE.
5502 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
5503 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
5504 (aarch64_expand_vec_perm_const_1): Handle polynomial
5505 d->perm.length () and d->perm elements.
5506 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
5507 Apply to_constant to d->perm elements.
5508 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
5509 polynomial CONST_VECTOR_NUNITS.
5510 (aarch64_move_pointer): Take amount as a poly_int64 rather
5512 (aarch64_progress_pointer): Avoid temporary variable.
5513 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
5514 the mode attribute instead of GET_MODE.
5516 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5517 Alan Hayward <alan.hayward@arm.com>
5518 David Sherwood <david.sherwood@arm.com>
5520 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
5521 x exists before using it.
5522 (aarch64_add_constant_internal): Rename to...
5523 (aarch64_add_offset_1): ...this. Replace regnum with separate
5524 src and dest rtxes. Handle the case in which they're different,
5525 including when the offset is zero. Replace scratchreg with an rtx.
5526 Use 2 additions if there is no spare register into which we can
5527 move a 16-bit constant.
5528 (aarch64_add_constant): Delete.
5529 (aarch64_add_offset): Replace reg with separate src and dest
5530 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
5531 Use aarch64_add_offset_1.
5532 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
5533 an rtx rather than an int. Take the delta as a poly_int64
5534 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
5535 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
5536 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
5537 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
5538 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
5540 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
5541 aarch64_add_constant.
5543 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5545 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
5546 Use scalar_float_mode.
5548 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5550 * config/aarch64/aarch64-simd.md
5551 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
5552 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
5553 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
5554 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
5555 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
5556 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
5557 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
5558 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
5559 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
5560 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
5562 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5565 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
5566 targ_options->x_arm_arch_string is non NULL.
5568 2018-01-11 Tamar Christina <tamar.christina@arm.com>
5570 * config/aarch64/aarch64.h
5571 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
5573 2018-01-11 Sudakshina Das <sudi.das@arm.com>
5576 * expmed.c (emit_store_flag_force): Swap if const op0
5577 and change VOIDmode to mode of op0.
5579 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5581 PR rtl-optimization/83761
5582 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
5583 than bytes to mode_for_size.
5585 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5588 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
5589 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
5592 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5595 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
5596 when in layout mode.
5597 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
5598 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
5601 2018-01-10 Michael Collison <michael.collison@arm.com>
5603 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
5604 * config/aarch64/aarch64-option-extension.def: Add
5605 AARCH64_OPT_EXTENSION of 'fp16fml'.
5606 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5607 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
5608 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
5609 * config/aarch64/constraints.md (Ui7): New constraint.
5610 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
5611 (VFMLA_SEL_W): Ditto.
5614 (VFMLA16_LOW): New int iterator.
5615 (VFMLA16_HIGH): Ditto.
5616 (UNSPEC_FMLAL): New unspec.
5617 (UNSPEC_FMLSL): Ditto.
5618 (UNSPEC_FMLAL2): Ditto.
5619 (UNSPEC_FMLSL2): Ditto.
5620 (f16mac): New code attribute.
5621 * config/aarch64/aarch64-simd-builtins.def
5622 (aarch64_fmlal_lowv2sf): Ditto.
5623 (aarch64_fmlsl_lowv2sf): Ditto.
5624 (aarch64_fmlalq_lowv4sf): Ditto.
5625 (aarch64_fmlslq_lowv4sf): Ditto.
5626 (aarch64_fmlal_highv2sf): Ditto.
5627 (aarch64_fmlsl_highv2sf): Ditto.
5628 (aarch64_fmlalq_highv4sf): Ditto.
5629 (aarch64_fmlslq_highv4sf): Ditto.
5630 (aarch64_fmlal_lane_lowv2sf): Ditto.
5631 (aarch64_fmlsl_lane_lowv2sf): Ditto.
5632 (aarch64_fmlal_laneq_lowv2sf): Ditto.
5633 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
5634 (aarch64_fmlalq_lane_lowv4sf): Ditto.
5635 (aarch64_fmlsl_lane_lowv4sf): Ditto.
5636 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
5637 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
5638 (aarch64_fmlal_lane_highv2sf): Ditto.
5639 (aarch64_fmlsl_lane_highv2sf): Ditto.
5640 (aarch64_fmlal_laneq_highv2sf): Ditto.
5641 (aarch64_fmlsl_laneq_highv2sf): Ditto.
5642 (aarch64_fmlalq_lane_highv4sf): Ditto.
5643 (aarch64_fmlsl_lane_highv4sf): Ditto.
5644 (aarch64_fmlalq_laneq_highv4sf): Ditto.
5645 (aarch64_fmlsl_laneq_highv4sf): Ditto.
5646 * config/aarch64/aarch64-simd.md:
5647 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
5648 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5649 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
5650 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5651 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
5652 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
5653 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
5654 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
5655 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
5656 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
5657 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
5658 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
5659 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
5660 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
5661 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
5662 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
5663 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
5664 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
5665 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
5666 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
5667 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
5668 (vfmlsl_low_u32): Ditto.
5669 (vfmlalq_low_u32): Ditto.
5670 (vfmlslq_low_u32): Ditto.
5671 (vfmlal_high_u32): Ditto.
5672 (vfmlsl_high_u32): Ditto.
5673 (vfmlalq_high_u32): Ditto.
5674 (vfmlslq_high_u32): Ditto.
5675 (vfmlal_lane_low_u32): Ditto.
5676 (vfmlsl_lane_low_u32): Ditto.
5677 (vfmlal_laneq_low_u32): Ditto.
5678 (vfmlsl_laneq_low_u32): Ditto.
5679 (vfmlalq_lane_low_u32): Ditto.
5680 (vfmlslq_lane_low_u32): Ditto.
5681 (vfmlalq_laneq_low_u32): Ditto.
5682 (vfmlslq_laneq_low_u32): Ditto.
5683 (vfmlal_lane_high_u32): Ditto.
5684 (vfmlsl_lane_high_u32): Ditto.
5685 (vfmlal_laneq_high_u32): Ditto.
5686 (vfmlsl_laneq_high_u32): Ditto.
5687 (vfmlalq_lane_high_u32): Ditto.
5688 (vfmlslq_lane_high_u32): Ditto.
5689 (vfmlalq_laneq_high_u32): Ditto.
5690 (vfmlslq_laneq_high_u32): Ditto.
5691 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
5692 (AARCH64_FL_FOR_ARCH8_4): New.
5693 (AARCH64_ISA_F16FML): New ISA flag.
5694 (TARGET_F16FML): New feature flag for fp16fml.
5695 (doc/invoke.texi): Document new fp16fml option.
5697 2018-01-10 Michael Collison <michael.collison@arm.com>
5699 * config/aarch64/aarch64-builtins.c:
5700 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
5701 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5702 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
5703 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
5704 (AARCH64_ISA_SHA3): New ISA flag.
5705 (TARGET_SHA3): New feature flag for sha3.
5706 * config/aarch64/iterators.md (sha512_op): New int attribute.
5707 (CRYPTO_SHA512): New int iterator.
5708 (UNSPEC_SHA512H): New unspec.
5709 (UNSPEC_SHA512H2): Ditto.
5710 (UNSPEC_SHA512SU0): Ditto.
5711 (UNSPEC_SHA512SU1): Ditto.
5712 * config/aarch64/aarch64-simd-builtins.def
5713 (aarch64_crypto_sha512hqv2di): New builtin.
5714 (aarch64_crypto_sha512h2qv2di): Ditto.
5715 (aarch64_crypto_sha512su0qv2di): Ditto.
5716 (aarch64_crypto_sha512su1qv2di): Ditto.
5717 (aarch64_eor3qv8hi): Ditto.
5718 (aarch64_rax1qv2di): Ditto.
5719 (aarch64_xarqv2di): Ditto.
5720 (aarch64_bcaxqv8hi): Ditto.
5721 * config/aarch64/aarch64-simd.md:
5722 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
5723 (aarch64_crypto_sha512su0qv2di): Ditto.
5724 (aarch64_crypto_sha512su1qv2di): Ditto.
5725 (aarch64_eor3qv8hi): Ditto.
5726 (aarch64_rax1qv2di): Ditto.
5727 (aarch64_xarqv2di): Ditto.
5728 (aarch64_bcaxqv8hi): Ditto.
5729 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
5730 (vsha512h2q_u64): Ditto.
5731 (vsha512su0q_u64): Ditto.
5732 (vsha512su1q_u64): Ditto.
5733 (veor3q_u16): Ditto.
5734 (vrax1q_u64): Ditto.
5736 (vbcaxq_u16): Ditto.
5737 * config/arm/types.md (crypto_sha512): New type attribute.
5738 (crypto_sha3): Ditto.
5739 (doc/invoke.texi): Document new sha3 option.
5741 2018-01-10 Michael Collison <michael.collison@arm.com>
5743 * config/aarch64/aarch64-builtins.c:
5744 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
5745 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5746 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
5747 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
5748 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
5749 (AARCH64_ISA_SM4): New ISA flag.
5750 (TARGET_SM4): New feature flag for sm4.
5751 * config/aarch64/aarch64-simd-builtins.def
5752 (aarch64_sm3ss1qv4si): Ditto.
5753 (aarch64_sm3tt1aq4si): Ditto.
5754 (aarch64_sm3tt1bq4si): Ditto.
5755 (aarch64_sm3tt2aq4si): Ditto.
5756 (aarch64_sm3tt2bq4si): Ditto.
5757 (aarch64_sm3partw1qv4si): Ditto.
5758 (aarch64_sm3partw2qv4si): Ditto.
5759 (aarch64_sm4eqv4si): Ditto.
5760 (aarch64_sm4ekeyqv4si): Ditto.
5761 * config/aarch64/aarch64-simd.md:
5762 (aarch64_sm3ss1qv4si): Ditto.
5763 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
5764 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
5765 (aarch64_sm4eqv4si): Ditto.
5766 (aarch64_sm4ekeyqv4si): Ditto.
5767 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
5768 (sm3part_op): Ditto.
5769 (CRYPTO_SM3TT): Ditto.
5770 (CRYPTO_SM3PART): Ditto.
5771 (UNSPEC_SM3SS1): New unspec.
5772 (UNSPEC_SM3TT1A): Ditto.
5773 (UNSPEC_SM3TT1B): Ditto.
5774 (UNSPEC_SM3TT2A): Ditto.
5775 (UNSPEC_SM3TT2B): Ditto.
5776 (UNSPEC_SM3PARTW1): Ditto.
5777 (UNSPEC_SM3PARTW2): Ditto.
5778 (UNSPEC_SM4E): Ditto.
5779 (UNSPEC_SM4EKEY): Ditto.
5780 * config/aarch64/constraints.md (Ui2): New constraint.
5781 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
5782 * config/arm/types.md (crypto_sm3): New type attribute.
5783 (crypto_sm4): Ditto.
5784 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
5785 (vsm3tt1aq_u32): Ditto.
5786 (vsm3tt1bq_u32): Ditto.
5787 (vsm3tt2aq_u32): Ditto.
5788 (vsm3tt2bq_u32): Ditto.
5789 (vsm3partw1q_u32): Ditto.
5790 (vsm3partw2q_u32): Ditto.
5791 (vsm4eq_u32): Ditto.
5792 (vsm4ekeyq_u32): Ditto.
5793 (doc/invoke.texi): Document new sm4 option.
5795 2018-01-10 Michael Collison <michael.collison@arm.com>
5797 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
5798 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
5799 (AARCH64_FL_FOR_ARCH8_4): New.
5800 (AARCH64_FL_V8_4): New flag.
5801 (doc/invoke.texi): Document new armv8.4-a option.
5803 2018-01-10 Michael Collison <michael.collison@arm.com>
5805 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5806 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
5807 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
5808 * config/aarch64/aarch64-option-extension.def: Add
5809 AARCH64_OPT_EXTENSION of 'sha2'.
5810 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
5811 (crypto): Disable sha2 and aes if crypto disabled.
5812 (crypto): Enable aes and sha2 if enabled.
5813 (simd): Disable sha2 and aes if simd disabled.
5814 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
5816 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
5817 (TARGET_SHA2): New feature flag for sha2.
5818 (TARGET_AES): New feature flag for aes.
5819 * config/aarch64/aarch64-simd.md:
5820 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
5821 conditional on TARGET_AES.
5822 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
5823 (aarch64_crypto_sha1hsi): Make pattern conditional
5825 (aarch64_crypto_sha1hv4si): Ditto.
5826 (aarch64_be_crypto_sha1hv4si): Ditto.
5827 (aarch64_crypto_sha1su1v4si): Ditto.
5828 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
5829 (aarch64_crypto_sha1su0v4si): Ditto.
5830 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
5831 (aarch64_crypto_sha256su0v4si): Ditto.
5832 (aarch64_crypto_sha256su1v4si): Ditto.
5833 (doc/invoke.texi): Document new aes and sha2 options.
5835 2018-01-10 Martin Sebor <msebor@redhat.com>
5837 PR tree-optimization/83781
5838 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
5841 2018-01-11 Martin Sebor <msebor@gmail.com>
5842 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5844 PR tree-optimization/83501
5845 PR tree-optimization/81703
5847 * tree-ssa-strlen.c (get_string_cst): Rename...
5848 (get_string_len): ...to this. Handle global constants.
5849 (handle_char_store): Adjust.
5851 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
5852 Jim Wilson <jimw@sifive.com>
5854 * config/riscv/riscv-protos.h (riscv_output_return): New.
5855 * config/riscv/riscv.c (struct machine_function): New naked_p field.
5856 (riscv_attribute_table, riscv_output_return),
5857 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
5858 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
5859 (riscv_compute_frame_info): Only compute frame->mask if not a naked
5861 (riscv_expand_prologue): Add early return for naked function.
5862 (riscv_expand_epilogue): Likewise.
5863 (riscv_function_ok_for_sibcall): Return false for naked function.
5864 (riscv_set_current_function): New.
5865 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
5866 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
5867 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
5868 * doc/extend.texi (RISC-V Function Attributes): New.
5870 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
5872 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
5873 check for 128-bit long double before checking TCmode.
5874 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
5875 128-bit long doubles before checking TFmode or TCmode.
5876 (FLOAT128_IBM_P): Likewise.
5878 2018-01-10 Martin Sebor <msebor@redhat.com>
5880 PR tree-optimization/83671
5881 * builtins.c (c_strlen): Unconditionally return zero for the empty
5883 Use -Warray-bounds for warnings.
5884 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
5885 for non-constant array indices with COMPONENT_REF, arrays of
5886 arrays, and pointers to arrays.
5887 (gimple_fold_builtin_strlen): Determine and set length range for
5888 non-constant character arrays.
5890 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
5893 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
5896 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
5898 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
5900 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5903 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
5904 VECTOR_MEM_ALTIVEC_OR_VSX_P.
5905 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
5906 indexed_or_indirect_operand predicate.
5907 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
5908 (*vsx_le_perm_load_v8hi): Likewise.
5909 (*vsx_le_perm_load_v16qi): Likewise.
5910 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
5911 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
5912 (*vsx_le_perm_store_v8hi): Likewise.
5913 (*vsx_le_perm_store_v16qi): Likewise.
5914 (eight unnamed splitters): Likewise.
5916 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5918 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
5919 * config/rs6000/emmintrin.h: Likewise.
5920 * config/rs6000/mmintrin.h: Likewise.
5921 * config/rs6000/xmmintrin.h: Likewise.
5923 2018-01-10 David Malcolm <dmalcolm@redhat.com>
5926 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
5928 * tree.c (tree_nop_conversion): Return true for location wrapper
5930 (maybe_wrap_with_location): New function.
5931 (selftest::check_strip_nops): New function.
5932 (selftest::test_location_wrappers): New function.
5933 (selftest::tree_c_tests): Call it.
5934 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
5935 (maybe_wrap_with_location): New decl.
5936 (EXPR_LOCATION_WRAPPER_P): New macro.
5937 (location_wrapper_p): New inline function.
5938 (tree_strip_any_location_wrapper): New inline function.
5940 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
5943 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
5944 stack_realign_offset for the largest alignment of stack slot
5946 (ix86_find_max_used_stack_alignment): New function.
5947 (ix86_finalize_stack_frame_flags): Use it. Set
5948 max_used_stack_alignment if we don't realign stack.
5949 * config/i386/i386.h (machine_function): Add
5950 max_used_stack_alignment.
5952 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
5954 * config/arm/arm.opt (-mbranch-cost): New option.
5955 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
5958 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
5961 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
5962 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
5964 2018-01-10 Richard Biener <rguenther@suse.de>
5967 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
5968 early out so it also covers the case where we have a non-NULL
5971 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
5973 PR tree-optimization/83753
5974 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
5975 for non-strided grouped accesses if the number of elements is 1.
5977 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5980 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
5981 * i386.h (TARGET_USE_GATHER): Define.
5982 * x86-tune.def (X86_TUNE_USE_GATHER): New.
5984 2018-01-10 Martin Liska <mliska@suse.cz>
5987 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
5988 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
5990 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
5991 CLEANUP_NO_PARTITIONING is not set.
5993 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
5995 * doc/rtl.texi: Remove documentation of (const ...) wrappers
5996 for vectors, as a partial revert of r254296.
5997 * rtl.h (const_vec_p): Delete.
5998 (const_vec_duplicate_p): Don't test for vector CONSTs.
5999 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
6000 * expmed.c (make_tree): Likewise.
6003 * common.md (E, F): Use CONSTANT_P instead of checking for
6005 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
6006 checking for CONST_VECTOR.
6008 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6011 * predict.c (force_edge_cold): Handle in more sane way edges
6014 2018-01-09 Carl Love <cel@us.ibm.com>
6016 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
6018 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
6019 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
6020 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
6021 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
6022 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
6023 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
6024 * config/rs6000/rs6000-protos.h: Add extern defition for
6025 rs6000_generate_float2_double_code.
6026 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
6028 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
6029 (float2_v2df): Add define_expand.
6031 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
6034 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
6035 op_mode in the force_to_mode call.
6037 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6039 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
6040 instead of checking each element individually.
6041 (aarch64_evpc_uzp): Likewise.
6042 (aarch64_evpc_zip): Likewise.
6043 (aarch64_evpc_ext): Likewise.
6044 (aarch64_evpc_rev): Likewise.
6045 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
6046 instead of checking each element individually. Return true without
6048 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
6049 whether all selected elements come from the same input, instead of
6050 checking each element individually. Remove calls to gen_rtx_REG,
6051 start_sequence and end_sequence and instead assert that no rtl is
6054 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6056 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
6057 order of HIGH and CONST checks.
6059 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6061 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
6062 if the destination isn't an SSA_NAME.
6064 2018-01-09 Richard Biener <rguenther@suse.de>
6066 PR tree-optimization/83668
6067 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
6069 (canonicalize_loop_form): ... here, renamed from ...
6070 (canonicalize_loop_closed_ssa_form): ... this and amended to
6071 swap successor edges for loop exit blocks to make us use
6072 the RPO order we need for initial schedule generation.
6074 2018-01-09 Joseph Myers <joseph@codesourcery.com>
6076 PR tree-optimization/64811
6077 * match.pd: When optimizing comparisons with Inf, avoid
6078 introducing or losing exceptions from comparisons with NaN.
6080 2018-01-09 Martin Liska <mliska@suse.cz>
6083 * asan.c (shadow_mem_size): Add gcc_assert.
6085 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
6087 Don't save registers in main().
6090 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
6091 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
6092 * config/avr/avr.c (avr_set_current_function): Don't error if
6093 naked, OS_task or OS_main are specified at the same time.
6094 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
6096 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
6098 * common/config/avr/avr-common.c (avr_option_optimization_table):
6099 Switch on -mmain-is-OS_task for optimizing compilations.
6101 2018-01-09 Richard Biener <rguenther@suse.de>
6103 PR tree-optimization/83572
6104 * graphite.c: Include cfganal.h.
6105 (graphite_transform_loops): Connect infinite loops to exit
6106 and remove fake edges at the end.
6108 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6110 * ipa-inline.c (edge_badness): Revert accidental checkin.
6112 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6115 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
6116 symbols; not inline clones.
6118 2018-01-09 Jakub Jelinek <jakub@redhat.com>
6121 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
6122 hard registers. Formatting fixes.
6124 PR preprocessor/83722
6125 * gcc.c (try_generate_repro): Pass
6126 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
6127 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
6130 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
6131 Kito Cheng <kito.cheng@gmail.com>
6133 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
6134 (riscv_leaf_function_p): Delete.
6135 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
6137 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6139 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
6141 (do_ifelse): New function.
6142 (do_isel): New function.
6143 (do_sub3): New function.
6144 (do_add3): New function.
6145 (do_load_mask_compare): New function.
6146 (do_overlap_load_compare): New function.
6147 (expand_compare_loop): New function.
6148 (expand_block_compare): Call expand_compare_loop() when appropriate.
6149 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
6151 (-mblock-compare-inline-loop-limit): New option.
6153 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6156 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
6157 Reverse order of second and third operands in first alternative.
6158 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
6159 of first and second elements in UNSPEC_VPERMR vector.
6160 (altivec_expand_vec_perm_le): Likewise.
6162 2018-01-08 Jeff Law <law@redhat.com>
6164 PR rtl-optimizatin/81308
6165 * tree-switch-conversion.c (cfg_altered): New file scoped static.
6166 (process_switch): If group_case_labels makes a change, then set
6168 (pass_convert_switch::execute): If a switch is converted, then
6169 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
6171 PR rtl-optimization/81308
6172 * recog.c (split_all_insns): Conditionally cleanup the CFG after
6175 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
6177 PR target/83663 - Revert r255946
6178 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
6179 generation for cases where splatting a value is not useful.
6180 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
6181 across a vec_duplicate and a paradoxical subreg forming a vector
6182 mode to a vec_concat.
6184 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6186 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
6187 -march=armv8.3-a variants.
6188 * config/arm/t-multilib: Likewise.
6189 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
6191 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6193 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
6195 (cceq_ior_compare_complement): Give it a name so I can use it, and
6196 change boolean_or_operator predicate to boolean_operator so it can
6197 be used to generate a crand.
6198 (eqne): New code iterator.
6199 (bd/bd_neg): New code_attrs.
6200 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
6201 a single define_insn.
6202 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
6203 decrement (bdnzt/bdnzf/bdzt/bdzf).
6204 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
6205 with the new names of the branch decrement patterns, and added the
6206 names of the branch decrement conditional patterns.
6208 2018-01-08 Richard Biener <rguenther@suse.de>
6210 PR tree-optimization/83563
6211 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
6214 2018-01-08 Richard Biener <rguenther@suse.de>
6217 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
6219 2018-01-08 Richard Biener <rguenther@suse.de>
6221 PR tree-optimization/83685
6222 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
6223 references to abnormals.
6225 2018-01-08 Richard Biener <rguenther@suse.de>
6228 * dwarf2out.c (output_indirect_strings): Handle empty
6229 skeleton_debug_str_hash.
6230 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
6232 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6234 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
6235 (emit_store_direct): Likewise.
6236 (arc_trampoline_adjust_address): Likewise.
6237 (arc_asm_trampoline_template): New function.
6238 (arc_initialize_trampoline): Use asm_trampoline_template.
6239 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
6240 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
6241 * config/arc/arc.md (flush_icache): Delete pattern.
6243 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6245 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
6246 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
6249 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6252 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
6253 by not USED_FOR_TARGET.
6254 (make_pass_resolve_sw_modes): Likewise.
6256 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6258 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
6261 2018-01-08 Richard Biener <rguenther@suse.de>
6264 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
6266 2018-01-08 Richard Biener <rguenther@suse.de>
6269 * match.pd ((t * 2) / 2) -> t): Add missing :c.
6271 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
6274 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
6275 basic blocks with a small number of successors.
6276 (convert_control_dep_chain_into_preds): Improve handling of
6278 (dump_predicates): Split apart into...
6279 (dump_pred_chain): ...here...
6280 (dump_pred_info): ...and here.
6281 (can_one_predicate_be_invalidated_p): Add debugging printfs.
6282 (can_chain_union_be_invalidated_p): Improve check for invalidation
6284 (uninit_uses_cannot_happen): Avoid unnecessary if
6285 convert_control_dep_chain_into_preds yielded nothing.
6287 2018-01-06 Martin Sebor <msebor@redhat.com>
6289 PR tree-optimization/83640
6290 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
6291 subtracting negative offset from size.
6292 (builtin_access::overlap): Adjust offset bounds of the access to fall
6293 within the size of the object if possible.
6295 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
6297 PR rtl-optimization/83699
6298 * expmed.c (extract_bit_field_1): Restrict the vector usage of
6299 extract_bit_field_as_subreg to cases in which the extracted
6300 value is also a vector.
6302 * lra-constraints.c (process_alt_operands): Test for the equivalence
6303 substitutions when detecting a possible reload cycle.
6305 2018-01-06 Jakub Jelinek <jakub@redhat.com>
6308 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
6309 by default if flag_selective_schedling{,2}. Formatting fixes.
6311 PR rtl-optimization/83682
6312 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
6313 if it has non-VECTOR_MODE element mode.
6314 (vec_duplicate_p): Likewise.
6317 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
6318 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
6320 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6323 * config/i386/i386-builtin.def
6324 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
6325 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
6326 Require also OPTION_MASK_ISA_AVX512F in addition to
6327 OPTION_MASK_ISA_GFNI.
6328 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
6329 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
6330 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
6331 to OPTION_MASK_ISA_GFNI.
6332 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
6333 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
6334 OPTION_MASK_ISA_AVX512BW.
6335 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
6336 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
6337 addition to OPTION_MASK_ISA_GFNI.
6338 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
6339 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
6340 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
6341 to OPTION_MASK_ISA_GFNI.
6342 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
6343 a requirement for all ISAs rather than any of them with a few
6345 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
6347 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
6348 bitmasks to be enabled with 3 exceptions, instead of requiring any
6349 enabled ISA with lots of exceptions.
6350 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
6351 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
6352 Change avx512bw in isa attribute to avx512f.
6353 * config/i386/sgxintrin.h: Add license boilerplate.
6354 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
6355 to __AVX512F__ and __AVX512VL to __AVX512VL__.
6356 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
6357 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
6359 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
6360 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
6361 temporarily sse2 rather than sse if not enabled already.
6364 * config/i386/sse.md (VI248_VLBW): Rename to ...
6365 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
6366 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
6367 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
6368 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
6369 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
6370 mode iterator instead of VI248_VLBW.
6372 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
6374 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
6375 (record_modified): Skip clobbers; add debug output.
6376 (param_change_prob): Use sreal frequencies.
6378 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6380 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
6381 punt for user-aligned variables.
6383 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6385 * tree-chrec.c (chrec_contains_symbols): Return true for
6388 2018-01-05 Sudakshina Das <sudi.das@arm.com>
6391 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
6392 of (x|y) == x for BICS pattern.
6394 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6396 PR tree-optimization/83605
6397 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
6398 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
6401 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
6403 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
6404 * config/epiphany/rtems.h: New file.
6406 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6407 Uros Bizjak <ubizjak@gmail.com>
6410 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
6411 QIreg_operand instead of register_operand predicate.
6412 * config/i386/i386.c (ix86_rop_should_change_byte_p,
6413 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
6414 comments instead of -fmitigate[-_]rop.
6416 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6419 * cgraphunit.c (symbol_table::compile): Switch to text_section
6420 before calling assembly_start debug hook.
6421 * run-rtl-passes.c (run_rtl_passes): Likewise.
6424 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6426 * tree-vrp.c (extract_range_from_binary_expr_1): Check
6427 range_int_cst_p rather than !symbolic_range_p before calling
6428 extract_range_from_multiplicative_op_1.
6430 2018-01-04 Jeff Law <law@redhat.com>
6432 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
6433 redundant test in assertion.
6435 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6437 * doc/rtl.texi: Document machine_mode wrapper classes.
6439 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6441 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
6444 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6446 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
6447 the VEC_PERM_EXPR fold to fail.
6449 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6452 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
6453 to switched_sections.
6455 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6458 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
6461 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
6464 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
6465 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
6467 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6470 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
6471 is BLKmode and bitpos not zero or mode change is needed.
6473 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6476 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
6479 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
6482 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
6483 instead of MULT rtx. Update all corresponding splitters.
6485 (*ssub<modesuffix>): Ditto.
6487 (*cmp_sadd_di): Update split patterns.
6488 (*cmp_sadd_si): Ditto.
6489 (*cmp_sadd_sidi): Ditto.
6490 (*cmp_ssub_di): Ditto.
6491 (*cmp_ssub_si): Ditto.
6492 (*cmp_ssub_sidi): Ditto.
6493 * config/alpha/predicates.md (const23_operand): New predicate.
6494 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
6495 Look for ASHIFT, not MULT inner operand.
6496 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
6498 2018-01-04 Martin Liska <mliska@suse.cz>
6500 PR gcov-profile/83669
6501 * gcov.c (output_intermediate_file): Add version to intermediate
6503 * doc/gcov.texi: Document new field 'version' in intermediate
6504 file format. Fix location of '-k' option of gcov command.
6506 2018-01-04 Martin Liska <mliska@suse.cz>
6509 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
6511 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6513 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
6515 2018-01-03 Martin Sebor <msebor@redhat.com>
6517 PR tree-optimization/83655
6518 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
6519 checking calls with invalid arguments.
6521 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6523 * tree-vect-stmts.c (vect_get_store_rhs): New function.
6524 (vectorizable_mask_load_store): Delete.
6525 (vectorizable_call): Return false for masked loads and stores.
6526 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
6527 instead of gimple_assign_rhs1.
6528 (vectorizable_load): Handle IFN_MASK_LOAD.
6529 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
6531 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6533 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
6535 (vectorizable_mask_load_store): ...here.
6536 (vectorizable_load): ...and here.
6538 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6540 * tree-vect-stmts.c (vect_build_all_ones_mask)
6541 (vect_build_zero_merge_argument): New functions, split out from...
6542 (vectorizable_load): ...here.
6544 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6546 * tree-vect-stmts.c (vect_check_store_rhs): New function,
6548 (vectorizable_mask_load_store): ...here.
6549 (vectorizable_store): ...and here.
6551 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6553 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
6555 (vectorizable_mask_load_store): ...here.
6557 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6559 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
6560 (vect_model_store_cost): Take a vec_load_store_type instead of a
6562 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
6563 (vect_model_store_cost): Take a vec_load_store_type instead of a
6565 (vectorizable_mask_load_store): Update accordingly.
6566 (vectorizable_store): Likewise.
6567 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
6569 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6571 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
6572 IFN_MASK_LOAD calls here rather than...
6573 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
6575 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6576 Alan Hayward <alan.hayward@arm.com>
6577 David Sherwood <david.sherwood@arm.com>
6579 * expmed.c (extract_bit_field_1): For vector extracts,
6580 fall back to extract_bit_field_as_subreg if vec_extract
6583 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6584 Alan Hayward <alan.hayward@arm.com>
6585 David Sherwood <david.sherwood@arm.com>
6587 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
6588 they are variable or constant sized.
6589 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
6590 slots for constant-sized data.
6592 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6593 Alan Hayward <alan.hayward@arm.com>
6594 David Sherwood <david.sherwood@arm.com>
6596 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
6597 handling COND_EXPRs with boolean comparisons, try to find a better
6598 basis for the mask type than the boolean itself.
6600 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6602 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
6603 is calculated and how it can be overridden.
6604 * genmodes.c (max_bitsize_mode_any_mode): New variable.
6605 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
6607 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
6610 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6611 Alan Hayward <alan.hayward@arm.com>
6612 David Sherwood <david.sherwood@arm.com>
6614 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
6615 Remove the mode argument.
6616 (aarch64_simd_valid_immediate): Remove the mode and inverse
6618 * config/aarch64/iterators.md (bitsize): New iterator.
6619 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
6620 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
6621 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
6622 aarch64_simd_valid_immediate.
6623 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
6624 (aarch64_reg_or_bic_imm): Likewise.
6625 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
6626 with an insn_type enum and msl with a modifier_type enum.
6627 Replace element_width with a scalar_mode. Change the shift
6628 to unsigned int. Add constructors for scalar_float_mode and
6629 scalar_int_mode elements.
6630 (aarch64_vect_float_const_representable_p): Delete.
6631 (aarch64_can_const_movi_rtx_p)
6632 (aarch64_simd_scalar_immediate_valid_for_move)
6633 (aarch64_simd_make_constant): Update call to
6634 aarch64_simd_valid_immediate.
6635 (aarch64_advsimd_valid_immediate_hs): New function.
6636 (aarch64_advsimd_valid_immediate): Likewise.
6637 (aarch64_simd_valid_immediate): Remove mode and inverse
6638 arguments. Rewrite to use the above. Use const_vec_duplicate_p
6639 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
6640 and aarch64_float_const_representable_p on the result.
6641 (aarch64_output_simd_mov_immediate): Remove mode argument.
6642 Update call to aarch64_simd_valid_immediate and use of
6643 simd_immediate_info.
6644 (aarch64_output_scalar_simd_mov_immediate): Update call
6647 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6648 Alan Hayward <alan.hayward@arm.com>
6649 David Sherwood <david.sherwood@arm.com>
6651 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
6652 (mode_nunits): Likewise CONST_MODE_NUNITS.
6653 * machmode.def (ADJUST_NUNITS): Document.
6654 * genmodes.c (mode_data::need_nunits_adj): New field.
6655 (blank_mode): Update accordingly.
6656 (adj_nunits): New variable.
6657 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
6659 (emit_mode_size_inline): Set need_bytesize_adj for all modes
6660 listed in adj_nunits.
6661 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
6662 listed in adj_nunits. Don't emit case statements for such modes.
6663 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
6664 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
6665 nothing if adj_nunits is nonnull.
6666 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
6667 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
6668 (emit_mode_fbit): Update use of print_maybe_const_decl.
6669 (emit_move_size): Likewise. Treat the array as non-const
6671 (emit_mode_adjustments): Handle adj_nunits.
6673 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6675 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
6676 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
6677 (VECTOR_MODES): Use it.
6678 (make_vector_modes): Take the prefix as an argument.
6680 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6681 Alan Hayward <alan.hayward@arm.com>
6682 David Sherwood <david.sherwood@arm.com>
6684 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
6685 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
6686 for MODE_VECTOR_BOOL.
6687 * machmode.def (VECTOR_BOOL_MODE): Document.
6688 * genmodes.c (VECTOR_BOOL_MODE): New macro.
6689 (make_vector_bool_mode): New function.
6690 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
6692 * lto-streamer-in.c (lto_input_mode_table): Likewise.
6693 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
6695 * stor-layout.c (int_mode_for_mode): Likewise.
6696 * tree.c (build_vector_type_for_mode): Likewise.
6697 * varasm.c (output_constant_pool_2): Likewise.
6698 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
6699 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
6700 for MODE_VECTOR_BOOL.
6701 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
6702 of mode class checks.
6703 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
6704 instead of a list of mode class checks.
6705 (expand_vector_scalar_condition): Likewise.
6706 (type_for_widest_vector_mode): Handle BImode as an inner mode.
6708 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6709 Alan Hayward <alan.hayward@arm.com>
6710 David Sherwood <david.sherwood@arm.com>
6712 * machmode.h (mode_size): Change from unsigned short to
6714 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
6715 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6716 or if measurement_type is not polynomial.
6717 (fixed_size_mode::includes_p): Check for constant-sized modes.
6718 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
6719 return a poly_uint16 rather than an unsigned short.
6720 (emit_mode_size): Change the type of mode_size from unsigned short
6721 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
6722 (emit_mode_adjustments): Cope with polynomial vector sizes.
6723 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6725 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6727 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
6728 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
6729 * caller-save.c (setup_save_areas): Likewise.
6730 (replace_reg_with_saved_mem): Likewise.
6731 * calls.c (emit_library_call_value_1): Likewise.
6732 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
6733 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
6734 (gen_lowpart_for_combine): Likewise.
6735 * convert.c (convert_to_integer_1): Likewise.
6736 * cse.c (equiv_constant, cse_insn): Likewise.
6737 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
6738 (cselib_subst_to_values): Likewise.
6739 * dce.c (word_dce_process_block): Likewise.
6740 * df-problems.c (df_word_lr_mark_ref): Likewise.
6741 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
6742 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
6743 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
6744 (rtl_for_decl_location): Likewise.
6745 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
6746 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
6747 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
6748 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
6749 (expand_expr_real_1): Likewise.
6750 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
6751 (pad_below): Likewise.
6752 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6753 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
6754 * ira.c (get_subreg_tracking_sizes): Likewise.
6755 * ira-build.c (ira_create_allocno_objects): Likewise.
6756 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
6757 (ira_sort_regnos_for_alter_reg): Likewise.
6758 * ira-costs.c (record_operand_costs): Likewise.
6759 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
6760 (resolve_simple_move): Likewise.
6761 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
6762 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
6763 (lra_constraints): Likewise.
6764 (CONST_POOL_OK_P): Reject variable-sized modes.
6765 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
6766 (add_pseudo_to_slot, lra_spill): Likewise.
6767 * omp-low.c (omp_clause_aligned_alignment): Likewise.
6768 * optabs-query.c (get_best_extraction_insn): Likewise.
6769 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6770 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
6771 (expand_mult_highpart, valid_multiword_target_p): Likewise.
6772 * recog.c (offsettable_address_addr_space_p): Likewise.
6773 * regcprop.c (maybe_mode_change): Likewise.
6774 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
6775 * regrename.c (build_def_use): Likewise.
6776 * regstat.c (dump_reg_info): Likewise.
6777 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
6778 (find_reloads, find_reloads_subreg_address): Likewise.
6779 * reload1.c (eliminate_regs_1): Likewise.
6780 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
6781 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
6782 (simplify_binary_operation_1, simplify_subreg): Likewise.
6783 * targhooks.c (default_function_arg_padding): Likewise.
6784 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
6785 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
6786 (verify_gimple_assign_ternary): Likewise.
6787 * tree-inline.c (estimate_move_cost): Likewise.
6788 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6789 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
6790 (get_address_cost_ainc): Likewise.
6791 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
6792 (vect_supportable_dr_alignment): Likewise.
6793 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6794 (vectorizable_reduction): Likewise.
6795 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
6796 (vectorizable_operation, vectorizable_load): Likewise.
6797 * tree.c (build_same_sized_truth_vector_type): Likewise.
6798 * valtrack.c (cleanup_auto_inc_dec): Likewise.
6799 * var-tracking.c (emit_note_insn_var_location): Likewise.
6800 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
6801 (ADDR_VEC_ALIGN): Likewise.
6803 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6804 Alan Hayward <alan.hayward@arm.com>
6805 David Sherwood <david.sherwood@arm.com>
6807 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
6809 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6810 or if measurement_type is polynomial.
6811 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
6812 * combine.c (make_extraction): Likewise.
6813 * dse.c (find_shift_sequence): Likewise.
6814 * dwarf2out.c (mem_loc_descriptor): Likewise.
6815 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
6816 (extract_bit_field, extract_low_bits): Likewise.
6817 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
6818 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
6819 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
6820 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
6821 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6822 * reload.c (find_reloads): Likewise.
6823 * reload1.c (alter_reg): Likewise.
6824 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
6825 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
6826 * tree-if-conv.c (predicate_mem_writes): Likewise.
6827 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
6828 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
6829 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
6830 * valtrack.c (dead_debug_insert_temp): Likewise.
6831 * varasm.c (mergeable_constant_section): Likewise.
6832 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
6834 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6835 Alan Hayward <alan.hayward@arm.com>
6836 David Sherwood <david.sherwood@arm.com>
6838 * expr.c (expand_assignment): Cope with polynomial mode sizes
6839 when assigning to a CONCAT.
6841 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6842 Alan Hayward <alan.hayward@arm.com>
6843 David Sherwood <david.sherwood@arm.com>
6845 * machmode.h (mode_precision): Change from unsigned short to
6847 (mode_to_precision): Return a poly_uint16 rather than an unsigned
6849 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
6850 or if measurement_type is not polynomial.
6851 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
6852 in which the mode is already known to be a scalar_int_mode.
6853 * genmodes.c (emit_mode_precision): Change the type of mode_precision
6854 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
6856 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6857 for GET_MODE_PRECISION.
6858 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6859 for GET_MODE_PRECISION.
6860 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
6862 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
6863 (expand_field_assignment, make_extraction): Likewise.
6864 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
6865 (get_last_value): Likewise.
6866 * convert.c (convert_to_integer_1): Likewise.
6867 * cse.c (cse_insn): Likewise.
6868 * expr.c (expand_expr_real_1): Likewise.
6869 * lra-constraints.c (simplify_operand_subreg): Likewise.
6870 * optabs-query.c (can_atomic_load_p): Likewise.
6871 * optabs.c (expand_atomic_load): Likewise.
6872 (expand_atomic_store): Likewise.
6873 * ree.c (combine_reaching_defs): Likewise.
6874 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
6875 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
6876 * tree.h (type_has_mode_precision_p): Likewise.
6877 * ubsan.c (instrument_si_overflow): Likewise.
6879 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6880 Alan Hayward <alan.hayward@arm.com>
6881 David Sherwood <david.sherwood@arm.com>
6883 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
6884 polynomial numbers of units.
6885 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
6886 (valid_vector_subparts_p): New function.
6887 (build_vector_type): Remove temporary shim and take the number
6888 of units as a poly_uint64 rather than an int.
6889 (build_opaque_vector_type): Take the number of units as a
6890 poly_uint64 rather than an int.
6891 * tree.c (build_vector_from_ctor): Handle polynomial
6892 TYPE_VECTOR_SUBPARTS.
6893 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
6894 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
6895 (build_vector_from_val): If the number of units is variable,
6896 use build_vec_duplicate_cst for constant operands and
6897 VEC_DUPLICATE_EXPR otherwise.
6898 (make_vector_type): Remove temporary is_constant ().
6899 (build_vector_type, build_opaque_vector_type): Take the number of
6900 units as a poly_uint64 rather than an int.
6901 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
6903 * cfgexpand.c (expand_debug_expr): Likewise.
6904 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
6905 (store_constructor, expand_expr_real_1): Likewise.
6906 (const_scalar_mask_from_tree): Likewise.
6907 * fold-const-call.c (fold_const_reduction): Likewise.
6908 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
6909 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
6910 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
6911 (fold_relational_const): Likewise.
6912 (native_interpret_vector): Likewise. Change the size from an
6913 int to an unsigned int.
6914 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
6915 TYPE_VECTOR_SUBPARTS.
6916 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
6917 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
6918 duplicating a non-constant operand into a variable-length vector.
6919 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
6920 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
6921 * ipa-icf.c (sem_variable::equals): Likewise.
6922 * match.pd: Likewise.
6923 * omp-simd-clone.c (simd_clone_subparts): Likewise.
6924 * print-tree.c (print_node): Likewise.
6925 * stor-layout.c (layout_type): Likewise.
6926 * targhooks.c (default_builtin_vectorization_cost): Likewise.
6927 * tree-cfg.c (verify_gimple_comparison): Likewise.
6928 (verify_gimple_assign_binary): Likewise.
6929 (verify_gimple_assign_ternary): Likewise.
6930 (verify_gimple_assign_single): Likewise.
6931 * tree-pretty-print.c (dump_generic_node): Likewise.
6932 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6933 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
6934 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
6935 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
6936 (vect_shift_permute_load_chain): Likewise.
6937 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
6938 (expand_vector_condition, optimize_vector_constructor): Likewise.
6939 (lower_vec_perm, get_compute_type): Likewise.
6940 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6941 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
6942 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
6943 (vect_recog_mask_conversion_pattern): Likewise.
6944 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
6945 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
6946 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
6947 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
6948 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
6949 (vectorizable_shift, vectorizable_operation, vectorizable_store)
6950 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
6951 (supportable_widening_operation): Likewise.
6952 (supportable_narrowing_operation): Likewise.
6953 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
6955 * varasm.c (output_constant): Likewise.
6957 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6958 Alan Hayward <alan.hayward@arm.com>
6959 David Sherwood <david.sherwood@arm.com>
6961 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
6962 so that both the length == 3 and length != 3 cases set up their
6963 own permute vectors. Add comments explaining why we know the
6964 number of elements is constant.
6965 (vect_permute_load_chain): Likewise.
6967 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6968 Alan Hayward <alan.hayward@arm.com>
6969 David Sherwood <david.sherwood@arm.com>
6971 * machmode.h (mode_nunits): Change from unsigned char to
6973 (ONLY_FIXED_SIZE_MODES): New macro.
6974 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
6975 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
6976 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
6978 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
6979 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
6980 or if measurement_type is not polynomial.
6981 * genmodes.c (ZERO_COEFFS): New macro.
6982 (emit_mode_nunits_inline): Make mode_nunits_inline return a
6984 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
6985 Use ZERO_COEFFS when emitting initializers.
6986 * data-streamer.h (bp_pack_poly_value): New function.
6987 (bp_unpack_poly_value): Likewise.
6988 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6989 for GET_MODE_NUNITS.
6990 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6991 for GET_MODE_NUNITS.
6992 * tree.c (make_vector_type): Remove temporary shim and make
6993 the real function take the number of units as a poly_uint64
6995 (build_vector_type_for_mode): Handle polynomial nunits.
6996 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
6997 * emit-rtl.c (const_vec_series_p_1): Likewise.
6998 (gen_rtx_CONST_VECTOR): Likewise.
6999 * fold-const.c (test_vec_duplicate_folding): Likewise.
7000 * genrecog.c (validate_pattern): Likewise.
7001 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
7002 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7003 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
7004 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
7005 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
7006 * rtlanal.c (subreg_get_info): Likewise.
7007 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7008 (vect_grouped_load_supported): Likewise.
7009 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
7010 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
7011 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
7012 (simplify_const_unary_operation, simplify_binary_operation_1)
7013 (simplify_const_binary_operation, simplify_ternary_operation)
7014 (test_vector_ops_duplicate, test_vector_ops): Likewise.
7015 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
7016 instead of CONST_VECTOR_NUNITS.
7017 * varasm.c (output_constant_pool_2): Likewise.
7018 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
7019 explicit-encoded elements in the XVEC for variable-length vectors.
7021 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7023 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
7025 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7026 Alan Hayward <alan.hayward@arm.com>
7027 David Sherwood <david.sherwood@arm.com>
7029 * coretypes.h (fixed_size_mode): Declare.
7030 (fixed_size_mode_pod): New typedef.
7031 * builtins.h (target_builtins::x_apply_args_mode)
7032 (target_builtins::x_apply_result_mode): Change type to
7033 fixed_size_mode_pod.
7034 * builtins.c (apply_args_size, apply_result_size, result_vector)
7035 (expand_builtin_apply_args_1, expand_builtin_apply)
7036 (expand_builtin_return): Update accordingly.
7038 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7040 * cse.c (hash_rtx_cb): Hash only the encoded elements.
7041 * cselib.c (cselib_hash_rtx): Likewise.
7042 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
7043 CONST_VECTOR encoding.
7045 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7046 Jeff Law <law@redhat.com>
7049 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
7050 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
7051 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
7052 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
7055 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
7056 explicitly probe *sp in a noreturn function if there were any callee
7057 register saves or frame pointer is needed.
7059 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7062 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
7063 BLKmode for ternary, binary or unary expressions.
7066 * var-tracking.c (delete_vta_debug_insn): New inline function.
7067 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
7068 insns from get_insns () to NULL instead of each bb separately.
7069 Use delete_vta_debug_insn. No longer static.
7070 (vt_debug_insns_local, variable_tracking_main_1): Adjust
7071 delete_vta_debug_insns callers.
7072 * rtl.h (delete_vta_debug_insns): Declare.
7073 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
7074 instead of variable_tracking_main.
7076 2018-01-03 Martin Sebor <msebor@redhat.com>
7078 PR tree-optimization/83603
7079 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
7080 arguments past the endof the argument list in functions declared
7081 without a prototype.
7082 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7083 Avoid checking when arguments are null.
7085 2018-01-03 Martin Sebor <msebor@redhat.com>
7088 * doc/extend.texi (attribute const): Fix a typo.
7089 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
7090 issuing -Wsuggest-attribute for void functions.
7092 2018-01-03 Martin Sebor <msebor@redhat.com>
7094 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
7095 offset_int::from instead of wide_int::to_shwi.
7096 (maybe_diag_overlap): Remove assertion.
7097 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
7098 * gimple-ssa-sprintf.c (format_directive): Same.
7099 (parse_directive): Same.
7100 (sprintf_dom_walker::compute_format_length): Same.
7101 (try_substitute_return_value): Same.
7103 2018-01-03 Jeff Law <law@redhat.com>
7106 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
7107 non-constant residual for zero at runtime and avoid probing in
7108 that case. Reorganize code for trailing problem to mirror handling
7111 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7113 PR tree-optimization/83501
7114 * tree-ssa-strlen.c (get_string_cst): New.
7115 (handle_char_store): Call get_string_cst.
7117 2018-01-03 Martin Liska <mliska@suse.cz>
7119 PR tree-optimization/83593
7120 * tree-ssa-strlen.c: Include tree-cfg.h.
7121 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
7122 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
7123 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
7125 (strlen_dom_walker::before_dom_children): Call
7126 gimple_purge_dead_eh_edges. Dump tranformation with details
7128 (strlen_dom_walker::before_dom_children): Update call by adding
7129 new argument cleanup_eh.
7130 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
7132 2018-01-03 Martin Liska <mliska@suse.cz>
7135 * cif-code.def (VARIADIC_THUNK): New enum value.
7136 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
7139 2018-01-03 Jan Beulich <jbeulich@suse.com>
7141 * sse.md (mov<mode>_internal): Tighten condition for when to use
7142 vmovdqu<ssescalarsize> for TI and OI modes.
7144 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7146 Update copyright years.
7148 2018-01-03 Martin Liska <mliska@suse.cz>
7151 * ipa-visibility.c (function_and_variable_visibility): Skip
7152 functions with noipa attribure.
7154 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7156 * gcc.c (process_command): Update copyright notice dates.
7157 * gcov-dump.c (print_version): Ditto.
7158 * gcov.c (print_version): Ditto.
7159 * gcov-tool.c (print_version): Ditto.
7160 * gengtype.c (create_file): Ditto.
7161 * doc/cpp.texi: Bump @copying's copyright year.
7162 * doc/cppinternals.texi: Ditto.
7163 * doc/gcc.texi: Ditto.
7164 * doc/gccint.texi: Ditto.
7165 * doc/gcov.texi: Ditto.
7166 * doc/install.texi: Ditto.
7167 * doc/invoke.texi: Ditto.
7169 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7171 * vector-builder.h (vector_builder::m_full_nelts): Change from
7172 unsigned int to poly_uint64.
7173 (vector_builder::full_nelts): Update prototype accordingly.
7174 (vector_builder::new_vector): Likewise.
7175 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
7176 (vector_builder::operator ==): Likewise.
7177 (vector_builder::finalize): Likewise.
7178 * int-vector-builder.h (int_vector_builder::int_vector_builder):
7179 Take the number of elements as a poly_uint64 rather than an
7181 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
7182 from unsigned int to poly_uint64.
7183 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
7184 (vec_perm_indices::new_vector): Likewise.
7185 (vec_perm_indices::length): Likewise.
7186 (vec_perm_indices::nelts_per_input): Likewise.
7187 (vec_perm_indices::input_nelts): Likewise.
7188 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
7189 number of elements per input as a poly_uint64 rather than an
7190 unsigned int. Use the original encoding for variable-length
7191 vectors, rather than clamping each individual element.
7192 For the second and subsequent elements in each pattern,
7193 clamp the step and base before clamping their sum.
7194 (vec_perm_indices::series_p): Handle polynomial element counts.
7195 (vec_perm_indices::all_in_range_p): Likewise.
7196 (vec_perm_indices_to_tree): Likewise.
7197 (vec_perm_indices_to_rtx): Likewise.
7198 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
7199 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
7200 (tree_vector_builder::new_binary_operation): Handle polynomial
7201 element counts. Return false if we need to know the number
7202 of elements at compile time.
7203 * fold-const.c (fold_vec_perm): Punt if the number of elements
7204 isn't known at compile time.
7206 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7208 * vec-perm-indices.h (vec_perm_builder): Change element type
7209 from HOST_WIDE_INT to poly_int64.
7210 (vec_perm_indices::element_type): Update accordingly.
7211 (vec_perm_indices::clamp): Handle polynomial element_types.
7212 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7213 (vec_perm_indices::all_in_range_p): Likewise.
7214 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
7216 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
7217 polynomial vec_perm_indices element types.
7218 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
7219 * fold-const.c (fold_vec_perm): Likewise.
7220 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
7221 * tree-vect-generic.c (lower_vec_perm): Likewise.
7222 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7223 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
7224 element type to HOST_WIDE_INT.
7226 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7227 Alan Hayward <alan.hayward@arm.com>
7228 David Sherwood <david.sherwood@arm.com>
7230 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
7231 rather than an int. Use plus_constant.
7232 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
7233 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
7235 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7236 Alan Hayward <alan.hayward@arm.com>
7237 David Sherwood <david.sherwood@arm.com>
7239 * calls.c (emit_call_1, expand_call): Change struct_value_size from
7240 a HOST_WIDE_INT to a poly_int64.
7242 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7243 Alan Hayward <alan.hayward@arm.com>
7244 David Sherwood <david.sherwood@arm.com>
7246 * calls.c (load_register_parameters): Cope with polynomial
7247 mode sizes. Require a constant size for BLKmode parameters
7248 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
7249 forces a parameter to be padded at the lsb end in order to
7250 fill a complete number of words, require the parameter size
7251 to be ordered wrt UNITS_PER_WORD.
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 * reload1.c (spill_stack_slot_width): Change element type
7258 from unsigned int to poly_uint64_pod.
7259 (alter_reg): Treat mode sizes as polynomial.
7261 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7262 Alan Hayward <alan.hayward@arm.com>
7263 David Sherwood <david.sherwood@arm.com>
7265 * reload.c (complex_word_subreg_p): New function.
7266 (reload_inner_reg_of_subreg, push_reload): Use it.
7268 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7269 Alan Hayward <alan.hayward@arm.com>
7270 David Sherwood <david.sherwood@arm.com>
7272 * lra-constraints.c (process_alt_operands): Reject matched
7273 operands whose sizes aren't ordered.
7274 (match_reload): Refer to this check here.
7276 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7277 Alan Hayward <alan.hayward@arm.com>
7278 David Sherwood <david.sherwood@arm.com>
7280 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
7281 that the mode size is in the set {1, 2, 4, 8, 16}.
7283 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7284 Alan Hayward <alan.hayward@arm.com>
7285 David Sherwood <david.sherwood@arm.com>
7287 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
7288 Use plus_constant instead of gen_rtx_PLUS.
7290 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7291 Alan Hayward <alan.hayward@arm.com>
7292 David Sherwood <david.sherwood@arm.com>
7294 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
7295 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
7296 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
7297 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
7298 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
7299 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
7300 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
7301 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
7302 * config/i386/i386.c (ix86_push_rounding): ...this new function.
7303 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
7305 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
7306 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
7307 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
7308 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
7309 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
7310 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
7311 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
7312 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
7313 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
7314 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
7316 * expr.c (emit_move_resolve_push): Treat the input and result
7317 of PUSH_ROUNDING as a poly_int64.
7318 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
7319 (emit_push_insn): Likewise.
7320 * lra-eliminations.c (mark_not_eliminable): Likewise.
7321 * recog.c (push_operand): Likewise.
7322 * reload1.c (elimination_effects): Likewise.
7323 * rtlanal.c (nonzero_bits1): Likewise.
7324 * calls.c (store_one_arg): Likewise. Require the padding to be
7325 known at compile time.
7327 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7328 Alan Hayward <alan.hayward@arm.com>
7329 David Sherwood <david.sherwood@arm.com>
7331 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
7332 Use plus_constant instead of gen_rtx_PLUS.
7334 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7335 Alan Hayward <alan.hayward@arm.com>
7336 David Sherwood <david.sherwood@arm.com>
7338 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
7341 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7342 Alan Hayward <alan.hayward@arm.com>
7343 David Sherwood <david.sherwood@arm.com>
7345 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
7346 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
7347 via stack temporaries. Treat the mode size as polynomial too.
7349 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7350 Alan Hayward <alan.hayward@arm.com>
7351 David Sherwood <david.sherwood@arm.com>
7353 * expr.c (expand_expr_real_2): When handling conversions involving
7354 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
7355 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
7356 as a poly_uint64 too.
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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
7364 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7365 Alan Hayward <alan.hayward@arm.com>
7366 David Sherwood <david.sherwood@arm.com>
7368 * combine.c (can_change_dest_mode): Handle polynomial
7369 REGMODE_NATURAL_SIZE.
7370 * expmed.c (store_bit_field_1): Likewise.
7371 * expr.c (store_constructor): Likewise.
7372 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
7373 and polynomial REGMODE_NATURAL_SIZE.
7374 (gen_lowpart_common): Likewise.
7375 * reginfo.c (record_subregs_of_mode): Likewise.
7376 * rtlanal.c (read_modify_subreg_p): Likewise.
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 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
7383 numbers of elements.
7385 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7386 Alan Hayward <alan.hayward@arm.com>
7387 David Sherwood <david.sherwood@arm.com>
7389 * match.pd: Cope with polynomial numbers of vector elements.
7391 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7392 Alan Hayward <alan.hayward@arm.com>
7393 David Sherwood <david.sherwood@arm.com>
7395 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
7396 in a POINTER_PLUS_EXPR.
7398 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7399 Alan Hayward <alan.hayward@arm.com>
7400 David Sherwood <david.sherwood@arm.com>
7402 * omp-simd-clone.c (simd_clone_subparts): New function.
7403 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
7404 (ipa_simd_modify_function_body): Likewise.
7406 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7407 Alan Hayward <alan.hayward@arm.com>
7408 David Sherwood <david.sherwood@arm.com>
7410 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
7411 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
7412 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
7413 (expand_vector_condition, vector_element): Likewise.
7414 (subparts_gt): New function.
7415 (get_compute_type): Use subparts_gt.
7416 (count_type_subparts): Delete.
7417 (expand_vector_operations_1): Use subparts_gt instead of
7418 count_type_subparts.
7420 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7421 Alan Hayward <alan.hayward@arm.com>
7422 David Sherwood <david.sherwood@arm.com>
7424 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
7425 (vect_compile_time_alias): ...this new function. Do the calculation
7426 on poly_ints rather than trees.
7427 (vect_prune_runtime_alias_test_list): Update call accordingly.
7429 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7430 Alan Hayward <alan.hayward@arm.com>
7431 David Sherwood <david.sherwood@arm.com>
7433 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
7435 (vect_schedule_slp_instance): Likewise.
7437 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7438 Alan Hayward <alan.hayward@arm.com>
7439 David Sherwood <david.sherwood@arm.com>
7441 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
7442 constant and extern definitions for variable-length vectors.
7443 (vect_get_constant_vectors): Note that the number of units
7444 is known to be constant.
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 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
7451 of units as polynomial. Choose between WIDE and NARROW based
7454 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7455 Alan Hayward <alan.hayward@arm.com>
7456 David Sherwood <david.sherwood@arm.com>
7458 * tree-vect-stmts.c (simd_clone_subparts): New function.
7459 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
7461 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7462 Alan Hayward <alan.hayward@arm.com>
7463 David Sherwood <david.sherwood@arm.com>
7465 * tree-vect-stmts.c (vectorizable_call): Treat the number of
7466 vectors as polynomial. Use build_index_vector for
7469 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7470 Alan Hayward <alan.hayward@arm.com>
7471 David Sherwood <david.sherwood@arm.com>
7473 * tree-vect-stmts.c (get_load_store_type): Treat the number of
7474 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
7475 for variable-length vectors.
7476 (vectorizable_mask_load_store): Treat the number of units as
7477 polynomial, asserting that it is constant if the condition has
7478 already been enforced.
7479 (vectorizable_store, vectorizable_load): 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-loop.c (vectorizable_live_operation): Treat the number
7486 of units as polynomial. Punt if we can't tell at compile time
7487 which vector contains the final result.
7489 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7490 Alan Hayward <alan.hayward@arm.com>
7491 David Sherwood <david.sherwood@arm.com>
7493 * tree-vect-loop.c (vectorizable_induction): Treat the number
7494 of units as polynomial. Punt on SLP inductions. Use an integer
7495 VEC_SERIES_EXPR for variable-length integer reductions. Use a
7496 cast of such a series for variable-length floating-point
7499 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7500 Alan Hayward <alan.hayward@arm.com>
7501 David Sherwood <david.sherwood@arm.com>
7503 * tree.h (build_index_vector): Declare.
7504 * tree.c (build_index_vector): New function.
7505 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
7506 of units as polynomial, forcibly converting it to a constant if
7507 vectorizable_reduction has already enforced the condition.
7508 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
7509 to create a {1,2,3,...} vector.
7510 (vectorizable_reduction): Treat the number of units as polynomial.
7511 Choose vectype_in based on the largest scalar element size rather
7512 than the smallest number of units. Enforce the restrictions
7515 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7516 Alan Hayward <alan.hayward@arm.com>
7517 David Sherwood <david.sherwood@arm.com>
7519 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
7520 number of units as polynomial.
7522 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7523 Alan Hayward <alan.hayward@arm.com>
7524 David Sherwood <david.sherwood@arm.com>
7526 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
7527 * target.def (autovectorize_vector_sizes): Return the vector sizes
7528 by pointer, using vector_sizes rather than a bitmask.
7529 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
7530 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
7531 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
7533 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
7534 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
7535 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
7536 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
7537 * omp-general.c (omp_max_vf): Likewise.
7538 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7539 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
7540 * tree-vect-loop.c (vect_analyze_loop): Likewise.
7541 * tree-vect-slp.c (vect_slp_bb): Likewise.
7542 * doc/tm.texi: Regenerate.
7543 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
7545 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
7546 the vector size as a poly_uint64 rather than an unsigned int.
7547 (current_vector_size): Change from an unsigned int to a poly_uint64.
7548 (get_vectype_for_scalar_type): Update accordingly.
7549 * tree.h (build_truth_vector_type): Take the size and number of
7550 units as a poly_uint64 rather than an unsigned int.
7551 (build_vector_type): Add a temporary overload that takes
7552 the number of units as a poly_uint64 rather than an unsigned int.
7553 * tree.c (make_vector_type): Likewise.
7554 (build_truth_vector_type): Take the number of units as a poly_uint64
7555 rather than an unsigned int.
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 * target.def (get_mask_mode): Take the number of units and length
7562 as poly_uint64s rather than unsigned ints.
7563 * targhooks.h (default_get_mask_mode): Update accordingly.
7564 * targhooks.c (default_get_mask_mode): Likewise.
7565 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
7566 * doc/tm.texi: Regenerate.
7568 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7569 Alan Hayward <alan.hayward@arm.com>
7570 David Sherwood <david.sherwood@arm.com>
7572 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
7573 * omp-general.c (omp_max_vf): Likewise.
7574 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
7575 (expand_omp_simd): Handle polynomial safelen.
7576 * omp-low.c (omplow_simd_context): Add a default constructor.
7577 (omplow_simd_context::max_vf): Change from int to poly_uint64.
7578 (lower_rec_simd_input_clauses): Update accordingly.
7579 (lower_rec_input_clauses): Likewise.
7581 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7582 Alan Hayward <alan.hayward@arm.com>
7583 David Sherwood <david.sherwood@arm.com>
7585 * tree-vectorizer.h (vect_nunits_for_cost): New function.
7586 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
7587 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
7588 (vect_analyze_slp_cost): Likewise.
7589 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
7590 (vect_model_load_cost): Likewise.
7592 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7593 Alan Hayward <alan.hayward@arm.com>
7594 David Sherwood <david.sherwood@arm.com>
7596 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
7597 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
7598 from an unsigned int * to a poly_uint64_pod *.
7599 (calculate_unrolling_factor): New function.
7600 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
7602 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7603 Alan Hayward <alan.hayward@arm.com>
7604 David Sherwood <david.sherwood@arm.com>
7606 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
7607 from an unsigned int to a poly_uint64.
7608 (_loop_vec_info::slp_unrolling_factor): Likewise.
7609 (_loop_vec_info::vectorization_factor): Change from an int
7611 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
7612 (vect_get_num_vectors): New function.
7613 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
7614 (vect_get_num_copies): Use vect_get_num_vectors.
7615 (vect_analyze_data_ref_dependences): Change max_vf from an int *
7616 to an unsigned int *.
7617 (vect_analyze_data_refs): Change min_vf from an int * to a
7619 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7620 than an unsigned HOST_WIDE_INT.
7621 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7622 (vect_analyze_data_ref_dependence): Change max_vf from an int *
7623 to an unsigned int *.
7624 (vect_analyze_data_ref_dependences): Likewise.
7625 (vect_compute_data_ref_alignment): Handle polynomial vf.
7626 (vect_enhance_data_refs_alignment): Likewise.
7627 (vect_prune_runtime_alias_test_list): Likewise.
7628 (vect_shift_permute_load_chain): Likewise.
7629 (vect_supportable_dr_alignment): Likewise.
7630 (dependence_distance_ge_vf): Take the vectorization factor as a
7631 poly_uint64 rather than an unsigned HOST_WIDE_INT.
7632 (vect_analyze_data_refs): Change min_vf from an int * to a
7634 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
7635 vfm1 as a poly_uint64 rather than an int. Make the same change
7636 for the returned bound_scalar.
7637 (vect_gen_vector_loop_niters): Handle polynomial vf.
7638 (vect_do_peeling): Likewise. Update call to
7639 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
7640 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
7642 * tree-vect-loop.c (vect_determine_vectorization_factor)
7643 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
7644 (vect_get_known_peeling_cost): Likewise.
7645 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
7646 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
7647 (vect_transform_loop): Likewise. Use the lowest possible VF when
7648 updating the upper bounds of the loop.
7649 (vect_min_worthwhile_factor): Make static. Return an unsigned int
7651 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
7652 polynomial unroll factors.
7653 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
7654 (vect_make_slp_decision): Likewise.
7655 (vect_supported_load_permutation_p): Likewise, and polynomial
7657 (vect_analyze_slp_cost): Handle polynomial vf.
7658 (vect_slp_analyze_node_operations): Likewise.
7659 (vect_slp_analyze_bb_1): Likewise.
7660 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7661 than an unsigned HOST_WIDE_INT.
7662 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
7663 (vectorizable_load): Handle polynomial vf.
7664 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
7666 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
7668 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7669 Alan Hayward <alan.hayward@arm.com>
7670 David Sherwood <david.sherwood@arm.com>
7672 * match.pd: Handle bit operations involving three constants
7673 and try to fold one pair.
7675 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7677 * tree-vect-loop-manip.c: Include gimple-fold.h.
7678 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
7679 niters_maybe_zero parameters. Handle other cases besides a step of 1.
7680 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
7681 Add a path that uses a step of VF instead of 1, but disable it
7683 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
7684 and niters_no_overflow parameters. Update calls to
7685 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
7686 Create a new SSA name if the latter choses to use a ste other
7687 than zero, and return it via niters_vector_mult_vf_var.
7688 * tree-vect-loop.c (vect_transform_loop): Update calls to
7689 vect_do_peeling, vect_gen_vector_loop_niters and
7690 slpeel_make_loop_iterate_ntimes.
7691 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
7692 (vect_gen_vector_loop_niters): Update declarations after above changes.
7694 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
7696 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
7697 128-bit round to integer instructions.
7698 (ceil<mode>2): Likewise.
7699 (btrunc<mode>2): Likewise.
7700 (round<mode>2): Likewise.
7702 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7704 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
7705 unaligned VSX load/store on P8/P9.
7706 (expand_block_clear): Allow the use of unaligned VSX
7707 load/store on P8/P9.
7709 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7711 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
7713 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
7714 swap associated with both a load and a store.
7716 2018-01-02 Andrew Waterman <andrew@sifive.com>
7718 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
7719 * config/riscv/riscv.md (clear_cache): Use it.
7721 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
7723 * web.c: Remove out-of-date comment.
7725 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7727 * expr.c (fixup_args_size_notes): Check that any existing
7728 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
7729 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
7730 (emit_single_push_insn): ...here.
7732 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7734 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
7735 (const_vector_encoded_nelts): New function.
7736 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
7737 (const_vector_int_elt, const_vector_elt): Declare.
7738 * emit-rtl.c (const_vector_int_elt_1): New function.
7739 (const_vector_elt): Likewise.
7740 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
7741 of CONST_VECTOR_ELT.
7743 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7745 * expr.c: Include rtx-vector-builder.h.
7746 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
7747 directly on the tree encoding.
7748 (const_vector_from_tree): Likewise.
7749 * optabs.c: Include rtx-vector-builder.h.
7750 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
7751 sequence of "u" values.
7752 * vec-perm-indices.c: Include rtx-vector-builder.h.
7753 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
7754 directly on the vec_perm_indices encoding.
7756 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7758 * doc/rtl.texi (const_vector): Describe new encoding scheme.
7759 * Makefile.in (OBJS): Add rtx-vector-builder.o.
7760 * rtx-vector-builder.h: New file.
7761 * rtx-vector-builder.c: Likewise.
7762 * rtl.h (rtx_def::u2): Add a const_vector field.
7763 (CONST_VECTOR_NPATTERNS): New macro.
7764 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
7765 (CONST_VECTOR_DUPLICATE_P): Likewise.
7766 (CONST_VECTOR_STEPPED_P): Likewise.
7767 (CONST_VECTOR_ENCODED_ELT): Likewise.
7768 (const_vec_duplicate_p): Check for a duplicated vector encoding.
7769 (unwrap_const_vec_duplicate): Likewise.
7770 (const_vec_series_p): Check for a non-duplicated vector encoding.
7771 Say that the function only returns true for integer vectors.
7772 * emit-rtl.c: Include rtx-vector-builder.h.
7773 (gen_const_vec_duplicate_1): Delete.
7774 (gen_const_vector): Call gen_const_vec_duplicate instead of
7775 gen_const_vec_duplicate_1.
7776 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
7777 (gen_const_vec_duplicate): Use rtx_vector_builder.
7778 (gen_const_vec_series): Likewise.
7779 (gen_rtx_CONST_VECTOR): Likewise.
7780 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
7781 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7782 Build a new vector rather than modifying a CONST_VECTOR in-place.
7783 (handle_special_swappables): Update call accordingly.
7784 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
7785 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7786 Build a new vector rather than modifying a CONST_VECTOR in-place.
7787 (handle_special_swappables): Update call accordingly.
7789 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7791 * simplify-rtx.c (simplify_const_binary_operation): Use
7792 CONST_VECTOR_ELT instead of XVECEXP.
7794 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7796 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
7797 the selector elements to be different from the data elements
7798 if the selector is a VECTOR_CST.
7799 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
7800 ssizetype for the selector.
7802 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7804 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
7805 before testing each element individually.
7806 * tree-vect-generic.c (lower_vec_perm): Likewise.
7808 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7810 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
7811 * selftest-run-tests.c (selftest::run_tests): Call it.
7812 * vector-builder.h (vector_builder::operator ==): New function.
7813 (vector_builder::operator !=): Likewise.
7814 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
7815 (vec_perm_indices::all_from_input_p): New function.
7816 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7817 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
7818 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
7819 instead of reading the VECTOR_CST directly. Detect whether both
7820 vector inputs are the same before constructing the vec_perm_indices,
7821 and update the number of inputs argument accordingly. Use the
7822 utility functions added above. Only construct sel2 if we need to.
7824 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7826 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
7827 the broadcast of the low byte.
7828 (expand_mult_highpart): Use an explicit encoding for the permutes.
7829 * optabs-query.c (can_mult_highpart_p): Likewise.
7830 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
7831 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7832 (vectorizable_bswap): Likewise.
7833 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
7834 explicit encoding for the power-of-2 permutes.
7835 (vect_permute_store_chain): Likewise.
7836 (vect_grouped_load_supported): Likewise.
7837 (vect_permute_load_chain): Likewise.
7839 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7841 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
7842 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
7843 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
7844 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7845 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
7846 (vect_gen_perm_mask_any): Likewise.
7848 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7850 * int-vector-builder.h: New file.
7851 * vec-perm-indices.h: Include int-vector-builder.h.
7852 (vec_perm_indices): Redefine as an int_vector_builder.
7853 (auto_vec_perm_indices): Delete.
7854 (vec_perm_builder): Redefine as a stand-alone class.
7855 (vec_perm_indices::vec_perm_indices): New function.
7856 (vec_perm_indices::clamp): Likewise.
7857 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
7858 (vec_perm_indices::new_vector): New function.
7859 (vec_perm_indices::new_expanded_vector): Update for new
7860 vec_perm_indices class.
7861 (vec_perm_indices::rotate_inputs): New function.
7862 (vec_perm_indices::all_in_range_p): Operate directly on the
7863 encoded form, without computing elided elements.
7864 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
7865 encoding. Update for new vec_perm_indices class.
7866 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
7867 the given vec_perm_builder.
7868 (expand_vec_perm_var): Update vec_perm_builder constructor.
7869 (expand_mult_highpart): Use vec_perm_builder instead of
7870 auto_vec_perm_indices.
7871 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
7872 vec_perm_indices instead of auto_vec_perm_indices. Use a single
7873 or double series encoding as appropriate.
7874 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
7875 vec_perm_indices instead of auto_vec_perm_indices.
7876 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7877 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7878 (vect_permute_store_chain): Likewise.
7879 (vect_grouped_load_supported): Likewise.
7880 (vect_permute_load_chain): Likewise.
7881 (vect_shift_permute_load_chain): Likewise.
7882 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
7883 (vect_transform_slp_perm_load): Likewise.
7884 (vect_schedule_slp_instance): Likewise.
7885 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7886 (vectorizable_mask_load_store): Likewise.
7887 (vectorizable_bswap): Likewise.
7888 (vectorizable_store): Likewise.
7889 (vectorizable_load): Likewise.
7890 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
7891 vec_perm_indices instead of auto_vec_perm_indices. Use
7892 tree_to_vec_perm_builder to read the vector from a tree.
7893 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
7894 vec_perm_builder instead of a vec_perm_indices.
7895 (have_whole_vector_shift): Use vec_perm_builder and
7896 vec_perm_indices instead of auto_vec_perm_indices. Leave the
7897 truncation to calc_vec_perm_mask_for_shift.
7898 (vect_create_epilog_for_reduction): Likewise.
7899 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
7900 from auto_vec_perm_indices to vec_perm_indices.
7901 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7902 instead of changing individual elements.
7903 (aarch64_vectorize_vec_perm_const): Use new_vector to install
7904 the vector in d.perm.
7905 * config/arm/arm.c (expand_vec_perm_d::perm): Change
7906 from auto_vec_perm_indices to vec_perm_indices.
7907 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7908 instead of changing individual elements.
7909 (arm_vectorize_vec_perm_const): Use new_vector to install
7910 the vector in d.perm.
7911 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
7912 Update vec_perm_builder constructor.
7913 (rs6000_expand_interleave): Likewise.
7914 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
7915 (rs6000_expand_interleave): Likewise.
7917 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7919 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
7920 to qimode could truncate the indices.
7921 * optabs.c (expand_vec_perm_var): Likewise.
7923 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7925 * Makefile.in (OBJS): Add vec-perm-indices.o.
7926 * vec-perm-indices.h: New file.
7927 * vec-perm-indices.c: Likewise.
7928 * target.h (vec_perm_indices): Replace with a forward class
7930 (auto_vec_perm_indices): Move to vec-perm-indices.h.
7931 * optabs.h: Include vec-perm-indices.h.
7932 (expand_vec_perm): Delete.
7933 (selector_fits_mode_p, expand_vec_perm_var): Declare.
7934 (expand_vec_perm_const): Declare.
7935 * target.def (vec_perm_const_ok): Replace with...
7936 (vec_perm_const): ...this new hook.
7937 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
7938 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
7939 * doc/tm.texi: Regenerate.
7940 * optabs.def (vec_perm_const): Delete.
7941 * doc/md.texi (vec_perm_const): Likewise.
7942 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
7943 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
7944 expand_vec_perm for constant permutation vectors. Assert that
7945 the mode of variable permutation vectors is the integer equivalent
7946 of the mode that is being permuted.
7947 * optabs-query.h (selector_fits_mode_p): Declare.
7948 * optabs-query.c: Include vec-perm-indices.h.
7949 (selector_fits_mode_p): New function.
7950 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
7951 is defined, instead of checking whether the vec_perm_const_optab
7952 exists. Use targetm.vectorize.vec_perm_const instead of
7953 targetm.vectorize.vec_perm_const_ok. Check whether the indices
7954 fit in the vector mode before using a variable permute.
7955 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
7956 vec_perm_indices instead of an rtx.
7957 (expand_vec_perm): Replace with...
7958 (expand_vec_perm_const): ...this new function. Take the selector
7959 as a vec_perm_indices rather than an rtx. Also take the mode of
7960 the selector. Update call to shift_amt_for_vec_perm_mask.
7961 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
7962 Use vec_perm_indices::new_expanded_vector to expand the original
7963 selector into bytes. Check whether the indices fit in the vector
7964 mode before using a variable permute.
7965 (expand_vec_perm_var): Make global.
7966 (expand_mult_highpart): Use expand_vec_perm_const.
7967 * fold-const.c: Includes vec-perm-indices.h.
7968 * tree-ssa-forwprop.c: Likewise.
7969 * tree-vect-data-refs.c: Likewise.
7970 * tree-vect-generic.c: Likewise.
7971 * tree-vect-loop.c: Likewise.
7972 * tree-vect-slp.c: Likewise.
7973 * tree-vect-stmts.c: Likewise.
7974 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
7976 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
7977 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
7978 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
7979 (aarch64_vectorize_vec_perm_const): ...this new function.
7980 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7981 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7982 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
7983 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
7984 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7985 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7986 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
7988 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
7989 check for NEON modes.
7990 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
7991 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
7992 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
7993 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
7995 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
7996 the old VEC_PERM_CONST conditions.
7997 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
7998 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
7999 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
8000 (ia64_vectorize_vec_perm_const_ok): Merge into...
8001 (ia64_vectorize_vec_perm_const): ...this new function.
8002 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
8003 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
8004 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
8005 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
8006 * config/mips/mips.c (mips_expand_vec_perm_const)
8007 (mips_vectorize_vec_perm_const_ok): Merge into...
8008 (mips_vectorize_vec_perm_const): ...this new function.
8009 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
8010 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
8011 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
8012 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
8013 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
8014 (rs6000_expand_vec_perm_const): Delete.
8015 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
8017 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8018 (altivec_expand_vec_perm_const_le): Take each operand individually.
8019 Operate on constant selectors rather than rtxes.
8020 (altivec_expand_vec_perm_const): Likewise. Update call to
8021 altivec_expand_vec_perm_const_le.
8022 (rs6000_expand_vec_perm_const): Delete.
8023 (rs6000_vectorize_vec_perm_const_ok): Delete.
8024 (rs6000_vectorize_vec_perm_const): New function.
8025 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8026 an element count and rtx array.
8027 (rs6000_expand_extract_even): Update call accordingly.
8028 (rs6000_expand_interleave): Likewise.
8029 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
8030 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
8031 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
8032 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
8033 (rs6000_expand_vec_perm_const): Delete.
8034 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8035 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8036 (altivec_expand_vec_perm_const_le): Take each operand individually.
8037 Operate on constant selectors rather than rtxes.
8038 (altivec_expand_vec_perm_const): Likewise. Update call to
8039 altivec_expand_vec_perm_const_le.
8040 (rs6000_expand_vec_perm_const): Delete.
8041 (rs6000_vectorize_vec_perm_const_ok): Delete.
8042 (rs6000_vectorize_vec_perm_const): New function. Remove stray
8043 reference to the SPE evmerge intructions.
8044 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8045 an element count and rtx array.
8046 (rs6000_expand_extract_even): Update call accordingly.
8047 (rs6000_expand_interleave): Likewise.
8048 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
8049 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
8051 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8053 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8055 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
8056 vector mode and that that mode matches the mode of the data
8058 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
8059 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
8060 directly using expand_vec_perm_1 when forcing selectors into
8062 (expand_vec_perm_var): New function, split out from expand_vec_perm.
8064 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8066 * optabs-query.h (can_vec_perm_p): Delete.
8067 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
8068 * optabs-query.c (can_vec_perm_p): Split into...
8069 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
8070 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
8071 particular selector is valid.
8072 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8073 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8074 (vect_grouped_load_supported): Likewise.
8075 (vect_shift_permute_load_chain): Likewise.
8076 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8077 (vect_transform_slp_perm_load): Likewise.
8078 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8079 (vectorizable_bswap): Likewise.
8080 (vect_gen_perm_mask_checked): Likewise.
8081 * fold-const.c (fold_ternary_loc): Likewise. Don't take
8082 implementations of variable permutation vectors into account
8083 when deciding which selector to use.
8084 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
8085 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
8086 with a false third argument.
8087 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
8088 to test whether the constant selector is valid and can_vec_perm_var_p
8089 to test whether a variable selector is valid.
8091 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8093 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
8094 * optabs-query.c (can_vec_perm_p): Likewise.
8095 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
8096 instead of vec_perm_indices.
8097 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
8098 (vect_gen_perm_mask_checked): Likewise,
8099 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
8100 (vect_gen_perm_mask_checked): Likewise,
8102 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8104 * optabs-query.h (qimode_for_vec_perm): Declare.
8105 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
8106 (qimode_for_vec_perm): ...this new function.
8107 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
8109 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8111 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
8112 does not have a conditional at the top.
8114 2018-01-02 Richard Biener <rguenther@suse.de>
8116 * ipa-inline.c (big_speedup_p): Fix expression.
8118 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8121 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
8124 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8128 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
8129 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
8130 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
8131 cond_taken_branch_cost 3->4.
8133 2018-01-01 Jakub Jelinek <jakub@redhat.com>
8135 PR tree-optimization/83581
8136 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
8137 TODO_cleanup_cfg if any changes have been made.
8140 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
8141 convert_modes if target mode has the right side, but different mode
8145 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
8146 last argument when extracting from CONCAT. If either from_real or
8147 from_imag is NULL, use expansion through memory. If result is not
8148 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
8149 the parts directly to inner mode, if even that fails, use expansion
8153 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
8154 check for bswap in mode rather than HImode and use that in expand_unop
8157 Copyright (C) 2018 Free Software Foundation, Inc.
8159 Copying and distribution of this file, with or without modification,
8160 are permitted in any medium without royalty provided the copyright
8161 notice and this notice are preserved.