1 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
4 Rename to ppc_speculation_barrier.
5 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
6 __builtin_ppc_speculation_barrier.
8 2018-03-05 Jakub Jelinek <jakub@redhat.com>
11 * combine.c (combine_simplify_rtx): Don't try to simplify if
12 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
15 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
17 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
18 to 32 bytes when compiling for POWER9.
20 2018-03-05 Jakub Jelinek <jakub@redhat.com>
23 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
24 regparm >= 3 with no arg reg available also for calls with
25 flag_force_indirect_call. Pass decl to ix86_function_regparm.
28 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
30 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
32 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
35 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
37 2018-03-05 Richard Biener <rguenther@suse.de>
39 PR tree-optimization/84486
40 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
41 When inserting a __builtin_assume_aligned call set the LHS
42 SSA name alignment info accordingly.
44 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
46 PR tree-optimization/84114
47 * config/aarch64/aarch64.c (aarch64_reassociation_width)
48 Avoid reassociation of FLOAT_MODE addition.
50 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
52 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
53 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
54 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
55 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
56 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
57 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
58 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
60 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
61 __builtin_ia32_wbinvd): New builtins.
63 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
65 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
66 (ix86_valid_target_attribute_inner_p): Ditto.
67 (ix86_init_mmx_sse_builtins): Add special_args2.
68 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
69 TARGET_WBNOINVD_P): New.
70 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
71 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
72 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
73 * config/i386/immintrin.h (_wbinvd): New intrinsic.
74 * config/i386/pconfigintrin.h: New file.
75 * config/i386/wbnoinvdintrin.h: Ditto.
76 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
77 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
79 2018-03-05 Richard Biener <rguenther@suse.de>
81 PR tree-optimization/84670
82 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
84 (BB_VISITED_WITH_VISITED_SUCCS): New define.
85 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
86 (compute_antic_aux): Only assert the number of values in ANTIC_IN
87 doesn't grow if all successors (recursively) were visited at least
90 2018-03-05 Richard Biener <rguenther@suse.de>
92 PR tree-optimization/84650
93 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
94 if executed in the loop pipeline.
96 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
98 * doc/configfiles.texi (Configuration Files): Move info about
99 conditionalizing $target-protos.h to...
100 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
101 differs from $target-protos.h.
103 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
104 Chung-Ju Wu <jasonwucj@gmail.com>
106 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
107 * config/nds32/nds32-multiple.md (setmemsi): Define.
108 * config/nds32/nds32-memory-manipulation.c
109 (nds32_gen_dup_4_byte_to_word_value): New.
110 (emit_setmem_word_loop): New.
111 (emit_setmem_byte_loop): New.
112 (nds32_expand_setmem_loop): New.
113 (nds32_expand_setmem_loop_v3m): New.
114 (nds32_expand_setmem_unroll): New.
115 (nds32_expand_setmem): New.
117 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
118 Chung-Ju Wu <jasonwucj@gmail.com>
120 * config/nds32/nds32-memory-manipulation.c
121 (nds32_emit_load_store): New.
122 (nds32_emit_post_inc_load_store): New.
123 (nds32_emit_mem_move): New.
124 (nds32_emit_mem_move_block): New.
125 (nds32_expand_movmemsi_loop_unknown_size): New.
126 (nds32_expand_movmemsi_loop_known_size): New.
127 (nds32_expand_movmemsi_loop): New.
128 (nds32_expand_movmemsi_unroll): New.
129 (nds32_expand_movmemqi): Rename ...
130 (nds32_expand_movmemsi): ... to this.
131 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
132 (movmemsi): ... to this.
133 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
134 (nds32_expand_movmemsi): ... to this.
136 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
137 Monk Chiang <sh.chiang04@gmail.com>
138 Chung-Ju Wu <jasonwucj@gmail.com>
140 * config/nds32/nds32-protos.h
141 (nds32_expand_load_multiple): New arguments.
142 (nds32_expand_store_multiple): Ditto.
143 (nds32_valid_multiple_load_store): Rename ...
144 (nds32_valid_multiple_load_store_p): ... to this.
145 * config/nds32/nds32-memory-manipulation.c
146 (nds32_expand_load_multiple): Refine implementation.
147 (nds32_expand_store_multiple): Ditto.
148 * config/nds32/nds32-multiple.md
149 (load_multiple): Update nds32_expand_load_multiple interface.
150 (store_multiple): Update nds32_expand_store_multiple interface.
151 * config/nds32/nds32-predicates.c
152 (nds32_valid_multiple_load_store): Rename ...
153 (nds32_valid_multiple_load_store_p): ... to this and refine
155 * config/nds32/predicates.md
156 (nds32_load_multiple_and_update_address_operation): New predicate.
157 (nds32_store_multiple_and_update_address_operation): New predicate.
159 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
160 Chung-Ju Wu <jasonwucj@gmail.com>
162 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
163 (combo): New attribute.
164 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
166 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
168 * config/nds32/nds32.opt: Change -mcmodel= default value.
170 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
171 Monk Chiang <sh.chiang04@gmail.com>
172 Chung-Ju Wu <jasonwucj@gmail.com>
174 * config/nds32/constants.md (unspec_element): New enum.
175 * config/nds32/constraints.md (Umw): New constraint.
176 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
177 * config/nds32/nds32-intrinsic.md: Likewise.
178 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
179 (nds32_valid_smw_lwm_base_p): New.
180 (nds32_output_smw_single_word): New.
181 (nds32_output_lmw_single_word): New.
182 (nds32_expand_unaligned_load): New.
183 (nds32_expand_unaligned_store): New.
184 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
185 (nds32_output_smw_single_word): Declare.
186 (nds32_output_lmw_single_word): Declare.
187 (nds32_expand_unaligned_load): Declare.
188 (nds32_expand_unaligned_store): Declare.
189 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
190 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
191 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
192 NDS32_BUILTIN_UASTORE_DW.
193 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
196 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
197 Kito Cheng <kito.cheng@gmail.com>
198 Chung-Ju Wu <jasonwucj@gmail.com>
200 * config/nds32/nds32-intrinsic.c
201 (nds32_expand_builtin_null_ftype_reg): Delete.
202 (nds32_expand_builtin_reg_ftype_imm): Ditto.
203 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
204 (nds32_read_argument): New.
205 (nds32_legitimize_target): Ditto.
206 (nds32_legitimize_argument): Ditto.
207 (nds32_check_constant_argument): Ditto.
208 (nds32_expand_unop_builtin): Ditto.
209 (nds32_expand_unopimm_builtin): Ditto.
210 (nds32_expand_binop_builtin): Ditto.
211 (nds32_builtin_decl_impl): Ditto.
212 (builtin_description): Ditto.
213 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
214 (nds32_init_builtins_impl): Ditto.
215 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
216 (nds32_builtin_decl): New.
217 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
218 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
220 2018-03-02 Jeff Law <law@redhat.com>
222 * reorg.c (stop_search_p): Handle DEBUG_INSN.
223 (redundant_insn, fill_simple_delay_slots): Likewise.
224 (fill_slots_from_thread): Likewise.
225 * resource.c (mark_referenced_resources): Likewise.
226 (mark_set_resources, find_dead_or_set_registers): Likewise.
228 2018-03-02 Jakub Jelinek <jakub@redhat.com>
230 * substring-locations.h (format_warning_va): Formatting fix for
232 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
234 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
235 * substring-locations.c: Include intl.h.
236 (format_warning_va): Turned into small wrapper around
237 format_warning_n_va, renamed to ...
238 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
239 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
241 (format_warning_at_substring_n): New function.
242 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
243 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
244 format_warning_at_substring with just a shorter name instead of
245 const function pointer.
246 (fmtwarn_n): New function.
247 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
248 appropriate, get rid of all the fmtstr temporaries, move conditionals
249 with G_() wrapped string literals directly into fmtwarn arguments,
250 cast dir.len to (int), formatting fixes.
252 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
254 * doc/invoke.texi: Remove "Cilk Plus" references.
256 2018-03-02 Jakub Jelinek <jakub@redhat.com>
257 Richard Biener <rguenther@suse.de>
260 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
261 for error or warning attributes if CALL_FROM_THUNK_P is set.
264 2018-03-02 Jakub Jelinek <jakub@redhat.com>
267 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
268 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
271 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
272 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
274 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
275 instead of -1U in last predictors element's probability member.
277 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
280 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
281 arguments if they are comparable.
283 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
285 PR tree-optimization/84634
286 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
287 masks and masked_loop_p with a single loop_masks, making sure it's
288 null for bb vectorization.
290 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
292 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
293 (vect_analyze_data_ref_access): Use loop->safe_len rather than
294 loop->force_vectorize to check whether there is no alias.
296 2018-03-02 Jakub Jelinek <jakub@redhat.com>
299 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
301 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
303 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
304 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
305 instead of a loop around prev_real_insn.
306 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
310 * config/i386/i386.c (ix86_print_operand): Use conditional
311 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
314 2018-03-02 Richard Biener <rguenther@suse.de>
316 PR tree-optimization/84427
317 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
318 (bitmap_set_subtract_values): Rewrite to handle multiple
321 (prune_clobbered_mems): Likewise.
322 (phi_translate): Take edge instead of pred/phiblock.
323 (phi_translate_1): Likewise.
324 (phi_translate_set): Likewise. Insert all translated
325 exprs for a value into the set, keeping possibly multiple
326 expressions per value.
327 (compute_antic_aux): Adjust for phi_translate changes.
328 When intersecting union the expressions and prune those
329 not in the final value set, keeping possibly multiple
330 expressions per value. Do not use value-insertion
331 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
332 all expressions. Add verification that the value-sets
333 only shrink during iteration.
334 (compute_partial_antic_aux): Adjust for the phi_translate changes.
335 (do_pre_regular_insertion): Likewise.
336 (do_pre_partial_partial_insertion): Likewise.
338 2018-03-02 Richard Biener <rguenther@suse.de>
341 * config/darwin.c (saved_debug_info_level): New static global.
342 (darwin_asm_lto_start): Disable debug info generation for LTO out.
343 (darwin_asm_lto_end): Restore debug info generation settings.
345 2018-03-01 Martin Liska <mliska@suse.cz>
348 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
351 2018-03-01 Richard Biener <rguenther@suse.de>
354 * dwarf2out.c (gen_variable_die): Properly handle late VLA
355 type annotation with LTO when debug was disabled at compile-time.
357 2018-03-01 Matthew Fortune <mfortune@gmail.com>
359 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
361 (mips_final_postscan_insn): Likewise.
363 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
365 PR rtl-optimization/84528
366 * alias.c (init_alias_target): Add commentary.
367 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
368 a unique base value if the frame pointer is not eliminated
369 to the stack pointer.
371 2018-03-01 Tom de Vries <tom@codesourcery.com>
373 PR rtl-optimization/83327
374 * lra-int.h (hard_regs_spilled_into): Declare.
375 * lra.c (hard_regs_spilled_into): Define.
376 (init_reg_info): Init hard_regs_spilled_into.
377 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
378 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
379 (process_bb_lives): Handle hard_regs_spilled_into.
380 (lra_create_live_ranges_1): Before doing liveness propagation, clear
381 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
383 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
385 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
386 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
387 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
388 * config/rs6000/aix72.h: New file.
390 2018-02-28 Jakub Jelinek <jakub@redhat.com>
392 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
393 instead of warning_at with conditional singular and plural messages
397 * stor-layout.c (update_alignment_for_field): For
398 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
399 && !DECL_PACKED (field), do the alignment update, just use
400 only desired_align instead of MAX (type_align, desired_align)
402 (place_field): Don't do known_align < desired_align handling
403 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
404 is non-NULL, instead do it after rli->prev_field handling and
405 only if not within a bitfield word. For DECL_PACKED (field)
406 use type_align of BITS_PER_UNIT.
408 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
410 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
411 superfluous parentheses and trailing spaces.
413 2018-02-28 Richard Biener <rguenther@suse.de>
415 PR tree-optimization/84584
416 * graphite-scop-detection.c (scop_detection::add_scop): Discard
417 SCoPs with fake exit edge.
419 2018-02-28 Martin Liska <mliska@suse.cz>
422 * timevar.c (timer::print): Fix format to properly print 100%
425 2018-02-28 Richard Biener <rguenther@suse.de>
428 * genmatch.c (capture_info::walk_match): Do not mark
429 captured expressions without operands as expr_p given
430 they act more like predicates and should be subject to
431 "lost tail" side-effect preserving.
433 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
435 PR rtl-optimization/81611
436 * auto-inc-dec.c (attempt_change): Move dead note from
437 mem_insn if it's the next use of regno
438 (find_address): Take address use of reg holding
439 non-incremented value. Add parm to limit search to the named
441 (merge_in_block): Attempt to use a mem insn that is the next
442 use of the original regno.
444 2018-02-27 Martin Sebor <msebor@redhat.com>
447 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
448 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
450 2018-02-27 Martin Sebor <msebor@redhat.com>
453 * diagnostic-core.h (warning_n, error_n, inform_n): Change
454 n argument to unsigned HOST_WIDE_INT.
455 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
456 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
457 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
458 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
460 2018-02-27 Richard Biener <rguenther@suse.de>
462 PR tree-optimization/84512
463 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
464 Do not use the estimate returned from record_stmt_cost for
465 the scalar iteration cost but sum properly using add_stmt_cost.
467 2018-02-27 Richard Biener <rguenther@suse.de>
469 PR tree-optimization/84466
470 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
471 Adjust last change to less strictly validate use operands.
473 2018-02-27 Martin Liska <mliska@suse.cz>
475 PR gcov-profile/84548
476 * gcov.c (process_file): Allow partial overlap and consider it
477 also as group functions.
478 (output_lines): Properly calculate range of lines for a group.
480 2018-02-27 Martin Liska <mliska@suse.cz>
482 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
483 'ggc' suffixes. Change first column width.
484 (timer::print): Fix formatting of the column.
486 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
488 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
489 preserve inline entry blocks for the sake of debug inline
490 entry point markers alone.
491 (remove_unused_locals): Suggest in comments a better place to
492 force the preservation of inline entry blocks that are
493 otherwise unused, but do not preserve them.
495 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
497 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
499 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
502 * config/i386/constraints.md (Bs): Replace
503 ix86_indirect_branch_register with
504 TARGET_INDIRECT_BRANCH_REGISTER.
506 * config/i386/i386.md (indirect_jump): Likewise.
507 (tablejump): Likewise.
508 (*sibcall_memory): Likewise.
509 (*sibcall_value_memory): Likewise.
510 Peepholes of indirect call and jump via memory: Likewise.
511 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
512 (*sibcall_value_GOT_32): Likewise.
513 * config/i386/predicates.md (indirect_branch_operand): Likewise.
514 (GOT_memory_operand): Likewise.
515 (call_insn_operand): Likewise.
516 (sibcall_insn_operand): Likewise.
517 (GOT32_symbol_operand): Likewise.
518 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
520 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
522 PR rtl-optimization/83496
523 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
524 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
525 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
526 redundant insn, if any.
527 (relax_delay_slots): Likewise.
528 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
530 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
532 PR tree-optimization/83965
533 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
534 that grouped statements are part of a reduction chain. Return
535 true if the statement is not marked as a reduction itself but
537 (vect_recog_dot_prod_pattern): Don't check whether the statement
538 is part of a group here.
539 (vect_recog_sad_pattern): Likewise.
540 (vect_recog_widen_sum_pattern): Likewise.
542 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
545 * final.c (rest_of_clean_state): Also look for calls inside sequences.
547 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
550 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
552 (ix86_output_indirect_function_return): New prototype.
553 (ix86_split_simple_return_pop_internal): Likewise.
554 * config/i386/i386.c (indirect_return_via_cx): New.
555 (indirect_return_via_cx_bnd): Likewise.
556 (indirect_thunk_name): Handle return va CX_REG.
557 (output_indirect_thunk_function): Create alias for
558 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
559 (ix86_output_indirect_jmp): Remove the bool argument.
560 (ix86_output_indirect_function_return): New function.
561 (ix86_split_simple_return_pop_internal): Likewise.
562 * config/i386/i386.md (*indirect_jump): Don't pass false
563 to ix86_output_indirect_jmp.
564 (*tablejump_1): Likewise.
565 (simple_return_pop_internal): Change it to define_insn_and_split.
566 Call ix86_split_simple_return_pop_internal to split it for
568 (simple_return_indirect_internal): Call
569 ix86_output_indirect_function_return instead of
570 ix86_output_indirect_jmp.
572 2018-02-26 Jakub Jelinek <jakub@redhat.com>
575 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
576 memset and value initialization afterwards.
578 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
580 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
582 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
585 * common/config/aarch64/aarch64-common.c
586 (aarch_option_optimization_table[]): Switch
587 off fomit-frame-pointer
589 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
590 Chung-Ju Wu <jasonwucj@gmail.com>
592 * config/nds32/nds32-multiple.md (load_multiple): Disallow
594 (store_multiple): Ditto.
596 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
598 * config.gcc: Add --with-cpu support for nds32 target.
599 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
600 * config/nds32/nds32.opt: Add -mcpu= option.
602 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
604 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
605 isel=yes): Warn for these deprecated options.
607 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
609 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
610 ISA_2_5_MASKS_EMBEDDED.
612 2018-02-23 Jakub Jelinek <jakub@redhat.com>
614 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
615 p->max as pointers rather than using iterative_hash_expr.
617 2018-02-23 Carl Love <cel@us.ibm.com>
619 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
620 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
622 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
623 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
624 P8V_BUILTIN_VEC_VUNSIGNED2.
626 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
629 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
630 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
632 * lra-constraints.c (curr_insn_transform): Set up
633 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
635 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
636 * lra-eliminations.c (spill_pseudos): Ditto.
637 (process_insn_for_elimination): Ditto.
638 * lra-lives.c (reg_early_clobber_p): Use the new macros.
639 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
640 LRA_NON_CLOBBERED_ALT.
642 2018-02-22 Martin Sebor <msebor@redhat.com>
644 PR tree-optimization/84480
645 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
646 to maybe_diag_stxncpy_trunc. Call it.
647 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
648 from gimple_fold_builtin_strcpy. Print inlining stack.
649 (handle_builtin_stxncpy): Print inlining stack.
650 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
652 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
655 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
656 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
657 and -fcheck-pointer-bounds are used together.
658 (indirect_thunk_prefix): New enum.
659 (indirect_thunk_need_prefix): New function.
660 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
661 "_nt" instead of "_bnd" for NOTRACK prefix.
662 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
663 (output_indirect_thunk_function): Likewise.
665 (ix86_code_end): Update output_indirect_thunk_function calls.
666 (ix86_output_indirect_branch_via_reg): Replace
667 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
668 (ix86_output_indirect_branch_via_push): Likewise.
669 (ix86_output_function_return): Likewise.
670 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
671 incompatible with -fcf-protection=branch and
672 -fcheck-pointer-bounds.
674 2018-02-22 Steve Ellcey <sellcey@cavium.com>
677 * config/aarch64/aarch64.c (aarch64_print_address_internal):
678 Change gcc_assert call to output_operand_lossage.
680 2018-02-22 Steve Ellcey <sellcey@cavium.com>
682 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
684 2018-02-22 DJ Delorie <dj@redhat.com>
685 Sebastian Perta <sebastian.perta@renesas.com>
686 Oleg Endo <olegendo@gcc.gnu.org>
688 * config/rx/rx.c (rx_rtx_costs): New function.
689 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
691 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
693 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
695 2018-02-22 Martin Liska <mliska@suse.cz>
698 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
699 Add "native" as a possible value.
701 2018-02-22 Martin Liska <mliska@suse.cz>
704 * config/i386/i386.c (ix86_option_override_internal):
705 Add "native" as a possible value for -march and -mtune.
707 2018-02-22 Jakub Jelinek <jakub@redhat.com>
710 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
711 to all type variants.
713 PR tree-optimization/84503
714 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
715 width as info->bitpos + info->bitsize - start.
716 (merged_store_group::merge_overlapping): Simplify width computation.
717 (check_no_overlap): New function.
718 (imm_store_chain_info::try_coalesce_bswap): Compute expected
719 start + width and last_order of the group, fail if check_no_overlap
721 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
722 to group if check_no_overlap fails.
724 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
726 * config/rs6000/altivec.md: Delete contraint arguments to
727 define_expand, define_split, and define_peephole2, and in
728 define_insn_and_split if always unused.
729 * config/rs6000/darwin.md: Ditto.
730 * config/rs6000/dfp.md: Ditto.
731 * config/rs6000/rs6000.md: Ditto.
732 * config/rs6000/sync.md: Ditto.
733 * config/rs6000/vector.md: Ditto.
734 * config/rs6000/vsx.md: Ditto.
736 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
738 * config/rs6000/altivec.md: Write output control strings as braced
739 blocks instead of double-quoted strings.
740 * config/rs6000/darwin.md: Ditto.
741 * config/rs6000/rs6000.md: Ditto.
742 * config/rs6000/vector.md: Ditto.
743 * config/rs6000/vsx.md: Ditto.
745 2018-02-21 Jason Merrill <jason@redhat.com>
747 PR c++/84314 - ICE with templates and fastcall attribute.
748 * attribs.c (build_type_attribute_qual_variant): Remove assert.
750 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
752 * ipa-cp.c (determine_versionability): Fix comment typos.
754 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
757 * ipa-cp.c (determine_versionability): Do not version functions caling
760 2018-02-21 Martin Liska <mliska@suse.cz>
763 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
764 Add "native" as a possible value.
765 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
766 the macro when native cpu detection is available.
768 2018-02-21 Martin Liska <mliska@suse.cz>
771 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
772 Add "native" as a possible value.
773 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
774 when native cpu detection is available.
776 2018-02-21 Jakub Jelinek <jakub@redhat.com>
777 Martin Sebor <msebor@redhat.com>
779 PR tree-optimization/84478
780 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
782 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
783 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
784 support which is conservatively correct, for 2 only stay conservative
785 for maxlen. Formatting and comment capitalization fixes. Add STRICT
786 argument to the 2 argument get_range_strlen, adjust 6 arg
787 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
789 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
790 (gimple_fold_builtin_strlen): Pass true as last argument to
793 2018-02-20 Martin Sebor <msebor@redhat.com>
796 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
797 (builtin_memref::set_base_and_offset): Same. Handle inner references.
798 (builtin_memref::builtin_memref): Factor out parts into
799 set_base_and_offset and call it.
801 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
804 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
805 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
806 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
807 search at the associated MODE_INT.
809 2018-02-20 Jeff Law <law@redhat.com>
812 PR tree-optimization/81592
814 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
815 for range data rather than using global data.
816 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
817 range data rather than using global data.
818 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
819 pass it to children as needed.
820 (struct directive::fmtresult): Similarly.
821 (struct directive::set_width): Similarly.
822 (struct directive::set_precision): Similarly.
823 (format_integer, format_directive, parse_directive): Similarly.
824 (format_none): Accept unnamed vr_values parameter.
825 (format_percent, format_floating, format_character): Similarly.
826 (format_string, format_plain): Similarly.
827 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
828 the EVRP range analyzer for range data rather than using global data.
829 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
830 gimple-ssa-evrp-analyze.h
831 (class sprintf_dom_walker): Add after_dom_children member function.
832 Add evrp_range_analyzer member.
833 (sprintf_dom_walker::before_dom_children): Call into the EVRP
834 range analyzer as needed.
835 (sprintf_dom_walker::after_dom_children): New member function.
836 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
838 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
839 (evrp_range_analyzer::pop_to_marker): Likewise.
841 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
843 PR tree-optimization/84419
844 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
845 with the required type if its current type is compatible but
848 2018-02-20 Jakub Jelinek <jakub@redhat.com>
851 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
854 2018-02-20 Martin Liska <mliska@suse.cz>
857 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
858 possible values if we don't have a hint.
860 2018-02-20 Martin Liska <mliska@suse.cz>
864 * final.c (shorten_branches): Build align_tab array with one
866 * opts.c (finish_options): Add alignment option limit check.
867 (MAX_CODE_ALIGN): Likewise.
868 (MAX_CODE_ALIGN_VALUE): Likewise.
869 * doc/invoke.texi: Document maximum allowed option value for
870 all -falign-* options.
872 2018-02-19 Jakub Jelinek <jakub@redhat.com>
875 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
876 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
877 * var-tracking.c (emit_note_insn_var_location): Remove all references
878 to NOTE_INSN_CALL_ARG_LOCATION.
879 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
880 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
881 Use copy_rtx_if_shared.
882 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
883 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
884 (dwarf2out_var_location): Remove handling of
885 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
887 * final.c (final_scan_insn): Remove all references to
888 NOTE_INSN_CALL_ARG_LOCATION.
889 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
890 before dumping final insns.
891 * except.c (emit_note_eh_region_end): Remove all references to
892 NOTE_INSN_CALL_ARG_LOCATION.
893 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
894 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
895 * config/arc/arc.c (hwloop_optimize): Likewise.
896 * config/arm/arm.c (create_fix_barrier): Likewise.
897 * config/s390/s390.c (s390_chunkify_start): Likewise.
898 * config/sh/sh.c (find_barrier): Likewise.
899 * config/i386/i386.c (rest_of_insert_endbranch,
900 ix86_seh_fixup_eh_fallthru): Likewise.
901 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
902 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
903 * config/frv/frv.c (frv_function_prologue): Likewise.
904 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
906 (note_outside_basic_block_p): Remove all references to
907 NOTE_INSN_CALL_ARG_LOCATION.
908 * gengtype.c (adjust_field_rtx_def): Likewise.
909 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
911 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
912 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
915 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
918 PR tree-optimization/84452
919 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
920 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
923 2018-02-19 Martin Liska <mliska@suse.cz>
926 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
928 2018-02-19 Martin Liska <mliska@suse.cz>
929 Richard Sandiford <richard.sandiford@linaro.org>
931 PR tree-optimization/82491
932 * gimple-fold.c (get_base_constructor): Make earlier bail out
935 2018-02-19 Carl Love <cel@us.ibm.com>
937 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
938 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
940 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
943 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
945 * config/rl78/rl78.md (movdf): New define expand.
947 2018-02-19 Martin Liska <mliska@suse.cz>
950 * doc/invoke.texi: Fix typo.
951 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
953 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
955 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
956 handle rs6000_single_float and rs6000_double_float specially for
959 2018-02-16 Jeff Law <law@redhat.com>
961 * config/rx/rx.c (add_pop_cfi_notes): New function.;
964 2018-02-16 Jakub Jelinek <jakub@redhat.com>
967 * ipa-inline.c (inline_small_functions): Fix a typo.
969 2018-02-16 Nathan Sidwell <nathan@acm.org>
971 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
973 2018-02-16 Carl Love <cel@us.ibm.com>
975 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
976 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
977 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
978 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
979 expansion to P8V_BUILTIN_VEC_FLOAT2.
981 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
983 PR rtl-optimization/70023
984 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
985 src_regno into account.
987 2018-02-16 Carl Love <cel@us.ibm.com>
989 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
990 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
991 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
992 * config/rs6000/rs6000.c: Remove case statements for
993 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
994 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
995 and P9V_BUILTIN_VEC_VINSERT4B.
996 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
997 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
998 * config/rs6000/vsx.md:
999 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
1002 2018-02-16 Carl Love <cel@us.ibm.com>
1004 * config/rs6000/altivec.h: Add builtin names vec_extract4b
1006 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
1008 * config/rs6000/rs6000-c.c: Add the definitions for
1009 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
1010 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
1011 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
1012 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
1013 definition for insert4b and define insn *insert3b_internal.
1014 * doc/extend.texi: Add documentation for vec_extract4b.
1016 2018-02-16 Nathan Sidwell <nathan@acm.org>
1018 * doc/extend.texi (Backwards Compatibility): Mention friend
1019 injection. Note for-scope is deprecated.
1020 * doc/invoke.texi (-ffriend-injection): Deprecate.
1022 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
1024 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
1025 that moved to I2, also allow destinations that are a paradoxical
1026 subreg (instead of a normal reg).
1028 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
1031 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
1034 2018-02-16 Richard Biener <rguenther@suse.de>
1036 PR tree-optimization/84037
1037 PR tree-optimization/84016
1039 * config/i386/i386.c (ix86_builtin_vectorization_cost):
1040 Adjust vec_construct for the fact we need additional higher latency
1041 128bit inserts for AVX256 and AVX512 vector builds.
1042 (ix86_add_stmt_cost): Scale vector construction cost for
1045 2018-02-16 Richard Biener <rguenther@suse.de>
1047 PR tree-optimization/84417
1048 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
1049 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
1050 (non_rewritable_lvalue_p): Likewise, use poly-ints.
1052 2018-02-16 Martin Liska <mliska@suse.cz>
1055 * internal-fn.def (ASAN_CHECK): Set proper flags.
1056 (ASAN_MARK): Likewise.
1058 2018-02-16 Julia Koval <julia.koval@intel.com>
1060 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
1061 from PTA_CANNONLAKE.
1063 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1066 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
1067 Use ++iter rather than iter++ for std::list iterators.
1068 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
1069 defer deleting them until all nodes in the forest are processed. Do
1070 free even leaf nodes. Change to_process into auto_vec.
1073 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
1074 * vec.h (vec_default_construct): Use memset instead of placement new
1075 if BROKEN_VALUE_INITIALIZATION is defined.
1076 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
1077 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
1080 PR rtl-optimization/83723
1081 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
1082 * lra.c (lra_substitute_pseudo): Likewise. If true, use
1083 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
1085 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
1087 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
1089 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
1091 PR rtl-optimization/81443
1092 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
1093 from inner REGs to paradoxical SUBREGs.
1095 2018-02-16 Richard Biener <rguenther@suse.de>
1097 PR tree-optimization/84399
1098 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1099 For operands we can analyze at their definition make sure we can
1100 analyze them at each use as well.
1102 2018-02-16 Richard Biener <rguenther@suse.de>
1104 PR tree-optimization/84190
1105 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
1106 volatile accesses if the decl isn't volatile.
1108 2018-02-15 Jason Merrill <jason@redhat.com>
1110 PR c++/84314 - ICE with templates and fastcall attribute.
1111 * attribs.c (build_type_attribute_qual_variant): Don't clobber
1112 TYPE_CANONICAL on an existing type.
1114 2018-02-15 Jakub Jelinek <jakub@redhat.com>
1116 PR tree-optimization/84383
1117 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
1118 dstoff nor call operand_equal_p if dstbase is NULL.
1120 PR tree-optimization/84334
1121 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
1122 also a CONSTANT_CLASS_P, punt.
1124 2018-02-14 Jim Wilson <jimw@sifive.com>
1126 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
1127 first SMALL_OPERAND check. New local min_second_step. Move assert
1128 to where locals are set. Add TARGET_RVC support.
1129 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
1131 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
1133 * doc/invoke.texi: Correct -Wformat-overflow code sample.
1135 2018-02-14 Martin Sebor <msebor@redhat.com>
1137 PR tree-optimization/83698
1138 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
1139 arrays constrain the offset range to their bounds.
1140 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
1141 (builtin_access::overlap): Avoid setting the size of overlap if it's
1143 (maybe_diag_overlap): Also consider arrays when deciding what values
1144 of offsets to include in diagnostics.
1146 2018-02-14 Martin Sebor <msebor@redhat.com>
1149 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
1150 that correspond to the kind of a declaration.
1152 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
1155 * config/pa/pa.md: Load address of PIC label using the linkage table
1156 if the label is nonlocal.
1158 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
1160 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
1161 warning message if user requests -maltivec=be.
1162 * doc/invoke.texi: Document deprecation of -maltivec=be.
1164 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
1167 * config/rs6000/rs6000-c.c: Update definitions for
1168 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
1169 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
1171 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1174 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
1175 add _get_ssp intrinsics. Remove argument from
1176 __builtin_ia32_rdssp[d|q].
1177 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
1178 * config/i386/i386-builtin.def: Remove argument from
1179 __builtin_ia32_rdssp[d|q].
1180 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
1181 ix86_expand_special_args_builtin for _rdssp[d|q].
1182 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
1183 Clear register before usage.
1184 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
1185 Add documentation for new _get_ssp and _inc_ssp intrinsics.
1187 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
1189 PR tree-optimization/84357
1190 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
1191 operand 1 of an ARRAY_REF too.
1193 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
1196 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
1197 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1199 (set_of_reg): New struct.
1200 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
1201 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
1202 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1204 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
1205 Split into bitclr, bitset, bitinvert patterns if appropriate.
1206 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
1207 use rx_fuse_in_memory_bitop.
1208 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
1209 to named insn, correct maximum insn length.
1211 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
1214 * machmode.def: Define a complex mode for PARTIAL_INT.
1215 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
1217 * doc/rtl.texi: Document CSPImode.
1218 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
1220 (msp430_hard_regno_nregs_with_padding): Likewise.
1222 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
1225 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
1227 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
1229 PR rtl-optimization/84169
1230 * combine.c (try_combine): New variable split_i2i3. Set it to true if
1231 we generated a parallel as new i3 and we split that to new i2 and i3
1232 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
1233 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
1234 those to i2, not i1. Partially rewrite this scan code.
1236 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1239 * stor-layout.c (place_field): For variable length fields, adjust
1240 offset_align afterwards not just based on the field's alignment,
1241 but also on the size.
1244 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
1245 of exps and logs in the use_exp2 case.
1247 2018-02-13 Jeff Law <law@redhat.com>
1249 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
1252 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
1255 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
1259 * common.opt (gas-loc-support, gas-locview-support): New.
1260 (ginline-points, ginternal-reset-location-views): New.
1261 * doc/invoke.texi: Document them. Use @itemx where intended.
1262 (gvariable-location-views): Adjust.
1263 * target.def (reset_location_view): New.
1264 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
1265 (TARGET_RESET_LOCATION_VIEW): New.
1266 * doc/tm.texi: Rebuilt.
1267 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
1268 (dwarf2out_default_as_locview_support): New.
1269 (output_asm_line_debug_info): Use option variables.
1270 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
1271 (output_loc_list): Likewise.
1272 (add_high_low_attributes): Check option variables.
1273 Don't output entry view attribute in strict mode.
1274 (gen_inlined_subroutine_die): Check option variables.
1275 (dwarf2out_inline_entry): Likewise.
1276 (init_sections_and_labels): Likewise.
1277 (dwarf2out_early_finish): Likewise.
1278 (maybe_reset_location_view): New, from...
1279 (dwarf2out_var_location): ... here. Call it.
1280 * debug.h (dwarf2out_default_as_loc_support): Declare.
1281 (dwarf2out_default_as_locview_support): Declare.
1282 * hooks.c (hook_int_rtx_insn_0): New.
1283 * hooks.h (hook_int_rtx_insn_0): Declare.
1284 * toplev.c (process_options): Take -gas-loc-support and
1285 -gas-locview-support from dwarf2out. Enable
1286 -gvariable-location-views by default only with locview
1287 assembler support. Enable -ginternal-reset-location-views by
1288 default only if the target defines the corresponding hook.
1289 Enable -ginline-points by default if location views are
1290 enabled; force it disabled if statement frontiers are
1292 * tree-inline.c (expand_call_inline): Check option variables.
1293 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1295 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
1297 PR tree-optimization/84321
1298 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
1299 handling. Also check whether the anti-range contains any values
1300 that satisfy the mask; switch to a VR_RANGE if not.
1302 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
1305 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
1307 2018-02-13 Martin Jambor <mjambor@suse.cz>
1310 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
1311 of call statements, also set location of a load to a temporary.
1313 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
1315 * config/rl78/rl78.c (add_vector_labels): New function.
1316 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
1317 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
1318 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
1319 which checks that no arguments are passed.
1320 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
1321 * doc/extend.texi: Documentation for the new attribute.
1323 2018-02-13 Andreas Schwab <schwab@suse.de>
1325 * config/riscv/linux.h (CPP_SPEC): Define.
1327 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1330 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
1331 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
1332 OPTION_MASK_ISA_AES as first argument to def_builtin_const
1333 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
1334 instead of OPTION_MASK_ISA_PCLMUL as first argument to
1335 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
1336 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
1337 temporarily for AES and PCLMUL builtins.
1339 PR tree-optimization/84339
1340 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
1341 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
1345 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
1346 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
1347 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
1349 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
1351 * omp-simd-clone.h: New file.
1352 * omp-simd-clone.c: Include omp-simd-clone.h.
1353 (expand_simd_clones): No longer static.
1354 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
1355 cgraph.h and omp-simd-clone.h.
1356 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
1357 (vect_recog_widen_shift_pattern): Formatting fix.
1358 (vect_pattern_recog_1): Don't check optab for calls.
1361 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
1362 operands[2] into a REG before using gen_lowpart on it.
1364 2018-02-12 Jeff Law <law@redhat.com>
1367 * config/sh/sh.c (find_barrier): Consider a sibling call
1370 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
1371 successfully back substituting a reg.
1373 2018-02-12 Richard Biener <rguenther@suse.de>
1375 PR tree-optimization/84037
1376 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
1377 parameter, move visited init to caller.
1378 (vect_slp_analyze_operations): Separate cost from validity
1379 check, initialize visited once for all instances.
1380 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
1382 * tree-vect-stmts.c (vect_model_simple_cost): Make early
1384 (vect_model_promotion_demotion_cost): Likewise.
1385 (vectorizable_bswap): Guard cost modeling with !slp_node
1386 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
1388 (vectorizable_call): Likewise.
1389 (vectorizable_conversion): Likewise.
1390 (vectorizable_assignment): Likewise.
1391 (vectorizable_shift): Likewise.
1392 (vectorizable_operation): Likewise.
1393 (vectorizable_store): Likewise.
1394 (vectorizable_load): Likewise.
1395 (vectorizable_condition): Likewise.
1396 (vectorizable_comparison): Likewise.
1398 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
1401 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
1402 (ASAN_MARK): Fix fnspec to account for return value, change pointer
1403 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
1405 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1408 * params.def (inline-min-speedup): Increase from 8 to 15.
1409 (max-inline-insns-auto): Decrease from 40 to 30.
1410 * ipa-split.c (consider_split): Add some buffer for function to
1411 be considered inlining candidate.
1412 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
1415 2018-02-12 Richard Biener <rguenther@suse.de>
1417 PR tree-optimization/84037
1418 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
1419 matched stmts if we cannot swap the non-matched ones.
1421 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1423 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
1424 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
1425 _mm_maskz_scalef_round_ss): New intrinsics.
1426 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
1427 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
1428 __builtin_ia32_scalefss_round): Remove.
1429 (__builtin_ia32_scalefsd_mask_round,
1430 __builtin_ia32_scalefss_mask_round): New intrinsics.
1431 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
1432 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
1433 ((match_operand:VF_128 2 "<round_nimm_predicate>"
1434 "<round_constraint>")): Changed to ...
1435 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
1436 "<round_scalar_constraint>")): ... this.
1437 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
1438 %0, %1, %2<round_op3>}"): Changed to ...
1439 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
1440 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
1441 %2<round_scalar_mask_op3>}"): ... this.
1442 * config/i386/subst.md (round_scalar_nimm_predicate): New.
1444 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1446 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
1447 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
1448 (_mm_maskz_sqrt_round_ss): New intrinsics.
1449 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
1450 (__builtin_ia32_sqrtsd_mask_round)
1451 (__builtin_ia32_sqrtss_mask_round): New builtins.
1452 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
1453 (__builtin_ia32_sqrtss_round): Remove.
1454 (__builtin_ia32_sqrtsd_mask_round)
1455 (__builtin_ia32_sqrtss_mask_round): New builtins.
1456 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
1457 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
1458 ((match_operand:VF_128 1 "vector_operand"
1459 "xBm,<round_constraint>")): Changed to ...
1460 ((match_operand:VF_128 1 "vector_operand"
1461 "xBm,<round_scalar_constraint>")): ... this.
1462 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
1463 %0, %2, %<iptr>1<round_op3>}): Changed to ...
1464 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
1465 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
1466 %<iptr>1<round_scalar_mask_op3>}): ... this.
1467 ((set_attr "prefix" "<round_prefix>")): Changed to ...
1468 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
1470 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
1473 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
1474 Cast vec_cmpeq result to correct type.
1475 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
1476 Cast vec_cmpgt result to correct type.
1478 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
1480 * final.c (final_scan_insn_1): Renamed from...
1481 (final_scan_insn): ... this. New wrapper, to recover
1482 seen from the outermost call in recursive ones.
1483 * config/sparc/sparc.c (output_return): Drop seen from call.
1484 (output_sibcall): Likewise.
1485 * config/visium/visium.c (output_branch): Likewise.
1487 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
1489 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
1492 2018-02-10 Alan Modra <amodra@gmail.com>
1495 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
1496 Specify LR as an input.
1498 2018-02-10 Jakub Jelinek <jakub@redhat.com>
1501 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
1502 remove_member_access_dummy_vars): New functions.
1503 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
1504 lower_omp_1, execute_lower_omp): Use them.
1506 PR rtl-optimization/84308
1507 * shrink-wrap.c (spread_components): Release todo vector.
1509 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
1511 PR rtl-optimization/57193
1512 * ira-color.c (struct allocno_color_data): Add member
1513 conflict_allocno_hard_prefs.
1514 (update_conflict_allocno_hard_prefs): New.
1515 (bucket_allocno_compare_func): Add a preference based on
1516 conflict_allocno_hard_prefs.
1517 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
1518 (color_allocnos): Remove a dead code. Initiate
1519 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
1521 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1524 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
1525 constraint from =wa to wa. Avoid a subreg on the output operand,
1526 instead use a pseudo and subreg it in a move.
1527 (p9_xxbrd_<mode>): Changed to ...
1528 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
1529 (p9_xxbrd_v2df): New expander.
1530 (p9_xxbrw_<mode>): Changed to ...
1531 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
1532 (p9_xxbrw_v4sf): New expander.
1534 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1536 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
1538 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
1541 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
1542 multiply in 32-bit mode.
1543 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
1544 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
1547 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1549 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
1550 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
1551 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
1552 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
1554 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
1557 * dwarf2out.c (is_trivial_indirect_ref): New function.
1558 (dwarf2out_late_global_decl): Do not generate a location
1559 attribute for variables that have a non-trivial DECL_VALUE_EXPR
1560 and that are not defined in the current unit.
1562 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1564 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
1565 instead of a libcall for UNORDERED.
1567 2018-02-09 Tamar Christina <tamar.christina@arm.com>
1570 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
1571 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
1573 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1576 * config/s390/s390.c (s390_set_current_function): Invoke
1577 s390_indirect_branch_settings also if fndecl didn't change.
1579 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1581 * config/rs6000/rs6000.md (blockage): Set length to zero.
1583 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1585 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
1587 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1590 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
1591 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
1595 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
1596 PARALLEL incoming that failed vt_get_decl_and_offset check.
1599 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
1600 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
1602 (get_variable_section): For decls in named .bss* sections pass true as
1603 second argument to bss_initializer_p.
1605 2018-02-09 Marek Polacek <polacek@redhat.com>
1606 Jakub Jelinek <jakub@redhat.com>
1609 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
1610 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
1611 Sync some changes from cxx_fold_indirect_ref.
1613 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1615 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
1617 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
1618 (BLOCK_INLINE_ENTRY_LABEL): New.
1619 (dwarf2out_var_location): Disregard inline entry markers.
1620 (inline_entry_data): New struct.
1621 (inline_entry_data_hasher): New hashtable type.
1622 (inline_entry_data_hasher::hash): New.
1623 (inline_entry_data_hasher::equal): New.
1624 (inline_entry_data_table): New variable.
1625 (add_high_low_attributes): Add DW_AT_entry_pc and
1626 DW_AT_GNU_entry_view attributes if a pending entry is found
1627 in inline_entry_data_table. Add old entry_pc attribute only
1628 if debug nonbinding markers are disabled.
1629 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
1630 markers are enabled.
1631 (block_within_block_p, dwarf2out_inline_entry): New.
1632 (dwarf2out_finish): Check that no entries remained in
1633 inline_entry_data_table.
1634 * final.c (reemit_insn_block_notes): Handle inline entry notes.
1635 (final_scan_insn, notice_source_line): Likewise.
1636 (rest_of_clean_state): Skip inline entry markers.
1637 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
1639 * gimple.c (gimple_build_debug_inline_entry): New.
1640 * gimple.h (enum gimple_debug_subcode): Add
1641 GIMPLE_DEBUG_INLINE_ENTRY.
1642 (gimple_build_debug_inline_entry): Declare.
1643 (gimple_debug_inline_entry_p): New.
1644 (gimple_debug_nonbind_marker_p): Adjust.
1645 * insn-notes.def (INLINE_ENTRY): New.
1646 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
1647 inline entry marker notes.
1648 (print_insn): Likewise.
1649 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
1650 (INSN_DEBUG_MARKER_KIND): Likewise.
1651 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
1652 * tree-inline.c (expand_call_inline): Build and insert
1653 debug_inline_entry stmt.
1654 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
1655 inline entry blocks early, if nonbind markers are enabled.
1656 (dump_scope_block): Dump fragment info.
1657 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
1658 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
1659 (gimple_build_debug_inline_entry): New.
1660 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
1661 Enable/disable inline entry points too.
1662 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
1663 (DEBUG_INSN): Describe inline entry markers.
1665 * common.opt (gvariable-location-views): New.
1666 (gvariable-location-views=incompat5): New.
1667 * config.in: Rebuilt.
1668 * configure: Rebuilt.
1669 * configure.ac: Test assembler for view support.
1670 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
1671 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
1672 * dwarf2out.c (var_loc_view): New typedef.
1673 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
1674 (dwarf2out_locviews_in_attribute): New.
1675 (dwarf2out_locviews_in_loclist): New.
1676 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
1677 (enum dw_line_info_opcode): Add LI_adv_address.
1678 (struct dw_line_info_table): Add view.
1679 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
1680 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
1681 (zero_view_p): New variable.
1682 (ZERO_VIEW_P): New macro.
1683 (output_asm_line_debug_info): New.
1684 (struct var_loc_node): Add view.
1685 (add_AT_view_list, AT_loc_list): New.
1686 (add_var_loc_to_decl): Add view param. Test it against last.
1687 (new_loc_list): Add view params. Record them.
1688 (AT_loc_list_ptr): Handle loc and view lists.
1689 (view_list_to_loc_list_val_node): New.
1690 (print_dw_val): Handle dw_val_class_view_list.
1691 (size_of_die): Likewise.
1692 (value_format): Likewise.
1693 (loc_list_has_views): New.
1694 (gen_llsym): Set vl_symbol too.
1695 (maybe_gen_llsym, skip_loc_list_entry): New.
1696 (dwarf2out_maybe_output_loclist_view_pair): New.
1697 (output_loc_list): Output view list or entries too.
1698 (output_view_list_offset): New.
1699 (output_die): Handle dw_val_class_view_list.
1700 (output_dwarf_version): New.
1701 (output_compilation_unit_header): Use it.
1702 (output_skeleton_debug_sections): Likewise.
1703 (output_rnglists, output_line_info): Likewise.
1704 (output_pubnames, output_aranges): Update version comments.
1705 (output_one_line_info_table): Output view numbers in asm comments.
1706 (dw_loc_list): Determine current endview, pass it to new_loc_list.
1707 Call maybe_gen_llsym.
1708 (loc_list_from_tree_1): Adjust.
1709 (add_AT_location_description): Create view list attribute if
1710 needed, check it's absent otherwise.
1711 (convert_cfa_to_fb_loc_list): Adjust.
1712 (maybe_emit_file): Call output_asm_line_debug_info for test.
1713 (dwarf2out_var_location): Reset views as needed. Precompute
1714 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
1715 attribute. Set view.
1716 (new_line_info_table): Reset next view.
1717 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
1718 (dwarf2out_source_line): Likewise. Output view resets and labels to
1719 the assembler, or select appropriate line info opcodes.
1720 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
1721 (optimize_string_length): Catch it. Adjust.
1722 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
1723 dw_val_class_view_list, and remove it if no longer needed.
1724 (hash_loc_list): Hash view numbers.
1725 (loc_list_hasher::equal): Compare them.
1726 (optimize_location_lists): Check whether a view list symbol is
1727 needed, and whether the locview attribute is present, and
1728 whether they match. Remove the locview attribute if no longer
1730 (index_location_lists): Call skip_loc_list_entry for test.
1731 (dwarf2out_finish): Call output_asm_line_debug_info for test.
1732 Use output_dwarf_version.
1733 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
1734 (struct dw_val_node): Add val_view_list.
1735 * final.c (SEEN_NEXT_VIEW): New.
1736 (set_next_view_needed): New.
1737 (clear_next_view_needed): New.
1738 (maybe_output_next_view): New.
1739 (final_start_function): Rename to...
1740 (final_start_function_1): ... this. Take pointer to FIRST,
1741 add SEEN parameter. Emit param bindings in the initial view.
1742 (final_start_function): Reintroduce SEEN-less interface.
1743 (final): Rename to...
1744 (final_1): ... this. Take SEEN parameter. Output final pending
1745 next view at the end.
1746 (final): Reintroduce seen-less interface.
1747 (final_scan_insn): Output pending next view before switching
1748 sections or ending a block. Mark the next view as needed when
1749 outputting variable locations. Notify debug backend of section
1750 changes, and of location view changes.
1751 (rest_of_handle_final): Adjust.
1752 * toplev.c (process_options): Autodetect value for debug variable
1753 location views option. Warn on incompat5 without -gdwarf-5.
1754 * doc/invoke.texi (gvariable-location-views): New.
1755 (gvariable-location-views=incompat5): New.
1756 (gno-variable-location-views): New.
1758 2018-02-08 David Malcolm <dmalcolm@redhat.com>
1760 PR tree-optimization/84136
1761 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
1762 that the result of find_edge is non-NULL.
1764 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1767 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
1768 storing integer register in SImode. Fix cost of 256 and 512
1769 byte aligned SSE register store.
1771 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1773 * config/i386/i386.c (ix86_multiplication_cost): Fix
1774 multiplication cost for TARGET_AVX512DQ.
1776 2018-02-08 Marek Polacek <polacek@redhat.com>
1778 PR tree-optimization/84238
1779 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
1782 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1784 PR tree-optimization/84265
1785 * tree-vect-stmts.c (vectorizable_store): Don't treat
1786 VMAT_CONTIGUOUS accesses as grouped.
1787 (vectorizable_load): Likewise.
1789 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1791 PR tree-optimization/81635
1792 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
1793 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
1794 (test_round_for_mask): New functions.
1795 (wide_int_cc_tests): Call test_round_for_mask.
1796 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
1797 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
1798 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
1799 range returned by get_range_info.
1801 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1804 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
1805 * symtab.c: Include builtins.h
1806 (symtab_node::output_to_lto_symbol_table_p): Move here
1807 from lto-streamer-out.c:output_symbol_p.
1808 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
1809 (output_symbol_p): Move all logic to symtab.c
1810 (produce_symtab): Update.
1812 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1814 * config/s390/s390-opts.h (enum indirect_branch): Define.
1815 * config/s390/s390-protos.h (s390_return_addr_from_memory)
1816 (s390_indirect_branch_via_thunk)
1817 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
1818 (enum s390_indirect_branch_type): Define.
1819 * config/s390/s390.c (struct s390_frame_layout, struct
1820 machine_function): Remove.
1821 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
1822 (indirect_branch_table_label_no, indirect_branch_table_name):
1824 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
1825 (enum s390_indirect_branch_option): Define.
1826 (s390_return_addr_from_memory): New function.
1827 (s390_handle_string_attribute): New function.
1828 (s390_attribute_table): Add new attribute handler.
1829 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
1830 (s390_indirect_branch_via_thunk): New function.
1831 (s390_indirect_branch_via_inline_thunk): New function.
1832 (s390_function_ok_for_sibcall): When jumping via thunk disallow
1833 sibling call optimization for non z10 compiles.
1834 (s390_emit_call): Force indirect branch target to be a single
1835 register. Add r1 clobber for non-z10 compiles.
1836 (s390_emit_epilogue): Emit return jump via return_use expander.
1837 (s390_reorg): Handle JUMP_INSNs as execute targets.
1838 (s390_option_override_internal): Perform validity checks for the
1839 new command line options.
1840 (s390_indirect_branch_attrvalue): New function.
1841 (s390_indirect_branch_settings): New function.
1842 (s390_set_current_function): Invoke s390_indirect_branch_settings.
1843 (s390_output_indirect_thunk_function): New function.
1844 (s390_code_end): Implement target hook.
1845 (s390_case_values_threshold): Implement target hook.
1846 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
1848 * config/s390/s390.h (struct s390_frame_layout)
1849 (struct machine_function): Move here from s390.c.
1850 (TARGET_INDIRECT_BRANCH_NOBP_RET)
1851 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
1852 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
1853 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
1854 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
1855 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
1856 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
1857 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
1858 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
1859 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
1860 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
1861 (mnemonic attribute): Add values which aren't recognized
1863 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
1864 pattern for branch conversion. Fix mnemonic attribute.
1865 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
1866 indirect branch via thunk if requested.
1867 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
1868 ("*indirect_jump"): Disable for branch conversion using out of
1870 ("indirect_jump_via_thunk<mode>_z10")
1871 ("indirect_jump_via_thunk<mode>")
1872 ("indirect_jump_via_inlinethunk<mode>_z10")
1873 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
1874 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
1875 ("casesi_jump_via_inlinethunk<mode>_z10")
1876 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
1877 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
1878 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
1879 ("*indirect2_jump"): Disable for branch conversion.
1880 ("casesi_jump"): Turn into expander and expand patterns for branch
1882 ("return_use"): New expander.
1883 ("*return"): Emit return via thunk and rename it to ...
1884 ("*return<mode>"): ... this one.
1885 * config/s390/s390.opt: Add new options and and enum for the
1888 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1890 * lra-constraints.c (match_reload): Unconditionally use
1891 gen_lowpart_SUBREG, rather than selecting between that
1892 and equivalent gen_rtx_SUBREG code.
1894 2018-02-08 Richard Biener <rguenther@suse.de>
1896 PR tree-optimization/84233
1897 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
1898 changed flag instead of boguously re-using phi_inserted.
1900 2018-02-08 Martin Jambor <mjambor@suse.cz>
1902 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
1903 static local variables.
1905 2018-02-08 Richard Biener <rguenther@suse.de>
1907 PR tree-optimization/84278
1908 * tree-vect-stmts.c (vectorizable_store): When looking for
1909 smaller vector types to perform grouped strided loads/stores
1910 make sure the mode is supported by the target.
1911 (vectorizable_load): Likewise.
1913 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1915 * config/aarch64/aarch64.c (aarch64_components_for_bb):
1916 Increase LDP/STP opportunities by adding adjacent callee-saves.
1918 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1920 PR rtl-optimization/84068
1921 PR rtl-optimization/83459
1922 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
1924 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
1926 PR tree-optimization/84224
1927 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
1928 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
1931 2018-02-07 Iain Sandoe <iain@codesourcery.com>
1934 * config/rs6000/altivec.md (*restore_world): Remove LR use.
1935 * config/rs6000/predicates.md (restore_world_operation): Adjust op
1936 count, remove one USE.
1938 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
1940 * doc/install.texi (Configuration): Document the
1941 --with-long-double-format={ibm,ieee} PowerPC configuration
1945 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
1946 Convert from define_expand to be define_insn_and_split. Rework
1947 float/double/_Float128 conversions to QI/HI/SImode to work with
1948 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
1949 conversions to QI/HImode types did a store and then a load to
1950 truncate the value. For conversions to VSX registers, don't split
1951 the insn, instead emit the code directly. Use the code iterator
1952 any_fix to combine signed and unsigned conversions.
1953 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
1954 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
1955 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
1956 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
1957 (fix_<mode>di2_hw): Likewise.
1958 (fixuns_<mode>di2_hw): Likewise.
1959 (fix_<mode>si2_hw): Likewise.
1960 (fixuns_<mode>si2_hw): Likewise.
1961 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
1962 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
1963 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
1964 fix<uns>_trunc<SFDF:mode>si2_p8.
1965 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
1967 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
1968 (fix<uns>_<mode>_mem): Likewise.
1969 (fctiw<u>z_<mode>_mem): Likewise.
1970 (fix<uns>_<mode>_mem): Likewise.
1971 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
1972 the register allocator from doing a direct move to the GPRs to do
1973 a store, and instead use the ISA 3.0 store byte/half-word from
1974 vector register instruction. For IEEE 128-bit floating point,
1975 also optimize stores of 32-bit ints.
1976 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
1978 2018-02-07 Alan Hayward <alan.hayward@arm.com>
1980 * genextract.c (push_pathstr_operand): New function to support
1982 (walk_rtx): Call push_pathstr_operand.
1983 (print_path): Support [a-zA-Z].
1985 2018-02-07 Richard Biener <rguenther@suse.de>
1987 PR tree-optimization/84037
1988 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
1989 (cse_and_gimplify_to_preheader): Declare.
1990 (vect_get_place_in_interleaving_chain): Likewise.
1991 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
1993 (_loop_vec_info::~_loop_vec_info): Delete it.
1994 (cse_and_gimplify_to_preheader): New function.
1995 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
1996 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
1997 (vectorizable_load): Likewise. For grouped stores always base
1998 the IV on the first element.
1999 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
2000 condition before gimplifying.
2002 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2004 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
2005 *DIV_EXPR and *MOD_EXPR.
2007 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
2010 * config/i386/i386.c (ix86_option_override_internal): Mask out
2011 the CF_SET bit when checking -fcf-protection.
2013 2018-02-07 Tom de Vries <tom@codesourcery.com>
2016 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
2019 2018-02-07 Richard Biener <rguenther@suse.de>
2021 PR tree-optimization/84204
2022 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2025 PR tree-optimization/84205
2026 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
2027 special-case isl_ast_op_zdiv_r.
2029 PR tree-optimization/84223
2030 * graphite-scop-detection.c (gather_bbs::before_dom_children):
2031 Only add conditions from within the region.
2032 (gather_bbs::after_dom_children): Adjust.
2034 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
2037 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
2038 * config/avr/avr.md: Only post-reload split REG-REG moves if
2039 either register is REGERAL_REG_P.
2041 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2043 PR tree-optimization/84235
2044 * tree-ssa-scopedtables.c
2045 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
2046 if the subtraction is performed in floating point type where NaNs are
2047 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
2048 build 1. Formatting fix.
2050 2018-02-06 Jakub Jelinek <jakub@redhat.com>
2053 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
2054 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
2055 and skip it regardless of bb boundaries. Use CALL_P macro,
2056 don't test INSN_P (insn) together with CALL_P or JUMP_P check
2057 unnecessarily, formatting fix.
2059 2018-02-06 Michael Collison <michael.collison@arm.com>
2061 * config/arm/thumb2.md:
2062 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
2063 (*thumb_mov_notscc): Ditto.
2065 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2068 * config/rs6000/rs6000.md (su code attribute): Use "u" for
2069 unsigned_fix, not "s".
2071 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2073 * configure.ac (gcc_fn_eh_frame_ro): New function.
2074 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
2075 correct .eh_frame permissions.
2076 * configure: Regenerate.
2078 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
2080 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
2083 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2085 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2086 Display warning message for -mno-speculate-indirect-jumps.
2088 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
2090 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
2092 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
2094 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
2096 PR tree-optimization/84225
2097 * tree-eh.c (find_trapping_overflow): Only call
2098 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
2100 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2103 * config/i386/i386.c: Reimplement the check of possible options
2104 -mibt/-mshstk conbination. Change error messages.
2105 * doc/invoke.texi: Fix a typo: remove extra '='.
2107 2018-02-06 Marek Polacek <polacek@redhat.com>
2109 PR tree-optimization/84228
2110 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
2112 2018-02-06 Tamar Christina <tamar.christina@arm.com>
2115 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
2116 emitted arch directives.
2117 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
2118 __ARM_FEATURE_COPROC before changing architectures.
2120 2018-02-06 Richard Biener <rguenther@suse.de>
2122 * config/i386/i386.c (print_reg): Fix typo.
2123 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
2125 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
2127 * configure: Regenerate.
2129 2018-02-05 Martin Sebor <msebor@redhat.com>
2131 PR tree-optimization/83369
2132 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
2135 2018-02-05 Martin Liska <mliska@suse.cz>
2137 * doc/invoke.texi: Cherry-pick upstream r323995.
2139 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
2141 * ira.c (ira_init_register_move_cost): Adjust comment.
2143 2018-02-05 Martin Liska <mliska@suse.cz>
2145 PR gcov-profile/84137
2146 * doc/gcov.texi: Fix typo in documentation.
2148 2018-02-05 Martin Liska <mliska@suse.cz>
2150 PR gcov-profile/83879
2151 * doc/gcov.texi: Document necessity of --dynamic-list-data when
2152 using dlopen functionality.
2154 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
2156 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
2157 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
2158 _mm_maskz_range_ss, _mm_mask_range_round_ss,
2159 _mm_maskz_range_round_ss): New intrinsics.
2160 (__builtin_ia32_rangesd128_round)
2161 (__builtin_ia32_rangess128_round): Remove.
2162 (__builtin_ia32_rangesd128_mask_round,
2163 __builtin_ia32_rangess128_mask_round): New builtins.
2164 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
2165 __builtin_ia32_rangess128_round): Remove.
2166 (__builtin_ia32_rangesd128_mask_round,
2167 __builtin_ia32_rangess128_mask_round): New builtins.
2168 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
2169 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
2170 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
2171 "<round_saeonly_constraint>")): Changed to ...
2172 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
2173 "<round_saeonly_scalar_constraint>")): ... this.
2174 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
2175 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
2176 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
2177 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
2178 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
2180 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
2182 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
2184 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
2185 Remove all values except native, 8540 and 8548.
2187 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
2189 * config/i386/i386.c (ix86_output_function_return): Pass
2190 INVALID_REGNUM, instead of -1, as invalid register number to
2191 indirect_thunk_name and output_indirect_thunk.
2193 2018-02-02 Julia Koval <julia.koval@intel.com>
2195 * config.gcc: Add -march=icelake.
2196 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
2197 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
2198 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
2199 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
2200 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
2201 (processor_target_table): Add icelake.
2202 (ix86_option_override_internal): Handle new PTAs.
2203 (get_builtin_code_for_version): Handle icelake.
2204 (M_INTEL_COREI7_ICELAKE): New.
2205 (fold_builtin_cpu): Handle icelake.
2206 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
2207 * doc/invoke.texi: Add -march=icelake.
2209 2018-02-02 Julia Koval <julia.koval@intel.com>
2211 * config/i386/i386.c (ix86_option_override_internal): Change flags type
2212 to wide_int_bitmask.
2213 * wide-int-bitmask.h: New.
2215 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2218 * config/i386/i386.md: Replace Pmode with word_mode in
2219 builtin_setjmp_setup and builtin_longjmp to support x32.
2221 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
2225 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
2227 (rs6000_supported_cpu_names): New static variable.
2228 (linux_cpu_translation_table): Likewise.
2229 (elf_platform) <cpu>: Define new static variable and use it.
2230 Translate kernel AT_PLATFORM name to canonical name if needed.
2231 Error if platform name is unknown.
2233 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
2236 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
2238 2018-02-01 Jeff Law <law@redhat.com>
2241 * config/i386/i386.c (release_scratch_register_on_entry): Add new
2242 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
2243 the scratch if RELEASE_VIA_POP is false.
2244 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
2245 If we have to save a temporary register, decrement SIZE appropriately.
2246 Pass new arguments to release_scratch_register_on_entry.
2247 (ix86_adjust_stack_and_probe): Likewise.
2248 (ix86_emit_probe_stack_range): Pass new arguments to
2249 release_scratch_register_on_entry.
2251 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
2253 PR rtl-optimization/84157
2254 * combine.c (change_zero_ext): Use REG_P predicate in
2255 front of HARD_REGISTER_P predicate.
2257 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
2259 * config/avr/avr.c (avr_option_override): Move disabling of
2260 -fdelete-null-pointer-checks to...
2261 * common/config/avr/avr-common.c (avr_option_optimization_table):
2264 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2266 PR tree-optimization/81635
2267 * tree-data-ref.c (split_constant_offset_1): For types that
2268 wrap on overflow, try to use range info to prove that wrapping
2271 2018-02-01 Renlin Li <renlin.li@arm.com>
2274 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
2276 (aarch64_register_move_cost): Likewise.
2277 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
2279 (REG_CLASS_NAMES): Likewise.
2280 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
2281 TAILCALL_ADDR_REGS. Remove IP registers.
2282 * config/aarch64/aarch64.md (Ucs): Update register constraint.
2284 2018-02-01 Richard Biener <rguenther@suse.de>
2286 * domwalk.h (dom_walker::dom_walker): Add additional constructor
2287 for specifying RPO order and allow NULL for that.
2288 * domwalk.c (dom_walker::dom_walker): Likewise.
2289 (dom_walker::walk): Handle NULL RPO order.
2290 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
2292 (rewrite_update_dom_walker): Likewise.
2293 (mark_def_dom_walker): Likewise.
2295 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2297 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
2298 (aarch64_maybe_expand_sve_subreg_move): Declare.
2299 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
2300 * config/aarch64/predicates.md (aarch64_any_register_operand): New
2302 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
2303 that are semantically a reverse operation.
2304 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
2305 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
2306 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
2308 (aarch64_can_change_mode_class): For big-endian, forbid changes
2309 between two SVE modes if they have different element sizes.
2311 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2313 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
2314 the TImode handling for big-endian targets.
2316 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2318 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
2319 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
2321 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
2322 Remove BSWAP handing for big-endian targets and use the form of
2323 LD1RQ appropariate for the mode.
2325 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2327 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
2328 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
2331 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2334 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
2335 check for operands that need to go through aarch64_sve_reload_be.
2337 2018-02-01 Jakub Jelinek <jakub@redhat.com>
2339 PR tree-optimization/81661
2340 PR tree-optimization/84117
2341 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
2342 * tree-eh.c: Include gimplify.h.
2343 (find_trapping_overflow, replace_trapping_overflow,
2344 rewrite_to_non_trapping_overflow): New functions.
2345 * tree-vect-loop.c: Include tree-eh.h.
2346 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
2347 * tree-data-ref.c: Include tree-eh.h.
2348 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
2350 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
2352 PR rtl-optimization/84123
2353 * combine.c (change_zero_ext): Check if hard register satisfies
2354 can_change_dest_mode before calling gen_lowpart_SUBREG.
2356 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
2359 * ira.c (ira_init_register_move_cost): Remove assert.
2361 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2363 PR rtl-optimization/84071
2364 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
2365 * doc/tm.texi: Regenerate.
2367 2018-01-31 Richard Biener <rguenther@suse.de>
2369 PR tree-optimization/84132
2370 * tree-data-ref.c (analyze_miv_subscript): Properly
2371 check whether evolution_function_is_affine_multivariate_p
2372 before calling gcd_of_steps_may_divide_p.
2374 2018-01-31 Julia Koval <julia.koval@intel.com>
2377 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
2378 * config/i386/i386.md (rdpid_rex64) New.
2379 (rdpid): Make 32bit only.
2381 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
2384 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
2385 an IDENTIFIER_NODE for FUNCTION_TYPE's.
2387 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2390 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
2392 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
2394 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2396 PR rtl-optimization/84071
2397 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
2398 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
2400 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2402 * config/arc/arc.c (arc_handle_aux_attribute): New function.
2403 (arc_attribute_table): Add 'aux' attribute.
2404 (arc_in_small_data_p): Consider aux like variables.
2405 (arc_is_aux_reg_p): New function.
2406 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
2407 (arc_get_aux_arg): New function.
2408 (prepare_move_operands): Handle aux-register access.
2409 (arc_handle_aux_attribute): New function.
2410 * doc/extend.texi (ARC Variable attributes): Add subsection.
2412 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2414 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
2415 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
2416 (arc_attribute_table): Add 'uncached' attribute.
2417 (arc_print_operand): Print '.di' flag for uncached memory
2419 (arc_in_small_data_p): Do not consider for small data the uncached
2421 (arc_is_uncached_mem_p): New function.
2422 * config/arc/predicates.md (compact_store_memory_operand): Check
2423 for uncached memory accesses.
2424 (nonvol_nonimm_operand): Likewise.
2425 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
2427 2018-01-31 Jakub Jelinek <jakub@redhat.com>
2430 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
2431 falign-loops=): Add Optimization flag.
2433 2018-01-30 Jeff Law <law@redhat.com>
2436 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
2437 INT_REGISTERS_SAVED. Check it prior to calling
2438 get_scratch_register_on_entry.
2439 (ix86_adjust_stack_and_probe): Similarly.
2440 (ix86_emit_probe_stack_range): Similarly.
2441 (ix86_expand_prologue): Corresponding changes.
2443 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2446 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
2447 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
2449 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
2452 * lra-constraints.c (curr_insn_transform): Process AND in the
2455 2018-01-30 Jakub Jelinek <jakub@redhat.com>
2457 PR rtl-optimization/83986
2458 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
2459 dependence against last_pending_memory_flush in addition to
2462 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
2464 PR tree-optimization/81611
2465 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
2468 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2471 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
2474 2018-01-30 Richard Biener <rguenther@suse.de>
2475 Jakub Jelinek <jakub@redhat.com>
2477 PR tree-optimization/84111
2478 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
2479 inner loops added during recursion, as they don't have up-to-date
2482 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2485 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
2486 (can_inline_edge_by_limits_p): ... here.
2487 (can_early_inline_edge_p, check_callers,
2488 update_caller_keys, update_callee_keys, recursive_inlining,
2489 add_new_edges_to_heap, speculation_useful_p,
2490 inline_small_functions,
2491 inline_small_functions, flatten_function,
2492 inline_to_all_callers_1): Update.
2494 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2496 * profile-count.c (profile_count::combine_with_ipa_count): Handle
2499 2018-01-30 Richard Biener <rguenther@suse.de>
2501 PR tree-optimization/83008
2502 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
2503 invariant and constant vector uses in stmts when they need
2506 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2509 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
2510 * configure: Regenerate.
2512 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2514 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
2516 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
2517 Use gen_rtx_REG rather than gen_lowpart.
2519 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2521 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
2522 rather than 0 when creating partial subregs.
2524 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2526 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
2529 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
2532 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
2533 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
2534 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
2535 flags. This restores the settings used before the 2017-07-24.
2536 Turning off pre increment/decrement/modify allows IVOPTS to
2537 optimize DF/SF loops where the index is an int.
2539 2018-01-29 Richard Biener <rguenther@suse.de>
2540 Kelvin Nilsen <kelvin@gcc.gnu.org>
2543 * tree-vect-stmts.c (vectorizable_call): Don't call
2544 targetm.vectorize_builtin_md_vectorized_function if callee is
2547 2018-01-22 Carl Love <cel@us.ibm.com>
2549 * doc/extend.tex: Fix typo in second arg in
2550 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
2552 2018-01-29 Richard Biener <rguenther@suse.de>
2554 PR tree-optimization/84086
2555 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
2556 (flush_ssaname_freelist): When SSA names were released reset
2557 the SCEV hash table.
2559 2018-01-29 Richard Biener <rguenther@suse.de>
2561 PR tree-optimization/84057
2562 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
2563 removed paths when removing edges.
2565 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
2567 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
2568 -mfunction-return=@var{choice}.
2570 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
2573 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
2575 (layout::print_source_line): Likewise.
2576 (test_get_line_width_without_trailing_whitespace): Add test cases.
2578 2018-01-27 Jakub Jelinek <jakub@redhat.com>
2581 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
2584 2018-01-26 Jim Wilson <jimw@sifive.com>
2586 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
2588 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
2591 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2593 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
2594 and CMP + SUB-immediate -> SUBS.
2596 2018-01-26 Martin Sebor <msebor@redhat.com>
2598 PR tree-optimization/83896
2599 * tree-ssa-strlen.c (get_string_len): Rename...
2600 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
2601 Avoid assuming length is constant.
2602 (handle_char_store): Use HOST_WIDE_INT for string length.
2604 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
2607 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
2608 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
2610 2018-01-26 Richard Biener <rguenther@suse.de>
2612 PR rtl-optimization/84003
2613 * dse.c (record_store): Only record redundant stores when
2614 the earlier store aliases at least all accesses the later one does.
2616 2018-01-26 Jakub Jelinek <jakub@redhat.com>
2618 PR rtl-optimization/83985
2619 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
2620 REG_CFA_RESTORE insns.
2621 (delete_unmarked_insns): Don't ignore separate shrink wrapping
2622 REG_CFA_RESTORE insns here.
2625 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
2626 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
2628 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2630 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
2631 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
2632 (arc_init): Likewise.
2633 (arc_override_options): Likewise.
2634 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
2636 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
2638 * config/arc/arc.h (TARGET_DBNZ): Define.
2639 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
2640 properly set the tune attribute.
2641 (dbnz): Use TARGET_DBNZ guard.
2642 * config/arc/arc.opt (mtune): Add core3 option.
2644 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2646 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
2647 recognize new pic like addresses.
2648 (arc_delegitimize_address): Clean up.
2650 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2652 * config/arc/arc-arches.def: Option mrf16 valid for all
2654 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
2655 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
2656 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
2657 * config/arc/arc-tables.opt: Regenerate.
2658 * config/arc/arc.c (arc_conditional_register_usage): Handle
2659 reduced register file case.
2660 (arc_file_start): Set must have build attributes.
2661 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
2663 * config/arc/arc.opt (mrf16): Add new option.
2664 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
2665 * config/arc/genmultilib.awk: Handle new mrf16 option.
2666 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
2667 * config/arc/t-multilib: Regenerate.
2668 * doc/invoke.texi (ARC Options): Document mrf16 option.
2670 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2672 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
2673 * config/arc/arc.c (arc_handle_secure_attribute): New function.
2674 (arc_attribute_table): Add 'secure_call' attribute.
2675 (arc_print_operand): Print secure call operand.
2676 (arc_function_ok_for_sibcall): Don't optimize tail calls when
2678 (arc_is_secure_call_p): New function. * config/arc/arc.md
2679 (call_i): Add support for sjli instruction.
2680 (call_value_i): Likewise.
2681 * config/arc/constraints.md (Csc): New constraint.
2683 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2684 John Eric Martin <John.Martin@emmicro-us.com>
2686 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
2687 * config/arc/arc.c (_arc_jli_section): New struct.
2688 (arc_jli_section): New type.
2689 (rc_jli_sections): New static variable.
2690 (arc_handle_jli_attribute): New function.
2691 (arc_attribute_table): Add jli_always and jli_fixed attribute.
2692 (arc_file_end): New function.
2693 (TARGET_ASM_FILE_END): Define.
2694 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
2695 (arc_add_jli_section): New function.
2696 (jli_call_scan): Likewise.
2697 (arc_reorg): Call jli_call_scan.
2698 (arc_output_addsi): Remove 'S' from printing asm operand.
2699 (arc_is_jli_call_p): New function.
2700 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
2702 (movhi_insn): Likewise.
2703 (movsi_insn): Likewise.
2704 (movsi_set_cc_insn): Likewise.
2705 (loadqi_update): Likewise.
2706 (load_zeroextendqisi_update): Likewise.
2707 (load_signextendqisi_update): Likewise.
2708 (loadhi_update): Likewise.
2709 (load_zeroextendhisi_update): Likewise.
2710 (load_signextendhisi_update): Likewise.
2711 (loadsi_update): Likewise.
2712 (loadsf_update): Likewise.
2713 (movsicc_insn): Likewise.
2714 (bset_insn): Likewise.
2715 (bxor_insn): Likewise.
2716 (bclr_insn): Likewise.
2717 (bmsk_insn): Likewise.
2718 (bicsi3_insn): Likewise.
2719 (cmpsi_cc_c_insn): Likewise.
2720 (movsi_ne): Likewise.
2721 (movsi_cond_exec): Likewise.
2722 (clrsbsi2): Likewise.
2733 (call_i): Remove 'S' asm letter, add jli instruction.
2734 (call_value_i): Likewise.
2735 * config/arc/arc.op (mjli-always): New option.
2736 * config/arc/constraints.md (Cji): New constraint.
2737 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
2739 (subsf3_fpx): Likewise.
2740 (mulsf3_fpx): Likewise.
2741 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
2743 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
2745 * doc/invoke.texi (ARC): Document mjli-always option.
2747 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
2749 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
2750 avoid addition with 0 and use incw and decw where possible.
2752 2018-01-26 Richard Biener <rguenther@suse.de>
2754 PR tree-optimization/81082
2755 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
2756 association if it requires casting to unsigned.
2757 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
2758 from fold_plusminus_mult_expr to catch important cases late when
2759 range info is available.
2761 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2763 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
2764 * configure.ac (hidden_linkonce): New test.
2765 * configure: Regenerate.
2766 * config.in: Regenerate.
2768 2018-01-26 Julia Koval <julia.koval@intel.com>
2770 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
2771 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
2772 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
2773 _mm_mask_bitshuffle_epi64_mask): Fix type.
2774 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
2775 USI_FTYPE_V4DI_V4DI_USI): Remove.
2776 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
2777 __builtin_ia32_vpshufbitqmb256_mask,
2778 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
2779 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
2780 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
2782 2018-01-26 Alan Modra <amodra@gmail.com>
2785 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
2786 UNSPEC_VBPERMQ. Sort other unspecs.
2788 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
2790 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
2792 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
2795 * predict.c (drop_profile): Do not push/pop cfun; update also
2797 (handle_missing_profiles): Fix logic looking for zero profiles.
2799 2018-01-25 Jakub Jelinek <jakub@redhat.com>
2802 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
2803 on functions with #pragma omp declare simd or functions with simd
2805 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
2806 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
2807 Remove trailing \n from warning_at calls.
2809 2018-01-25 Tom de Vries <tom@codesourcery.com>
2812 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2813 for neutered workers.
2815 2018-01-24 Joseph Myers <joseph@codesourcery.com>
2818 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
2819 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
2821 2018-01-24 Jeff Law <law@redhat.com>
2824 * i386.c (get_probe_interval): Move to earlier point.
2825 (ix86_compute_frame_layout): If -fstack-clash-protection and
2826 the frame is larger than the probe interval, then use pushes
2827 to save registers rather than reg->mem moves.
2828 (ix86_expand_prologue): Remove conditional for int_registers_saved
2831 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
2834 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
2835 min/max for never referenced object.
2837 2018-01-24 Jakub Jelinek <jakub@redhat.com>
2840 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
2842 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
2843 attributes from DECL_ATTRIBUTES (decl) without affecting
2844 DECL_ATTRIBUTES (current_function_decl).
2845 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
2846 functions with non-NULL DECL_ABSTRACT_ORIGIN.
2848 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
2850 PR tree-optimization/83979
2851 * fold-const.c (fold_comparison): Use constant_boolean_node
2852 instead of boolean_{true,false}_node.
2854 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
2856 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
2859 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2861 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
2862 Simplify the clause that sets the length attribute.
2863 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
2864 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
2865 clause that sets the length attribute.
2866 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
2868 2018-01-24 Tom de Vries <tom@codesourcery.com>
2871 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
2872 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
2873 Add strict parameter.
2874 (prevent_branch_around_nothing): Insert dummy insn between branch to
2875 label and label with no ptx insn inbetween.
2876 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
2878 2018-01-24 Tom de Vries <tom@codesourcery.com>
2881 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2882 for neutered threads in warp.
2883 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
2885 2018-01-24 Richard Biener <rguenther@suse.de>
2887 PR tree-optimization/83176
2888 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
2891 2018-01-24 Richard Biener <rguenther@suse.de>
2893 PR tree-optimization/82819
2894 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
2895 code generating pluses that are no-ops in the target precision.
2897 2018-01-24 Richard Biener <rguenther@suse.de>
2900 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
2902 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2904 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
2905 to merge probabilities.
2906 * predict.c (probably_never_executed): Also mark as cold functions
2907 with global 0 profile and guessed local profile.
2908 * profile-count.c (profile_probability::combine_with_count): New
2910 * profile-count.h (profile_probability::operator*,
2911 profile_probability::operator*=, profile_probability::operator/,
2912 profile_probability::operator/=): Reduce precision to adjusted
2913 and set value to guessed on contradictory divisions.
2914 (profile_probability::combine_with_freq): Remove.
2915 (profile_probability::combine_wiht_count): Declare.
2916 (profile_count::force_nonzero):: Set to adjusted.
2917 (profile_count::probability_in):: Set quality to adjusted.
2918 * tree-ssa-tail-merge.c (replace_block_by): Use
2921 2018-01-23 Andrew Waterman <andrew@sifive.com>
2922 Jim Wilson <jimw@sifive.com>
2924 * config/riscv/riscv.c (riscv_stack_boundary): New.
2925 (riscv_option_override): Set riscv_stack_boundary. Handle
2926 riscv_preferred_stack_boundary_arg.
2927 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
2928 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
2929 (STACK_BOUNDARY): Set to riscv_stack_boundary.
2930 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
2931 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
2932 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
2934 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
2937 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
2938 of struct ix86_frame.
2939 (ix86_expand_epilogue): Likewise. Add a local variable for
2940 the reg_save_offset field in struct ix86_frame.
2942 2018-01-23 Bin Cheng <bin.cheng@arm.com>
2944 PR tree-optimization/82604
2945 * tree-loop-distribution.c (enum partition_kind): New enum item
2946 PKIND_PARTIAL_MEMSET.
2947 (partition_builtin_p): Support above new enum item.
2948 (generate_code_for_partition): Ditto.
2949 (compute_access_range): Differentiate cases that equality can be
2950 proven at all loops, the innermost loops or no loops.
2951 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
2952 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
2953 (finalize_partitions, distribute_loop): Don't fuse partition of
2954 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
2955 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
2958 2018-01-23 Martin Liska <mliska@suse.cz>
2960 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
2961 order to ignore the predictor.
2962 (PRED_POLYMORPHIC_CALL): Likewise.
2963 (PRED_RECURSIVE_CALL): Likewise.
2965 2018-01-23 Martin Liska <mliska@suse.cz>
2967 * tree-profile.c (tree_profiling): Print function header to
2968 aware reader which function we are working on.
2969 * value-prof.c (gimple_find_values_to_profile): Do not print
2970 not interesting value histograms.
2972 2018-01-23 Martin Liska <mliska@suse.cz>
2974 * profile-count.h (enum profile_quality): Add
2975 profile_uninitialized as the first value. Do not number values
2976 as they are zero based.
2977 (profile_count::verify): Update sanity check.
2978 (profile_probability::verify): Likewise.
2980 2018-01-23 Nathan Sidwell <nathan@acm.org>
2982 * doc/invoke.texi (ffor-scope): Deprecate.
2984 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2986 PR tree-optimization/83510
2987 * domwalk.c (set_all_edges_as_executable): New function.
2988 (dom_walker::dom_walker): Convert bool param
2989 "skip_unreachable_blocks" to enum reachability. Move setup of
2990 edge flags to set_all_edges_as_executable and only do it when
2991 reachability is REACHABLE_BLOCKS.
2992 * domwalk.h (enum dom_walker::reachability): New enum.
2993 (dom_walker::dom_walker): Convert bool param
2994 "skip_unreachable_blocks" to enum reachability.
2995 (set_all_edges_as_executable): New decl.
2996 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
2997 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
2999 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
3000 but converting true to REACHABLE_BLOCKS.
3001 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
3003 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
3004 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
3005 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
3007 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
3008 if check_all_array_refs will be called.
3010 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3012 * tree.c (selftest::test_location_wrappers): Add more test
3015 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3017 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
3018 (selftest::test_bit_in_range): Likewise.
3020 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3023 * doc/sourcebuild.texi (vect_float): Say that the selector
3024 only describes the situation when -funsafe-math-optimizations is on.
3025 (vect_float_strict): Document.
3027 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3029 PR tree-optimization/83965
3030 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3031 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
3032 instead of checking only for a reduction.
3033 (vect_recog_widen_sum_pattern): Likewise.
3035 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3037 * predict.c (probably_never_executed): Only use precise profile info.
3038 (compute_function_frequency): Skip after inlining hack since we now
3039 have quality checking.
3041 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3043 * profile-count.h (profile_probability::very_unlikely,
3044 profile_probability::unlikely, profile_probability::even): Set
3045 precision to guessed.
3047 2018-01-23 Richard Biener <rguenther@suse.de>
3049 PR tree-optimization/83963
3050 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
3051 Properly terminate dominator walk when crossing the exit edge not
3052 when visiting its source block.
3054 2018-01-23 Jakub Jelinek <jakub@redhat.com>
3057 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
3058 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
3060 2018-01-22 Jakub Jelinek <jakub@redhat.com>
3062 PR tree-optimization/83957
3063 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
3064 semicolon after for body surrounded by braces.
3066 PR tree-optimization/83081
3067 * profile-count.h (profile_probability::split): New method.
3068 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
3069 Use profile_probability::split.
3070 (do_compare_rtx_and_jump): Fix adjustment of probabilities
3071 when splitting a single conditional jump into 2.
3073 2018-01-22 David Malcolm <dmalcolm@redhat.com>
3075 PR tree-optimization/69452
3076 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
3079 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3081 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
3082 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
3083 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
3085 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3087 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
3088 * config/rl78/rl78.md (movdi): New define_expand.
3089 * config/rl78/rl78.c (rl78_split_movdi): New function.
3091 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
3094 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
3096 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
3097 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
3098 128-bit to produce an UNSPEC move to get the double word with the
3099 signbit and then a shift directly to do signbit.
3100 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
3101 implementation with a new version that just does either a direct
3102 move or a regular move. Move memory interface to separate insns.
3103 Move insns so they are next to the expander.
3104 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
3105 with signbit move. Split big and little endian case.
3106 (signbit<mode>2_dm_mem_le): Likewise.
3107 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
3108 (signbit<mode>2_dm2): Likewise.
3110 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3112 * config/rl78/rl78.md (anddi3): New define_expand.
3114 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3116 * config/rl78/rl78.md (umindi3): New define_expand.
3118 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3120 * config/rl78/rl78.md (smindi3): New define_expand.
3122 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3124 * config/rl78/rl78.md (smaxdi3): New define_expand.
3126 2018-01-22 Carl Love <cel@us.ibm.com>
3128 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
3129 LVX_V1TI): Add macro expansion.
3130 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
3131 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
3132 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
3133 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
3134 Change check to determine if the instruction is a byte reversing
3135 entry. Fix typo in comment.
3136 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
3137 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
3138 Add def_builtin calls for new builtins.
3139 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
3140 Add define_insn expansion.
3142 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3144 * config/rl78/rl78.md (umaxdi3): New define_expand.
3146 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3148 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
3149 for non-QImode registers.
3151 2018-01-22 Richard Biener <rguenther@suse.de>
3153 PR tree-optimization/83963
3154 * graphite-scop-detection.c (scop_detection::get_sese): Delay
3155 including the loop exit block.
3156 (scop_detection::merge_sese): Likewise.
3157 (scop_detection::add_scop): Do it here instead.
3159 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3161 * doc/sourcebuild.texi (arm_softfloat): Document.
3163 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
3166 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
3167 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
3168 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
3170 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3171 David Edelsohn <dje.gcc@gmail.com>
3174 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3175 Change "crset eq" to "crset 2".
3176 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3177 (*call_indirect_aix<mode>_nospec): Likewise.
3178 (*call_value_indirect_aix<mode>_nospec): Likewise.
3179 (*call_indirect_elfv2<mode>_nospec): Likewise.
3180 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
3181 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
3182 change assembly output from . to $.
3183 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3184 (indirect_jump<mode>_nospec): Change assembly output from . to $.
3185 (*tablejump<mode>_internal1_nospec): Likewise.
3187 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
3190 * config/sh/sh_optimize_sett_clrt.cc:
3191 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
3193 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
3195 PR tree-optimization/83940
3196 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
3197 offset_dt to vect_constant_def rather than vect_unknown_def_type.
3198 (vect_check_load_store_mask): Add a mask_dt_out parameter and
3199 use it to pass back the definition type.
3200 (vect_check_store_rhs): Likewise rhs_dt_out.
3201 (vect_build_gather_load_calls): Add a mask_dt argument and use
3202 it instead of a call to vect_is_simple_use.
3203 (vectorizable_store): Update calls to vect_check_load_store_mask
3204 and vect_check_store_rhs. Use the dt returned by the latter instead
3205 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
3206 instead of calls to vect_is_simple_use. Pass the scalar rather
3207 than the vector operand to vect_is_simple_use when handling
3208 second and subsequent copies of an rhs value.
3209 (vectorizable_load): Update calls to vect_check_load_store_mask
3210 and vect_build_gather_load_calls. Use the cached mask_dt and
3211 gs_info.offset_dt instead of calls to vect_is_simple_use.
3213 2018-01-20 Jakub Jelinek <jakub@redhat.com>
3216 * tree-emutls.c: Include gimplify.h.
3217 (lower_emutls_2): New function.
3218 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
3219 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
3220 it before further processing.
3223 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
3224 UINTVAL (trueop1) instead of INTVAL (op1).
3226 2018-01-19 Jakub Jelinek <jakub@redhat.com>
3230 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
3231 INCOMING_FRAME_SP_OFFSET if not defined.
3232 (scan_trace): Add ENTRY argument. If true and
3233 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
3234 emit a note to adjust the CFA offset.
3235 (create_cfi_notes): Adjust scan_trace callers.
3236 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
3237 INCOMING_FRAME_SP_OFFSET in the CIE.
3238 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
3239 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
3241 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
3242 * doc/tm.texi: Regenerated.
3244 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3246 PR rtl-optimization/83147
3247 * lra-constraints.c (remove_inheritance_pseudos): Use
3248 lra_substitute_pseudo_within_insn.
3250 2018-01-19 Tom de Vries <tom@codesourcery.com>
3251 Cesar Philippidis <cesar@codesourcery.com>
3254 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
3256 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
3259 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
3260 spaces for function labels.
3262 2018-01-19 Martin Liska <mliska@suse.cz>
3264 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
3265 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
3266 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
3267 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
3268 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
3269 (PRED_CONST_RETURN): Change from 69 to 65.
3270 (PRED_NULL_RETURN): Change from 91 to 71.
3271 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
3272 (PRED_LOOP_GUARD): Change from 66 to 73.
3274 2018-01-19 Martin Liska <mliska@suse.cz>
3276 * predict.c (predict_insn_def): Add new assert.
3277 (struct branch_predictor): Change type to signed integer.
3278 (test_prediction_value_range): Amend test to cover
3280 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
3281 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
3282 (PRED_LOOP_ITERATIONS_MAX): Likewise.
3283 (PRED_LOOP_IV_COMPARE): Likewise.
3284 * predict.h (PROB_UNINITIALIZED): Define new constant.
3286 2018-01-19 Martin Liska <mliska@suse.cz>
3288 * predict.c (dump_prediction): Add new format for
3289 analyze_brprob.py script which is enabled with -details
3291 * profile-count.h (precise_p): New function.
3293 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3295 PR tree-optimization/83922
3296 * tree-vect-loop.c (vect_verify_full_masking): Return false if
3297 there are no statements that need masking.
3298 (vect_active_double_reduction_p): New function.
3299 (vect_analyze_loop_operations): Use it when handling phis that
3300 are not in the loop header.
3302 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3304 PR tree-optimization/83914
3305 * tree-vect-loop.c (vectorizable_induction): Don't convert
3306 init_expr or apply the peeling adjustment for inductions
3307 that are nested within the vectorized loop.
3309 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3311 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
3314 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3318 * function.h (gimplify_parameters): Add gimple_seq * argument.
3319 * function.c: Include gimple.h and options.h.
3320 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
3321 for the added local temporaries if needed.
3322 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
3323 if there are any parameter cleanups, wrap whole body into a
3324 try/finally with the cleanups.
3326 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
3329 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
3330 Use GET_MODE_CLASS for scalar floating point.
3332 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3336 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3337 Fix call of call_cgraph_insertion_hooks.
3339 2018-01-18 Martin Sebor <msebor@redhat.com>
3341 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
3343 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3346 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
3349 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
3352 * common.opt: (-ffile-prefix-map): New option.
3353 * opts.c (common_handle_option): Defer it.
3354 * opts-global.c (handle_common_deferred_options): Handle it.
3355 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
3356 * file-prefix-map.h: New file.
3357 (remap_debug_filename, add_debug_prefix_map): ...here.
3358 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
3359 * final.c (debug_prefix_map, add_debug_prefix_map
3360 remap_debug_filename): Move to...
3361 * file-prefix-map.c: New file.
3362 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
3363 generalize, get rid of alloca(), use strrchr() instead of strchr().
3364 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
3365 Implement in terms of add_prefix_map().
3366 (remap_macro_filename, remap_debug_filename): Implement in term of
3368 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
3369 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
3370 * dbxout.c: Include file-prefix-map.h.
3371 * varasm.c: Likewise.
3372 * vmsdbgout.c: Likewise.
3373 * xcoffout.c: Likewise.
3374 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
3375 * doc/cppopts.texi (-fmacro-prefix-map): Document.
3376 * doc/invoke.texi (-ffile-prefix-map): Document.
3377 (-fdebug-prefix-map): Update description.
3379 2018-01-18 Martin Liska <mliska@suse.cz>
3381 * config/i386/i386.c (indirect_thunk_name): Document that also
3383 (output_indirect_thunk): Document why both instructions
3384 (pause and lfence) are generated.
3386 2018-01-18 Richard Biener <rguenther@suse.de>
3388 PR tree-optimization/83887
3389 * graphite-scop-detection.c
3390 (scop_detection::get_nearest_dom_with_single_entry): Remove.
3391 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
3392 (scop_detection::merge_sese): Re-implement with a flood-fill
3393 algorithm that properly finds a SESE region if it exists.
3395 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3398 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
3399 pointer_diff optimizations use view_convert instead of convert.
3401 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3403 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3404 Generate different code for -mno-speculate-indirect-jumps.
3405 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3406 (*call_indirect_aix<mode>): Disable for
3407 -mno-speculate-indirect-jumps.
3408 (*call_indirect_aix<mode>_nospec): New define_insn.
3409 (*call_value_indirect_aix<mode>): Disable for
3410 -mno-speculate-indirect-jumps.
3411 (*call_value_indirect_aix<mode>_nospec): New define_insn.
3412 (*sibcall_nonlocal_sysv<mode>): Generate different code for
3413 -mno-speculate-indirect-jumps.
3414 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3416 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3418 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
3419 long double type, set the flags for noting the default long double
3420 type, even if we don't pass or return a long double type.
3422 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
3425 * ipa-inline.c (flatten_function): Do not overwrite final inlining
3428 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
3430 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
3431 support for merge[hl].
3432 (fold_mergehl_helper): New helper function.
3433 (tree-vector-builder.h): New #include for tree_vector_builder usage.
3434 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
3435 (altivec_vmrglw_direct): Add xxmrglw insn.
3437 2018-01-17 Andrew Waterman <andrew@sifive.com>
3439 * config/riscv/riscv.c (riscv_conditional_register_usage): If
3440 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
3442 2018-01-17 David Malcolm <dmalcolm@redhat.com>
3445 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
3446 call the lto_location_cache before reading the
3447 DECL_SOURCE_LOCATION of the types.
3449 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3450 Richard Sandiford <richard.sandiford@linaro.org>
3452 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
3453 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
3454 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
3455 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
3456 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
3458 * config/aarch64/constraints.md (aarch64_movti_operand):
3460 * config/aarch64/predicates.md (Uti): Add new constraint.
3462 2018-01-17 Carl Love <cel@us.ibm.com>
3464 * config/rs6000/vsx.md (define_expand xl_len_r,
3465 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
3466 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
3468 (define_expand, define_insn): Move the shift left from the
3469 define_insn to the define_expand for lxvl and stxvl instructions.
3470 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
3471 and XL_LEN_R definitions to PURE.
3473 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
3475 * config/i386/i386.c (indirect_thunk_name): Declare regno
3476 as unsigned int. Compare regno with INVALID_REGNUM.
3477 (output_indirect_thunk): Ditto.
3478 (output_indirect_thunk_function): Ditto.
3479 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
3480 in the call to output_indirect_thunk_function.
3482 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
3485 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
3486 rather than the size of inner_type to determine the stack slot size
3487 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
3489 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
3492 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
3495 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
3497 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
3498 endian Linux systems to optionally enable multilibs for selecting
3499 the long double type if the user configured an explicit type.
3500 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
3501 have no long double multilibs if not defined.
3502 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
3503 warn if the user used -mabi={ieee,ibm}longdouble and we built
3504 multilibs for long double.
3505 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
3506 appropriate multilib option.
3507 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
3509 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
3510 for building long double multilibs.
3511 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
3513 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
3515 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
3518 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
3520 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
3523 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
3526 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
3529 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3531 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
3532 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
3534 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3536 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
3537 different rtl trees depending on TARGET_64BIT.
3538 (rs6000_gen_lvx): Likewise.
3540 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3542 * config/visium/visium.md (nop): Tweak comment.
3543 (hazard_nop): Likewise.
3545 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3547 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
3548 -mspeculate-indirect-jumps.
3549 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
3550 for -mno-speculate-indirect-jumps.
3551 (*call_indirect_elfv2<mode>_nospec): New define_insn.
3552 (*call_value_indirect_elfv2<mode>): Disable for
3553 -mno-speculate-indirect-jumps.
3554 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
3555 (indirect_jump): Emit different RTL for
3556 -mno-speculate-indirect-jumps.
3557 (*indirect_jump<mode>): Disable for
3558 -mno-speculate-indirect-jumps.
3559 (*indirect_jump<mode>_nospec): New define_insn.
3560 (tablejump): Emit different RTL for
3561 -mno-speculate-indirect-jumps.
3562 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
3563 (tablejumpsi_nospec): New define_expand.
3564 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
3565 (tablejumpdi_nospec): New define_expand.
3566 (*tablejump<mode>_internal1): Disable for
3567 -mno-speculate-indirect-jumps.
3568 (*tablejump<mode>_internal1_nospec): New define_insn.
3569 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
3572 2018-01-16 Artyom Skrobov tyomitch@gmail.com
3574 * caller-save.c (insert_save): Drop unnecessary parameter. All
3577 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3578 Richard Biener <rguenth@suse.de>
3581 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
3582 return early, inline manually is_gimple_sizepos. Make sure if we
3583 call gimplify_expr we don't end up with a gimple constant.
3584 * tree.c (variably_modified_type_p): Don't return true for
3585 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
3586 * gimplify.h (is_gimple_sizepos): Remove.
3588 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3590 PR tree-optimization/83857
3591 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
3592 vectorizable_live_operation for pure SLP statements.
3593 (vectorizable_live_operation): Handle PHIs.
3595 2018-01-16 Richard Biener <rguenther@suse.de>
3597 PR tree-optimization/83867
3598 * tree-vect-stmts.c (vect_transform_stmt): Precompute
3599 nested_in_vect_loop_p since the scalar stmt may get invalidated.
3601 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3604 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
3605 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
3606 If off is not INTEGER_CST, issue a may not be aligned warning
3607 rather than isn't aligned. Use isn%'t rather than isn't.
3608 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
3610 <case MULT_EXPR>: Improve the case when bottom and one of the
3611 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
3612 operand, in that case check if the other operand is multiple of
3613 bottom divided by the INTEGER_CST operand.
3615 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3618 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
3619 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
3620 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
3621 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
3622 * config/pa/pa.c (pa_function_arg_advance): Likewise.
3623 (pa_function_arg, pa_arg_partial_bytes): Likewise.
3624 (pa_function_arg_size): New function.
3626 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3628 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
3629 in a separate statement.
3631 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3633 PR tree-optimization/83847
3634 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
3635 group gathers and scatters.
3637 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3639 PR rtl-optimization/86620
3640 * params.def (max-sched-ready-insns): Bump minimum value to 1.
3642 PR rtl-optimization/83213
3643 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
3644 to last if both are JUMP_INSNs.
3646 PR tree-optimization/83843
3647 * gimple-ssa-store-merging.c
3648 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
3649 store_immediate_info for bswap/nop orig_stores.
3651 2018-01-15 Andrew Waterman <andrew@sifive.com>
3653 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
3655 <UDIV>: Increase cost if !TARGET_DIV.
3657 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
3659 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
3660 (define_attr "cr_logical_3op"): New.
3661 (cceq_ior_compare): Adjust.
3662 (cceq_ior_compare_complement): Adjust.
3663 (*cceq_rev_compare): Adjust.
3664 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
3665 (is_cracked_insn): Adjust.
3666 (insn_must_be_first_in_group): Adjust.
3667 * config/rs6000/40x.md: Adjust.
3668 * config/rs6000/440.md: Adjust.
3669 * config/rs6000/476.md: Adjust.
3670 * config/rs6000/601.md: Adjust.
3671 * config/rs6000/603.md: Adjust.
3672 * config/rs6000/6xx.md: Adjust.
3673 * config/rs6000/7450.md: Adjust.
3674 * config/rs6000/7xx.md: Adjust.
3675 * config/rs6000/8540.md: Adjust.
3676 * config/rs6000/cell.md: Adjust.
3677 * config/rs6000/e300c2c3.md: Adjust.
3678 * config/rs6000/e500mc.md: Adjust.
3679 * config/rs6000/e500mc64.md: Adjust.
3680 * config/rs6000/e5500.md: Adjust.
3681 * config/rs6000/e6500.md: Adjust.
3682 * config/rs6000/mpc.md: Adjust.
3683 * config/rs6000/power4.md: Adjust.
3684 * config/rs6000/power5.md: Adjust.
3685 * config/rs6000/power6.md: Adjust.
3686 * config/rs6000/power7.md: Adjust.
3687 * config/rs6000/power8.md: Adjust.
3688 * config/rs6000/power9.md: Adjust.
3689 * config/rs6000/rs64.md: Adjust.
3690 * config/rs6000/titan.md: Adjust.
3692 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3694 * config/i386/predicates.md (indirect_branch_operand): Rewrite
3695 ix86_indirect_branch_register logic.
3697 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3699 * config/i386/constraints.md (Bs): Update
3700 ix86_indirect_branch_register check. Don't check
3701 ix86_indirect_branch_register with GOT_memory_operand.
3703 * config/i386/predicates.md (GOT_memory_operand): Don't check
3704 ix86_indirect_branch_register here.
3705 (GOT32_symbol_operand): Likewise.
3707 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3709 * config/i386/predicates.md (constant_call_address_operand):
3710 Rewrite ix86_indirect_branch_register logic.
3711 (sibcall_insn_operand): Likewise.
3713 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3715 * config/i386/constraints.md (Bs): Replace
3716 ix86_indirect_branch_thunk_register with
3717 ix86_indirect_branch_register.
3719 * config/i386/i386.md (indirect_jump): Likewise.
3720 (tablejump): Likewise.
3721 (*sibcall_memory): Likewise.
3722 (*sibcall_value_memory): Likewise.
3723 Peepholes of indirect call and jump via memory: Likewise.
3724 * config/i386/i386.opt: Likewise.
3725 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3726 (GOT_memory_operand): Likewise.
3727 (call_insn_operand): Likewise.
3728 (sibcall_insn_operand): Likewise.
3729 (GOT32_symbol_operand): Likewise.
3731 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3734 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
3735 type rather than type addr's type points to.
3736 (expand_omp_atomic_mutex): Likewise.
3737 (expand_omp_atomic): Likewise.
3739 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3742 * config/i386/i386.c (output_indirect_thunk_function): Use
3743 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
3744 for __x86_return_thunk.
3746 2018-01-15 Richard Biener <rguenther@suse.de>
3749 * expmed.c (extract_bit_field_1): Fix typo.
3751 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3754 * config/arm/iterators.md (VF): New mode iterator.
3755 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
3756 Remove integer-related logic from pattern.
3757 (neon_vabd<mode>_3): Likewise.
3759 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3762 * common.opt (fstrict-overflow): No longer an alias.
3763 (fwrapv-pointer): New option.
3764 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
3765 also for pointer types based on flag_wrapv_pointer.
3766 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
3767 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
3768 opts->x_flag_wrapv got set.
3769 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
3770 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
3771 POINTER_TYPE_OVERFLOW_UNDEFINED.
3772 * match.pd: Likewise in address comparison pattern.
3773 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
3775 2018-01-15 Richard Biener <rguenther@suse.de>
3778 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
3779 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
3780 Reset type names to their identifier if their TYPE_DECL doesn't
3781 have linkage (and thus is used for ODR and devirt).
3782 (save_debug_info_for_decl): Remove.
3783 (save_debug_info_for_type): Likewise.
3784 (add_tree_to_fld_list): Adjust.
3785 * tree-pretty-print.c (dump_generic_node): Make dumping of
3786 type names more robust.
3788 2018-01-15 Richard Biener <rguenther@suse.de>
3790 * BASE-VER: Bump to 8.0.1.
3792 2018-01-14 Martin Sebor <msebor@redhat.com>
3795 * builtins.c (check_access): Avoid warning when the no-warning bit
3798 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
3800 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
3801 * ira-color (allocno_hard_regs_compare): Likewise.
3803 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
3806 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
3807 Use .pushsection/.popsection.
3809 2018-01-14 Martin Sebor <msebor@redhat.com>
3812 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
3814 2018-01-14 Jakub Jelinek <jakub@redhat.com>
3816 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
3817 entry from extra_headers.
3818 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
3819 extra_headers, make the list bitwise identical to the i?86-*-* one.
3821 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3823 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
3824 -mcmodel=large with -mindirect-branch=thunk,
3825 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
3826 -mfunction-return=thunk-extern.
3827 * doc/invoke.texi: Document -mcmodel=large is incompatible with
3828 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
3829 -mfunction-return=thunk and -mfunction-return=thunk-extern.
3831 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3833 * config/i386/i386.c (print_reg): Print the name of the full
3834 integer register without '%'.
3835 (ix86_print_operand): Handle 'V'.
3836 * doc/extend.texi: Document 'V' modifier.
3838 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3840 * config/i386/constraints.md (Bs): Disallow memory operand for
3841 -mindirect-branch-register.
3843 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3844 (GOT_memory_operand): Likewise.
3845 (call_insn_operand): Likewise.
3846 (sibcall_insn_operand): Likewise.
3847 (GOT32_symbol_operand): Likewise.
3848 * config/i386/i386.md (indirect_jump): Call convert_memory_address
3849 for -mindirect-branch-register.
3850 (tablejump): Likewise.
3851 (*sibcall_memory): Likewise.
3852 (*sibcall_value_memory): Likewise.
3853 Disallow peepholes of indirect call and jump via memory for
3854 -mindirect-branch-register.
3855 (*call_pop): Replace m with Bw.
3856 (*call_value_pop): Likewise.
3857 (*sibcall_pop_memory): Replace m with Bs.
3858 * config/i386/i386.opt (mindirect-branch-register): New option.
3859 * doc/invoke.texi: Document -mindirect-branch-register option.
3861 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3863 * config/i386/i386-protos.h (ix86_output_function_return): New.
3864 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
3865 set function_return_type.
3866 (indirect_thunk_name): Add ret_p to indicate thunk for function
3868 (output_indirect_thunk_function): Pass false to
3869 indirect_thunk_name.
3870 (ix86_output_indirect_branch_via_reg): Likewise.
3871 (ix86_output_indirect_branch_via_push): Likewise.
3872 (output_indirect_thunk_function): Create alias for function
3873 return thunk if regno < 0.
3874 (ix86_output_function_return): New function.
3875 (ix86_handle_fndecl_attribute): Handle function_return.
3876 (ix86_attribute_table): Add function_return.
3877 * config/i386/i386.h (machine_function): Add
3878 function_return_type.
3879 * config/i386/i386.md (simple_return_internal): Use
3880 ix86_output_function_return.
3881 (simple_return_internal_long): Likewise.
3882 * config/i386/i386.opt (mfunction-return=): New option.
3883 (indirect_branch): Mention -mfunction-return=.
3884 * doc/extend.texi: Document function_return function attribute.
3885 * doc/invoke.texi: Document -mfunction-return= option.
3887 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3889 * config/i386/i386-opts.h (indirect_branch): New.
3890 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
3891 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
3892 with local indirect jump when converting indirect call and jump.
3893 (ix86_set_indirect_branch_type): New.
3894 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
3895 (indirectlabelno): New.
3896 (indirect_thunk_needed): Likewise.
3897 (indirect_thunk_bnd_needed): Likewise.
3898 (indirect_thunks_used): Likewise.
3899 (indirect_thunks_bnd_used): Likewise.
3900 (INDIRECT_LABEL): Likewise.
3901 (indirect_thunk_name): Likewise.
3902 (output_indirect_thunk): Likewise.
3903 (output_indirect_thunk_function): Likewise.
3904 (ix86_output_indirect_branch_via_reg): Likewise.
3905 (ix86_output_indirect_branch_via_push): Likewise.
3906 (ix86_output_indirect_branch): Likewise.
3907 (ix86_output_indirect_jmp): Likewise.
3908 (ix86_code_end): Call output_indirect_thunk_function if needed.
3909 (ix86_output_call_insn): Call ix86_output_indirect_branch if
3911 (ix86_handle_fndecl_attribute): Handle indirect_branch.
3912 (ix86_attribute_table): Add indirect_branch.
3913 * config/i386/i386.h (machine_function): Add indirect_branch_type
3914 and has_local_indirect_jump.
3915 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
3917 (tablejump): Likewise.
3918 (*indirect_jump): Use ix86_output_indirect_jmp.
3919 (*tablejump_1): Likewise.
3920 (simple_return_indirect_internal): Likewise.
3921 * config/i386/i386.opt (mindirect-branch=): New option.
3922 (indirect_branch): New.
3925 (thunk-inline): Likewise.
3926 (thunk-extern): Likewise.
3927 * doc/extend.texi: Document indirect_branch function attribute.
3928 * doc/invoke.texi: Document -mindirect-branch= option.
3930 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
3933 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
3935 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
3937 * ipa-inline.c (want_inline_small_function_p): Return false if
3938 inlining has already failed with CIF_FINAL_ERROR.
3939 (update_caller_keys): Call want_inline_small_function_p before
3941 (update_callee_keys): Likewise.
3943 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3945 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
3947 (rs6000_quadword_masked_address_p): Likewise.
3948 (quad_aligned_load_p): Likewise.
3949 (quad_aligned_store_p): Likewise.
3950 (const_load_sequence_p): Add comment to describe the outer-most loop.
3951 (mimic_memory_attributes_and_flags): New function.
3952 (rs6000_gen_stvx): Likewise.
3953 (replace_swapped_aligned_store): Likewise.
3954 (rs6000_gen_lvx): Likewise.
3955 (replace_swapped_aligned_load): Likewise.
3956 (replace_swapped_load_constant): Capitalize argument name in
3957 comment describing this function.
3958 (rs6000_analyze_swaps): Add a third pass to search for vector loads
3959 and stores that access quad-word aligned addresses and replace
3960 with stvx or lvx instructions when appropriate.
3961 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
3962 New function prototype.
3963 (rs6000_quadword_masked_address_p): Likewise.
3964 (rs6000_gen_lvx): Likewise.
3965 (rs6000_gen_stvx): Likewise.
3966 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
3967 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
3968 when memory address is aligned.
3969 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
3970 this split to select lvx instruction when memory address is aligned.
3971 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
3972 instruction when memory address is aligned.
3973 (*vsx_le_perm_load_v16qi): Likewise.
3974 (four unnamed splitters): Modify to select the stvx instruction
3975 when memory is aligned.
3977 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
3979 * predict.c (determine_unlikely_bbs): Handle correctly BBs
3980 which appears in the queue multiple times.
3982 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3983 Alan Hayward <alan.hayward@arm.com>
3984 David Sherwood <david.sherwood@arm.com>
3986 * tree-vectorizer.h (vec_lower_bound): New structure.
3987 (_loop_vec_info): Add check_nonzero and lower_bounds.
3988 (LOOP_VINFO_CHECK_NONZERO): New macro.
3989 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
3990 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
3991 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
3992 fields. Make seg_len the distance travelled, not including the
3994 (dr_direction_indicator): Declare.
3995 (dr_zero_step_indicator): Likewise.
3996 (dr_known_forward_stride_p): Likewise.
3997 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
3999 (runtime_alias_check_p): Allow runtime alias checks with
4001 (operator ==): Compare access_size and align.
4002 (prune_runtime_alias_test_list): Rework for new distinction between
4003 the access_size and seg_len.
4004 (create_intersect_range_checks_index): Likewise. Cope with polynomial
4006 (get_segment_min_max): New function.
4007 (create_intersect_range_checks): Use it.
4008 (dr_step_indicator): New function.
4009 (dr_direction_indicator): Likewise.
4010 (dr_zero_step_indicator): Likewise.
4011 (dr_known_forward_stride_p): Likewise.
4012 * tree-loop-distribution.c (data_ref_segment_size): Return
4013 DR_STEP * (niters - 1).
4014 (compute_alias_check_pairs): Update call to the dr_with_seg_len
4016 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
4017 (vect_preserves_scalar_order_p): New function, split out from...
4018 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
4019 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
4020 (vect_vfa_access_size): New function.
4021 (vect_vfa_align): Likewise.
4022 (vect_compile_time_alias): Take access_size_a and access_b arguments.
4023 (dump_lower_bound): New function.
4024 (vect_check_lower_bound): Likewise.
4025 (vect_small_gap_p): Likewise.
4026 (vectorizable_with_step_bound_p): Likewise.
4027 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
4028 depencies if the vectorization factor is 1. Convert the checks
4029 for nonzero steps into checks on the bounds of DR_STEP. Try using
4030 a bunds check for variable steps if the minimum required step is
4031 relatively small. Update calls to the dr_with_seg_len
4032 constructor and to vect_compile_time_alias.
4033 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
4035 (vect_loop_versioning): Call it.
4036 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
4038 (vect_estimate_min_profitable_iters): Account for any bounds checks.
4040 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4041 Alan Hayward <alan.hayward@arm.com>
4042 David Sherwood <david.sherwood@arm.com>
4044 * doc/sourcebuild.texi (vect_scatter_store): Document.
4045 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
4047 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
4049 * genopinit.c (main): Add supports_vec_scatter_store and
4050 supports_vec_scatter_store_cached to target_optabs.
4051 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
4052 IFN_MASK_SCATTER_STORE.
4053 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
4055 * internal-fn.h (internal_store_fn_p): Declare.
4056 (internal_fn_stored_value_index): Likewise.
4057 * internal-fn.c (scatter_store_direct): New macro.
4058 (expand_scatter_store_optab_fn): New function.
4059 (direct_scatter_store_optab_supported_p): New macro.
4060 (internal_store_fn_p): New function.
4061 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
4062 IFN_MASK_SCATTER_STORE.
4063 (internal_fn_mask_index): Likewise.
4064 (internal_fn_stored_value_index): New function.
4065 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
4067 * optabs-query.h (supports_vec_scatter_store_p): Declare.
4068 * optabs-query.c (supports_vec_scatter_store_p): New function.
4069 * tree-vectorizer.h (vect_get_store_rhs): Declare.
4070 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
4071 true for scatter stores.
4072 (vect_gather_scatter_fn_p): Handle scatter stores too.
4073 (vect_check_gather_scatter): Consider using scatter stores if
4074 supports_vec_scatter_store_p.
4075 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
4077 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4078 internal_fn_stored_value_index.
4079 (check_load_store_masking): Handle scatter stores too.
4080 (vect_get_store_rhs): Make public.
4081 (vectorizable_call): Use internal_store_fn_p.
4082 (vectorizable_store): Handle scatter store internal functions.
4083 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
4084 when deciding whether the end of the group has been reached.
4085 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
4086 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
4087 (mask_scatter_store<mode>): New insns.
4089 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4090 Alan Hayward <alan.hayward@arm.com>
4091 David Sherwood <david.sherwood@arm.com>
4093 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
4094 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
4095 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
4097 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
4098 Use vect_truncate_gather_scatter_offset if we can't treat the
4099 operation as a normal gather load or scatter store.
4100 (get_group_load_store_type): Take the gather_scatter_info
4101 as argument. Try using a gather load or scatter store for
4102 single-element groups.
4103 (get_load_store_type): Update calls to get_group_load_store_type
4104 and vect_use_strided_gather_scatters_p.
4106 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4107 Alan Hayward <alan.hayward@arm.com>
4108 David Sherwood <david.sherwood@arm.com>
4110 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
4111 optional tree argument.
4112 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
4114 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
4115 but continue to use the current value as a fallback.
4116 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
4117 to compare the updates.
4118 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
4119 (get_load_store_type): Use it when handling a strided access.
4120 (vect_get_strided_load_store_ops): New function.
4121 (vect_get_data_ptr_increment): Likewise.
4122 (vectorizable_load): Handle strided gather loads. Always pass
4123 a step to vect_create_data_ref_ptr and bump_vector_ptr.
4125 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4126 Alan Hayward <alan.hayward@arm.com>
4127 David Sherwood <david.sherwood@arm.com>
4129 * doc/md.texi (gather_load@var{m}): Document.
4130 (mask_gather_load@var{m}): Likewise.
4131 * genopinit.c (main): Add supports_vec_gather_load and
4132 supports_vec_gather_load_cached to target_optabs.
4133 * optabs-tree.c (init_tree_optimization_optabs): Use
4134 ggc_cleared_alloc to allocate target_optabs.
4135 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
4136 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
4138 * internal-fn.h (internal_load_fn_p): Declare.
4139 (internal_gather_scatter_fn_p): Likewise.
4140 (internal_fn_mask_index): Likewise.
4141 (internal_gather_scatter_fn_supported_p): Likewise.
4142 * internal-fn.c (gather_load_direct): New macro.
4143 (expand_gather_load_optab_fn): New function.
4144 (direct_gather_load_optab_supported_p): New macro.
4145 (direct_internal_fn_optab): New function.
4146 (internal_load_fn_p): Likewise.
4147 (internal_gather_scatter_fn_p): Likewise.
4148 (internal_fn_mask_index): Likewise.
4149 (internal_gather_scatter_fn_supported_p): Likewise.
4150 * optabs-query.c (supports_at_least_one_mode_p): New function.
4151 (supports_vec_gather_load_p): Likewise.
4152 * optabs-query.h (supports_vec_gather_load_p): Declare.
4153 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
4154 and memory_type field.
4155 (NUM_PATTERNS): Bump to 15.
4156 * tree-vect-data-refs.c: Include internal-fn.h.
4157 (vect_gather_scatter_fn_p): New function.
4158 (vect_describe_gather_scatter_call): Likewise.
4159 (vect_check_gather_scatter): Try using internal functions for
4160 gather loads. Recognize existing calls to a gather load function.
4161 (vect_analyze_data_refs): Consider using gather loads if
4162 supports_vec_gather_load_p.
4163 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
4164 (vect_get_gather_scatter_offset_type): Likewise.
4165 (vect_convert_mask_for_vectype): Likewise.
4166 (vect_add_conversion_to_patterm): Likewise.
4167 (vect_try_gather_scatter_pattern): Likewise.
4168 (vect_recog_gather_scatter_pattern): New pattern recognizer.
4169 (vect_vect_recog_func_ptrs): Add it.
4170 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4171 internal_fn_mask_index and internal_gather_scatter_fn_p.
4172 (check_load_store_masking): Take the gather_scatter_info as an
4173 argument and handle gather loads.
4174 (vect_get_gather_scatter_ops): New function.
4175 (vectorizable_call): Check internal_load_fn_p.
4176 (vectorizable_load): Likewise. Handle gather load internal
4178 (vectorizable_store): Update call to check_load_store_masking.
4179 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
4180 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
4181 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
4182 (aarch64_gather_scale_operand_d): New predicates.
4183 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
4184 (mask_gather_load<mode>): New insns.
4186 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4187 Alan Hayward <alan.hayward@arm.com>
4188 David Sherwood <david.sherwood@arm.com>
4190 * optabs.def (fold_left_plus_optab): New optab.
4191 * doc/md.texi (fold_left_plus_@var{m}): Document.
4192 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
4193 * internal-fn.c (fold_left_direct): Define.
4194 (expand_fold_left_optab_fn): Likewise.
4195 (direct_fold_left_optab_supported_p): Likewise.
4196 * fold-const-call.c (fold_const_fold_left): New function.
4197 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
4198 * tree-parloops.c (valid_reduction_p): New function.
4199 (gather_scalar_reductions): Use it.
4200 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
4201 (vect_finish_replace_stmt): Declare.
4202 * tree-vect-loop.c (fold_left_reduction_fn): New function.
4203 (needs_fold_left_reduction_p): New function, split out from...
4204 (vect_is_simple_reduction): ...here. Accept reductions that
4205 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
4206 (vect_force_simple_reduction): Also store the reduction type in
4207 the assignment's STMT_VINFO_REDUC_TYPE.
4208 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
4209 (merge_with_identity): New function.
4210 (vect_expand_fold_left): Likewise.
4211 (vectorize_fold_left_reduction): Likewise.
4212 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
4213 scalar phi in place for it. Check for target support and reject
4214 cases that would reassociate the operation. Defer the transform
4215 phase to vectorize_fold_left_reduction.
4216 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
4217 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
4218 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
4220 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4222 * tree-if-conv.c (predicate_mem_writes): Remove redundant
4223 call to ifc_temp_var.
4225 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4226 Alan Hayward <alan.hayward@arm.com>
4227 David Sherwood <david.sherwood@arm.com>
4229 * target.def (legitimize_address_displacement): Take the original
4230 offset as a poly_int.
4231 * targhooks.h (default_legitimize_address_displacement): Update
4233 * targhooks.c (default_legitimize_address_displacement): Likewise.
4234 * doc/tm.texi: Regenerate.
4235 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
4236 as an argument, moving assert of ad->disp == ad->disp_term to...
4237 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
4238 Try calling targetm.legitimize_address_displacement before expanding
4239 the address rather than afterwards, and adjust for the new interface.
4240 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
4241 Match the new hook interface. Handle SVE addresses.
4242 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
4245 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4247 * Makefile.in (OBJS): Add early-remat.o.
4248 * target.def (select_early_remat_modes): New hook.
4249 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
4250 * doc/tm.texi: Regenerate.
4251 * targhooks.h (default_select_early_remat_modes): Declare.
4252 * targhooks.c (default_select_early_remat_modes): New function.
4253 * timevar.def (TV_EARLY_REMAT): New timevar.
4254 * passes.def (pass_early_remat): New pass.
4255 * tree-pass.h (make_pass_early_remat): Declare.
4256 * early-remat.c: New file.
4257 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
4259 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
4261 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4262 Alan Hayward <alan.hayward@arm.com>
4263 David Sherwood <david.sherwood@arm.com>
4265 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
4266 vfm1 with a bound_epilog parameter.
4267 (vect_do_peeling): Update calls accordingly, and move the prologue
4268 call earlier in the function. Treat the base bound_epilog as 0 for
4269 fully-masked loops and retain vf - 1 for other loops. Add 1 to
4270 this base when peeling for gaps.
4271 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
4272 with fully-masked loops.
4273 (vect_estimate_min_profitable_iters): Handle the single peeled
4274 iteration in that case.
4276 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4277 Alan Hayward <alan.hayward@arm.com>
4278 David Sherwood <david.sherwood@arm.com>
4280 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
4281 single-element interleaving even if the size is not a power of 2.
4282 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
4283 accesses for single-element interleaving if the group size is
4286 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4287 Alan Hayward <alan.hayward@arm.com>
4288 David Sherwood <david.sherwood@arm.com>
4290 * doc/md.texi (fold_extract_last_@var{m}): Document.
4291 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
4292 * optabs.def (fold_extract_last_optab): New optab.
4293 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
4294 * internal-fn.c (fold_extract_direct): New macro.
4295 (expand_fold_extract_optab_fn): Likewise.
4296 (direct_fold_extract_optab_supported_p): Likewise.
4297 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
4298 * tree-vect-loop.c (vect_model_reduction_cost): Handle
4299 EXTRACT_LAST_REDUCTION.
4300 (get_initial_def_for_reduction): Do not create an initial vector
4301 for EXTRACT_LAST_REDUCTION reductions.
4302 (vectorizable_reduction): Leave the scalar phi in place for
4303 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
4304 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
4305 epilogue code for EXTRACT_LAST_REDUCTION and defer the
4306 transform phase to vectorizable_condition.
4307 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
4309 (vect_finish_stmt_generation): ...here.
4310 (vect_finish_replace_stmt): New function.
4311 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
4312 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
4314 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
4316 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4317 Alan Hayward <alan.hayward@arm.com>
4318 David Sherwood <david.sherwood@arm.com>
4320 * doc/md.texi (extract_last_@var{m}): Document.
4321 * optabs.def (extract_last_optab): New optab.
4322 * internal-fn.def (EXTRACT_LAST): New internal function.
4323 * internal-fn.c (cond_unary_direct): New macro.
4324 (expand_cond_unary_optab_fn): Likewise.
4325 (direct_cond_unary_optab_supported_p): Likewise.
4326 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
4327 loops using EXTRACT_LAST.
4328 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
4329 (extract_last_<mode>): ...this optab.
4330 (vec_extract<mode><Vel>): Update accordingly.
4332 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4333 Alan Hayward <alan.hayward@arm.com>
4334 David Sherwood <david.sherwood@arm.com>
4336 * target.def (empty_mask_is_expensive): New hook.
4337 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
4338 * doc/tm.texi: Regenerate.
4339 * targhooks.h (default_empty_mask_is_expensive): Declare.
4340 * targhooks.c (default_empty_mask_is_expensive): New function.
4341 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
4342 if the target says that empty masks are expensive.
4343 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
4345 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
4347 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4348 Alan Hayward <alan.hayward@arm.com>
4349 David Sherwood <david.sherwood@arm.com>
4351 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
4352 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
4353 (vect_use_loop_mask_for_alignment_p): New function.
4354 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
4355 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
4356 niters_skip argument. Make sure that the first niters_skip elements
4357 of the first iteration are inactive.
4358 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
4359 Update call to vect_set_loop_masks_directly.
4360 (get_misalign_in_elems): New function, split out from...
4361 (vect_gen_prolog_loop_niters): ...here.
4362 (vect_update_init_of_dr): Take a code argument that specifies whether
4363 the adjustment should be added or subtracted.
4364 (vect_update_init_of_drs): Likewise.
4365 (vect_prepare_for_masked_peels): New function.
4366 (vect_do_peeling): Skip prologue peeling if we're using a mask
4367 instead. Update call to vect_update_inits_of_drs.
4368 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4370 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
4371 alignment. Do not include the number of peeled iterations in
4372 the minimum threshold in that case.
4373 (vectorizable_induction): Adjust the start value down by
4374 LOOP_VINFO_MASK_SKIP_NITERS iterations.
4375 (vect_transform_loop): Call vect_prepare_for_masked_peels.
4376 Take the number of skipped iterations into account when calculating
4378 * tree-vect-stmts.c (vect_gen_while_not): New function.
4380 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4381 Alan Hayward <alan.hayward@arm.com>
4382 David Sherwood <david.sherwood@arm.com>
4384 * doc/sourcebuild.texi (vect_fully_masked): Document.
4385 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
4387 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
4389 (vect_analyze_loop_2): ...here. Don't check the vectorization
4390 factor against the number of loop iterations if the loop is
4393 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4394 Alan Hayward <alan.hayward@arm.com>
4395 David Sherwood <david.sherwood@arm.com>
4397 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
4398 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
4399 (dump_groups): Update accordingly.
4400 (iv_use::mem_type): New member variable.
4401 (address_p): New function.
4402 (record_use): Add a mem_type argument and initialize the new
4404 (record_group_use): Add a mem_type argument. Use address_p.
4405 Remove obsolete null checks of base_object. Update call to record_use.
4406 (find_interesting_uses_op): Update call to record_group_use.
4407 (find_interesting_uses_cond): Likewise.
4408 (find_interesting_uses_address): Likewise.
4409 (get_mem_type_for_internal_fn): New function.
4410 (find_address_like_use): Likewise.
4411 (find_interesting_uses_stmt): Try find_address_like_use before
4412 calling find_interesting_uses_op.
4413 (addr_offset_valid_p): Use the iv mem_type field as the type
4414 of the addressed memory.
4415 (add_autoinc_candidates): Likewise.
4416 (get_address_cost): Likewise.
4417 (split_small_address_groups_p): Use address_p.
4418 (split_address_groups): Likewise.
4419 (add_iv_candidate_for_use): Likewise.
4420 (autoinc_possible_for_pair): Likewise.
4421 (rewrite_groups): Likewise.
4422 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
4423 (determine_group_iv_cost): Update after split of USE_ADDRESS.
4424 (get_alias_ptr_type_for_ptr_address): New function.
4425 (rewrite_use_address): Rewrite address uses in calls that were
4426 identified by find_address_like_use.
4428 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4429 Alan Hayward <alan.hayward@arm.com>
4430 David Sherwood <david.sherwood@arm.com>
4432 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
4434 * gimple-expr.h (is_gimple_addressable: Likewise.
4435 * gimple-expr.c (is_gimple_address): Likewise.
4436 * internal-fn.c (expand_call_mem_ref): New function.
4437 (expand_mask_load_optab_fn): Use it.
4438 (expand_mask_store_optab_fn): Likewise.
4440 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4441 Alan Hayward <alan.hayward@arm.com>
4442 David Sherwood <david.sherwood@arm.com>
4444 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
4445 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
4446 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
4447 (cond_umax@var{mode}): Document.
4448 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
4449 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
4450 (cond_umin_optab, cond_umax_optab): New optabs.
4451 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
4452 (COND_IOR, COND_XOR): New internal functions.
4453 * internal-fn.h (get_conditional_internal_fn): Declare.
4454 * internal-fn.c (cond_binary_direct): New macro.
4455 (expand_cond_binary_optab_fn): Likewise.
4456 (direct_cond_binary_optab_supported_p): Likewise.
4457 (get_conditional_internal_fn): New function.
4458 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
4459 Cope with reduction statements that are vectorized as calls rather
4461 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
4462 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
4463 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
4464 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4465 (UNSPEC_COND_EOR): New unspecs.
4466 (optab): Add mappings for them.
4467 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
4468 (sve_int_op, sve_fp_op): New int attributes.
4470 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4471 Alan Hayward <alan.hayward@arm.com>
4472 David Sherwood <david.sherwood@arm.com>
4474 * optabs.def (while_ult_optab): New optab.
4475 * doc/md.texi (while_ult@var{m}@var{n}): Document.
4476 * internal-fn.def (WHILE_ULT): New internal function.
4477 * internal-fn.h (direct_internal_fn_supported_p): New override
4478 that takes two types as argument.
4479 * internal-fn.c (while_direct): New macro.
4480 (expand_while_optab_fn): New function.
4481 (convert_optab_supported_p): Likewise.
4482 (direct_while_optab_supported_p): New macro.
4483 * wide-int.h (wi::udiv_ceil): New function.
4484 * tree-vectorizer.h (rgroup_masks): New structure.
4485 (vec_loop_masks): New typedef.
4486 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
4488 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
4489 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
4490 (vect_max_vf): New function.
4491 (slpeel_make_loop_iterate_ntimes): Delete.
4492 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
4493 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
4494 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
4495 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
4496 internal-fn.h, stor-layout.h and optabs-query.h.
4497 (vect_set_loop_mask): New function.
4498 (add_preheader_seq): Likewise.
4499 (add_header_seq): Likewise.
4500 (interleave_supported_p): Likewise.
4501 (vect_maybe_permute_loop_masks): Likewise.
4502 (vect_set_loop_masks_directly): Likewise.
4503 (vect_set_loop_condition_masked): Likewise.
4504 (vect_set_loop_condition_unmasked): New function, split out from
4505 slpeel_make_loop_iterate_ntimes.
4506 (slpeel_make_loop_iterate_ntimes): Rename to..
4507 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
4508 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
4509 (vect_do_peeling): Update call accordingly.
4510 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
4512 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4513 mask_compare_type, can_fully_mask_p and fully_masked_p.
4514 (release_vec_loop_masks): New function.
4515 (_loop_vec_info): Use it to free the loop masks.
4516 (can_produce_all_loop_masks_p): New function.
4517 (vect_get_max_nscalars_per_iter): Likewise.
4518 (vect_verify_full_masking): Likewise.
4519 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
4520 retries, and free the mask rgroups before retrying. Check loop-wide
4521 reasons for disallowing fully-masked loops. Make the final decision
4522 about whether use a fully-masked loop or not.
4523 (vect_estimate_min_profitable_iters): Do not assume that peeling
4524 for the number of iterations will be needed for fully-masked loops.
4525 (vectorizable_reduction): Disable fully-masked loops.
4526 (vectorizable_live_operation): Likewise.
4527 (vect_halve_mask_nunits): New function.
4528 (vect_double_mask_nunits): Likewise.
4529 (vect_record_loop_mask): Likewise.
4530 (vect_get_loop_mask): Likewise.
4531 (vect_transform_loop): Handle the case in which the final loop
4532 iteration might handle a partial vector. Call vect_set_loop_condition
4533 instead of slpeel_make_loop_iterate_ntimes.
4534 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
4535 (check_load_store_masking): New function.
4536 (prepare_load_store_mask): Likewise.
4537 (vectorizable_store): Handle fully-masked loops.
4538 (vectorizable_load): Likewise.
4539 (supportable_widening_operation): Use vect_halve_mask_nunits for
4541 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
4542 (vect_gen_while): New function.
4543 * config/aarch64/aarch64.md (umax<mode>3): New expander.
4544 (aarch64_uqdec<mode>): New insn.
4546 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4547 Alan Hayward <alan.hayward@arm.com>
4548 David Sherwood <david.sherwood@arm.com>
4550 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
4551 (reduc_xor_scal_optab): New optabs.
4552 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
4553 (reduc_xor_scal_@var{m}): Document.
4554 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
4555 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
4557 * fold-const-call.c (fold_const_call): Handle them.
4558 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
4559 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
4560 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
4561 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
4562 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
4563 (UNSPEC_XORV): New unspecs.
4564 (optab): Add entries for them.
4565 (BITWISEV): New int iterator.
4566 (bit_reduc_op): New int attributes.
4568 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4569 Alan Hayward <alan.hayward@arm.com>
4570 David Sherwood <david.sherwood@arm.com>
4572 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
4573 * internal-fn.def (VEC_SHL_INSERT): New internal function.
4574 * optabs.def (vec_shl_insert_optab): New optab.
4575 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
4576 (duplicate_and_interleave): Likewise.
4577 * tree-vect-loop.c: Include internal-fn.h.
4578 (neutral_op_for_slp_reduction): New function, split out from
4579 get_initial_defs_for_reduction.
4580 (get_initial_def_for_reduction): Handle option 2 for variable-length
4581 vectors by loading the neutral value into a vector and then shifting
4582 the initial value into element 0.
4583 (get_initial_defs_for_reduction): Replace the code argument with
4584 the neutral value calculated by neutral_op_for_slp_reduction.
4585 Use gimple_build_vector for constant-length vectors.
4586 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
4587 but the first group_size elements have a neutral value.
4588 Use duplicate_and_interleave otherwise.
4589 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
4590 Update call to get_initial_defs_for_reduction. Handle SLP
4591 reductions for variable-length vectors by creating one vector
4592 result for each scalar result, with the elements associated
4593 with other scalar results stubbed out with the neutral value.
4594 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
4595 Require IFN_VEC_SHL_INSERT for double reductions on
4596 variable-length vectors, or SLP reductions that have
4597 a neutral value. Require can_duplicate_and_interleave_p
4598 support for variable-length unchained SLP reductions if there
4599 is no neutral value, such as for MIN/MAX reductions. Also require
4600 the number of vector elements to be a multiple of the number of
4601 SLP statements when doing variable-length unchained SLP reductions.
4602 Update call to vect_create_epilog_for_reduction.
4603 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
4604 and remove initial values.
4605 (duplicate_and_interleave): Make public.
4606 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
4607 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
4609 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4610 Alan Hayward <alan.hayward@arm.com>
4611 David Sherwood <david.sherwood@arm.com>
4613 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
4614 (can_duplicate_and_interleave_p): New function.
4615 (vect_get_and_check_slp_defs): Take the vector of statements
4616 rather than just the current one. Remove excess parentheses.
4617 Restriction rejectinon of vect_constant_def and vect_external_def
4618 for variable-length vectors to boolean types, or types for which
4619 can_duplicate_and_interleave_p is false.
4620 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
4621 (duplicate_and_interleave): New function.
4622 (vect_get_constant_vectors): Use gimple_build_vector for
4623 constant-length vectors and suitable variable-length constant
4624 vectors. Use duplicate_and_interleave for other variable-length
4625 vectors. Don't defer the update when inserting new statements.
4627 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4628 Alan Hayward <alan.hayward@arm.com>
4629 David Sherwood <david.sherwood@arm.com>
4631 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
4632 min_profitable_iters doesn't go negative.
4634 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4635 Alan Hayward <alan.hayward@arm.com>
4636 David Sherwood <david.sherwood@arm.com>
4638 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
4639 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
4640 * optabs.def (vec_mask_load_lanes_optab): New optab.
4641 (vec_mask_store_lanes_optab): Likewise.
4642 * internal-fn.def (MASK_LOAD_LANES): New internal function.
4643 (MASK_STORE_LANES): Likewise.
4644 * internal-fn.c (mask_load_lanes_direct): New macro.
4645 (mask_store_lanes_direct): Likewise.
4646 (expand_mask_load_optab_fn): Handle masked operations.
4647 (expand_mask_load_lanes_optab_fn): New macro.
4648 (expand_mask_store_optab_fn): Handle masked operations.
4649 (expand_mask_store_lanes_optab_fn): New macro.
4650 (direct_mask_load_lanes_optab_supported_p): Likewise.
4651 (direct_mask_store_lanes_optab_supported_p): Likewise.
4652 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
4654 (vect_load_lanes_supported): Likewise.
4655 * tree-vect-data-refs.c (strip_conversion): New function.
4656 (can_group_stmts_p): Likewise.
4657 (vect_analyze_data_ref_accesses): Use it instead of checking
4658 for a pair of assignments.
4659 (vect_store_lanes_supported): Take a masked_p parameter.
4660 (vect_load_lanes_supported): Likewise.
4661 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
4662 vect_store_lanes_supported and vect_load_lanes_supported.
4663 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4664 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
4665 parameter. Don't allow gaps for masked accesses.
4666 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
4667 and vect_load_lanes_supported.
4668 (get_load_store_type): Take a masked_p parameter and update
4669 call to get_group_load_store_type.
4670 (vectorizable_store): Update call to get_load_store_type.
4671 Handle IFN_MASK_STORE_LANES.
4672 (vectorizable_load): Update call to get_load_store_type.
4673 Handle IFN_MASK_LOAD_LANES.
4675 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4676 Alan Hayward <alan.hayward@arm.com>
4677 David Sherwood <david.sherwood@arm.com>
4679 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
4681 * config/aarch64/aarch64-protos.h
4682 (aarch64_sve_struct_memory_operand_p): Declare.
4683 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
4684 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
4685 (VPRED, vpred): Handle SVE structure modes.
4686 * config/aarch64/constraints.md (Utx): New constraint.
4687 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
4688 (aarch64_sve_struct_nonimmediate_operand): New predicates.
4689 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
4690 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
4691 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
4692 structure modes. Split into pieces after RA.
4693 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
4694 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
4696 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
4697 SVE structure modes.
4698 (aarch64_classify_address): Likewise.
4699 (sizetochar): Move earlier in file.
4700 (aarch64_print_operand): Handle SVE register lists.
4701 (aarch64_array_mode): New function.
4702 (aarch64_sve_struct_memory_operand_p): Likewise.
4703 (TARGET_ARRAY_MODE): Redefine.
4705 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4706 Alan Hayward <alan.hayward@arm.com>
4707 David Sherwood <david.sherwood@arm.com>
4709 * target.def (array_mode): New target hook.
4710 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
4711 * doc/tm.texi: Regenerate.
4712 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
4713 * hooks.c (hook_optmode_mode_uhwi_none): New function.
4714 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
4716 * stor-layout.c (mode_for_array): Likewise. Support polynomial
4719 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4720 Alan Hayward <alan.hayward@arm.com>
4721 David Sherwood <david.sherwood@arm.com>
4723 * fold-const.c (fold_binary_loc): Check the argument types
4724 rather than the result type when testing for a vector operation.
4726 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4728 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
4729 * doc/tm.texi: Regenerate.
4731 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4732 Alan Hayward <alan.hayward@arm.com>
4733 David Sherwood <david.sherwood@arm.com>
4735 * doc/invoke.texi (-msve-vector-bits=): Document new option.
4736 (sve): Document new AArch64 extension.
4737 * doc/md.texi (w): Extend the description of the AArch64
4738 constraint to include SVE vectors.
4739 (Upl, Upa): Document new AArch64 predicate constraints.
4740 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
4742 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
4743 (msve-vector-bits=): New option.
4744 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
4745 SVE when these are disabled.
4746 (sve): New extension.
4747 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
4748 modes. Adjust their number of units based on aarch64_sve_vg.
4749 (MAX_BITSIZE_MODE_ANY_MODE): Define.
4750 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
4751 aarch64_addr_query_type.
4752 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
4753 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
4754 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
4755 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
4756 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
4757 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
4758 (aarch64_simd_imm_zero_p): Delete.
4759 (aarch64_check_zero_based_sve_index_immediate): Declare.
4760 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4761 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4762 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4763 (aarch64_sve_float_mul_immediate_p): Likewise.
4764 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4766 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
4767 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
4768 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
4769 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
4770 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
4771 (aarch64_regmode_natural_size): Likewise.
4772 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
4773 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
4775 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
4776 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
4777 for VG and the SVE predicate registers.
4778 (V_ALIASES): Add a "z"-prefixed alias.
4779 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
4780 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
4781 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
4782 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
4783 (REG_CLASS_NAMES): Add entries for them.
4784 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
4785 and the predicate registers.
4786 (aarch64_sve_vg): Declare.
4787 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
4788 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
4789 (REGMODE_NATURAL_SIZE): Define.
4790 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
4792 * config/aarch64/aarch64.c: Include cfgrtl.h.
4793 (simd_immediate_info): Add a constructor for series vectors,
4794 and an associated step field.
4795 (aarch64_sve_vg): New variable.
4796 (aarch64_dbx_register_number): Handle VG and the predicate registers.
4797 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
4798 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
4799 (VEC_ANY_DATA, VEC_STRUCT): New constants.
4800 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
4801 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
4802 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
4803 (aarch64_get_mask_mode): New functions.
4804 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
4805 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
4806 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
4807 predicate modes and predicate registers. Explicitly restrict
4808 GPRs to modes of 16 bytes or smaller. Only allow FP registers
4809 to store a vector mode if it is recognized by
4810 aarch64_classify_vector_mode.
4811 (aarch64_regmode_natural_size): New function.
4812 (aarch64_hard_regno_caller_save_mode): Return the original mode
4814 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
4815 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
4816 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
4817 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
4819 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
4820 does not overlap dest if the function is frame-related. Handle
4822 (aarch64_split_add_offset): New function.
4823 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
4824 them aarch64_add_offset.
4825 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
4826 and update call to aarch64_sub_sp.
4827 (aarch64_add_cfa_expression): New function.
4828 (aarch64_expand_prologue): Pass extra temporary registers to the
4829 functions above. Handle the case in which we need to emit new
4830 DW_CFA_expressions for registers that were originally saved
4831 relative to the stack pointer, but now have to be expressed
4832 relative to the frame pointer.
4833 (aarch64_output_mi_thunk): Pass extra temporary registers to the
4835 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
4836 IP0 and IP1 values for SVE frames.
4837 (aarch64_expand_vec_series): New function.
4838 (aarch64_expand_sve_widened_duplicate): Likewise.
4839 (aarch64_expand_sve_const_vector): Likewise.
4840 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
4841 Handle SVE constants. Use emit_move_insn to move a force_const_mem
4842 into the register, rather than emitting a SET directly.
4843 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
4844 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
4845 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
4846 (offset_9bit_signed_scaled_p): New functions.
4847 (aarch64_replicate_bitmask_imm): New function.
4848 (aarch64_bitmask_imm): Use it.
4849 (aarch64_cannot_force_const_mem): Reject expressions involving
4850 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
4851 (aarch64_classify_index): Handle SVE indices, by requiring
4852 a plain register index with a scale that matches the element size.
4853 (aarch64_classify_address): Handle SVE addresses. Assert that
4854 the mode of the address is VOIDmode or an integer mode.
4855 Update call to aarch64_classify_symbol.
4856 (aarch64_classify_symbolic_expression): Update call to
4857 aarch64_classify_symbol.
4858 (aarch64_const_vec_all_in_range_p): New function.
4859 (aarch64_print_vector_float_operand): Likewise.
4860 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
4861 "vN" for FP registers with SVE modes. Handle (const ...) vectors
4862 and the FP immediates 1.0 and 0.5.
4863 (aarch64_print_address_internal): Handle SVE addresses.
4864 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
4865 (aarch64_regno_regclass): Handle predicate registers.
4866 (aarch64_secondary_reload): Handle big-endian reloads of SVE
4868 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
4869 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
4870 (aarch64_convert_sve_vector_bits): New function.
4871 (aarch64_override_options): Use it to handle -msve-vector-bits=.
4872 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4874 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
4875 Handle SVE vector and predicate modes. Accept VL-based constants
4876 that need only one temporary register, and VL offsets that require
4877 no temporary registers.
4878 (aarch64_conditional_register_usage): Mark the predicate registers
4879 as fixed if SVE isn't available.
4880 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
4881 Return true for SVE vector and predicate modes.
4882 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
4883 rather than an unsigned int. Handle SVE modes.
4884 (aarch64_preferred_simd_mode): Update call accordingly. Handle
4886 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
4888 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4889 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4890 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4891 (aarch64_sve_float_mul_immediate_p): New functions.
4892 (aarch64_sve_valid_immediate): New function.
4893 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
4894 Explicitly reject structure modes. Check for INDEX constants.
4895 Handle PTRUE and PFALSE constants.
4896 (aarch64_check_zero_based_sve_index_immediate): New function.
4897 (aarch64_simd_imm_zero_p): Delete.
4898 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
4899 vector modes. Accept constants in the range of CNT[BHWD].
4900 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
4901 ask for an Advanced SIMD mode.
4902 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
4903 (aarch64_simd_vector_alignment): Handle SVE predicates.
4904 (aarch64_vectorize_preferred_vector_alignment): New function.
4905 (aarch64_simd_vector_alignment_reachable): Use it instead of
4907 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
4908 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
4910 (MAX_VECT_LEN): Delete.
4911 (expand_vec_perm_d): Add a vec_flags field.
4912 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
4913 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
4914 (aarch64_evpc_ext): Don't apply a big-endian lane correction
4916 (aarch64_evpc_rev): Rename to...
4917 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
4918 (aarch64_evpc_rev_global): New function.
4919 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
4920 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
4922 (aarch64_evpc_sve_tbl): New function.
4923 (aarch64_expand_vec_perm_const_1): Update after rename of
4924 aarch64_evpc_rev. Handle SVE permutes too, trying
4925 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
4926 than aarch64_evpc_tbl.
4927 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
4928 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
4929 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
4930 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
4931 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
4932 (aarch64_expand_sve_vcond): New functions.
4933 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
4934 of aarch64_vector_mode_p.
4935 (aarch64_dwarf_poly_indeterminate_value): New function.
4936 (aarch64_compute_pressure_classes): Likewise.
4937 (aarch64_can_change_mode_class): Likewise.
4938 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
4939 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
4940 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
4941 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
4942 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
4943 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
4944 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
4945 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
4947 (Dn, Dl, Dr): Accept const as well as const_vector.
4948 (Dz): Likewise. Compare against CONST0_RTX.
4949 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
4950 of "vector" where appropriate.
4951 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
4952 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
4953 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
4954 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
4955 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
4956 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
4957 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
4958 (v_int_equiv): Extend to SVE modes.
4959 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
4961 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
4962 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
4963 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
4964 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
4965 (SVE_COND_FP_CMP): New int iterators.
4966 (perm_hilo): Handle the new unpack unspecs.
4967 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
4969 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
4970 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
4971 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
4972 (aarch64_equality_operator, aarch64_constant_vector_operand)
4973 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
4974 (aarch64_sve_nonimmediate_operand): Likewise.
4975 (aarch64_sve_general_operand): Likewise.
4976 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
4977 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
4978 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
4979 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
4980 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
4981 (aarch64_sve_float_arith_immediate): Likewise.
4982 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
4983 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
4984 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
4985 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
4986 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
4987 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
4988 (aarch64_sve_float_arith_operand): Likewise.
4989 (aarch64_sve_float_arith_with_sub_operand): Likewise.
4990 (aarch64_sve_float_mul_operand): Likewise.
4991 (aarch64_sve_vec_perm_operand): Likewise.
4992 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
4993 (aarch64_mov_operand): Accept const_poly_int and const_vector.
4994 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
4995 as well as const_vector.
4996 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
4997 in file. Use CONST0_RTX and CONSTM1_RTX.
4998 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
4999 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
5000 Use aarch64_simd_imm_zero.
5001 * config/aarch64/aarch64-sve.md: New file.
5002 * config/aarch64/aarch64.md: Include it.
5003 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
5004 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
5005 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
5006 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
5007 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
5008 (sve): New attribute.
5009 (enabled): Disable instructions with the sve attribute unless
5011 (movqi, movhi): Pass CONST_POLY_INT operaneds through
5012 aarch64_expand_mov_immediate.
5013 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
5014 CNT[BHSD] immediates.
5015 (movti): Split CONST_POLY_INT moves into two halves.
5016 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
5017 Split additions that need a temporary here if the destination
5018 is the stack pointer.
5019 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
5020 (*add<mode>3_poly_1): New instruction.
5021 (set_clobber_cc): New expander.
5023 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5025 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
5026 parameter and use it instead of GET_MODE_SIZE (innermode). Use
5027 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
5028 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
5029 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
5030 Change innermode from fixed_mode_size to machine_mode.
5031 (simplify_subreg): Update call accordingly. Handle a constant-sized
5032 subreg of a variable-length CONST_VECTOR.
5034 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5035 Alan Hayward <alan.hayward@arm.com>
5036 David Sherwood <david.sherwood@arm.com>
5038 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
5039 (add_offset_to_base): New function, split out from...
5040 (create_mem_ref): ...here. When handling a scale other than 1,
5041 check first whether the address is valid without the offset.
5042 Add it into the base if so, leaving the index and scale as-is.
5044 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5047 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
5048 fold_for_warn before checking if arg2 is INTEGER_CST.
5050 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
5052 * config/rs6000/predicates.md (load_multiple_operation): Delete.
5053 (store_multiple_operation): Delete.
5054 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
5055 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
5056 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
5057 guarded by TARGET_STRING.
5058 (rs6000_output_load_multiple): Delete.
5059 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
5060 OPTION_MASK_STRING / TARGET_STRING handling.
5061 (print_operand) <'N', 'O'>: Add comment that these are unused now.
5062 (const rs6000_opt_masks) <"string">: Change mask to 0.
5063 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
5064 (MASK_STRING): Delete.
5065 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
5067 (load_multiple): Delete.
5074 (store_multiple): Delete.
5081 (movmemsi_8reg): Delete.
5082 (corresponding unnamed define_insn): Delete.
5083 (movmemsi_6reg): Delete.
5084 (corresponding unnamed define_insn): Delete.
5085 (movmemsi_4reg): Delete.
5086 (corresponding unnamed define_insn): Delete.
5087 (movmemsi_2reg): Delete.
5088 (corresponding unnamed define_insn): Delete.
5089 (movmemsi_1reg): Delete.
5090 (corresponding unnamed define_insn): Delete.
5091 * config/rs6000/rs6000.opt (mno-string): New.
5092 (mstring): Replace by deprecation warning stub.
5093 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
5095 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5097 * regrename.c (regrename_do_replace): If replacing the same
5098 reg multiple times, try to reuse last created gen_raw_REG.
5101 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
5102 main to workaround a bug in GDB.
5104 2018-01-12 Tom de Vries <tom@codesourcery.com>
5107 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
5109 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
5111 PR rtl-optimization/80481
5112 * ira-color.c (get_cap_member): New function.
5113 (allocnos_conflict_by_live_ranges_p): Use it.
5114 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
5115 (setup_slot_coalesced_allocno_live_ranges): Ditto.
5117 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
5120 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
5121 (*saddl_se_1): Ditto.
5123 (*ssubl_se_1): Ditto.
5125 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5127 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
5128 rather than wi::to_widest for DR_INITs.
5129 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
5130 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
5131 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
5133 (vect_analyze_group_access_1): Note that here.
5135 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5137 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
5138 polynomial type sizes.
5140 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5142 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
5143 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
5144 (gimple_add_tmp_var): Likewise.
5146 2018-01-12 Martin Liska <mliska@suse.cz>
5148 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
5149 (gimple_alloc_sizes): Likewise.
5150 (dump_gimple_statistics): Use PRIu64 in printf format.
5151 * gimple.h: Change uint64_t to int.
5153 2018-01-12 Martin Liska <mliska@suse.cz>
5155 * tree-core.h: Use uint64_t instead of int.
5156 * tree.c (tree_node_counts): Likewise.
5157 (tree_node_sizes): Likewise.
5158 (dump_tree_statistics): Use PRIu64 in printf format.
5160 2018-01-12 Martin Liska <mliska@suse.cz>
5162 * Makefile.in: As qsort_chk is implemented in vec.c, add
5163 vec.o to linkage of gencfn-macros.
5164 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
5165 passing the info to record_node_allocation_statistics.
5166 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
5168 * ggc-common.c (struct ggc_usage): Add operator== and use
5169 it in operator< and compare function.
5170 * mem-stats.h (struct mem_usage): Likewise.
5171 * vec.c (struct vec_usage): Remove operator< and compare
5172 function. Can be simply inherited.
5174 2018-01-12 Martin Jambor <mjambor@suse.cz>
5177 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
5178 * tree-ssa-math-opts.c: Include domwalk.h.
5179 (convert_mult_to_fma_1): New function.
5180 (fma_transformation_info): New type.
5181 (fma_deferring_state): Likewise.
5182 (cancel_fma_deferring): New function.
5183 (result_of_phi): Likewise.
5184 (last_fma_candidate_feeds_initial_phi): Likewise.
5185 (convert_mult_to_fma): Added deferring logic, split actual
5186 transformation to convert_mult_to_fma_1.
5187 (math_opts_dom_walker): New type.
5188 (math_opts_dom_walker::after_dom_children): New method, body moved
5189 here from pass_optimize_widening_mul::execute, added deferring logic
5191 (pass_optimize_widening_mul::execute): Moved most of code to
5192 math_opts_dom_walker::after_dom_children.
5193 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
5194 * config/i386/i386.c (ix86_option_override_internal): Added
5195 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
5197 2018-01-12 Richard Biener <rguenther@suse.de>
5200 * dwarf2out.c (gen_variable_die): Do not reset old_die for
5201 inline instance vars.
5203 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
5206 * config/rx/rx.c (rx_is_restricted_memory_address):
5209 2018-01-12 Richard Biener <rguenther@suse.de>
5211 PR tree-optimization/80846
5212 * target.def (split_reduction): New target hook.
5213 * targhooks.c (default_split_reduction): New function.
5214 * targhooks.h (default_split_reduction): Declare.
5215 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
5216 target requests first reduce vectors by combining low and high
5218 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
5219 (get_vectype_for_scalar_type_and_size): Export.
5220 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
5221 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
5222 * doc/tm.texi: Regenerate.
5223 * config/i386/i386.c (ix86_split_reduction): Implement
5224 TARGET_VECTORIZE_SPLIT_REDUCTION.
5226 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5229 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
5230 in PIC mode except for TARGET_VXWORKS_RTP.
5231 * config/sparc/sparc.c: Include cfgrtl.h.
5232 (TARGET_INIT_PIC_REG): Define.
5233 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
5234 (sparc_pic_register_p): New predicate.
5235 (sparc_legitimate_address_p): Use it.
5236 (sparc_legitimize_pic_address): Likewise.
5237 (sparc_delegitimize_address): Likewise.
5238 (sparc_mode_dependent_address_p): Likewise.
5239 (gen_load_pcrel_sym): Remove 4th parameter.
5240 (load_got_register): Adjust call to above. Remove obsolete stuff.
5241 (sparc_expand_prologue): Do not call load_got_register here.
5242 (sparc_flat_expand_prologue): Likewise.
5243 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
5244 (sparc_use_pseudo_pic_reg): New function.
5245 (sparc_init_pic_reg): Likewise.
5246 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5247 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
5249 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
5251 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
5252 Add item for branch_cost.
5254 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5256 PR rtl-optimization/83565
5257 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
5258 not extend the result to a larger mode for rotate operations.
5259 (num_sign_bit_copies1): Likewise.
5261 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5264 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
5266 Use values-Xc.o for -pedantic.
5267 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
5269 2018-01-12 Martin Liska <mliska@suse.cz>
5272 * ipa-devirt.c (final_warning_record::grow_type_warnings):
5274 (possible_polymorphic_call_targets): Use it.
5275 (ipa_devirt): Likewise.
5277 2018-01-12 Martin Liska <mliska@suse.cz>
5279 * profile-count.h (enum profile_quality): Use 0 as invalid
5280 enum value of profile_quality.
5282 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
5284 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
5285 -mext-string options.
5287 2018-01-12 Richard Biener <rguenther@suse.de>
5289 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
5290 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
5291 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
5293 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
5295 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5297 * configure.ac (--with-long-double-format): Add support for the
5298 configuration option to change the default long double format on
5300 * config.gcc (powerpc*-linux*-*): Likewise.
5301 * configure: Regenerate.
5302 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
5303 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
5304 used without modification.
5306 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5308 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
5309 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
5310 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
5311 MISC_BUILTIN_SPEC_BARRIER.
5312 (rs6000_init_builtins): Likewise.
5313 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
5315 (speculation_barrier): New define_insn.
5316 * doc/extend.texi: Document __builtin_speculation_barrier.
5318 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5321 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
5322 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
5323 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
5325 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
5326 integral modes instead of "ss" and "sd".
5327 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
5328 vectors with 32-bit and 64-bit elements.
5329 (vecdupssescalarmodesuffix): New mode attribute.
5330 (vec_dup<mode>): Use it.
5332 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
5335 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
5336 frame if argument is passed on stack.
5338 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5341 * ree.c (combine_reaching_defs): Optimize also
5342 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
5343 reg2=any_extend(exp); reg1=reg2;, formatting fix.
5345 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5348 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
5350 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5353 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
5354 after they are computed.
5356 2018-01-11 Bin Cheng <bin.cheng@arm.com>
5358 PR tree-optimization/83695
5359 * gimple-loop-linterchange.cc
5360 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
5361 reset cached scev information after interchange.
5362 (pass_linterchange::execute): Remove call to scev_reset_htab.
5364 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5366 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
5367 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
5368 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
5369 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
5370 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
5371 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
5372 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
5373 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
5374 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
5375 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
5376 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
5377 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
5378 (V_lane_reg): Likewise.
5379 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
5381 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
5382 (vfmal_lane_low<mode>_intrinsic,
5383 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
5384 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
5385 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
5386 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
5387 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
5388 vfmsl_lane_high<mode>_intrinsic): New define_insns.
5390 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5392 * config/arm/arm-cpus.in (fp16fml): New feature.
5393 (ALL_SIMD): Add fp16fml.
5394 (armv8.2-a): Add fp16fml as an option.
5395 (armv8.3-a): Likewise.
5396 (armv8.4-a): Add fp16fml as part of fp16.
5397 * config/arm/arm.h (TARGET_FP16FML): Define.
5398 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
5400 * config/arm/arm-modes.def (V2HF): Define.
5401 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
5402 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
5403 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
5404 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
5405 vfmsl_low, vfmsl_high): New set of builtins.
5406 * config/arm/iterators.md (PLUSMINUS): New code iterator.
5407 (vfml_op): New code attribute.
5408 (VFMLHALVES): New int iterator.
5409 (VFML, VFMLSEL): New mode attributes.
5410 (V_reg): Define mapping for V2HF.
5411 (V_hi, V_lo): New mode attributes.
5412 (VF_constraint): Likewise.
5413 (vfml_half, vfml_half_selector): New int attributes.
5414 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
5416 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
5417 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
5419 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
5420 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
5421 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
5422 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
5424 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
5425 Document new effective target and option set.
5427 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5429 * config/arm/arm-cpus.in (armv8_4): New feature.
5430 (ARMv8_4a): New fgroup.
5431 (armv8.4-a): New arch.
5432 * config/arm/arm-tables.opt: Regenerate.
5433 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
5434 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
5435 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
5436 Add matching rules for -march=armv8.4-a and extensions.
5437 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
5439 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
5442 * config/rx/rx.md (BW): New mode attribute.
5443 (sync_lock_test_and_setsi): Add mode suffix to insn output.
5445 2018-01-11 Richard Biener <rguenther@suse.de>
5447 PR tree-optimization/83435
5448 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
5449 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
5450 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
5452 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5453 Alan Hayward <alan.hayward@arm.com>
5454 David Sherwood <david.sherwood@arm.com>
5456 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
5458 (aarch64_classify_address): Initialize it. Track polynomial offsets.
5459 (aarch64_print_address_internal): Use it to check for a zero offset.
5461 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5462 Alan Hayward <alan.hayward@arm.com>
5463 David Sherwood <david.sherwood@arm.com>
5465 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
5466 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
5467 Return a poly_int64 rather than a HOST_WIDE_INT.
5468 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
5469 rather than a HOST_WIDE_INT.
5470 * config/aarch64/aarch64.h (aarch64_frame): Protect with
5471 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
5472 hard_fp_offset, frame_size, initial_adjust, callee_offset and
5473 final_offset from HOST_WIDE_INT to poly_int64.
5474 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
5475 to_constant when getting the number of units in an Advanced SIMD
5477 (aarch64_builtin_vectorized_function): Check for a constant number
5479 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
5481 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
5482 attribute instead of GET_MODE_NUNITS.
5483 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
5484 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
5485 GET_MODE_SIZE for fixed-size registers.
5486 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
5487 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
5488 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
5489 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
5490 (aarch64_print_operand, aarch64_print_address_internal)
5491 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
5492 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
5493 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
5494 Handle polynomial GET_MODE_SIZE.
5495 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
5496 wider than SImode without modification.
5497 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
5498 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
5499 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
5500 passing and returning SVE modes.
5501 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
5502 rather than GEN_INT.
5503 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
5504 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
5505 (aarch64_allocate_and_probe_stack_space): Likewise.
5506 (aarch64_layout_frame): Cope with polynomial offsets.
5507 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
5508 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
5510 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
5511 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
5512 poly_int64 rather than a HOST_WIDE_INT.
5513 (aarch64_get_separate_components, aarch64_process_components)
5514 (aarch64_expand_prologue, aarch64_expand_epilogue)
5515 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
5516 (aarch64_anchor_offset): New function, split out from...
5517 (aarch64_legitimize_address): ...here.
5518 (aarch64_builtin_vectorization_cost): Handle polynomial
5519 TYPE_VECTOR_SUBPARTS.
5520 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
5522 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
5523 number of elements from the PARALLEL rather than the mode.
5524 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
5525 rather than GET_MODE_BITSIZE.
5526 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
5527 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
5528 (aarch64_expand_vec_perm_const_1): Handle polynomial
5529 d->perm.length () and d->perm elements.
5530 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
5531 Apply to_constant to d->perm elements.
5532 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
5533 polynomial CONST_VECTOR_NUNITS.
5534 (aarch64_move_pointer): Take amount as a poly_int64 rather
5536 (aarch64_progress_pointer): Avoid temporary variable.
5537 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
5538 the mode attribute instead of GET_MODE.
5540 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5541 Alan Hayward <alan.hayward@arm.com>
5542 David Sherwood <david.sherwood@arm.com>
5544 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
5545 x exists before using it.
5546 (aarch64_add_constant_internal): Rename to...
5547 (aarch64_add_offset_1): ...this. Replace regnum with separate
5548 src and dest rtxes. Handle the case in which they're different,
5549 including when the offset is zero. Replace scratchreg with an rtx.
5550 Use 2 additions if there is no spare register into which we can
5551 move a 16-bit constant.
5552 (aarch64_add_constant): Delete.
5553 (aarch64_add_offset): Replace reg with separate src and dest
5554 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
5555 Use aarch64_add_offset_1.
5556 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
5557 an rtx rather than an int. Take the delta as a poly_int64
5558 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
5559 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
5560 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
5561 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
5562 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
5564 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
5565 aarch64_add_constant.
5567 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5569 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
5570 Use scalar_float_mode.
5572 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5574 * config/aarch64/aarch64-simd.md
5575 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
5576 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
5577 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
5578 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
5579 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
5580 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
5581 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
5582 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
5583 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
5584 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
5586 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5589 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
5590 targ_options->x_arm_arch_string is non NULL.
5592 2018-01-11 Tamar Christina <tamar.christina@arm.com>
5594 * config/aarch64/aarch64.h
5595 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
5597 2018-01-11 Sudakshina Das <sudi.das@arm.com>
5600 * expmed.c (emit_store_flag_force): Swap if const op0
5601 and change VOIDmode to mode of op0.
5603 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5605 PR rtl-optimization/83761
5606 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
5607 than bytes to mode_for_size.
5609 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5612 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
5613 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
5616 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5619 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
5620 when in layout mode.
5621 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
5622 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
5625 2018-01-10 Michael Collison <michael.collison@arm.com>
5627 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
5628 * config/aarch64/aarch64-option-extension.def: Add
5629 AARCH64_OPT_EXTENSION of 'fp16fml'.
5630 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5631 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
5632 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
5633 * config/aarch64/constraints.md (Ui7): New constraint.
5634 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
5635 (VFMLA_SEL_W): Ditto.
5638 (VFMLA16_LOW): New int iterator.
5639 (VFMLA16_HIGH): Ditto.
5640 (UNSPEC_FMLAL): New unspec.
5641 (UNSPEC_FMLSL): Ditto.
5642 (UNSPEC_FMLAL2): Ditto.
5643 (UNSPEC_FMLSL2): Ditto.
5644 (f16mac): New code attribute.
5645 * config/aarch64/aarch64-simd-builtins.def
5646 (aarch64_fmlal_lowv2sf): Ditto.
5647 (aarch64_fmlsl_lowv2sf): Ditto.
5648 (aarch64_fmlalq_lowv4sf): Ditto.
5649 (aarch64_fmlslq_lowv4sf): Ditto.
5650 (aarch64_fmlal_highv2sf): Ditto.
5651 (aarch64_fmlsl_highv2sf): Ditto.
5652 (aarch64_fmlalq_highv4sf): Ditto.
5653 (aarch64_fmlslq_highv4sf): Ditto.
5654 (aarch64_fmlal_lane_lowv2sf): Ditto.
5655 (aarch64_fmlsl_lane_lowv2sf): Ditto.
5656 (aarch64_fmlal_laneq_lowv2sf): Ditto.
5657 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
5658 (aarch64_fmlalq_lane_lowv4sf): Ditto.
5659 (aarch64_fmlsl_lane_lowv4sf): Ditto.
5660 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
5661 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
5662 (aarch64_fmlal_lane_highv2sf): Ditto.
5663 (aarch64_fmlsl_lane_highv2sf): Ditto.
5664 (aarch64_fmlal_laneq_highv2sf): Ditto.
5665 (aarch64_fmlsl_laneq_highv2sf): Ditto.
5666 (aarch64_fmlalq_lane_highv4sf): Ditto.
5667 (aarch64_fmlsl_lane_highv4sf): Ditto.
5668 (aarch64_fmlalq_laneq_highv4sf): Ditto.
5669 (aarch64_fmlsl_laneq_highv4sf): Ditto.
5670 * config/aarch64/aarch64-simd.md:
5671 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
5672 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5673 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
5674 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5675 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
5676 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
5677 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
5678 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
5679 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
5680 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
5681 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
5682 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
5683 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
5684 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
5685 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
5686 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
5687 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
5688 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
5689 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
5690 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
5691 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
5692 (vfmlsl_low_u32): Ditto.
5693 (vfmlalq_low_u32): Ditto.
5694 (vfmlslq_low_u32): Ditto.
5695 (vfmlal_high_u32): Ditto.
5696 (vfmlsl_high_u32): Ditto.
5697 (vfmlalq_high_u32): Ditto.
5698 (vfmlslq_high_u32): Ditto.
5699 (vfmlal_lane_low_u32): Ditto.
5700 (vfmlsl_lane_low_u32): Ditto.
5701 (vfmlal_laneq_low_u32): Ditto.
5702 (vfmlsl_laneq_low_u32): Ditto.
5703 (vfmlalq_lane_low_u32): Ditto.
5704 (vfmlslq_lane_low_u32): Ditto.
5705 (vfmlalq_laneq_low_u32): Ditto.
5706 (vfmlslq_laneq_low_u32): Ditto.
5707 (vfmlal_lane_high_u32): Ditto.
5708 (vfmlsl_lane_high_u32): Ditto.
5709 (vfmlal_laneq_high_u32): Ditto.
5710 (vfmlsl_laneq_high_u32): Ditto.
5711 (vfmlalq_lane_high_u32): Ditto.
5712 (vfmlslq_lane_high_u32): Ditto.
5713 (vfmlalq_laneq_high_u32): Ditto.
5714 (vfmlslq_laneq_high_u32): Ditto.
5715 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
5716 (AARCH64_FL_FOR_ARCH8_4): New.
5717 (AARCH64_ISA_F16FML): New ISA flag.
5718 (TARGET_F16FML): New feature flag for fp16fml.
5719 (doc/invoke.texi): Document new fp16fml option.
5721 2018-01-10 Michael Collison <michael.collison@arm.com>
5723 * config/aarch64/aarch64-builtins.c:
5724 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
5725 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5726 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
5727 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
5728 (AARCH64_ISA_SHA3): New ISA flag.
5729 (TARGET_SHA3): New feature flag for sha3.
5730 * config/aarch64/iterators.md (sha512_op): New int attribute.
5731 (CRYPTO_SHA512): New int iterator.
5732 (UNSPEC_SHA512H): New unspec.
5733 (UNSPEC_SHA512H2): Ditto.
5734 (UNSPEC_SHA512SU0): Ditto.
5735 (UNSPEC_SHA512SU1): Ditto.
5736 * config/aarch64/aarch64-simd-builtins.def
5737 (aarch64_crypto_sha512hqv2di): New builtin.
5738 (aarch64_crypto_sha512h2qv2di): Ditto.
5739 (aarch64_crypto_sha512su0qv2di): Ditto.
5740 (aarch64_crypto_sha512su1qv2di): Ditto.
5741 (aarch64_eor3qv8hi): Ditto.
5742 (aarch64_rax1qv2di): Ditto.
5743 (aarch64_xarqv2di): Ditto.
5744 (aarch64_bcaxqv8hi): Ditto.
5745 * config/aarch64/aarch64-simd.md:
5746 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
5747 (aarch64_crypto_sha512su0qv2di): Ditto.
5748 (aarch64_crypto_sha512su1qv2di): Ditto.
5749 (aarch64_eor3qv8hi): Ditto.
5750 (aarch64_rax1qv2di): Ditto.
5751 (aarch64_xarqv2di): Ditto.
5752 (aarch64_bcaxqv8hi): Ditto.
5753 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
5754 (vsha512h2q_u64): Ditto.
5755 (vsha512su0q_u64): Ditto.
5756 (vsha512su1q_u64): Ditto.
5757 (veor3q_u16): Ditto.
5758 (vrax1q_u64): Ditto.
5760 (vbcaxq_u16): Ditto.
5761 * config/arm/types.md (crypto_sha512): New type attribute.
5762 (crypto_sha3): Ditto.
5763 (doc/invoke.texi): Document new sha3 option.
5765 2018-01-10 Michael Collison <michael.collison@arm.com>
5767 * config/aarch64/aarch64-builtins.c:
5768 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
5769 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5770 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
5771 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
5772 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
5773 (AARCH64_ISA_SM4): New ISA flag.
5774 (TARGET_SM4): New feature flag for sm4.
5775 * config/aarch64/aarch64-simd-builtins.def
5776 (aarch64_sm3ss1qv4si): Ditto.
5777 (aarch64_sm3tt1aq4si): Ditto.
5778 (aarch64_sm3tt1bq4si): Ditto.
5779 (aarch64_sm3tt2aq4si): Ditto.
5780 (aarch64_sm3tt2bq4si): Ditto.
5781 (aarch64_sm3partw1qv4si): Ditto.
5782 (aarch64_sm3partw2qv4si): Ditto.
5783 (aarch64_sm4eqv4si): Ditto.
5784 (aarch64_sm4ekeyqv4si): Ditto.
5785 * config/aarch64/aarch64-simd.md:
5786 (aarch64_sm3ss1qv4si): Ditto.
5787 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
5788 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
5789 (aarch64_sm4eqv4si): Ditto.
5790 (aarch64_sm4ekeyqv4si): Ditto.
5791 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
5792 (sm3part_op): Ditto.
5793 (CRYPTO_SM3TT): Ditto.
5794 (CRYPTO_SM3PART): Ditto.
5795 (UNSPEC_SM3SS1): New unspec.
5796 (UNSPEC_SM3TT1A): Ditto.
5797 (UNSPEC_SM3TT1B): Ditto.
5798 (UNSPEC_SM3TT2A): Ditto.
5799 (UNSPEC_SM3TT2B): Ditto.
5800 (UNSPEC_SM3PARTW1): Ditto.
5801 (UNSPEC_SM3PARTW2): Ditto.
5802 (UNSPEC_SM4E): Ditto.
5803 (UNSPEC_SM4EKEY): Ditto.
5804 * config/aarch64/constraints.md (Ui2): New constraint.
5805 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
5806 * config/arm/types.md (crypto_sm3): New type attribute.
5807 (crypto_sm4): Ditto.
5808 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
5809 (vsm3tt1aq_u32): Ditto.
5810 (vsm3tt1bq_u32): Ditto.
5811 (vsm3tt2aq_u32): Ditto.
5812 (vsm3tt2bq_u32): Ditto.
5813 (vsm3partw1q_u32): Ditto.
5814 (vsm3partw2q_u32): Ditto.
5815 (vsm4eq_u32): Ditto.
5816 (vsm4ekeyq_u32): Ditto.
5817 (doc/invoke.texi): Document new sm4 option.
5819 2018-01-10 Michael Collison <michael.collison@arm.com>
5821 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
5822 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
5823 (AARCH64_FL_FOR_ARCH8_4): New.
5824 (AARCH64_FL_V8_4): New flag.
5825 (doc/invoke.texi): Document new armv8.4-a option.
5827 2018-01-10 Michael Collison <michael.collison@arm.com>
5829 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5830 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
5831 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
5832 * config/aarch64/aarch64-option-extension.def: Add
5833 AARCH64_OPT_EXTENSION of 'sha2'.
5834 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
5835 (crypto): Disable sha2 and aes if crypto disabled.
5836 (crypto): Enable aes and sha2 if enabled.
5837 (simd): Disable sha2 and aes if simd disabled.
5838 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
5840 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
5841 (TARGET_SHA2): New feature flag for sha2.
5842 (TARGET_AES): New feature flag for aes.
5843 * config/aarch64/aarch64-simd.md:
5844 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
5845 conditional on TARGET_AES.
5846 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
5847 (aarch64_crypto_sha1hsi): Make pattern conditional
5849 (aarch64_crypto_sha1hv4si): Ditto.
5850 (aarch64_be_crypto_sha1hv4si): Ditto.
5851 (aarch64_crypto_sha1su1v4si): Ditto.
5852 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
5853 (aarch64_crypto_sha1su0v4si): Ditto.
5854 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
5855 (aarch64_crypto_sha256su0v4si): Ditto.
5856 (aarch64_crypto_sha256su1v4si): Ditto.
5857 (doc/invoke.texi): Document new aes and sha2 options.
5859 2018-01-10 Martin Sebor <msebor@redhat.com>
5861 PR tree-optimization/83781
5862 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
5865 2018-01-11 Martin Sebor <msebor@gmail.com>
5866 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5868 PR tree-optimization/83501
5869 PR tree-optimization/81703
5871 * tree-ssa-strlen.c (get_string_cst): Rename...
5872 (get_string_len): ...to this. Handle global constants.
5873 (handle_char_store): Adjust.
5875 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
5876 Jim Wilson <jimw@sifive.com>
5878 * config/riscv/riscv-protos.h (riscv_output_return): New.
5879 * config/riscv/riscv.c (struct machine_function): New naked_p field.
5880 (riscv_attribute_table, riscv_output_return),
5881 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
5882 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
5883 (riscv_compute_frame_info): Only compute frame->mask if not a naked
5885 (riscv_expand_prologue): Add early return for naked function.
5886 (riscv_expand_epilogue): Likewise.
5887 (riscv_function_ok_for_sibcall): Return false for naked function.
5888 (riscv_set_current_function): New.
5889 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
5890 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
5891 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
5892 * doc/extend.texi (RISC-V Function Attributes): New.
5894 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
5896 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
5897 check for 128-bit long double before checking TCmode.
5898 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
5899 128-bit long doubles before checking TFmode or TCmode.
5900 (FLOAT128_IBM_P): Likewise.
5902 2018-01-10 Martin Sebor <msebor@redhat.com>
5904 PR tree-optimization/83671
5905 * builtins.c (c_strlen): Unconditionally return zero for the empty
5907 Use -Warray-bounds for warnings.
5908 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
5909 for non-constant array indices with COMPONENT_REF, arrays of
5910 arrays, and pointers to arrays.
5911 (gimple_fold_builtin_strlen): Determine and set length range for
5912 non-constant character arrays.
5914 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
5917 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
5920 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
5922 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
5924 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5927 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
5928 VECTOR_MEM_ALTIVEC_OR_VSX_P.
5929 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
5930 indexed_or_indirect_operand predicate.
5931 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
5932 (*vsx_le_perm_load_v8hi): Likewise.
5933 (*vsx_le_perm_load_v16qi): Likewise.
5934 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
5935 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
5936 (*vsx_le_perm_store_v8hi): Likewise.
5937 (*vsx_le_perm_store_v16qi): Likewise.
5938 (eight unnamed splitters): Likewise.
5940 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5942 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
5943 * config/rs6000/emmintrin.h: Likewise.
5944 * config/rs6000/mmintrin.h: Likewise.
5945 * config/rs6000/xmmintrin.h: Likewise.
5947 2018-01-10 David Malcolm <dmalcolm@redhat.com>
5950 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
5952 * tree.c (tree_nop_conversion): Return true for location wrapper
5954 (maybe_wrap_with_location): New function.
5955 (selftest::check_strip_nops): New function.
5956 (selftest::test_location_wrappers): New function.
5957 (selftest::tree_c_tests): Call it.
5958 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
5959 (maybe_wrap_with_location): New decl.
5960 (EXPR_LOCATION_WRAPPER_P): New macro.
5961 (location_wrapper_p): New inline function.
5962 (tree_strip_any_location_wrapper): New inline function.
5964 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
5967 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
5968 stack_realign_offset for the largest alignment of stack slot
5970 (ix86_find_max_used_stack_alignment): New function.
5971 (ix86_finalize_stack_frame_flags): Use it. Set
5972 max_used_stack_alignment if we don't realign stack.
5973 * config/i386/i386.h (machine_function): Add
5974 max_used_stack_alignment.
5976 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
5978 * config/arm/arm.opt (-mbranch-cost): New option.
5979 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
5982 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
5985 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
5986 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
5988 2018-01-10 Richard Biener <rguenther@suse.de>
5991 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
5992 early out so it also covers the case where we have a non-NULL
5995 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
5997 PR tree-optimization/83753
5998 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
5999 for non-strided grouped accesses if the number of elements is 1.
6001 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6004 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
6005 * i386.h (TARGET_USE_GATHER): Define.
6006 * x86-tune.def (X86_TUNE_USE_GATHER): New.
6008 2018-01-10 Martin Liska <mliska@suse.cz>
6011 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
6012 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
6014 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
6015 CLEANUP_NO_PARTITIONING is not set.
6017 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6019 * doc/rtl.texi: Remove documentation of (const ...) wrappers
6020 for vectors, as a partial revert of r254296.
6021 * rtl.h (const_vec_p): Delete.
6022 (const_vec_duplicate_p): Don't test for vector CONSTs.
6023 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
6024 * expmed.c (make_tree): Likewise.
6027 * common.md (E, F): Use CONSTANT_P instead of checking for
6029 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
6030 checking for CONST_VECTOR.
6032 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6035 * predict.c (force_edge_cold): Handle in more sane way edges
6038 2018-01-09 Carl Love <cel@us.ibm.com>
6040 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
6042 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
6043 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
6044 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
6045 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
6046 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
6047 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
6048 * config/rs6000/rs6000-protos.h: Add extern defition for
6049 rs6000_generate_float2_double_code.
6050 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
6052 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
6053 (float2_v2df): Add define_expand.
6055 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
6058 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
6059 op_mode in the force_to_mode call.
6061 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6063 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
6064 instead of checking each element individually.
6065 (aarch64_evpc_uzp): Likewise.
6066 (aarch64_evpc_zip): Likewise.
6067 (aarch64_evpc_ext): Likewise.
6068 (aarch64_evpc_rev): Likewise.
6069 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
6070 instead of checking each element individually. Return true without
6072 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
6073 whether all selected elements come from the same input, instead of
6074 checking each element individually. Remove calls to gen_rtx_REG,
6075 start_sequence and end_sequence and instead assert that no rtl is
6078 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6080 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
6081 order of HIGH and CONST checks.
6083 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6085 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
6086 if the destination isn't an SSA_NAME.
6088 2018-01-09 Richard Biener <rguenther@suse.de>
6090 PR tree-optimization/83668
6091 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
6093 (canonicalize_loop_form): ... here, renamed from ...
6094 (canonicalize_loop_closed_ssa_form): ... this and amended to
6095 swap successor edges for loop exit blocks to make us use
6096 the RPO order we need for initial schedule generation.
6098 2018-01-09 Joseph Myers <joseph@codesourcery.com>
6100 PR tree-optimization/64811
6101 * match.pd: When optimizing comparisons with Inf, avoid
6102 introducing or losing exceptions from comparisons with NaN.
6104 2018-01-09 Martin Liska <mliska@suse.cz>
6107 * asan.c (shadow_mem_size): Add gcc_assert.
6109 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
6111 Don't save registers in main().
6114 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
6115 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
6116 * config/avr/avr.c (avr_set_current_function): Don't error if
6117 naked, OS_task or OS_main are specified at the same time.
6118 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
6120 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
6122 * common/config/avr/avr-common.c (avr_option_optimization_table):
6123 Switch on -mmain-is-OS_task for optimizing compilations.
6125 2018-01-09 Richard Biener <rguenther@suse.de>
6127 PR tree-optimization/83572
6128 * graphite.c: Include cfganal.h.
6129 (graphite_transform_loops): Connect infinite loops to exit
6130 and remove fake edges at the end.
6132 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6134 * ipa-inline.c (edge_badness): Revert accidental checkin.
6136 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6139 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
6140 symbols; not inline clones.
6142 2018-01-09 Jakub Jelinek <jakub@redhat.com>
6145 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
6146 hard registers. Formatting fixes.
6148 PR preprocessor/83722
6149 * gcc.c (try_generate_repro): Pass
6150 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
6151 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
6154 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
6155 Kito Cheng <kito.cheng@gmail.com>
6157 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
6158 (riscv_leaf_function_p): Delete.
6159 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
6161 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6163 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
6165 (do_ifelse): New function.
6166 (do_isel): New function.
6167 (do_sub3): New function.
6168 (do_add3): New function.
6169 (do_load_mask_compare): New function.
6170 (do_overlap_load_compare): New function.
6171 (expand_compare_loop): New function.
6172 (expand_block_compare): Call expand_compare_loop() when appropriate.
6173 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
6175 (-mblock-compare-inline-loop-limit): New option.
6177 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6180 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
6181 Reverse order of second and third operands in first alternative.
6182 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
6183 of first and second elements in UNSPEC_VPERMR vector.
6184 (altivec_expand_vec_perm_le): Likewise.
6186 2018-01-08 Jeff Law <law@redhat.com>
6188 PR rtl-optimizatin/81308
6189 * tree-switch-conversion.c (cfg_altered): New file scoped static.
6190 (process_switch): If group_case_labels makes a change, then set
6192 (pass_convert_switch::execute): If a switch is converted, then
6193 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
6195 PR rtl-optimization/81308
6196 * recog.c (split_all_insns): Conditionally cleanup the CFG after
6199 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
6201 PR target/83663 - Revert r255946
6202 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
6203 generation for cases where splatting a value is not useful.
6204 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
6205 across a vec_duplicate and a paradoxical subreg forming a vector
6206 mode to a vec_concat.
6208 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6210 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
6211 -march=armv8.3-a variants.
6212 * config/arm/t-multilib: Likewise.
6213 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
6215 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6217 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
6219 (cceq_ior_compare_complement): Give it a name so I can use it, and
6220 change boolean_or_operator predicate to boolean_operator so it can
6221 be used to generate a crand.
6222 (eqne): New code iterator.
6223 (bd/bd_neg): New code_attrs.
6224 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
6225 a single define_insn.
6226 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
6227 decrement (bdnzt/bdnzf/bdzt/bdzf).
6228 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
6229 with the new names of the branch decrement patterns, and added the
6230 names of the branch decrement conditional patterns.
6232 2018-01-08 Richard Biener <rguenther@suse.de>
6234 PR tree-optimization/83563
6235 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
6238 2018-01-08 Richard Biener <rguenther@suse.de>
6241 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
6243 2018-01-08 Richard Biener <rguenther@suse.de>
6245 PR tree-optimization/83685
6246 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
6247 references to abnormals.
6249 2018-01-08 Richard Biener <rguenther@suse.de>
6252 * dwarf2out.c (output_indirect_strings): Handle empty
6253 skeleton_debug_str_hash.
6254 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
6256 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6258 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
6259 (emit_store_direct): Likewise.
6260 (arc_trampoline_adjust_address): Likewise.
6261 (arc_asm_trampoline_template): New function.
6262 (arc_initialize_trampoline): Use asm_trampoline_template.
6263 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
6264 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
6265 * config/arc/arc.md (flush_icache): Delete pattern.
6267 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6269 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
6270 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
6273 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6276 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
6277 by not USED_FOR_TARGET.
6278 (make_pass_resolve_sw_modes): Likewise.
6280 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6282 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
6285 2018-01-08 Richard Biener <rguenther@suse.de>
6288 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
6290 2018-01-08 Richard Biener <rguenther@suse.de>
6293 * match.pd ((t * 2) / 2) -> t): Add missing :c.
6295 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
6298 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
6299 basic blocks with a small number of successors.
6300 (convert_control_dep_chain_into_preds): Improve handling of
6302 (dump_predicates): Split apart into...
6303 (dump_pred_chain): ...here...
6304 (dump_pred_info): ...and here.
6305 (can_one_predicate_be_invalidated_p): Add debugging printfs.
6306 (can_chain_union_be_invalidated_p): Improve check for invalidation
6308 (uninit_uses_cannot_happen): Avoid unnecessary if
6309 convert_control_dep_chain_into_preds yielded nothing.
6311 2018-01-06 Martin Sebor <msebor@redhat.com>
6313 PR tree-optimization/83640
6314 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
6315 subtracting negative offset from size.
6316 (builtin_access::overlap): Adjust offset bounds of the access to fall
6317 within the size of the object if possible.
6319 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
6321 PR rtl-optimization/83699
6322 * expmed.c (extract_bit_field_1): Restrict the vector usage of
6323 extract_bit_field_as_subreg to cases in which the extracted
6324 value is also a vector.
6326 * lra-constraints.c (process_alt_operands): Test for the equivalence
6327 substitutions when detecting a possible reload cycle.
6329 2018-01-06 Jakub Jelinek <jakub@redhat.com>
6332 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
6333 by default if flag_selective_schedling{,2}. Formatting fixes.
6335 PR rtl-optimization/83682
6336 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
6337 if it has non-VECTOR_MODE element mode.
6338 (vec_duplicate_p): Likewise.
6341 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
6342 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
6344 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6347 * config/i386/i386-builtin.def
6348 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
6349 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
6350 Require also OPTION_MASK_ISA_AVX512F in addition to
6351 OPTION_MASK_ISA_GFNI.
6352 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
6353 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
6354 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
6355 to OPTION_MASK_ISA_GFNI.
6356 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
6357 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
6358 OPTION_MASK_ISA_AVX512BW.
6359 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
6360 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
6361 addition to OPTION_MASK_ISA_GFNI.
6362 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
6363 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
6364 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
6365 to OPTION_MASK_ISA_GFNI.
6366 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
6367 a requirement for all ISAs rather than any of them with a few
6369 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
6371 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
6372 bitmasks to be enabled with 3 exceptions, instead of requiring any
6373 enabled ISA with lots of exceptions.
6374 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
6375 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
6376 Change avx512bw in isa attribute to avx512f.
6377 * config/i386/sgxintrin.h: Add license boilerplate.
6378 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
6379 to __AVX512F__ and __AVX512VL to __AVX512VL__.
6380 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
6381 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
6383 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
6384 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
6385 temporarily sse2 rather than sse if not enabled already.
6388 * config/i386/sse.md (VI248_VLBW): Rename to ...
6389 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
6390 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
6391 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
6392 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
6393 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
6394 mode iterator instead of VI248_VLBW.
6396 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
6398 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
6399 (record_modified): Skip clobbers; add debug output.
6400 (param_change_prob): Use sreal frequencies.
6402 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6404 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
6405 punt for user-aligned variables.
6407 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6409 * tree-chrec.c (chrec_contains_symbols): Return true for
6412 2018-01-05 Sudakshina Das <sudi.das@arm.com>
6415 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
6416 of (x|y) == x for BICS pattern.
6418 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6420 PR tree-optimization/83605
6421 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
6422 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
6425 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
6427 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
6428 * config/epiphany/rtems.h: New file.
6430 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6431 Uros Bizjak <ubizjak@gmail.com>
6434 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
6435 QIreg_operand instead of register_operand predicate.
6436 * config/i386/i386.c (ix86_rop_should_change_byte_p,
6437 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
6438 comments instead of -fmitigate[-_]rop.
6440 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6443 * cgraphunit.c (symbol_table::compile): Switch to text_section
6444 before calling assembly_start debug hook.
6445 * run-rtl-passes.c (run_rtl_passes): Likewise.
6448 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6450 * tree-vrp.c (extract_range_from_binary_expr_1): Check
6451 range_int_cst_p rather than !symbolic_range_p before calling
6452 extract_range_from_multiplicative_op_1.
6454 2018-01-04 Jeff Law <law@redhat.com>
6456 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
6457 redundant test in assertion.
6459 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6461 * doc/rtl.texi: Document machine_mode wrapper classes.
6463 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6465 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
6468 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6470 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
6471 the VEC_PERM_EXPR fold to fail.
6473 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6476 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
6477 to switched_sections.
6479 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6482 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
6485 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
6488 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
6489 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
6491 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6494 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
6495 is BLKmode and bitpos not zero or mode change is needed.
6497 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6500 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
6503 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
6506 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
6507 instead of MULT rtx. Update all corresponding splitters.
6509 (*ssub<modesuffix>): Ditto.
6511 (*cmp_sadd_di): Update split patterns.
6512 (*cmp_sadd_si): Ditto.
6513 (*cmp_sadd_sidi): Ditto.
6514 (*cmp_ssub_di): Ditto.
6515 (*cmp_ssub_si): Ditto.
6516 (*cmp_ssub_sidi): Ditto.
6517 * config/alpha/predicates.md (const23_operand): New predicate.
6518 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
6519 Look for ASHIFT, not MULT inner operand.
6520 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
6522 2018-01-04 Martin Liska <mliska@suse.cz>
6524 PR gcov-profile/83669
6525 * gcov.c (output_intermediate_file): Add version to intermediate
6527 * doc/gcov.texi: Document new field 'version' in intermediate
6528 file format. Fix location of '-k' option of gcov command.
6530 2018-01-04 Martin Liska <mliska@suse.cz>
6533 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
6535 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6537 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
6539 2018-01-03 Martin Sebor <msebor@redhat.com>
6541 PR tree-optimization/83655
6542 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
6543 checking calls with invalid arguments.
6545 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6547 * tree-vect-stmts.c (vect_get_store_rhs): New function.
6548 (vectorizable_mask_load_store): Delete.
6549 (vectorizable_call): Return false for masked loads and stores.
6550 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
6551 instead of gimple_assign_rhs1.
6552 (vectorizable_load): Handle IFN_MASK_LOAD.
6553 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
6555 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6557 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
6559 (vectorizable_mask_load_store): ...here.
6560 (vectorizable_load): ...and here.
6562 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6564 * tree-vect-stmts.c (vect_build_all_ones_mask)
6565 (vect_build_zero_merge_argument): New functions, split out from...
6566 (vectorizable_load): ...here.
6568 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6570 * tree-vect-stmts.c (vect_check_store_rhs): New function,
6572 (vectorizable_mask_load_store): ...here.
6573 (vectorizable_store): ...and here.
6575 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6577 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
6579 (vectorizable_mask_load_store): ...here.
6581 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6583 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
6584 (vect_model_store_cost): Take a vec_load_store_type instead of a
6586 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
6587 (vect_model_store_cost): Take a vec_load_store_type instead of a
6589 (vectorizable_mask_load_store): Update accordingly.
6590 (vectorizable_store): Likewise.
6591 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
6593 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6595 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
6596 IFN_MASK_LOAD calls here rather than...
6597 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
6599 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6600 Alan Hayward <alan.hayward@arm.com>
6601 David Sherwood <david.sherwood@arm.com>
6603 * expmed.c (extract_bit_field_1): For vector extracts,
6604 fall back to extract_bit_field_as_subreg if vec_extract
6607 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6608 Alan Hayward <alan.hayward@arm.com>
6609 David Sherwood <david.sherwood@arm.com>
6611 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
6612 they are variable or constant sized.
6613 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
6614 slots for constant-sized data.
6616 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6617 Alan Hayward <alan.hayward@arm.com>
6618 David Sherwood <david.sherwood@arm.com>
6620 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
6621 handling COND_EXPRs with boolean comparisons, try to find a better
6622 basis for the mask type than the boolean itself.
6624 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6626 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
6627 is calculated and how it can be overridden.
6628 * genmodes.c (max_bitsize_mode_any_mode): New variable.
6629 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
6631 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
6634 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6635 Alan Hayward <alan.hayward@arm.com>
6636 David Sherwood <david.sherwood@arm.com>
6638 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
6639 Remove the mode argument.
6640 (aarch64_simd_valid_immediate): Remove the mode and inverse
6642 * config/aarch64/iterators.md (bitsize): New iterator.
6643 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
6644 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
6645 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
6646 aarch64_simd_valid_immediate.
6647 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
6648 (aarch64_reg_or_bic_imm): Likewise.
6649 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
6650 with an insn_type enum and msl with a modifier_type enum.
6651 Replace element_width with a scalar_mode. Change the shift
6652 to unsigned int. Add constructors for scalar_float_mode and
6653 scalar_int_mode elements.
6654 (aarch64_vect_float_const_representable_p): Delete.
6655 (aarch64_can_const_movi_rtx_p)
6656 (aarch64_simd_scalar_immediate_valid_for_move)
6657 (aarch64_simd_make_constant): Update call to
6658 aarch64_simd_valid_immediate.
6659 (aarch64_advsimd_valid_immediate_hs): New function.
6660 (aarch64_advsimd_valid_immediate): Likewise.
6661 (aarch64_simd_valid_immediate): Remove mode and inverse
6662 arguments. Rewrite to use the above. Use const_vec_duplicate_p
6663 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
6664 and aarch64_float_const_representable_p on the result.
6665 (aarch64_output_simd_mov_immediate): Remove mode argument.
6666 Update call to aarch64_simd_valid_immediate and use of
6667 simd_immediate_info.
6668 (aarch64_output_scalar_simd_mov_immediate): Update call
6671 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6672 Alan Hayward <alan.hayward@arm.com>
6673 David Sherwood <david.sherwood@arm.com>
6675 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
6676 (mode_nunits): Likewise CONST_MODE_NUNITS.
6677 * machmode.def (ADJUST_NUNITS): Document.
6678 * genmodes.c (mode_data::need_nunits_adj): New field.
6679 (blank_mode): Update accordingly.
6680 (adj_nunits): New variable.
6681 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
6683 (emit_mode_size_inline): Set need_bytesize_adj for all modes
6684 listed in adj_nunits.
6685 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
6686 listed in adj_nunits. Don't emit case statements for such modes.
6687 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
6688 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
6689 nothing if adj_nunits is nonnull.
6690 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
6691 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
6692 (emit_mode_fbit): Update use of print_maybe_const_decl.
6693 (emit_move_size): Likewise. Treat the array as non-const
6695 (emit_mode_adjustments): Handle adj_nunits.
6697 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6699 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
6700 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
6701 (VECTOR_MODES): Use it.
6702 (make_vector_modes): Take the prefix as an argument.
6704 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6705 Alan Hayward <alan.hayward@arm.com>
6706 David Sherwood <david.sherwood@arm.com>
6708 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
6709 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
6710 for MODE_VECTOR_BOOL.
6711 * machmode.def (VECTOR_BOOL_MODE): Document.
6712 * genmodes.c (VECTOR_BOOL_MODE): New macro.
6713 (make_vector_bool_mode): New function.
6714 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
6716 * lto-streamer-in.c (lto_input_mode_table): Likewise.
6717 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
6719 * stor-layout.c (int_mode_for_mode): Likewise.
6720 * tree.c (build_vector_type_for_mode): Likewise.
6721 * varasm.c (output_constant_pool_2): Likewise.
6722 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
6723 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
6724 for MODE_VECTOR_BOOL.
6725 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
6726 of mode class checks.
6727 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
6728 instead of a list of mode class checks.
6729 (expand_vector_scalar_condition): Likewise.
6730 (type_for_widest_vector_mode): Handle BImode as an inner mode.
6732 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6733 Alan Hayward <alan.hayward@arm.com>
6734 David Sherwood <david.sherwood@arm.com>
6736 * machmode.h (mode_size): Change from unsigned short to
6738 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
6739 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6740 or if measurement_type is not polynomial.
6741 (fixed_size_mode::includes_p): Check for constant-sized modes.
6742 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
6743 return a poly_uint16 rather than an unsigned short.
6744 (emit_mode_size): Change the type of mode_size from unsigned short
6745 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
6746 (emit_mode_adjustments): Cope with polynomial vector sizes.
6747 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6749 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6751 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
6752 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
6753 * caller-save.c (setup_save_areas): Likewise.
6754 (replace_reg_with_saved_mem): Likewise.
6755 * calls.c (emit_library_call_value_1): Likewise.
6756 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
6757 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
6758 (gen_lowpart_for_combine): Likewise.
6759 * convert.c (convert_to_integer_1): Likewise.
6760 * cse.c (equiv_constant, cse_insn): Likewise.
6761 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
6762 (cselib_subst_to_values): Likewise.
6763 * dce.c (word_dce_process_block): Likewise.
6764 * df-problems.c (df_word_lr_mark_ref): Likewise.
6765 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
6766 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
6767 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
6768 (rtl_for_decl_location): Likewise.
6769 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
6770 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
6771 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
6772 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
6773 (expand_expr_real_1): Likewise.
6774 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
6775 (pad_below): Likewise.
6776 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6777 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
6778 * ira.c (get_subreg_tracking_sizes): Likewise.
6779 * ira-build.c (ira_create_allocno_objects): Likewise.
6780 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
6781 (ira_sort_regnos_for_alter_reg): Likewise.
6782 * ira-costs.c (record_operand_costs): Likewise.
6783 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
6784 (resolve_simple_move): Likewise.
6785 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
6786 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
6787 (lra_constraints): Likewise.
6788 (CONST_POOL_OK_P): Reject variable-sized modes.
6789 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
6790 (add_pseudo_to_slot, lra_spill): Likewise.
6791 * omp-low.c (omp_clause_aligned_alignment): Likewise.
6792 * optabs-query.c (get_best_extraction_insn): Likewise.
6793 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6794 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
6795 (expand_mult_highpart, valid_multiword_target_p): Likewise.
6796 * recog.c (offsettable_address_addr_space_p): Likewise.
6797 * regcprop.c (maybe_mode_change): Likewise.
6798 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
6799 * regrename.c (build_def_use): Likewise.
6800 * regstat.c (dump_reg_info): Likewise.
6801 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
6802 (find_reloads, find_reloads_subreg_address): Likewise.
6803 * reload1.c (eliminate_regs_1): Likewise.
6804 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
6805 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
6806 (simplify_binary_operation_1, simplify_subreg): Likewise.
6807 * targhooks.c (default_function_arg_padding): Likewise.
6808 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
6809 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
6810 (verify_gimple_assign_ternary): Likewise.
6811 * tree-inline.c (estimate_move_cost): Likewise.
6812 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6813 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
6814 (get_address_cost_ainc): Likewise.
6815 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
6816 (vect_supportable_dr_alignment): Likewise.
6817 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6818 (vectorizable_reduction): Likewise.
6819 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
6820 (vectorizable_operation, vectorizable_load): Likewise.
6821 * tree.c (build_same_sized_truth_vector_type): Likewise.
6822 * valtrack.c (cleanup_auto_inc_dec): Likewise.
6823 * var-tracking.c (emit_note_insn_var_location): Likewise.
6824 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
6825 (ADDR_VEC_ALIGN): Likewise.
6827 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6828 Alan Hayward <alan.hayward@arm.com>
6829 David Sherwood <david.sherwood@arm.com>
6831 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
6833 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6834 or if measurement_type is polynomial.
6835 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
6836 * combine.c (make_extraction): Likewise.
6837 * dse.c (find_shift_sequence): Likewise.
6838 * dwarf2out.c (mem_loc_descriptor): Likewise.
6839 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
6840 (extract_bit_field, extract_low_bits): Likewise.
6841 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
6842 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
6843 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
6844 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
6845 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6846 * reload.c (find_reloads): Likewise.
6847 * reload1.c (alter_reg): Likewise.
6848 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
6849 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
6850 * tree-if-conv.c (predicate_mem_writes): Likewise.
6851 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
6852 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
6853 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
6854 * valtrack.c (dead_debug_insert_temp): Likewise.
6855 * varasm.c (mergeable_constant_section): Likewise.
6856 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
6858 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6859 Alan Hayward <alan.hayward@arm.com>
6860 David Sherwood <david.sherwood@arm.com>
6862 * expr.c (expand_assignment): Cope with polynomial mode sizes
6863 when assigning to a CONCAT.
6865 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6866 Alan Hayward <alan.hayward@arm.com>
6867 David Sherwood <david.sherwood@arm.com>
6869 * machmode.h (mode_precision): Change from unsigned short to
6871 (mode_to_precision): Return a poly_uint16 rather than an unsigned
6873 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
6874 or if measurement_type is not polynomial.
6875 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
6876 in which the mode is already known to be a scalar_int_mode.
6877 * genmodes.c (emit_mode_precision): Change the type of mode_precision
6878 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
6880 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6881 for GET_MODE_PRECISION.
6882 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6883 for GET_MODE_PRECISION.
6884 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
6886 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
6887 (expand_field_assignment, make_extraction): Likewise.
6888 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
6889 (get_last_value): Likewise.
6890 * convert.c (convert_to_integer_1): Likewise.
6891 * cse.c (cse_insn): Likewise.
6892 * expr.c (expand_expr_real_1): Likewise.
6893 * lra-constraints.c (simplify_operand_subreg): Likewise.
6894 * optabs-query.c (can_atomic_load_p): Likewise.
6895 * optabs.c (expand_atomic_load): Likewise.
6896 (expand_atomic_store): Likewise.
6897 * ree.c (combine_reaching_defs): Likewise.
6898 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
6899 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
6900 * tree.h (type_has_mode_precision_p): Likewise.
6901 * ubsan.c (instrument_si_overflow): Likewise.
6903 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6904 Alan Hayward <alan.hayward@arm.com>
6905 David Sherwood <david.sherwood@arm.com>
6907 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
6908 polynomial numbers of units.
6909 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
6910 (valid_vector_subparts_p): New function.
6911 (build_vector_type): Remove temporary shim and take the number
6912 of units as a poly_uint64 rather than an int.
6913 (build_opaque_vector_type): Take the number of units as a
6914 poly_uint64 rather than an int.
6915 * tree.c (build_vector_from_ctor): Handle polynomial
6916 TYPE_VECTOR_SUBPARTS.
6917 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
6918 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
6919 (build_vector_from_val): If the number of units is variable,
6920 use build_vec_duplicate_cst for constant operands and
6921 VEC_DUPLICATE_EXPR otherwise.
6922 (make_vector_type): Remove temporary is_constant ().
6923 (build_vector_type, build_opaque_vector_type): Take the number of
6924 units as a poly_uint64 rather than an int.
6925 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
6927 * cfgexpand.c (expand_debug_expr): Likewise.
6928 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
6929 (store_constructor, expand_expr_real_1): Likewise.
6930 (const_scalar_mask_from_tree): Likewise.
6931 * fold-const-call.c (fold_const_reduction): Likewise.
6932 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
6933 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
6934 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
6935 (fold_relational_const): Likewise.
6936 (native_interpret_vector): Likewise. Change the size from an
6937 int to an unsigned int.
6938 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
6939 TYPE_VECTOR_SUBPARTS.
6940 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
6941 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
6942 duplicating a non-constant operand into a variable-length vector.
6943 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
6944 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
6945 * ipa-icf.c (sem_variable::equals): Likewise.
6946 * match.pd: Likewise.
6947 * omp-simd-clone.c (simd_clone_subparts): Likewise.
6948 * print-tree.c (print_node): Likewise.
6949 * stor-layout.c (layout_type): Likewise.
6950 * targhooks.c (default_builtin_vectorization_cost): Likewise.
6951 * tree-cfg.c (verify_gimple_comparison): Likewise.
6952 (verify_gimple_assign_binary): Likewise.
6953 (verify_gimple_assign_ternary): Likewise.
6954 (verify_gimple_assign_single): Likewise.
6955 * tree-pretty-print.c (dump_generic_node): Likewise.
6956 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6957 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
6958 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
6959 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
6960 (vect_shift_permute_load_chain): Likewise.
6961 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
6962 (expand_vector_condition, optimize_vector_constructor): Likewise.
6963 (lower_vec_perm, get_compute_type): Likewise.
6964 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6965 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
6966 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
6967 (vect_recog_mask_conversion_pattern): Likewise.
6968 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
6969 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
6970 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
6971 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
6972 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
6973 (vectorizable_shift, vectorizable_operation, vectorizable_store)
6974 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
6975 (supportable_widening_operation): Likewise.
6976 (supportable_narrowing_operation): Likewise.
6977 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
6979 * varasm.c (output_constant): Likewise.
6981 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6982 Alan Hayward <alan.hayward@arm.com>
6983 David Sherwood <david.sherwood@arm.com>
6985 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
6986 so that both the length == 3 and length != 3 cases set up their
6987 own permute vectors. Add comments explaining why we know the
6988 number of elements is constant.
6989 (vect_permute_load_chain): Likewise.
6991 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6992 Alan Hayward <alan.hayward@arm.com>
6993 David Sherwood <david.sherwood@arm.com>
6995 * machmode.h (mode_nunits): Change from unsigned char to
6997 (ONLY_FIXED_SIZE_MODES): New macro.
6998 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
6999 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
7000 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
7002 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
7003 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
7004 or if measurement_type is not polynomial.
7005 * genmodes.c (ZERO_COEFFS): New macro.
7006 (emit_mode_nunits_inline): Make mode_nunits_inline return a
7008 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
7009 Use ZERO_COEFFS when emitting initializers.
7010 * data-streamer.h (bp_pack_poly_value): New function.
7011 (bp_unpack_poly_value): Likewise.
7012 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7013 for GET_MODE_NUNITS.
7014 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7015 for GET_MODE_NUNITS.
7016 * tree.c (make_vector_type): Remove temporary shim and make
7017 the real function take the number of units as a poly_uint64
7019 (build_vector_type_for_mode): Handle polynomial nunits.
7020 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
7021 * emit-rtl.c (const_vec_series_p_1): Likewise.
7022 (gen_rtx_CONST_VECTOR): Likewise.
7023 * fold-const.c (test_vec_duplicate_folding): Likewise.
7024 * genrecog.c (validate_pattern): Likewise.
7025 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
7026 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7027 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
7028 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
7029 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
7030 * rtlanal.c (subreg_get_info): Likewise.
7031 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7032 (vect_grouped_load_supported): Likewise.
7033 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
7034 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
7035 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
7036 (simplify_const_unary_operation, simplify_binary_operation_1)
7037 (simplify_const_binary_operation, simplify_ternary_operation)
7038 (test_vector_ops_duplicate, test_vector_ops): Likewise.
7039 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
7040 instead of CONST_VECTOR_NUNITS.
7041 * varasm.c (output_constant_pool_2): Likewise.
7042 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
7043 explicit-encoded elements in the XVEC for variable-length vectors.
7045 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7047 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
7049 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7050 Alan Hayward <alan.hayward@arm.com>
7051 David Sherwood <david.sherwood@arm.com>
7053 * coretypes.h (fixed_size_mode): Declare.
7054 (fixed_size_mode_pod): New typedef.
7055 * builtins.h (target_builtins::x_apply_args_mode)
7056 (target_builtins::x_apply_result_mode): Change type to
7057 fixed_size_mode_pod.
7058 * builtins.c (apply_args_size, apply_result_size, result_vector)
7059 (expand_builtin_apply_args_1, expand_builtin_apply)
7060 (expand_builtin_return): Update accordingly.
7062 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7064 * cse.c (hash_rtx_cb): Hash only the encoded elements.
7065 * cselib.c (cselib_hash_rtx): Likewise.
7066 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
7067 CONST_VECTOR encoding.
7069 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7070 Jeff Law <law@redhat.com>
7073 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
7074 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
7075 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
7076 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
7079 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
7080 explicitly probe *sp in a noreturn function if there were any callee
7081 register saves or frame pointer is needed.
7083 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7086 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
7087 BLKmode for ternary, binary or unary expressions.
7090 * var-tracking.c (delete_vta_debug_insn): New inline function.
7091 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
7092 insns from get_insns () to NULL instead of each bb separately.
7093 Use delete_vta_debug_insn. No longer static.
7094 (vt_debug_insns_local, variable_tracking_main_1): Adjust
7095 delete_vta_debug_insns callers.
7096 * rtl.h (delete_vta_debug_insns): Declare.
7097 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
7098 instead of variable_tracking_main.
7100 2018-01-03 Martin Sebor <msebor@redhat.com>
7102 PR tree-optimization/83603
7103 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
7104 arguments past the endof the argument list in functions declared
7105 without a prototype.
7106 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7107 Avoid checking when arguments are null.
7109 2018-01-03 Martin Sebor <msebor@redhat.com>
7112 * doc/extend.texi (attribute const): Fix a typo.
7113 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
7114 issuing -Wsuggest-attribute for void functions.
7116 2018-01-03 Martin Sebor <msebor@redhat.com>
7118 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
7119 offset_int::from instead of wide_int::to_shwi.
7120 (maybe_diag_overlap): Remove assertion.
7121 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
7122 * gimple-ssa-sprintf.c (format_directive): Same.
7123 (parse_directive): Same.
7124 (sprintf_dom_walker::compute_format_length): Same.
7125 (try_substitute_return_value): Same.
7127 2018-01-03 Jeff Law <law@redhat.com>
7130 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
7131 non-constant residual for zero at runtime and avoid probing in
7132 that case. Reorganize code for trailing problem to mirror handling
7135 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7137 PR tree-optimization/83501
7138 * tree-ssa-strlen.c (get_string_cst): New.
7139 (handle_char_store): Call get_string_cst.
7141 2018-01-03 Martin Liska <mliska@suse.cz>
7143 PR tree-optimization/83593
7144 * tree-ssa-strlen.c: Include tree-cfg.h.
7145 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
7146 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
7147 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
7149 (strlen_dom_walker::before_dom_children): Call
7150 gimple_purge_dead_eh_edges. Dump tranformation with details
7152 (strlen_dom_walker::before_dom_children): Update call by adding
7153 new argument cleanup_eh.
7154 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
7156 2018-01-03 Martin Liska <mliska@suse.cz>
7159 * cif-code.def (VARIADIC_THUNK): New enum value.
7160 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
7163 2018-01-03 Jan Beulich <jbeulich@suse.com>
7165 * sse.md (mov<mode>_internal): Tighten condition for when to use
7166 vmovdqu<ssescalarsize> for TI and OI modes.
7168 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7170 Update copyright years.
7172 2018-01-03 Martin Liska <mliska@suse.cz>
7175 * ipa-visibility.c (function_and_variable_visibility): Skip
7176 functions with noipa attribure.
7178 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7180 * gcc.c (process_command): Update copyright notice dates.
7181 * gcov-dump.c (print_version): Ditto.
7182 * gcov.c (print_version): Ditto.
7183 * gcov-tool.c (print_version): Ditto.
7184 * gengtype.c (create_file): Ditto.
7185 * doc/cpp.texi: Bump @copying's copyright year.
7186 * doc/cppinternals.texi: Ditto.
7187 * doc/gcc.texi: Ditto.
7188 * doc/gccint.texi: Ditto.
7189 * doc/gcov.texi: Ditto.
7190 * doc/install.texi: Ditto.
7191 * doc/invoke.texi: Ditto.
7193 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7195 * vector-builder.h (vector_builder::m_full_nelts): Change from
7196 unsigned int to poly_uint64.
7197 (vector_builder::full_nelts): Update prototype accordingly.
7198 (vector_builder::new_vector): Likewise.
7199 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
7200 (vector_builder::operator ==): Likewise.
7201 (vector_builder::finalize): Likewise.
7202 * int-vector-builder.h (int_vector_builder::int_vector_builder):
7203 Take the number of elements as a poly_uint64 rather than an
7205 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
7206 from unsigned int to poly_uint64.
7207 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
7208 (vec_perm_indices::new_vector): Likewise.
7209 (vec_perm_indices::length): Likewise.
7210 (vec_perm_indices::nelts_per_input): Likewise.
7211 (vec_perm_indices::input_nelts): Likewise.
7212 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
7213 number of elements per input as a poly_uint64 rather than an
7214 unsigned int. Use the original encoding for variable-length
7215 vectors, rather than clamping each individual element.
7216 For the second and subsequent elements in each pattern,
7217 clamp the step and base before clamping their sum.
7218 (vec_perm_indices::series_p): Handle polynomial element counts.
7219 (vec_perm_indices::all_in_range_p): Likewise.
7220 (vec_perm_indices_to_tree): Likewise.
7221 (vec_perm_indices_to_rtx): Likewise.
7222 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
7223 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
7224 (tree_vector_builder::new_binary_operation): Handle polynomial
7225 element counts. Return false if we need to know the number
7226 of elements at compile time.
7227 * fold-const.c (fold_vec_perm): Punt if the number of elements
7228 isn't known at compile time.
7230 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7232 * vec-perm-indices.h (vec_perm_builder): Change element type
7233 from HOST_WIDE_INT to poly_int64.
7234 (vec_perm_indices::element_type): Update accordingly.
7235 (vec_perm_indices::clamp): Handle polynomial element_types.
7236 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7237 (vec_perm_indices::all_in_range_p): Likewise.
7238 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
7240 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
7241 polynomial vec_perm_indices element types.
7242 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
7243 * fold-const.c (fold_vec_perm): Likewise.
7244 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
7245 * tree-vect-generic.c (lower_vec_perm): Likewise.
7246 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7247 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
7248 element type to HOST_WIDE_INT.
7250 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7251 Alan Hayward <alan.hayward@arm.com>
7252 David Sherwood <david.sherwood@arm.com>
7254 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
7255 rather than an int. Use plus_constant.
7256 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
7257 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
7259 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7260 Alan Hayward <alan.hayward@arm.com>
7261 David Sherwood <david.sherwood@arm.com>
7263 * calls.c (emit_call_1, expand_call): Change struct_value_size from
7264 a HOST_WIDE_INT to a poly_int64.
7266 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7267 Alan Hayward <alan.hayward@arm.com>
7268 David Sherwood <david.sherwood@arm.com>
7270 * calls.c (load_register_parameters): Cope with polynomial
7271 mode sizes. Require a constant size for BLKmode parameters
7272 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
7273 forces a parameter to be padded at the lsb end in order to
7274 fill a complete number of words, require the parameter size
7275 to be ordered wrt UNITS_PER_WORD.
7277 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7278 Alan Hayward <alan.hayward@arm.com>
7279 David Sherwood <david.sherwood@arm.com>
7281 * reload1.c (spill_stack_slot_width): Change element type
7282 from unsigned int to poly_uint64_pod.
7283 (alter_reg): Treat mode sizes as polynomial.
7285 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7286 Alan Hayward <alan.hayward@arm.com>
7287 David Sherwood <david.sherwood@arm.com>
7289 * reload.c (complex_word_subreg_p): New function.
7290 (reload_inner_reg_of_subreg, push_reload): Use it.
7292 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7293 Alan Hayward <alan.hayward@arm.com>
7294 David Sherwood <david.sherwood@arm.com>
7296 * lra-constraints.c (process_alt_operands): Reject matched
7297 operands whose sizes aren't ordered.
7298 (match_reload): Refer to this check here.
7300 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7301 Alan Hayward <alan.hayward@arm.com>
7302 David Sherwood <david.sherwood@arm.com>
7304 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
7305 that the mode size is in the set {1, 2, 4, 8, 16}.
7307 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7308 Alan Hayward <alan.hayward@arm.com>
7309 David Sherwood <david.sherwood@arm.com>
7311 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
7312 Use plus_constant instead of gen_rtx_PLUS.
7314 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7315 Alan Hayward <alan.hayward@arm.com>
7316 David Sherwood <david.sherwood@arm.com>
7318 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
7319 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
7320 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
7321 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
7322 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
7323 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
7324 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
7325 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
7326 * config/i386/i386.c (ix86_push_rounding): ...this new function.
7327 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
7329 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
7330 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
7331 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
7332 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
7333 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
7334 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
7335 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
7336 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
7337 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
7338 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
7340 * expr.c (emit_move_resolve_push): Treat the input and result
7341 of PUSH_ROUNDING as a poly_int64.
7342 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
7343 (emit_push_insn): Likewise.
7344 * lra-eliminations.c (mark_not_eliminable): Likewise.
7345 * recog.c (push_operand): Likewise.
7346 * reload1.c (elimination_effects): Likewise.
7347 * rtlanal.c (nonzero_bits1): Likewise.
7348 * calls.c (store_one_arg): Likewise. Require the padding to be
7349 known at compile time.
7351 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7352 Alan Hayward <alan.hayward@arm.com>
7353 David Sherwood <david.sherwood@arm.com>
7355 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
7356 Use plus_constant instead of gen_rtx_PLUS.
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 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
7365 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7366 Alan Hayward <alan.hayward@arm.com>
7367 David Sherwood <david.sherwood@arm.com>
7369 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
7370 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
7371 via stack temporaries. Treat the mode size as polynomial too.
7373 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7374 Alan Hayward <alan.hayward@arm.com>
7375 David Sherwood <david.sherwood@arm.com>
7377 * expr.c (expand_expr_real_2): When handling conversions involving
7378 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
7379 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
7380 as a poly_uint64 too.
7382 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7383 Alan Hayward <alan.hayward@arm.com>
7384 David Sherwood <david.sherwood@arm.com>
7386 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
7388 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7389 Alan Hayward <alan.hayward@arm.com>
7390 David Sherwood <david.sherwood@arm.com>
7392 * combine.c (can_change_dest_mode): Handle polynomial
7393 REGMODE_NATURAL_SIZE.
7394 * expmed.c (store_bit_field_1): Likewise.
7395 * expr.c (store_constructor): Likewise.
7396 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
7397 and polynomial REGMODE_NATURAL_SIZE.
7398 (gen_lowpart_common): Likewise.
7399 * reginfo.c (record_subregs_of_mode): Likewise.
7400 * rtlanal.c (read_modify_subreg_p): Likewise.
7402 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7403 Alan Hayward <alan.hayward@arm.com>
7404 David Sherwood <david.sherwood@arm.com>
7406 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
7407 numbers of elements.
7409 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7410 Alan Hayward <alan.hayward@arm.com>
7411 David Sherwood <david.sherwood@arm.com>
7413 * match.pd: Cope with polynomial numbers of vector elements.
7415 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7416 Alan Hayward <alan.hayward@arm.com>
7417 David Sherwood <david.sherwood@arm.com>
7419 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
7420 in a POINTER_PLUS_EXPR.
7422 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7423 Alan Hayward <alan.hayward@arm.com>
7424 David Sherwood <david.sherwood@arm.com>
7426 * omp-simd-clone.c (simd_clone_subparts): New function.
7427 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
7428 (ipa_simd_modify_function_body): Likewise.
7430 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7431 Alan Hayward <alan.hayward@arm.com>
7432 David Sherwood <david.sherwood@arm.com>
7434 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
7435 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
7436 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
7437 (expand_vector_condition, vector_element): Likewise.
7438 (subparts_gt): New function.
7439 (get_compute_type): Use subparts_gt.
7440 (count_type_subparts): Delete.
7441 (expand_vector_operations_1): Use subparts_gt instead of
7442 count_type_subparts.
7444 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7445 Alan Hayward <alan.hayward@arm.com>
7446 David Sherwood <david.sherwood@arm.com>
7448 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
7449 (vect_compile_time_alias): ...this new function. Do the calculation
7450 on poly_ints rather than trees.
7451 (vect_prune_runtime_alias_test_list): Update call accordingly.
7453 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7454 Alan Hayward <alan.hayward@arm.com>
7455 David Sherwood <david.sherwood@arm.com>
7457 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
7459 (vect_schedule_slp_instance): Likewise.
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-slp.c (vect_get_and_check_slp_defs): Reject
7466 constant and extern definitions for variable-length vectors.
7467 (vect_get_constant_vectors): Note that the number of units
7468 is known to be constant.
7470 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7471 Alan Hayward <alan.hayward@arm.com>
7472 David Sherwood <david.sherwood@arm.com>
7474 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
7475 of units as polynomial. Choose between WIDE and NARROW based
7478 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7479 Alan Hayward <alan.hayward@arm.com>
7480 David Sherwood <david.sherwood@arm.com>
7482 * tree-vect-stmts.c (simd_clone_subparts): New function.
7483 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
7485 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7486 Alan Hayward <alan.hayward@arm.com>
7487 David Sherwood <david.sherwood@arm.com>
7489 * tree-vect-stmts.c (vectorizable_call): Treat the number of
7490 vectors as polynomial. Use build_index_vector for
7493 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7494 Alan Hayward <alan.hayward@arm.com>
7495 David Sherwood <david.sherwood@arm.com>
7497 * tree-vect-stmts.c (get_load_store_type): Treat the number of
7498 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
7499 for variable-length vectors.
7500 (vectorizable_mask_load_store): Treat the number of units as
7501 polynomial, asserting that it is constant if the condition has
7502 already been enforced.
7503 (vectorizable_store, vectorizable_load): Likewise.
7505 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7506 Alan Hayward <alan.hayward@arm.com>
7507 David Sherwood <david.sherwood@arm.com>
7509 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
7510 of units as polynomial. Punt if we can't tell at compile time
7511 which vector contains the final result.
7513 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7514 Alan Hayward <alan.hayward@arm.com>
7515 David Sherwood <david.sherwood@arm.com>
7517 * tree-vect-loop.c (vectorizable_induction): Treat the number
7518 of units as polynomial. Punt on SLP inductions. Use an integer
7519 VEC_SERIES_EXPR for variable-length integer reductions. Use a
7520 cast of such a series for variable-length floating-point
7523 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7524 Alan Hayward <alan.hayward@arm.com>
7525 David Sherwood <david.sherwood@arm.com>
7527 * tree.h (build_index_vector): Declare.
7528 * tree.c (build_index_vector): New function.
7529 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
7530 of units as polynomial, forcibly converting it to a constant if
7531 vectorizable_reduction has already enforced the condition.
7532 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
7533 to create a {1,2,3,...} vector.
7534 (vectorizable_reduction): Treat the number of units as polynomial.
7535 Choose vectype_in based on the largest scalar element size rather
7536 than the smallest number of units. Enforce the restrictions
7539 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7540 Alan Hayward <alan.hayward@arm.com>
7541 David Sherwood <david.sherwood@arm.com>
7543 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
7544 number of units as polynomial.
7546 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7547 Alan Hayward <alan.hayward@arm.com>
7548 David Sherwood <david.sherwood@arm.com>
7550 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
7551 * target.def (autovectorize_vector_sizes): Return the vector sizes
7552 by pointer, using vector_sizes rather than a bitmask.
7553 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
7554 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
7555 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
7557 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
7558 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
7559 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
7560 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
7561 * omp-general.c (omp_max_vf): Likewise.
7562 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7563 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
7564 * tree-vect-loop.c (vect_analyze_loop): Likewise.
7565 * tree-vect-slp.c (vect_slp_bb): Likewise.
7566 * doc/tm.texi: Regenerate.
7567 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
7569 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
7570 the vector size as a poly_uint64 rather than an unsigned int.
7571 (current_vector_size): Change from an unsigned int to a poly_uint64.
7572 (get_vectype_for_scalar_type): Update accordingly.
7573 * tree.h (build_truth_vector_type): Take the size and number of
7574 units as a poly_uint64 rather than an unsigned int.
7575 (build_vector_type): Add a temporary overload that takes
7576 the number of units as a poly_uint64 rather than an unsigned int.
7577 * tree.c (make_vector_type): Likewise.
7578 (build_truth_vector_type): Take the number of units as a poly_uint64
7579 rather than an unsigned int.
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 * target.def (get_mask_mode): Take the number of units and length
7586 as poly_uint64s rather than unsigned ints.
7587 * targhooks.h (default_get_mask_mode): Update accordingly.
7588 * targhooks.c (default_get_mask_mode): Likewise.
7589 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
7590 * doc/tm.texi: Regenerate.
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 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
7597 * omp-general.c (omp_max_vf): Likewise.
7598 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
7599 (expand_omp_simd): Handle polynomial safelen.
7600 * omp-low.c (omplow_simd_context): Add a default constructor.
7601 (omplow_simd_context::max_vf): Change from int to poly_uint64.
7602 (lower_rec_simd_input_clauses): Update accordingly.
7603 (lower_rec_input_clauses): Likewise.
7605 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7606 Alan Hayward <alan.hayward@arm.com>
7607 David Sherwood <david.sherwood@arm.com>
7609 * tree-vectorizer.h (vect_nunits_for_cost): New function.
7610 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
7611 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
7612 (vect_analyze_slp_cost): Likewise.
7613 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
7614 (vect_model_load_cost): Likewise.
7616 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7617 Alan Hayward <alan.hayward@arm.com>
7618 David Sherwood <david.sherwood@arm.com>
7620 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
7621 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
7622 from an unsigned int * to a poly_uint64_pod *.
7623 (calculate_unrolling_factor): New function.
7624 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
7626 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7627 Alan Hayward <alan.hayward@arm.com>
7628 David Sherwood <david.sherwood@arm.com>
7630 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
7631 from an unsigned int to a poly_uint64.
7632 (_loop_vec_info::slp_unrolling_factor): Likewise.
7633 (_loop_vec_info::vectorization_factor): Change from an int
7635 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
7636 (vect_get_num_vectors): New function.
7637 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
7638 (vect_get_num_copies): Use vect_get_num_vectors.
7639 (vect_analyze_data_ref_dependences): Change max_vf from an int *
7640 to an unsigned int *.
7641 (vect_analyze_data_refs): Change min_vf from an int * to a
7643 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7644 than an unsigned HOST_WIDE_INT.
7645 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7646 (vect_analyze_data_ref_dependence): Change max_vf from an int *
7647 to an unsigned int *.
7648 (vect_analyze_data_ref_dependences): Likewise.
7649 (vect_compute_data_ref_alignment): Handle polynomial vf.
7650 (vect_enhance_data_refs_alignment): Likewise.
7651 (vect_prune_runtime_alias_test_list): Likewise.
7652 (vect_shift_permute_load_chain): Likewise.
7653 (vect_supportable_dr_alignment): Likewise.
7654 (dependence_distance_ge_vf): Take the vectorization factor as a
7655 poly_uint64 rather than an unsigned HOST_WIDE_INT.
7656 (vect_analyze_data_refs): Change min_vf from an int * to a
7658 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
7659 vfm1 as a poly_uint64 rather than an int. Make the same change
7660 for the returned bound_scalar.
7661 (vect_gen_vector_loop_niters): Handle polynomial vf.
7662 (vect_do_peeling): Likewise. Update call to
7663 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
7664 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
7666 * tree-vect-loop.c (vect_determine_vectorization_factor)
7667 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
7668 (vect_get_known_peeling_cost): Likewise.
7669 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
7670 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
7671 (vect_transform_loop): Likewise. Use the lowest possible VF when
7672 updating the upper bounds of the loop.
7673 (vect_min_worthwhile_factor): Make static. Return an unsigned int
7675 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
7676 polynomial unroll factors.
7677 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
7678 (vect_make_slp_decision): Likewise.
7679 (vect_supported_load_permutation_p): Likewise, and polynomial
7681 (vect_analyze_slp_cost): Handle polynomial vf.
7682 (vect_slp_analyze_node_operations): Likewise.
7683 (vect_slp_analyze_bb_1): Likewise.
7684 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7685 than an unsigned HOST_WIDE_INT.
7686 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
7687 (vectorizable_load): Handle polynomial vf.
7688 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
7690 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
7692 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7693 Alan Hayward <alan.hayward@arm.com>
7694 David Sherwood <david.sherwood@arm.com>
7696 * match.pd: Handle bit operations involving three constants
7697 and try to fold one pair.
7699 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7701 * tree-vect-loop-manip.c: Include gimple-fold.h.
7702 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
7703 niters_maybe_zero parameters. Handle other cases besides a step of 1.
7704 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
7705 Add a path that uses a step of VF instead of 1, but disable it
7707 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
7708 and niters_no_overflow parameters. Update calls to
7709 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
7710 Create a new SSA name if the latter choses to use a ste other
7711 than zero, and return it via niters_vector_mult_vf_var.
7712 * tree-vect-loop.c (vect_transform_loop): Update calls to
7713 vect_do_peeling, vect_gen_vector_loop_niters and
7714 slpeel_make_loop_iterate_ntimes.
7715 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
7716 (vect_gen_vector_loop_niters): Update declarations after above changes.
7718 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
7720 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
7721 128-bit round to integer instructions.
7722 (ceil<mode>2): Likewise.
7723 (btrunc<mode>2): Likewise.
7724 (round<mode>2): Likewise.
7726 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7728 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
7729 unaligned VSX load/store on P8/P9.
7730 (expand_block_clear): Allow the use of unaligned VSX
7731 load/store on P8/P9.
7733 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7735 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
7737 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
7738 swap associated with both a load and a store.
7740 2018-01-02 Andrew Waterman <andrew@sifive.com>
7742 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
7743 * config/riscv/riscv.md (clear_cache): Use it.
7745 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
7747 * web.c: Remove out-of-date comment.
7749 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7751 * expr.c (fixup_args_size_notes): Check that any existing
7752 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
7753 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
7754 (emit_single_push_insn): ...here.
7756 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7758 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
7759 (const_vector_encoded_nelts): New function.
7760 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
7761 (const_vector_int_elt, const_vector_elt): Declare.
7762 * emit-rtl.c (const_vector_int_elt_1): New function.
7763 (const_vector_elt): Likewise.
7764 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
7765 of CONST_VECTOR_ELT.
7767 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7769 * expr.c: Include rtx-vector-builder.h.
7770 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
7771 directly on the tree encoding.
7772 (const_vector_from_tree): Likewise.
7773 * optabs.c: Include rtx-vector-builder.h.
7774 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
7775 sequence of "u" values.
7776 * vec-perm-indices.c: Include rtx-vector-builder.h.
7777 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
7778 directly on the vec_perm_indices encoding.
7780 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7782 * doc/rtl.texi (const_vector): Describe new encoding scheme.
7783 * Makefile.in (OBJS): Add rtx-vector-builder.o.
7784 * rtx-vector-builder.h: New file.
7785 * rtx-vector-builder.c: Likewise.
7786 * rtl.h (rtx_def::u2): Add a const_vector field.
7787 (CONST_VECTOR_NPATTERNS): New macro.
7788 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
7789 (CONST_VECTOR_DUPLICATE_P): Likewise.
7790 (CONST_VECTOR_STEPPED_P): Likewise.
7791 (CONST_VECTOR_ENCODED_ELT): Likewise.
7792 (const_vec_duplicate_p): Check for a duplicated vector encoding.
7793 (unwrap_const_vec_duplicate): Likewise.
7794 (const_vec_series_p): Check for a non-duplicated vector encoding.
7795 Say that the function only returns true for integer vectors.
7796 * emit-rtl.c: Include rtx-vector-builder.h.
7797 (gen_const_vec_duplicate_1): Delete.
7798 (gen_const_vector): Call gen_const_vec_duplicate instead of
7799 gen_const_vec_duplicate_1.
7800 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
7801 (gen_const_vec_duplicate): Use rtx_vector_builder.
7802 (gen_const_vec_series): Likewise.
7803 (gen_rtx_CONST_VECTOR): Likewise.
7804 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
7805 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7806 Build a new vector rather than modifying a CONST_VECTOR in-place.
7807 (handle_special_swappables): Update call accordingly.
7808 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
7809 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7810 Build a new vector rather than modifying a CONST_VECTOR in-place.
7811 (handle_special_swappables): Update call accordingly.
7813 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7815 * simplify-rtx.c (simplify_const_binary_operation): Use
7816 CONST_VECTOR_ELT instead of XVECEXP.
7818 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7820 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
7821 the selector elements to be different from the data elements
7822 if the selector is a VECTOR_CST.
7823 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
7824 ssizetype for the selector.
7826 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7828 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
7829 before testing each element individually.
7830 * tree-vect-generic.c (lower_vec_perm): Likewise.
7832 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7834 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
7835 * selftest-run-tests.c (selftest::run_tests): Call it.
7836 * vector-builder.h (vector_builder::operator ==): New function.
7837 (vector_builder::operator !=): Likewise.
7838 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
7839 (vec_perm_indices::all_from_input_p): New function.
7840 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7841 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
7842 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
7843 instead of reading the VECTOR_CST directly. Detect whether both
7844 vector inputs are the same before constructing the vec_perm_indices,
7845 and update the number of inputs argument accordingly. Use the
7846 utility functions added above. Only construct sel2 if we need to.
7848 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7850 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
7851 the broadcast of the low byte.
7852 (expand_mult_highpart): Use an explicit encoding for the permutes.
7853 * optabs-query.c (can_mult_highpart_p): Likewise.
7854 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
7855 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7856 (vectorizable_bswap): Likewise.
7857 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
7858 explicit encoding for the power-of-2 permutes.
7859 (vect_permute_store_chain): Likewise.
7860 (vect_grouped_load_supported): Likewise.
7861 (vect_permute_load_chain): Likewise.
7863 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7865 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
7866 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
7867 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
7868 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7869 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
7870 (vect_gen_perm_mask_any): Likewise.
7872 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7874 * int-vector-builder.h: New file.
7875 * vec-perm-indices.h: Include int-vector-builder.h.
7876 (vec_perm_indices): Redefine as an int_vector_builder.
7877 (auto_vec_perm_indices): Delete.
7878 (vec_perm_builder): Redefine as a stand-alone class.
7879 (vec_perm_indices::vec_perm_indices): New function.
7880 (vec_perm_indices::clamp): Likewise.
7881 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
7882 (vec_perm_indices::new_vector): New function.
7883 (vec_perm_indices::new_expanded_vector): Update for new
7884 vec_perm_indices class.
7885 (vec_perm_indices::rotate_inputs): New function.
7886 (vec_perm_indices::all_in_range_p): Operate directly on the
7887 encoded form, without computing elided elements.
7888 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
7889 encoding. Update for new vec_perm_indices class.
7890 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
7891 the given vec_perm_builder.
7892 (expand_vec_perm_var): Update vec_perm_builder constructor.
7893 (expand_mult_highpart): Use vec_perm_builder instead of
7894 auto_vec_perm_indices.
7895 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
7896 vec_perm_indices instead of auto_vec_perm_indices. Use a single
7897 or double series encoding as appropriate.
7898 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
7899 vec_perm_indices instead of auto_vec_perm_indices.
7900 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7901 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7902 (vect_permute_store_chain): Likewise.
7903 (vect_grouped_load_supported): Likewise.
7904 (vect_permute_load_chain): Likewise.
7905 (vect_shift_permute_load_chain): Likewise.
7906 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
7907 (vect_transform_slp_perm_load): Likewise.
7908 (vect_schedule_slp_instance): Likewise.
7909 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7910 (vectorizable_mask_load_store): Likewise.
7911 (vectorizable_bswap): Likewise.
7912 (vectorizable_store): Likewise.
7913 (vectorizable_load): Likewise.
7914 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
7915 vec_perm_indices instead of auto_vec_perm_indices. Use
7916 tree_to_vec_perm_builder to read the vector from a tree.
7917 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
7918 vec_perm_builder instead of a vec_perm_indices.
7919 (have_whole_vector_shift): Use vec_perm_builder and
7920 vec_perm_indices instead of auto_vec_perm_indices. Leave the
7921 truncation to calc_vec_perm_mask_for_shift.
7922 (vect_create_epilog_for_reduction): Likewise.
7923 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
7924 from auto_vec_perm_indices to vec_perm_indices.
7925 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7926 instead of changing individual elements.
7927 (aarch64_vectorize_vec_perm_const): Use new_vector to install
7928 the vector in d.perm.
7929 * config/arm/arm.c (expand_vec_perm_d::perm): Change
7930 from auto_vec_perm_indices to vec_perm_indices.
7931 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7932 instead of changing individual elements.
7933 (arm_vectorize_vec_perm_const): Use new_vector to install
7934 the vector in d.perm.
7935 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
7936 Update vec_perm_builder constructor.
7937 (rs6000_expand_interleave): Likewise.
7938 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
7939 (rs6000_expand_interleave): Likewise.
7941 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7943 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
7944 to qimode could truncate the indices.
7945 * optabs.c (expand_vec_perm_var): Likewise.
7947 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7949 * Makefile.in (OBJS): Add vec-perm-indices.o.
7950 * vec-perm-indices.h: New file.
7951 * vec-perm-indices.c: Likewise.
7952 * target.h (vec_perm_indices): Replace with a forward class
7954 (auto_vec_perm_indices): Move to vec-perm-indices.h.
7955 * optabs.h: Include vec-perm-indices.h.
7956 (expand_vec_perm): Delete.
7957 (selector_fits_mode_p, expand_vec_perm_var): Declare.
7958 (expand_vec_perm_const): Declare.
7959 * target.def (vec_perm_const_ok): Replace with...
7960 (vec_perm_const): ...this new hook.
7961 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
7962 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
7963 * doc/tm.texi: Regenerate.
7964 * optabs.def (vec_perm_const): Delete.
7965 * doc/md.texi (vec_perm_const): Likewise.
7966 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
7967 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
7968 expand_vec_perm for constant permutation vectors. Assert that
7969 the mode of variable permutation vectors is the integer equivalent
7970 of the mode that is being permuted.
7971 * optabs-query.h (selector_fits_mode_p): Declare.
7972 * optabs-query.c: Include vec-perm-indices.h.
7973 (selector_fits_mode_p): New function.
7974 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
7975 is defined, instead of checking whether the vec_perm_const_optab
7976 exists. Use targetm.vectorize.vec_perm_const instead of
7977 targetm.vectorize.vec_perm_const_ok. Check whether the indices
7978 fit in the vector mode before using a variable permute.
7979 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
7980 vec_perm_indices instead of an rtx.
7981 (expand_vec_perm): Replace with...
7982 (expand_vec_perm_const): ...this new function. Take the selector
7983 as a vec_perm_indices rather than an rtx. Also take the mode of
7984 the selector. Update call to shift_amt_for_vec_perm_mask.
7985 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
7986 Use vec_perm_indices::new_expanded_vector to expand the original
7987 selector into bytes. Check whether the indices fit in the vector
7988 mode before using a variable permute.
7989 (expand_vec_perm_var): Make global.
7990 (expand_mult_highpart): Use expand_vec_perm_const.
7991 * fold-const.c: Includes vec-perm-indices.h.
7992 * tree-ssa-forwprop.c: Likewise.
7993 * tree-vect-data-refs.c: Likewise.
7994 * tree-vect-generic.c: Likewise.
7995 * tree-vect-loop.c: Likewise.
7996 * tree-vect-slp.c: Likewise.
7997 * tree-vect-stmts.c: Likewise.
7998 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
8000 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
8001 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
8002 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
8003 (aarch64_vectorize_vec_perm_const): ...this new function.
8004 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8005 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8006 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
8007 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
8008 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8009 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8010 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
8012 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
8013 check for NEON modes.
8014 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
8015 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
8016 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
8017 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
8019 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
8020 the old VEC_PERM_CONST conditions.
8021 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
8022 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
8023 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
8024 (ia64_vectorize_vec_perm_const_ok): Merge into...
8025 (ia64_vectorize_vec_perm_const): ...this new function.
8026 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
8027 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
8028 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
8029 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
8030 * config/mips/mips.c (mips_expand_vec_perm_const)
8031 (mips_vectorize_vec_perm_const_ok): Merge into...
8032 (mips_vectorize_vec_perm_const): ...this new function.
8033 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
8034 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
8035 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
8036 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
8037 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
8038 (rs6000_expand_vec_perm_const): Delete.
8039 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
8041 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8042 (altivec_expand_vec_perm_const_le): Take each operand individually.
8043 Operate on constant selectors rather than rtxes.
8044 (altivec_expand_vec_perm_const): Likewise. Update call to
8045 altivec_expand_vec_perm_const_le.
8046 (rs6000_expand_vec_perm_const): Delete.
8047 (rs6000_vectorize_vec_perm_const_ok): Delete.
8048 (rs6000_vectorize_vec_perm_const): New function.
8049 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8050 an element count and rtx array.
8051 (rs6000_expand_extract_even): Update call accordingly.
8052 (rs6000_expand_interleave): Likewise.
8053 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
8054 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
8055 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
8056 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
8057 (rs6000_expand_vec_perm_const): Delete.
8058 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8059 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8060 (altivec_expand_vec_perm_const_le): Take each operand individually.
8061 Operate on constant selectors rather than rtxes.
8062 (altivec_expand_vec_perm_const): Likewise. Update call to
8063 altivec_expand_vec_perm_const_le.
8064 (rs6000_expand_vec_perm_const): Delete.
8065 (rs6000_vectorize_vec_perm_const_ok): Delete.
8066 (rs6000_vectorize_vec_perm_const): New function. Remove stray
8067 reference to the SPE evmerge intructions.
8068 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8069 an element count and rtx array.
8070 (rs6000_expand_extract_even): Update call accordingly.
8071 (rs6000_expand_interleave): Likewise.
8072 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
8073 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
8075 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8077 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8079 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
8080 vector mode and that that mode matches the mode of the data
8082 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
8083 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
8084 directly using expand_vec_perm_1 when forcing selectors into
8086 (expand_vec_perm_var): New function, split out from expand_vec_perm.
8088 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8090 * optabs-query.h (can_vec_perm_p): Delete.
8091 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
8092 * optabs-query.c (can_vec_perm_p): Split into...
8093 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
8094 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
8095 particular selector is valid.
8096 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8097 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8098 (vect_grouped_load_supported): Likewise.
8099 (vect_shift_permute_load_chain): Likewise.
8100 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8101 (vect_transform_slp_perm_load): Likewise.
8102 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8103 (vectorizable_bswap): Likewise.
8104 (vect_gen_perm_mask_checked): Likewise.
8105 * fold-const.c (fold_ternary_loc): Likewise. Don't take
8106 implementations of variable permutation vectors into account
8107 when deciding which selector to use.
8108 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
8109 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
8110 with a false third argument.
8111 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
8112 to test whether the constant selector is valid and can_vec_perm_var_p
8113 to test whether a variable selector is valid.
8115 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8117 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
8118 * optabs-query.c (can_vec_perm_p): Likewise.
8119 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
8120 instead of vec_perm_indices.
8121 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
8122 (vect_gen_perm_mask_checked): Likewise,
8123 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
8124 (vect_gen_perm_mask_checked): Likewise,
8126 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8128 * optabs-query.h (qimode_for_vec_perm): Declare.
8129 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
8130 (qimode_for_vec_perm): ...this new function.
8131 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
8133 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8135 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
8136 does not have a conditional at the top.
8138 2018-01-02 Richard Biener <rguenther@suse.de>
8140 * ipa-inline.c (big_speedup_p): Fix expression.
8142 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8145 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
8148 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8152 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
8153 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
8154 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
8155 cond_taken_branch_cost 3->4.
8157 2018-01-01 Jakub Jelinek <jakub@redhat.com>
8159 PR tree-optimization/83581
8160 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
8161 TODO_cleanup_cfg if any changes have been made.
8164 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
8165 convert_modes if target mode has the right side, but different mode
8169 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
8170 last argument when extracting from CONCAT. If either from_real or
8171 from_imag is NULL, use expansion through memory. If result is not
8172 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
8173 the parts directly to inner mode, if even that fails, use expansion
8177 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
8178 check for bswap in mode rather than HImode and use that in expand_unop
8181 Copyright (C) 2018 Free Software Foundation, Inc.
8183 Copying and distribution of this file, with or without modification,
8184 are permitted in any medium without royalty provided the copyright
8185 notice and this notice are preserved.