1 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4 * multiple_target.c: Include tree-inline.h and intl.h.
5 (expand_target_clones): Diagnose and fail if node->definition and
6 !tree_versionable_function_p (node->decl).
8 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
10 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
12 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
13 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
14 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
16 2018-03-06 Jakub Jelinek <jakub@redhat.com>
19 * combine.c (try_combine): Use reg_or_subregno instead of handling
20 just paradoxical SUBREGs and REGs.
22 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
24 * config/arc/arc.c (arc_finalize_pic): Remove function.
25 (arc_must_save_register): We use single base PIC register, remove
26 checks to save/restore the PIC register.
27 (arc_expand_prologue): Likewise.
28 * config/arc/arc-protos.h (arc_set_default_type_attributes):
30 (arc_verify_short): Likewise.
31 (arc_attr_type): Likewise.
32 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
33 (walk_stores): Likewise.
34 (arc_address_cost): Make it static.
35 (arc_verify_short): Likewise.
36 (branch_dest): Likewise.
37 (arc_attr_type): Likewise.
38 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
39 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
40 (arc_final_prescan_insn): Remove inserting the nops due to
41 hardware hazards. It is done in reorg step.
42 (insn_length_variant_t): Remove.
43 (insn_length_parameters_t): Likewise.
44 (arc_insn_length_parameters): Likewise.
45 (arc_get_insn_variants): Likewise.
46 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
48 2018-03-06 Jakub Jelinek <jakub@redhat.com>
51 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
54 PR tree-optimization/84687
55 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
57 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
59 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
61 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
62 Rename to ppc_speculation_barrier.
63 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
64 __builtin_ppc_speculation_barrier.
66 2018-03-05 Jakub Jelinek <jakub@redhat.com>
69 * combine.c (combine_simplify_rtx): Don't try to simplify if
70 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
73 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
75 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
76 to 32 bytes when compiling for POWER9.
78 2018-03-05 Jakub Jelinek <jakub@redhat.com>
81 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
82 regparm >= 3 with no arg reg available also for calls with
83 flag_force_indirect_call. Pass decl to ix86_function_regparm.
86 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
88 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
90 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
93 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
95 2018-03-05 Richard Biener <rguenther@suse.de>
97 PR tree-optimization/84486
98 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
99 When inserting a __builtin_assume_aligned call set the LHS
100 SSA name alignment info accordingly.
102 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
104 PR tree-optimization/84114
105 * config/aarch64/aarch64.c (aarch64_reassociation_width)
106 Avoid reassociation of FLOAT_MODE addition.
108 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
110 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
111 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
112 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
113 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
114 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
115 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
116 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
118 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
119 __builtin_ia32_wbinvd): New builtins.
120 (SPECIAL_ARGS2): New.
121 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
122 (SPECIAL_ARGS2): New.
123 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
124 (ix86_valid_target_attribute_inner_p): Ditto.
125 (ix86_init_mmx_sse_builtins): Add special_args2.
126 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
127 TARGET_WBNOINVD_P): New.
128 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
129 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
130 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
131 * config/i386/immintrin.h (_wbinvd): New intrinsic.
132 * config/i386/pconfigintrin.h: New file.
133 * config/i386/wbnoinvdintrin.h: Ditto.
134 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
135 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
137 2018-03-05 Richard Biener <rguenther@suse.de>
139 PR tree-optimization/84670
140 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
142 (BB_VISITED_WITH_VISITED_SUCCS): New define.
143 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
144 (compute_antic_aux): Only assert the number of values in ANTIC_IN
145 doesn't grow if all successors (recursively) were visited at least
148 2018-03-05 Richard Biener <rguenther@suse.de>
150 PR tree-optimization/84650
151 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
152 if executed in the loop pipeline.
154 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
156 * doc/configfiles.texi (Configuration Files): Move info about
157 conditionalizing $target-protos.h to...
158 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
159 differs from $target-protos.h.
161 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
162 Chung-Ju Wu <jasonwucj@gmail.com>
164 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
165 * config/nds32/nds32-multiple.md (setmemsi): Define.
166 * config/nds32/nds32-memory-manipulation.c
167 (nds32_gen_dup_4_byte_to_word_value): New.
168 (emit_setmem_word_loop): New.
169 (emit_setmem_byte_loop): New.
170 (nds32_expand_setmem_loop): New.
171 (nds32_expand_setmem_loop_v3m): New.
172 (nds32_expand_setmem_unroll): New.
173 (nds32_expand_setmem): New.
175 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
176 Chung-Ju Wu <jasonwucj@gmail.com>
178 * config/nds32/nds32-memory-manipulation.c
179 (nds32_emit_load_store): New.
180 (nds32_emit_post_inc_load_store): New.
181 (nds32_emit_mem_move): New.
182 (nds32_emit_mem_move_block): New.
183 (nds32_expand_movmemsi_loop_unknown_size): New.
184 (nds32_expand_movmemsi_loop_known_size): New.
185 (nds32_expand_movmemsi_loop): New.
186 (nds32_expand_movmemsi_unroll): New.
187 (nds32_expand_movmemqi): Rename ...
188 (nds32_expand_movmemsi): ... to this.
189 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
190 (movmemsi): ... to this.
191 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
192 (nds32_expand_movmemsi): ... to this.
194 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
195 Monk Chiang <sh.chiang04@gmail.com>
196 Chung-Ju Wu <jasonwucj@gmail.com>
198 * config/nds32/nds32-protos.h
199 (nds32_expand_load_multiple): New arguments.
200 (nds32_expand_store_multiple): Ditto.
201 (nds32_valid_multiple_load_store): Rename ...
202 (nds32_valid_multiple_load_store_p): ... to this.
203 * config/nds32/nds32-memory-manipulation.c
204 (nds32_expand_load_multiple): Refine implementation.
205 (nds32_expand_store_multiple): Ditto.
206 * config/nds32/nds32-multiple.md
207 (load_multiple): Update nds32_expand_load_multiple interface.
208 (store_multiple): Update nds32_expand_store_multiple interface.
209 * config/nds32/nds32-predicates.c
210 (nds32_valid_multiple_load_store): Rename ...
211 (nds32_valid_multiple_load_store_p): ... to this and refine
213 * config/nds32/predicates.md
214 (nds32_load_multiple_and_update_address_operation): New predicate.
215 (nds32_store_multiple_and_update_address_operation): New predicate.
217 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
218 Chung-Ju Wu <jasonwucj@gmail.com>
220 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
221 (combo): New attribute.
222 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
224 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
226 * config/nds32/nds32.opt: Change -mcmodel= default value.
228 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
229 Monk Chiang <sh.chiang04@gmail.com>
230 Chung-Ju Wu <jasonwucj@gmail.com>
232 * config/nds32/constants.md (unspec_element): New enum.
233 * config/nds32/constraints.md (Umw): New constraint.
234 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
235 * config/nds32/nds32-intrinsic.md: Likewise.
236 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
237 (nds32_valid_smw_lwm_base_p): New.
238 (nds32_output_smw_single_word): New.
239 (nds32_output_lmw_single_word): New.
240 (nds32_expand_unaligned_load): New.
241 (nds32_expand_unaligned_store): New.
242 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
243 (nds32_output_smw_single_word): Declare.
244 (nds32_output_lmw_single_word): Declare.
245 (nds32_expand_unaligned_load): Declare.
246 (nds32_expand_unaligned_store): Declare.
247 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
248 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
249 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
250 NDS32_BUILTIN_UASTORE_DW.
251 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
254 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
255 Kito Cheng <kito.cheng@gmail.com>
256 Chung-Ju Wu <jasonwucj@gmail.com>
258 * config/nds32/nds32-intrinsic.c
259 (nds32_expand_builtin_null_ftype_reg): Delete.
260 (nds32_expand_builtin_reg_ftype_imm): Ditto.
261 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
262 (nds32_read_argument): New.
263 (nds32_legitimize_target): Ditto.
264 (nds32_legitimize_argument): Ditto.
265 (nds32_check_constant_argument): Ditto.
266 (nds32_expand_unop_builtin): Ditto.
267 (nds32_expand_unopimm_builtin): Ditto.
268 (nds32_expand_binop_builtin): Ditto.
269 (nds32_builtin_decl_impl): Ditto.
270 (builtin_description): Ditto.
271 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
272 (nds32_init_builtins_impl): Ditto.
273 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
274 (nds32_builtin_decl): New.
275 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
276 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
278 2018-03-02 Jeff Law <law@redhat.com>
280 * reorg.c (stop_search_p): Handle DEBUG_INSN.
281 (redundant_insn, fill_simple_delay_slots): Likewise.
282 (fill_slots_from_thread): Likewise.
283 * resource.c (mark_referenced_resources): Likewise.
284 (mark_set_resources, find_dead_or_set_registers): Likewise.
286 2018-03-02 Jakub Jelinek <jakub@redhat.com>
288 * substring-locations.h (format_warning_va): Formatting fix for
290 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
292 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
293 * substring-locations.c: Include intl.h.
294 (format_warning_va): Turned into small wrapper around
295 format_warning_n_va, renamed to ...
296 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
297 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
299 (format_warning_at_substring_n): New function.
300 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
301 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
302 format_warning_at_substring with just a shorter name instead of
303 const function pointer.
304 (fmtwarn_n): New function.
305 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
306 appropriate, get rid of all the fmtstr temporaries, move conditionals
307 with G_() wrapped string literals directly into fmtwarn arguments,
308 cast dir.len to (int), formatting fixes.
310 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
312 * doc/invoke.texi: Remove "Cilk Plus" references.
314 2018-03-02 Jakub Jelinek <jakub@redhat.com>
315 Richard Biener <rguenther@suse.de>
318 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
319 for error or warning attributes if CALL_FROM_THUNK_P is set.
322 2018-03-02 Jakub Jelinek <jakub@redhat.com>
325 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
326 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
329 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
330 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
332 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
333 instead of -1U in last predictors element's probability member.
335 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
338 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
339 arguments if they are comparable.
341 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
343 PR tree-optimization/84634
344 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
345 masks and masked_loop_p with a single loop_masks, making sure it's
346 null for bb vectorization.
348 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
350 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
351 (vect_analyze_data_ref_access): Use loop->safe_len rather than
352 loop->force_vectorize to check whether there is no alias.
354 2018-03-02 Jakub Jelinek <jakub@redhat.com>
357 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
359 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
361 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
362 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
363 instead of a loop around prev_real_insn.
364 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
368 * config/i386/i386.c (ix86_print_operand): Use conditional
369 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
372 2018-03-02 Richard Biener <rguenther@suse.de>
374 PR tree-optimization/84427
375 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
376 (bitmap_set_subtract_values): Rewrite to handle multiple
379 (prune_clobbered_mems): Likewise.
380 (phi_translate): Take edge instead of pred/phiblock.
381 (phi_translate_1): Likewise.
382 (phi_translate_set): Likewise. Insert all translated
383 exprs for a value into the set, keeping possibly multiple
384 expressions per value.
385 (compute_antic_aux): Adjust for phi_translate changes.
386 When intersecting union the expressions and prune those
387 not in the final value set, keeping possibly multiple
388 expressions per value. Do not use value-insertion
389 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
390 all expressions. Add verification that the value-sets
391 only shrink during iteration.
392 (compute_partial_antic_aux): Adjust for the phi_translate changes.
393 (do_pre_regular_insertion): Likewise.
394 (do_pre_partial_partial_insertion): Likewise.
396 2018-03-02 Richard Biener <rguenther@suse.de>
399 * config/darwin.c (saved_debug_info_level): New static global.
400 (darwin_asm_lto_start): Disable debug info generation for LTO out.
401 (darwin_asm_lto_end): Restore debug info generation settings.
403 2018-03-01 Martin Liska <mliska@suse.cz>
406 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
409 2018-03-01 Richard Biener <rguenther@suse.de>
412 * dwarf2out.c (gen_variable_die): Properly handle late VLA
413 type annotation with LTO when debug was disabled at compile-time.
415 2018-03-01 Matthew Fortune <mfortune@gmail.com>
417 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
419 (mips_final_postscan_insn): Likewise.
421 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
423 PR rtl-optimization/84528
424 * alias.c (init_alias_target): Add commentary.
425 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
426 a unique base value if the frame pointer is not eliminated
427 to the stack pointer.
429 2018-03-01 Tom de Vries <tom@codesourcery.com>
431 PR rtl-optimization/83327
432 * lra-int.h (hard_regs_spilled_into): Declare.
433 * lra.c (hard_regs_spilled_into): Define.
434 (init_reg_info): Init hard_regs_spilled_into.
435 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
436 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
437 (process_bb_lives): Handle hard_regs_spilled_into.
438 (lra_create_live_ranges_1): Before doing liveness propagation, clear
439 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
441 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
443 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
444 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
445 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
446 * config/rs6000/aix72.h: New file.
448 2018-02-28 Jakub Jelinek <jakub@redhat.com>
450 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
451 instead of warning_at with conditional singular and plural messages
455 * stor-layout.c (update_alignment_for_field): For
456 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
457 && !DECL_PACKED (field), do the alignment update, just use
458 only desired_align instead of MAX (type_align, desired_align)
460 (place_field): Don't do known_align < desired_align handling
461 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
462 is non-NULL, instead do it after rli->prev_field handling and
463 only if not within a bitfield word. For DECL_PACKED (field)
464 use type_align of BITS_PER_UNIT.
466 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
468 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
469 superfluous parentheses and trailing spaces.
471 2018-02-28 Richard Biener <rguenther@suse.de>
473 PR tree-optimization/84584
474 * graphite-scop-detection.c (scop_detection::add_scop): Discard
475 SCoPs with fake exit edge.
477 2018-02-28 Martin Liska <mliska@suse.cz>
480 * timevar.c (timer::print): Fix format to properly print 100%
483 2018-02-28 Richard Biener <rguenther@suse.de>
486 * genmatch.c (capture_info::walk_match): Do not mark
487 captured expressions without operands as expr_p given
488 they act more like predicates and should be subject to
489 "lost tail" side-effect preserving.
491 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
493 PR rtl-optimization/81611
494 * auto-inc-dec.c (attempt_change): Move dead note from
495 mem_insn if it's the next use of regno
496 (find_address): Take address use of reg holding
497 non-incremented value. Add parm to limit search to the named
499 (merge_in_block): Attempt to use a mem insn that is the next
500 use of the original regno.
502 2018-02-27 Martin Sebor <msebor@redhat.com>
505 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
506 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
508 2018-02-27 Martin Sebor <msebor@redhat.com>
511 * diagnostic-core.h (warning_n, error_n, inform_n): Change
512 n argument to unsigned HOST_WIDE_INT.
513 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
514 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
515 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
516 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
518 2018-02-27 Richard Biener <rguenther@suse.de>
520 PR tree-optimization/84512
521 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
522 Do not use the estimate returned from record_stmt_cost for
523 the scalar iteration cost but sum properly using add_stmt_cost.
525 2018-02-27 Richard Biener <rguenther@suse.de>
527 PR tree-optimization/84466
528 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
529 Adjust last change to less strictly validate use operands.
531 2018-02-27 Martin Liska <mliska@suse.cz>
533 PR gcov-profile/84548
534 * gcov.c (process_file): Allow partial overlap and consider it
535 also as group functions.
536 (output_lines): Properly calculate range of lines for a group.
538 2018-02-27 Martin Liska <mliska@suse.cz>
540 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
541 'ggc' suffixes. Change first column width.
542 (timer::print): Fix formatting of the column.
544 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
546 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
547 preserve inline entry blocks for the sake of debug inline
548 entry point markers alone.
549 (remove_unused_locals): Suggest in comments a better place to
550 force the preservation of inline entry blocks that are
551 otherwise unused, but do not preserve them.
553 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
555 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
557 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
560 * config/i386/constraints.md (Bs): Replace
561 ix86_indirect_branch_register with
562 TARGET_INDIRECT_BRANCH_REGISTER.
564 * config/i386/i386.md (indirect_jump): Likewise.
565 (tablejump): Likewise.
566 (*sibcall_memory): Likewise.
567 (*sibcall_value_memory): Likewise.
568 Peepholes of indirect call and jump via memory: Likewise.
569 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
570 (*sibcall_value_GOT_32): Likewise.
571 * config/i386/predicates.md (indirect_branch_operand): Likewise.
572 (GOT_memory_operand): Likewise.
573 (call_insn_operand): Likewise.
574 (sibcall_insn_operand): Likewise.
575 (GOT32_symbol_operand): Likewise.
576 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
578 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
580 PR rtl-optimization/83496
581 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
582 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
583 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
584 redundant insn, if any.
585 (relax_delay_slots): Likewise.
586 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
588 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
590 PR tree-optimization/83965
591 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
592 that grouped statements are part of a reduction chain. Return
593 true if the statement is not marked as a reduction itself but
595 (vect_recog_dot_prod_pattern): Don't check whether the statement
596 is part of a group here.
597 (vect_recog_sad_pattern): Likewise.
598 (vect_recog_widen_sum_pattern): Likewise.
600 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
603 * final.c (rest_of_clean_state): Also look for calls inside sequences.
605 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
608 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
610 (ix86_output_indirect_function_return): New prototype.
611 (ix86_split_simple_return_pop_internal): Likewise.
612 * config/i386/i386.c (indirect_return_via_cx): New.
613 (indirect_return_via_cx_bnd): Likewise.
614 (indirect_thunk_name): Handle return va CX_REG.
615 (output_indirect_thunk_function): Create alias for
616 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
617 (ix86_output_indirect_jmp): Remove the bool argument.
618 (ix86_output_indirect_function_return): New function.
619 (ix86_split_simple_return_pop_internal): Likewise.
620 * config/i386/i386.md (*indirect_jump): Don't pass false
621 to ix86_output_indirect_jmp.
622 (*tablejump_1): Likewise.
623 (simple_return_pop_internal): Change it to define_insn_and_split.
624 Call ix86_split_simple_return_pop_internal to split it for
626 (simple_return_indirect_internal): Call
627 ix86_output_indirect_function_return instead of
628 ix86_output_indirect_jmp.
630 2018-02-26 Jakub Jelinek <jakub@redhat.com>
633 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
634 memset and value initialization afterwards.
636 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
638 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
640 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
643 * common/config/aarch64/aarch64-common.c
644 (aarch_option_optimization_table[]): Switch
645 off fomit-frame-pointer
647 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
648 Chung-Ju Wu <jasonwucj@gmail.com>
650 * config/nds32/nds32-multiple.md (load_multiple): Disallow
652 (store_multiple): Ditto.
654 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
656 * config.gcc: Add --with-cpu support for nds32 target.
657 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
658 * config/nds32/nds32.opt: Add -mcpu= option.
660 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
662 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
663 isel=yes): Warn for these deprecated options.
665 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
667 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
668 ISA_2_5_MASKS_EMBEDDED.
670 2018-02-23 Jakub Jelinek <jakub@redhat.com>
672 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
673 p->max as pointers rather than using iterative_hash_expr.
675 2018-02-23 Carl Love <cel@us.ibm.com>
677 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
678 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
680 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
681 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
682 P8V_BUILTIN_VEC_VUNSIGNED2.
684 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
687 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
688 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
690 * lra-constraints.c (curr_insn_transform): Set up
691 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
693 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
694 * lra-eliminations.c (spill_pseudos): Ditto.
695 (process_insn_for_elimination): Ditto.
696 * lra-lives.c (reg_early_clobber_p): Use the new macros.
697 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
698 LRA_NON_CLOBBERED_ALT.
700 2018-02-22 Martin Sebor <msebor@redhat.com>
702 PR tree-optimization/84480
703 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
704 to maybe_diag_stxncpy_trunc. Call it.
705 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
706 from gimple_fold_builtin_strcpy. Print inlining stack.
707 (handle_builtin_stxncpy): Print inlining stack.
708 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
710 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
713 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
714 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
715 and -fcheck-pointer-bounds are used together.
716 (indirect_thunk_prefix): New enum.
717 (indirect_thunk_need_prefix): New function.
718 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
719 "_nt" instead of "_bnd" for NOTRACK prefix.
720 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
721 (output_indirect_thunk_function): Likewise.
723 (ix86_code_end): Update output_indirect_thunk_function calls.
724 (ix86_output_indirect_branch_via_reg): Replace
725 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
726 (ix86_output_indirect_branch_via_push): Likewise.
727 (ix86_output_function_return): Likewise.
728 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
729 incompatible with -fcf-protection=branch and
730 -fcheck-pointer-bounds.
732 2018-02-22 Steve Ellcey <sellcey@cavium.com>
735 * config/aarch64/aarch64.c (aarch64_print_address_internal):
736 Change gcc_assert call to output_operand_lossage.
738 2018-02-22 Steve Ellcey <sellcey@cavium.com>
740 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
742 2018-02-22 DJ Delorie <dj@redhat.com>
743 Sebastian Perta <sebastian.perta@renesas.com>
744 Oleg Endo <olegendo@gcc.gnu.org>
746 * config/rx/rx.c (rx_rtx_costs): New function.
747 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
749 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
751 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
753 2018-02-22 Martin Liska <mliska@suse.cz>
756 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
757 Add "native" as a possible value.
759 2018-02-22 Martin Liska <mliska@suse.cz>
762 * config/i386/i386.c (ix86_option_override_internal):
763 Add "native" as a possible value for -march and -mtune.
765 2018-02-22 Jakub Jelinek <jakub@redhat.com>
768 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
769 to all type variants.
771 PR tree-optimization/84503
772 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
773 width as info->bitpos + info->bitsize - start.
774 (merged_store_group::merge_overlapping): Simplify width computation.
775 (check_no_overlap): New function.
776 (imm_store_chain_info::try_coalesce_bswap): Compute expected
777 start + width and last_order of the group, fail if check_no_overlap
779 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
780 to group if check_no_overlap fails.
782 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
784 * config/rs6000/altivec.md: Delete contraint arguments to
785 define_expand, define_split, and define_peephole2, and in
786 define_insn_and_split if always unused.
787 * config/rs6000/darwin.md: Ditto.
788 * config/rs6000/dfp.md: Ditto.
789 * config/rs6000/rs6000.md: Ditto.
790 * config/rs6000/sync.md: Ditto.
791 * config/rs6000/vector.md: Ditto.
792 * config/rs6000/vsx.md: Ditto.
794 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
796 * config/rs6000/altivec.md: Write output control strings as braced
797 blocks instead of double-quoted strings.
798 * config/rs6000/darwin.md: Ditto.
799 * config/rs6000/rs6000.md: Ditto.
800 * config/rs6000/vector.md: Ditto.
801 * config/rs6000/vsx.md: Ditto.
803 2018-02-21 Jason Merrill <jason@redhat.com>
805 PR c++/84314 - ICE with templates and fastcall attribute.
806 * attribs.c (build_type_attribute_qual_variant): Remove assert.
808 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
810 * ipa-cp.c (determine_versionability): Fix comment typos.
812 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
815 * ipa-cp.c (determine_versionability): Do not version functions caling
818 2018-02-21 Martin Liska <mliska@suse.cz>
821 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
822 Add "native" as a possible value.
823 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
824 the macro when native cpu detection is available.
826 2018-02-21 Martin Liska <mliska@suse.cz>
829 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
830 Add "native" as a possible value.
831 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
832 when native cpu detection is available.
834 2018-02-21 Jakub Jelinek <jakub@redhat.com>
835 Martin Sebor <msebor@redhat.com>
837 PR tree-optimization/84478
838 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
840 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
841 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
842 support which is conservatively correct, for 2 only stay conservative
843 for maxlen. Formatting and comment capitalization fixes. Add STRICT
844 argument to the 2 argument get_range_strlen, adjust 6 arg
845 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
847 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
848 (gimple_fold_builtin_strlen): Pass true as last argument to
851 2018-02-20 Martin Sebor <msebor@redhat.com>
854 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
855 (builtin_memref::set_base_and_offset): Same. Handle inner references.
856 (builtin_memref::builtin_memref): Factor out parts into
857 set_base_and_offset and call it.
859 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
862 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
863 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
864 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
865 search at the associated MODE_INT.
867 2018-02-20 Jeff Law <law@redhat.com>
870 PR tree-optimization/81592
872 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
873 for range data rather than using global data.
874 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
875 range data rather than using global data.
876 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
877 pass it to children as needed.
878 (struct directive::fmtresult): Similarly.
879 (struct directive::set_width): Similarly.
880 (struct directive::set_precision): Similarly.
881 (format_integer, format_directive, parse_directive): Similarly.
882 (format_none): Accept unnamed vr_values parameter.
883 (format_percent, format_floating, format_character): Similarly.
884 (format_string, format_plain): Similarly.
885 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
886 the EVRP range analyzer for range data rather than using global data.
887 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
888 gimple-ssa-evrp-analyze.h
889 (class sprintf_dom_walker): Add after_dom_children member function.
890 Add evrp_range_analyzer member.
891 (sprintf_dom_walker::before_dom_children): Call into the EVRP
892 range analyzer as needed.
893 (sprintf_dom_walker::after_dom_children): New member function.
894 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
896 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
897 (evrp_range_analyzer::pop_to_marker): Likewise.
899 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
901 PR tree-optimization/84419
902 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
903 with the required type if its current type is compatible but
906 2018-02-20 Jakub Jelinek <jakub@redhat.com>
909 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
912 2018-02-20 Martin Liska <mliska@suse.cz>
915 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
916 possible values if we don't have a hint.
918 2018-02-20 Martin Liska <mliska@suse.cz>
922 * final.c (shorten_branches): Build align_tab array with one
924 * opts.c (finish_options): Add alignment option limit check.
925 (MAX_CODE_ALIGN): Likewise.
926 (MAX_CODE_ALIGN_VALUE): Likewise.
927 * doc/invoke.texi: Document maximum allowed option value for
928 all -falign-* options.
930 2018-02-19 Jakub Jelinek <jakub@redhat.com>
933 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
934 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
935 * var-tracking.c (emit_note_insn_var_location): Remove all references
936 to NOTE_INSN_CALL_ARG_LOCATION.
937 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
938 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
939 Use copy_rtx_if_shared.
940 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
941 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
942 (dwarf2out_var_location): Remove handling of
943 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
945 * final.c (final_scan_insn): Remove all references to
946 NOTE_INSN_CALL_ARG_LOCATION.
947 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
948 before dumping final insns.
949 * except.c (emit_note_eh_region_end): Remove all references to
950 NOTE_INSN_CALL_ARG_LOCATION.
951 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
952 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
953 * config/arc/arc.c (hwloop_optimize): Likewise.
954 * config/arm/arm.c (create_fix_barrier): Likewise.
955 * config/s390/s390.c (s390_chunkify_start): Likewise.
956 * config/sh/sh.c (find_barrier): Likewise.
957 * config/i386/i386.c (rest_of_insert_endbranch,
958 ix86_seh_fixup_eh_fallthru): Likewise.
959 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
960 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
961 * config/frv/frv.c (frv_function_prologue): Likewise.
962 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
964 (note_outside_basic_block_p): Remove all references to
965 NOTE_INSN_CALL_ARG_LOCATION.
966 * gengtype.c (adjust_field_rtx_def): Likewise.
967 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
969 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
970 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
973 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
976 PR tree-optimization/84452
977 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
978 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
981 2018-02-19 Martin Liska <mliska@suse.cz>
984 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
986 2018-02-19 Martin Liska <mliska@suse.cz>
987 Richard Sandiford <richard.sandiford@linaro.org>
989 PR tree-optimization/82491
990 * gimple-fold.c (get_base_constructor): Make earlier bail out
993 2018-02-19 Carl Love <cel@us.ibm.com>
995 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
996 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
998 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
1001 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
1003 * config/rl78/rl78.md (movdf): New define expand.
1005 2018-02-19 Martin Liska <mliska@suse.cz>
1008 * doc/invoke.texi: Fix typo.
1009 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
1011 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
1013 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
1014 handle rs6000_single_float and rs6000_double_float specially for
1017 2018-02-16 Jeff Law <law@redhat.com>
1019 * config/rx/rx.c (add_pop_cfi_notes): New function.;
1022 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1025 * ipa-inline.c (inline_small_functions): Fix a typo.
1027 2018-02-16 Nathan Sidwell <nathan@acm.org>
1029 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
1031 2018-02-16 Carl Love <cel@us.ibm.com>
1033 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
1034 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
1035 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
1036 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
1037 expansion to P8V_BUILTIN_VEC_FLOAT2.
1039 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
1041 PR rtl-optimization/70023
1042 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
1043 src_regno into account.
1045 2018-02-16 Carl Love <cel@us.ibm.com>
1047 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
1048 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
1049 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
1050 * config/rs6000/rs6000.c: Remove case statements for
1051 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
1052 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
1053 and P9V_BUILTIN_VEC_VINSERT4B.
1054 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
1055 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
1056 * config/rs6000/vsx.md:
1057 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
1060 2018-02-16 Carl Love <cel@us.ibm.com>
1062 * config/rs6000/altivec.h: Add builtin names vec_extract4b
1064 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
1066 * config/rs6000/rs6000-c.c: Add the definitions for
1067 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
1068 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
1069 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
1070 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
1071 definition for insert4b and define insn *insert3b_internal.
1072 * doc/extend.texi: Add documentation for vec_extract4b.
1074 2018-02-16 Nathan Sidwell <nathan@acm.org>
1076 * doc/extend.texi (Backwards Compatibility): Mention friend
1077 injection. Note for-scope is deprecated.
1078 * doc/invoke.texi (-ffriend-injection): Deprecate.
1080 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
1082 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
1083 that moved to I2, also allow destinations that are a paradoxical
1084 subreg (instead of a normal reg).
1086 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
1089 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
1092 2018-02-16 Richard Biener <rguenther@suse.de>
1094 PR tree-optimization/84037
1095 PR tree-optimization/84016
1097 * config/i386/i386.c (ix86_builtin_vectorization_cost):
1098 Adjust vec_construct for the fact we need additional higher latency
1099 128bit inserts for AVX256 and AVX512 vector builds.
1100 (ix86_add_stmt_cost): Scale vector construction cost for
1103 2018-02-16 Richard Biener <rguenther@suse.de>
1105 PR tree-optimization/84417
1106 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
1107 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
1108 (non_rewritable_lvalue_p): Likewise, use poly-ints.
1110 2018-02-16 Martin Liska <mliska@suse.cz>
1113 * internal-fn.def (ASAN_CHECK): Set proper flags.
1114 (ASAN_MARK): Likewise.
1116 2018-02-16 Julia Koval <julia.koval@intel.com>
1118 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
1119 from PTA_CANNONLAKE.
1121 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1124 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
1125 Use ++iter rather than iter++ for std::list iterators.
1126 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
1127 defer deleting them until all nodes in the forest are processed. Do
1128 free even leaf nodes. Change to_process into auto_vec.
1131 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
1132 * vec.h (vec_default_construct): Use memset instead of placement new
1133 if BROKEN_VALUE_INITIALIZATION is defined.
1134 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
1135 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
1138 PR rtl-optimization/83723
1139 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
1140 * lra.c (lra_substitute_pseudo): Likewise. If true, use
1141 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
1143 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
1145 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
1147 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
1149 PR rtl-optimization/81443
1150 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
1151 from inner REGs to paradoxical SUBREGs.
1153 2018-02-16 Richard Biener <rguenther@suse.de>
1155 PR tree-optimization/84399
1156 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1157 For operands we can analyze at their definition make sure we can
1158 analyze them at each use as well.
1160 2018-02-16 Richard Biener <rguenther@suse.de>
1162 PR tree-optimization/84190
1163 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
1164 volatile accesses if the decl isn't volatile.
1166 2018-02-15 Jason Merrill <jason@redhat.com>
1168 PR c++/84314 - ICE with templates and fastcall attribute.
1169 * attribs.c (build_type_attribute_qual_variant): Don't clobber
1170 TYPE_CANONICAL on an existing type.
1172 2018-02-15 Jakub Jelinek <jakub@redhat.com>
1174 PR tree-optimization/84383
1175 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
1176 dstoff nor call operand_equal_p if dstbase is NULL.
1178 PR tree-optimization/84334
1179 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
1180 also a CONSTANT_CLASS_P, punt.
1182 2018-02-14 Jim Wilson <jimw@sifive.com>
1184 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
1185 first SMALL_OPERAND check. New local min_second_step. Move assert
1186 to where locals are set. Add TARGET_RVC support.
1187 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
1189 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
1191 * doc/invoke.texi: Correct -Wformat-overflow code sample.
1193 2018-02-14 Martin Sebor <msebor@redhat.com>
1195 PR tree-optimization/83698
1196 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
1197 arrays constrain the offset range to their bounds.
1198 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
1199 (builtin_access::overlap): Avoid setting the size of overlap if it's
1201 (maybe_diag_overlap): Also consider arrays when deciding what values
1202 of offsets to include in diagnostics.
1204 2018-02-14 Martin Sebor <msebor@redhat.com>
1207 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
1208 that correspond to the kind of a declaration.
1210 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
1213 * config/pa/pa.md: Load address of PIC label using the linkage table
1214 if the label is nonlocal.
1216 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
1218 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
1219 warning message if user requests -maltivec=be.
1220 * doc/invoke.texi: Document deprecation of -maltivec=be.
1222 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
1225 * config/rs6000/rs6000-c.c: Update definitions for
1226 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
1227 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
1229 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1232 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
1233 add _get_ssp intrinsics. Remove argument from
1234 __builtin_ia32_rdssp[d|q].
1235 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
1236 * config/i386/i386-builtin.def: Remove argument from
1237 __builtin_ia32_rdssp[d|q].
1238 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
1239 ix86_expand_special_args_builtin for _rdssp[d|q].
1240 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
1241 Clear register before usage.
1242 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
1243 Add documentation for new _get_ssp and _inc_ssp intrinsics.
1245 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
1247 PR tree-optimization/84357
1248 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
1249 operand 1 of an ARRAY_REF too.
1251 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
1254 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
1255 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1257 (set_of_reg): New struct.
1258 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
1259 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
1260 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1262 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
1263 Split into bitclr, bitset, bitinvert patterns if appropriate.
1264 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
1265 use rx_fuse_in_memory_bitop.
1266 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
1267 to named insn, correct maximum insn length.
1269 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
1272 * machmode.def: Define a complex mode for PARTIAL_INT.
1273 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
1275 * doc/rtl.texi: Document CSPImode.
1276 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
1278 (msp430_hard_regno_nregs_with_padding): Likewise.
1280 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
1283 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
1285 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
1287 PR rtl-optimization/84169
1288 * combine.c (try_combine): New variable split_i2i3. Set it to true if
1289 we generated a parallel as new i3 and we split that to new i2 and i3
1290 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
1291 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
1292 those to i2, not i1. Partially rewrite this scan code.
1294 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1297 * stor-layout.c (place_field): For variable length fields, adjust
1298 offset_align afterwards not just based on the field's alignment,
1299 but also on the size.
1302 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
1303 of exps and logs in the use_exp2 case.
1305 2018-02-13 Jeff Law <law@redhat.com>
1307 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
1310 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
1313 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
1317 * common.opt (gas-loc-support, gas-locview-support): New.
1318 (ginline-points, ginternal-reset-location-views): New.
1319 * doc/invoke.texi: Document them. Use @itemx where intended.
1320 (gvariable-location-views): Adjust.
1321 * target.def (reset_location_view): New.
1322 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
1323 (TARGET_RESET_LOCATION_VIEW): New.
1324 * doc/tm.texi: Rebuilt.
1325 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
1326 (dwarf2out_default_as_locview_support): New.
1327 (output_asm_line_debug_info): Use option variables.
1328 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
1329 (output_loc_list): Likewise.
1330 (add_high_low_attributes): Check option variables.
1331 Don't output entry view attribute in strict mode.
1332 (gen_inlined_subroutine_die): Check option variables.
1333 (dwarf2out_inline_entry): Likewise.
1334 (init_sections_and_labels): Likewise.
1335 (dwarf2out_early_finish): Likewise.
1336 (maybe_reset_location_view): New, from...
1337 (dwarf2out_var_location): ... here. Call it.
1338 * debug.h (dwarf2out_default_as_loc_support): Declare.
1339 (dwarf2out_default_as_locview_support): Declare.
1340 * hooks.c (hook_int_rtx_insn_0): New.
1341 * hooks.h (hook_int_rtx_insn_0): Declare.
1342 * toplev.c (process_options): Take -gas-loc-support and
1343 -gas-locview-support from dwarf2out. Enable
1344 -gvariable-location-views by default only with locview
1345 assembler support. Enable -ginternal-reset-location-views by
1346 default only if the target defines the corresponding hook.
1347 Enable -ginline-points by default if location views are
1348 enabled; force it disabled if statement frontiers are
1350 * tree-inline.c (expand_call_inline): Check option variables.
1351 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1353 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
1355 PR tree-optimization/84321
1356 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
1357 handling. Also check whether the anti-range contains any values
1358 that satisfy the mask; switch to a VR_RANGE if not.
1360 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
1363 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
1365 2018-02-13 Martin Jambor <mjambor@suse.cz>
1368 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
1369 of call statements, also set location of a load to a temporary.
1371 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
1373 * config/rl78/rl78.c (add_vector_labels): New function.
1374 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
1375 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
1376 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
1377 which checks that no arguments are passed.
1378 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
1379 * doc/extend.texi: Documentation for the new attribute.
1381 2018-02-13 Andreas Schwab <schwab@suse.de>
1383 * config/riscv/linux.h (CPP_SPEC): Define.
1385 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1388 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
1389 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
1390 OPTION_MASK_ISA_AES as first argument to def_builtin_const
1391 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
1392 instead of OPTION_MASK_ISA_PCLMUL as first argument to
1393 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
1394 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
1395 temporarily for AES and PCLMUL builtins.
1397 PR tree-optimization/84339
1398 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
1399 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
1403 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
1404 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
1405 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
1407 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
1409 * omp-simd-clone.h: New file.
1410 * omp-simd-clone.c: Include omp-simd-clone.h.
1411 (expand_simd_clones): No longer static.
1412 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
1413 cgraph.h and omp-simd-clone.h.
1414 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
1415 (vect_recog_widen_shift_pattern): Formatting fix.
1416 (vect_pattern_recog_1): Don't check optab for calls.
1419 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
1420 operands[2] into a REG before using gen_lowpart on it.
1422 2018-02-12 Jeff Law <law@redhat.com>
1425 * config/sh/sh.c (find_barrier): Consider a sibling call
1428 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
1429 successfully back substituting a reg.
1431 2018-02-12 Richard Biener <rguenther@suse.de>
1433 PR tree-optimization/84037
1434 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
1435 parameter, move visited init to caller.
1436 (vect_slp_analyze_operations): Separate cost from validity
1437 check, initialize visited once for all instances.
1438 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
1440 * tree-vect-stmts.c (vect_model_simple_cost): Make early
1442 (vect_model_promotion_demotion_cost): Likewise.
1443 (vectorizable_bswap): Guard cost modeling with !slp_node
1444 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
1446 (vectorizable_call): Likewise.
1447 (vectorizable_conversion): Likewise.
1448 (vectorizable_assignment): Likewise.
1449 (vectorizable_shift): Likewise.
1450 (vectorizable_operation): Likewise.
1451 (vectorizable_store): Likewise.
1452 (vectorizable_load): Likewise.
1453 (vectorizable_condition): Likewise.
1454 (vectorizable_comparison): Likewise.
1456 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
1459 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
1460 (ASAN_MARK): Fix fnspec to account for return value, change pointer
1461 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
1463 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1466 * params.def (inline-min-speedup): Increase from 8 to 15.
1467 (max-inline-insns-auto): Decrease from 40 to 30.
1468 * ipa-split.c (consider_split): Add some buffer for function to
1469 be considered inlining candidate.
1470 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
1473 2018-02-12 Richard Biener <rguenther@suse.de>
1475 PR tree-optimization/84037
1476 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
1477 matched stmts if we cannot swap the non-matched ones.
1479 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1481 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
1482 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
1483 _mm_maskz_scalef_round_ss): New intrinsics.
1484 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
1485 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
1486 __builtin_ia32_scalefss_round): Remove.
1487 (__builtin_ia32_scalefsd_mask_round,
1488 __builtin_ia32_scalefss_mask_round): New intrinsics.
1489 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
1490 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
1491 ((match_operand:VF_128 2 "<round_nimm_predicate>"
1492 "<round_constraint>")): Changed to ...
1493 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
1494 "<round_scalar_constraint>")): ... this.
1495 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
1496 %0, %1, %2<round_op3>}"): Changed to ...
1497 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
1498 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
1499 %2<round_scalar_mask_op3>}"): ... this.
1500 * config/i386/subst.md (round_scalar_nimm_predicate): New.
1502 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1504 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
1505 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
1506 (_mm_maskz_sqrt_round_ss): New intrinsics.
1507 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
1508 (__builtin_ia32_sqrtsd_mask_round)
1509 (__builtin_ia32_sqrtss_mask_round): New builtins.
1510 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
1511 (__builtin_ia32_sqrtss_round): Remove.
1512 (__builtin_ia32_sqrtsd_mask_round)
1513 (__builtin_ia32_sqrtss_mask_round): New builtins.
1514 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
1515 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
1516 ((match_operand:VF_128 1 "vector_operand"
1517 "xBm,<round_constraint>")): Changed to ...
1518 ((match_operand:VF_128 1 "vector_operand"
1519 "xBm,<round_scalar_constraint>")): ... this.
1520 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
1521 %0, %2, %<iptr>1<round_op3>}): Changed to ...
1522 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
1523 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
1524 %<iptr>1<round_scalar_mask_op3>}): ... this.
1525 ((set_attr "prefix" "<round_prefix>")): Changed to ...
1526 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
1528 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
1531 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
1532 Cast vec_cmpeq result to correct type.
1533 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
1534 Cast vec_cmpgt result to correct type.
1536 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
1538 * final.c (final_scan_insn_1): Renamed from...
1539 (final_scan_insn): ... this. New wrapper, to recover
1540 seen from the outermost call in recursive ones.
1541 * config/sparc/sparc.c (output_return): Drop seen from call.
1542 (output_sibcall): Likewise.
1543 * config/visium/visium.c (output_branch): Likewise.
1545 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
1547 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
1550 2018-02-10 Alan Modra <amodra@gmail.com>
1553 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
1554 Specify LR as an input.
1556 2018-02-10 Jakub Jelinek <jakub@redhat.com>
1559 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
1560 remove_member_access_dummy_vars): New functions.
1561 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
1562 lower_omp_1, execute_lower_omp): Use them.
1564 PR rtl-optimization/84308
1565 * shrink-wrap.c (spread_components): Release todo vector.
1567 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
1569 PR rtl-optimization/57193
1570 * ira-color.c (struct allocno_color_data): Add member
1571 conflict_allocno_hard_prefs.
1572 (update_conflict_allocno_hard_prefs): New.
1573 (bucket_allocno_compare_func): Add a preference based on
1574 conflict_allocno_hard_prefs.
1575 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
1576 (color_allocnos): Remove a dead code. Initiate
1577 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
1579 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1582 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
1583 constraint from =wa to wa. Avoid a subreg on the output operand,
1584 instead use a pseudo and subreg it in a move.
1585 (p9_xxbrd_<mode>): Changed to ...
1586 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
1587 (p9_xxbrd_v2df): New expander.
1588 (p9_xxbrw_<mode>): Changed to ...
1589 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
1590 (p9_xxbrw_v4sf): New expander.
1592 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1594 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
1596 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
1599 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
1600 multiply in 32-bit mode.
1601 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
1602 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
1605 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1607 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
1608 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
1609 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
1610 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
1612 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
1615 * dwarf2out.c (is_trivial_indirect_ref): New function.
1616 (dwarf2out_late_global_decl): Do not generate a location
1617 attribute for variables that have a non-trivial DECL_VALUE_EXPR
1618 and that are not defined in the current unit.
1620 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1622 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
1623 instead of a libcall for UNORDERED.
1625 2018-02-09 Tamar Christina <tamar.christina@arm.com>
1628 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
1629 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
1631 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1634 * config/s390/s390.c (s390_set_current_function): Invoke
1635 s390_indirect_branch_settings also if fndecl didn't change.
1637 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1639 * config/rs6000/rs6000.md (blockage): Set length to zero.
1641 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1643 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
1645 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1648 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
1649 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
1653 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
1654 PARALLEL incoming that failed vt_get_decl_and_offset check.
1657 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
1658 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
1660 (get_variable_section): For decls in named .bss* sections pass true as
1661 second argument to bss_initializer_p.
1663 2018-02-09 Marek Polacek <polacek@redhat.com>
1664 Jakub Jelinek <jakub@redhat.com>
1667 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
1668 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
1669 Sync some changes from cxx_fold_indirect_ref.
1671 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1673 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
1675 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
1676 (BLOCK_INLINE_ENTRY_LABEL): New.
1677 (dwarf2out_var_location): Disregard inline entry markers.
1678 (inline_entry_data): New struct.
1679 (inline_entry_data_hasher): New hashtable type.
1680 (inline_entry_data_hasher::hash): New.
1681 (inline_entry_data_hasher::equal): New.
1682 (inline_entry_data_table): New variable.
1683 (add_high_low_attributes): Add DW_AT_entry_pc and
1684 DW_AT_GNU_entry_view attributes if a pending entry is found
1685 in inline_entry_data_table. Add old entry_pc attribute only
1686 if debug nonbinding markers are disabled.
1687 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
1688 markers are enabled.
1689 (block_within_block_p, dwarf2out_inline_entry): New.
1690 (dwarf2out_finish): Check that no entries remained in
1691 inline_entry_data_table.
1692 * final.c (reemit_insn_block_notes): Handle inline entry notes.
1693 (final_scan_insn, notice_source_line): Likewise.
1694 (rest_of_clean_state): Skip inline entry markers.
1695 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
1697 * gimple.c (gimple_build_debug_inline_entry): New.
1698 * gimple.h (enum gimple_debug_subcode): Add
1699 GIMPLE_DEBUG_INLINE_ENTRY.
1700 (gimple_build_debug_inline_entry): Declare.
1701 (gimple_debug_inline_entry_p): New.
1702 (gimple_debug_nonbind_marker_p): Adjust.
1703 * insn-notes.def (INLINE_ENTRY): New.
1704 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
1705 inline entry marker notes.
1706 (print_insn): Likewise.
1707 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
1708 (INSN_DEBUG_MARKER_KIND): Likewise.
1709 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
1710 * tree-inline.c (expand_call_inline): Build and insert
1711 debug_inline_entry stmt.
1712 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
1713 inline entry blocks early, if nonbind markers are enabled.
1714 (dump_scope_block): Dump fragment info.
1715 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
1716 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
1717 (gimple_build_debug_inline_entry): New.
1718 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
1719 Enable/disable inline entry points too.
1720 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
1721 (DEBUG_INSN): Describe inline entry markers.
1723 * common.opt (gvariable-location-views): New.
1724 (gvariable-location-views=incompat5): New.
1725 * config.in: Rebuilt.
1726 * configure: Rebuilt.
1727 * configure.ac: Test assembler for view support.
1728 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
1729 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
1730 * dwarf2out.c (var_loc_view): New typedef.
1731 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
1732 (dwarf2out_locviews_in_attribute): New.
1733 (dwarf2out_locviews_in_loclist): New.
1734 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
1735 (enum dw_line_info_opcode): Add LI_adv_address.
1736 (struct dw_line_info_table): Add view.
1737 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
1738 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
1739 (zero_view_p): New variable.
1740 (ZERO_VIEW_P): New macro.
1741 (output_asm_line_debug_info): New.
1742 (struct var_loc_node): Add view.
1743 (add_AT_view_list, AT_loc_list): New.
1744 (add_var_loc_to_decl): Add view param. Test it against last.
1745 (new_loc_list): Add view params. Record them.
1746 (AT_loc_list_ptr): Handle loc and view lists.
1747 (view_list_to_loc_list_val_node): New.
1748 (print_dw_val): Handle dw_val_class_view_list.
1749 (size_of_die): Likewise.
1750 (value_format): Likewise.
1751 (loc_list_has_views): New.
1752 (gen_llsym): Set vl_symbol too.
1753 (maybe_gen_llsym, skip_loc_list_entry): New.
1754 (dwarf2out_maybe_output_loclist_view_pair): New.
1755 (output_loc_list): Output view list or entries too.
1756 (output_view_list_offset): New.
1757 (output_die): Handle dw_val_class_view_list.
1758 (output_dwarf_version): New.
1759 (output_compilation_unit_header): Use it.
1760 (output_skeleton_debug_sections): Likewise.
1761 (output_rnglists, output_line_info): Likewise.
1762 (output_pubnames, output_aranges): Update version comments.
1763 (output_one_line_info_table): Output view numbers in asm comments.
1764 (dw_loc_list): Determine current endview, pass it to new_loc_list.
1765 Call maybe_gen_llsym.
1766 (loc_list_from_tree_1): Adjust.
1767 (add_AT_location_description): Create view list attribute if
1768 needed, check it's absent otherwise.
1769 (convert_cfa_to_fb_loc_list): Adjust.
1770 (maybe_emit_file): Call output_asm_line_debug_info for test.
1771 (dwarf2out_var_location): Reset views as needed. Precompute
1772 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
1773 attribute. Set view.
1774 (new_line_info_table): Reset next view.
1775 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
1776 (dwarf2out_source_line): Likewise. Output view resets and labels to
1777 the assembler, or select appropriate line info opcodes.
1778 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
1779 (optimize_string_length): Catch it. Adjust.
1780 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
1781 dw_val_class_view_list, and remove it if no longer needed.
1782 (hash_loc_list): Hash view numbers.
1783 (loc_list_hasher::equal): Compare them.
1784 (optimize_location_lists): Check whether a view list symbol is
1785 needed, and whether the locview attribute is present, and
1786 whether they match. Remove the locview attribute if no longer
1788 (index_location_lists): Call skip_loc_list_entry for test.
1789 (dwarf2out_finish): Call output_asm_line_debug_info for test.
1790 Use output_dwarf_version.
1791 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
1792 (struct dw_val_node): Add val_view_list.
1793 * final.c (SEEN_NEXT_VIEW): New.
1794 (set_next_view_needed): New.
1795 (clear_next_view_needed): New.
1796 (maybe_output_next_view): New.
1797 (final_start_function): Rename to...
1798 (final_start_function_1): ... this. Take pointer to FIRST,
1799 add SEEN parameter. Emit param bindings in the initial view.
1800 (final_start_function): Reintroduce SEEN-less interface.
1801 (final): Rename to...
1802 (final_1): ... this. Take SEEN parameter. Output final pending
1803 next view at the end.
1804 (final): Reintroduce seen-less interface.
1805 (final_scan_insn): Output pending next view before switching
1806 sections or ending a block. Mark the next view as needed when
1807 outputting variable locations. Notify debug backend of section
1808 changes, and of location view changes.
1809 (rest_of_handle_final): Adjust.
1810 * toplev.c (process_options): Autodetect value for debug variable
1811 location views option. Warn on incompat5 without -gdwarf-5.
1812 * doc/invoke.texi (gvariable-location-views): New.
1813 (gvariable-location-views=incompat5): New.
1814 (gno-variable-location-views): New.
1816 2018-02-08 David Malcolm <dmalcolm@redhat.com>
1818 PR tree-optimization/84136
1819 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
1820 that the result of find_edge is non-NULL.
1822 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1825 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
1826 storing integer register in SImode. Fix cost of 256 and 512
1827 byte aligned SSE register store.
1829 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1831 * config/i386/i386.c (ix86_multiplication_cost): Fix
1832 multiplication cost for TARGET_AVX512DQ.
1834 2018-02-08 Marek Polacek <polacek@redhat.com>
1836 PR tree-optimization/84238
1837 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
1840 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1842 PR tree-optimization/84265
1843 * tree-vect-stmts.c (vectorizable_store): Don't treat
1844 VMAT_CONTIGUOUS accesses as grouped.
1845 (vectorizable_load): Likewise.
1847 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1849 PR tree-optimization/81635
1850 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
1851 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
1852 (test_round_for_mask): New functions.
1853 (wide_int_cc_tests): Call test_round_for_mask.
1854 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
1855 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
1856 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
1857 range returned by get_range_info.
1859 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1862 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
1863 * symtab.c: Include builtins.h
1864 (symtab_node::output_to_lto_symbol_table_p): Move here
1865 from lto-streamer-out.c:output_symbol_p.
1866 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
1867 (output_symbol_p): Move all logic to symtab.c
1868 (produce_symtab): Update.
1870 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1872 * config/s390/s390-opts.h (enum indirect_branch): Define.
1873 * config/s390/s390-protos.h (s390_return_addr_from_memory)
1874 (s390_indirect_branch_via_thunk)
1875 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
1876 (enum s390_indirect_branch_type): Define.
1877 * config/s390/s390.c (struct s390_frame_layout, struct
1878 machine_function): Remove.
1879 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
1880 (indirect_branch_table_label_no, indirect_branch_table_name):
1882 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
1883 (enum s390_indirect_branch_option): Define.
1884 (s390_return_addr_from_memory): New function.
1885 (s390_handle_string_attribute): New function.
1886 (s390_attribute_table): Add new attribute handler.
1887 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
1888 (s390_indirect_branch_via_thunk): New function.
1889 (s390_indirect_branch_via_inline_thunk): New function.
1890 (s390_function_ok_for_sibcall): When jumping via thunk disallow
1891 sibling call optimization for non z10 compiles.
1892 (s390_emit_call): Force indirect branch target to be a single
1893 register. Add r1 clobber for non-z10 compiles.
1894 (s390_emit_epilogue): Emit return jump via return_use expander.
1895 (s390_reorg): Handle JUMP_INSNs as execute targets.
1896 (s390_option_override_internal): Perform validity checks for the
1897 new command line options.
1898 (s390_indirect_branch_attrvalue): New function.
1899 (s390_indirect_branch_settings): New function.
1900 (s390_set_current_function): Invoke s390_indirect_branch_settings.
1901 (s390_output_indirect_thunk_function): New function.
1902 (s390_code_end): Implement target hook.
1903 (s390_case_values_threshold): Implement target hook.
1904 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
1906 * config/s390/s390.h (struct s390_frame_layout)
1907 (struct machine_function): Move here from s390.c.
1908 (TARGET_INDIRECT_BRANCH_NOBP_RET)
1909 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
1910 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
1911 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
1912 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
1913 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
1914 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
1915 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
1916 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
1917 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
1918 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
1919 (mnemonic attribute): Add values which aren't recognized
1921 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
1922 pattern for branch conversion. Fix mnemonic attribute.
1923 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
1924 indirect branch via thunk if requested.
1925 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
1926 ("*indirect_jump"): Disable for branch conversion using out of
1928 ("indirect_jump_via_thunk<mode>_z10")
1929 ("indirect_jump_via_thunk<mode>")
1930 ("indirect_jump_via_inlinethunk<mode>_z10")
1931 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
1932 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
1933 ("casesi_jump_via_inlinethunk<mode>_z10")
1934 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
1935 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
1936 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
1937 ("*indirect2_jump"): Disable for branch conversion.
1938 ("casesi_jump"): Turn into expander and expand patterns for branch
1940 ("return_use"): New expander.
1941 ("*return"): Emit return via thunk and rename it to ...
1942 ("*return<mode>"): ... this one.
1943 * config/s390/s390.opt: Add new options and and enum for the
1946 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1948 * lra-constraints.c (match_reload): Unconditionally use
1949 gen_lowpart_SUBREG, rather than selecting between that
1950 and equivalent gen_rtx_SUBREG code.
1952 2018-02-08 Richard Biener <rguenther@suse.de>
1954 PR tree-optimization/84233
1955 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
1956 changed flag instead of boguously re-using phi_inserted.
1958 2018-02-08 Martin Jambor <mjambor@suse.cz>
1960 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
1961 static local variables.
1963 2018-02-08 Richard Biener <rguenther@suse.de>
1965 PR tree-optimization/84278
1966 * tree-vect-stmts.c (vectorizable_store): When looking for
1967 smaller vector types to perform grouped strided loads/stores
1968 make sure the mode is supported by the target.
1969 (vectorizable_load): Likewise.
1971 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1973 * config/aarch64/aarch64.c (aarch64_components_for_bb):
1974 Increase LDP/STP opportunities by adding adjacent callee-saves.
1976 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1978 PR rtl-optimization/84068
1979 PR rtl-optimization/83459
1980 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
1982 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
1984 PR tree-optimization/84224
1985 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
1986 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
1989 2018-02-07 Iain Sandoe <iain@codesourcery.com>
1992 * config/rs6000/altivec.md (*restore_world): Remove LR use.
1993 * config/rs6000/predicates.md (restore_world_operation): Adjust op
1994 count, remove one USE.
1996 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
1998 * doc/install.texi (Configuration): Document the
1999 --with-long-double-format={ibm,ieee} PowerPC configuration
2003 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
2004 Convert from define_expand to be define_insn_and_split. Rework
2005 float/double/_Float128 conversions to QI/HI/SImode to work with
2006 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
2007 conversions to QI/HImode types did a store and then a load to
2008 truncate the value. For conversions to VSX registers, don't split
2009 the insn, instead emit the code directly. Use the code iterator
2010 any_fix to combine signed and unsigned conversions.
2011 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
2012 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
2013 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2014 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
2015 (fix_<mode>di2_hw): Likewise.
2016 (fixuns_<mode>di2_hw): Likewise.
2017 (fix_<mode>si2_hw): Likewise.
2018 (fixuns_<mode>si2_hw): Likewise.
2019 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
2020 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2021 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
2022 fix<uns>_trunc<SFDF:mode>si2_p8.
2023 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
2025 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
2026 (fix<uns>_<mode>_mem): Likewise.
2027 (fctiw<u>z_<mode>_mem): Likewise.
2028 (fix<uns>_<mode>_mem): Likewise.
2029 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
2030 the register allocator from doing a direct move to the GPRs to do
2031 a store, and instead use the ISA 3.0 store byte/half-word from
2032 vector register instruction. For IEEE 128-bit floating point,
2033 also optimize stores of 32-bit ints.
2034 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
2036 2018-02-07 Alan Hayward <alan.hayward@arm.com>
2038 * genextract.c (push_pathstr_operand): New function to support
2040 (walk_rtx): Call push_pathstr_operand.
2041 (print_path): Support [a-zA-Z].
2043 2018-02-07 Richard Biener <rguenther@suse.de>
2045 PR tree-optimization/84037
2046 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
2047 (cse_and_gimplify_to_preheader): Declare.
2048 (vect_get_place_in_interleaving_chain): Likewise.
2049 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2051 (_loop_vec_info::~_loop_vec_info): Delete it.
2052 (cse_and_gimplify_to_preheader): New function.
2053 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
2054 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
2055 (vectorizable_load): Likewise. For grouped stores always base
2056 the IV on the first element.
2057 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
2058 condition before gimplifying.
2060 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2062 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
2063 *DIV_EXPR and *MOD_EXPR.
2065 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
2068 * config/i386/i386.c (ix86_option_override_internal): Mask out
2069 the CF_SET bit when checking -fcf-protection.
2071 2018-02-07 Tom de Vries <tom@codesourcery.com>
2074 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
2077 2018-02-07 Richard Biener <rguenther@suse.de>
2079 PR tree-optimization/84204
2080 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2083 PR tree-optimization/84205
2084 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
2085 special-case isl_ast_op_zdiv_r.
2087 PR tree-optimization/84223
2088 * graphite-scop-detection.c (gather_bbs::before_dom_children):
2089 Only add conditions from within the region.
2090 (gather_bbs::after_dom_children): Adjust.
2092 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
2095 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
2096 * config/avr/avr.md: Only post-reload split REG-REG moves if
2097 either register is GENERAL_REG_P.
2099 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2101 PR tree-optimization/84235
2102 * tree-ssa-scopedtables.c
2103 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
2104 if the subtraction is performed in floating point type where NaNs are
2105 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
2106 build 1. Formatting fix.
2108 2018-02-06 Jakub Jelinek <jakub@redhat.com>
2111 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
2112 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
2113 and skip it regardless of bb boundaries. Use CALL_P macro,
2114 don't test INSN_P (insn) together with CALL_P or JUMP_P check
2115 unnecessarily, formatting fix.
2117 2018-02-06 Michael Collison <michael.collison@arm.com>
2119 * config/arm/thumb2.md:
2120 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
2121 (*thumb_mov_notscc): Ditto.
2123 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2126 * config/rs6000/rs6000.md (su code attribute): Use "u" for
2127 unsigned_fix, not "s".
2129 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2131 * configure.ac (gcc_fn_eh_frame_ro): New function.
2132 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
2133 correct .eh_frame permissions.
2134 * configure: Regenerate.
2136 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
2138 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
2141 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2143 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2144 Display warning message for -mno-speculate-indirect-jumps.
2146 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
2148 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
2150 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
2152 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
2154 PR tree-optimization/84225
2155 * tree-eh.c (find_trapping_overflow): Only call
2156 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
2158 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2161 * config/i386/i386.c: Reimplement the check of possible options
2162 -mibt/-mshstk conbination. Change error messages.
2163 * doc/invoke.texi: Fix a typo: remove extra '='.
2165 2018-02-06 Marek Polacek <polacek@redhat.com>
2167 PR tree-optimization/84228
2168 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
2170 2018-02-06 Tamar Christina <tamar.christina@arm.com>
2173 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
2174 emitted arch directives.
2175 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
2176 __ARM_FEATURE_COPROC before changing architectures.
2178 2018-02-06 Richard Biener <rguenther@suse.de>
2180 * config/i386/i386.c (print_reg): Fix typo.
2181 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
2183 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
2185 * configure: Regenerate.
2187 2018-02-05 Martin Sebor <msebor@redhat.com>
2189 PR tree-optimization/83369
2190 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
2193 2018-02-05 Martin Liska <mliska@suse.cz>
2195 * doc/invoke.texi: Cherry-pick upstream r323995.
2197 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
2199 * ira.c (ira_init_register_move_cost): Adjust comment.
2201 2018-02-05 Martin Liska <mliska@suse.cz>
2203 PR gcov-profile/84137
2204 * doc/gcov.texi: Fix typo in documentation.
2206 2018-02-05 Martin Liska <mliska@suse.cz>
2208 PR gcov-profile/83879
2209 * doc/gcov.texi: Document necessity of --dynamic-list-data when
2210 using dlopen functionality.
2212 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
2214 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
2215 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
2216 _mm_maskz_range_ss, _mm_mask_range_round_ss,
2217 _mm_maskz_range_round_ss): New intrinsics.
2218 (__builtin_ia32_rangesd128_round)
2219 (__builtin_ia32_rangess128_round): Remove.
2220 (__builtin_ia32_rangesd128_mask_round,
2221 __builtin_ia32_rangess128_mask_round): New builtins.
2222 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
2223 __builtin_ia32_rangess128_round): Remove.
2224 (__builtin_ia32_rangesd128_mask_round,
2225 __builtin_ia32_rangess128_mask_round): New builtins.
2226 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
2227 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
2228 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
2229 "<round_saeonly_constraint>")): Changed to ...
2230 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
2231 "<round_saeonly_scalar_constraint>")): ... this.
2232 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
2233 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
2234 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
2235 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
2236 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
2238 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
2240 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
2242 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
2243 Remove all values except native, 8540 and 8548.
2245 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
2247 * config/i386/i386.c (ix86_output_function_return): Pass
2248 INVALID_REGNUM, instead of -1, as invalid register number to
2249 indirect_thunk_name and output_indirect_thunk.
2251 2018-02-02 Julia Koval <julia.koval@intel.com>
2253 * config.gcc: Add -march=icelake.
2254 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
2255 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
2256 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
2257 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
2258 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
2259 (processor_target_table): Add icelake.
2260 (ix86_option_override_internal): Handle new PTAs.
2261 (get_builtin_code_for_version): Handle icelake.
2262 (M_INTEL_COREI7_ICELAKE): New.
2263 (fold_builtin_cpu): Handle icelake.
2264 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
2265 * doc/invoke.texi: Add -march=icelake.
2267 2018-02-02 Julia Koval <julia.koval@intel.com>
2269 * config/i386/i386.c (ix86_option_override_internal): Change flags type
2270 to wide_int_bitmask.
2271 * wide-int-bitmask.h: New.
2273 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2276 * config/i386/i386.md: Replace Pmode with word_mode in
2277 builtin_setjmp_setup and builtin_longjmp to support x32.
2279 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
2283 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
2285 (rs6000_supported_cpu_names): New static variable.
2286 (linux_cpu_translation_table): Likewise.
2287 (elf_platform) <cpu>: Define new static variable and use it.
2288 Translate kernel AT_PLATFORM name to canonical name if needed.
2289 Error if platform name is unknown.
2291 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
2294 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
2296 2018-02-01 Jeff Law <law@redhat.com>
2299 * config/i386/i386.c (release_scratch_register_on_entry): Add new
2300 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
2301 the scratch if RELEASE_VIA_POP is false.
2302 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
2303 If we have to save a temporary register, decrement SIZE appropriately.
2304 Pass new arguments to release_scratch_register_on_entry.
2305 (ix86_adjust_stack_and_probe): Likewise.
2306 (ix86_emit_probe_stack_range): Pass new arguments to
2307 release_scratch_register_on_entry.
2309 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
2311 PR rtl-optimization/84157
2312 * combine.c (change_zero_ext): Use REG_P predicate in
2313 front of HARD_REGISTER_P predicate.
2315 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
2317 * config/avr/avr.c (avr_option_override): Move disabling of
2318 -fdelete-null-pointer-checks to...
2319 * common/config/avr/avr-common.c (avr_option_optimization_table):
2322 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2324 PR tree-optimization/81635
2325 * tree-data-ref.c (split_constant_offset_1): For types that
2326 wrap on overflow, try to use range info to prove that wrapping
2329 2018-02-01 Renlin Li <renlin.li@arm.com>
2332 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
2334 (aarch64_register_move_cost): Likewise.
2335 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
2337 (REG_CLASS_NAMES): Likewise.
2338 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
2339 TAILCALL_ADDR_REGS. Remove IP registers.
2340 * config/aarch64/aarch64.md (Ucs): Update register constraint.
2342 2018-02-01 Richard Biener <rguenther@suse.de>
2344 * domwalk.h (dom_walker::dom_walker): Add additional constructor
2345 for specifying RPO order and allow NULL for that.
2346 * domwalk.c (dom_walker::dom_walker): Likewise.
2347 (dom_walker::walk): Handle NULL RPO order.
2348 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
2350 (rewrite_update_dom_walker): Likewise.
2351 (mark_def_dom_walker): Likewise.
2353 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2355 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
2356 (aarch64_maybe_expand_sve_subreg_move): Declare.
2357 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
2358 * config/aarch64/predicates.md (aarch64_any_register_operand): New
2360 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
2361 that are semantically a reverse operation.
2362 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
2363 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
2364 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
2366 (aarch64_can_change_mode_class): For big-endian, forbid changes
2367 between two SVE modes if they have different element sizes.
2369 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2371 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
2372 the TImode handling for big-endian targets.
2374 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2376 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
2377 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
2379 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
2380 Remove BSWAP handing for big-endian targets and use the form of
2381 LD1RQ appropariate for the mode.
2383 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2385 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
2386 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
2389 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2392 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
2393 check for operands that need to go through aarch64_sve_reload_be.
2395 2018-02-01 Jakub Jelinek <jakub@redhat.com>
2397 PR tree-optimization/81661
2398 PR tree-optimization/84117
2399 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
2400 * tree-eh.c: Include gimplify.h.
2401 (find_trapping_overflow, replace_trapping_overflow,
2402 rewrite_to_non_trapping_overflow): New functions.
2403 * tree-vect-loop.c: Include tree-eh.h.
2404 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
2405 * tree-data-ref.c: Include tree-eh.h.
2406 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
2408 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
2410 PR rtl-optimization/84123
2411 * combine.c (change_zero_ext): Check if hard register satisfies
2412 can_change_dest_mode before calling gen_lowpart_SUBREG.
2414 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
2417 * ira.c (ira_init_register_move_cost): Remove assert.
2419 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2421 PR rtl-optimization/84071
2422 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
2423 * doc/tm.texi: Regenerate.
2425 2018-01-31 Richard Biener <rguenther@suse.de>
2427 PR tree-optimization/84132
2428 * tree-data-ref.c (analyze_miv_subscript): Properly
2429 check whether evolution_function_is_affine_multivariate_p
2430 before calling gcd_of_steps_may_divide_p.
2432 2018-01-31 Julia Koval <julia.koval@intel.com>
2435 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
2436 * config/i386/i386.md (rdpid_rex64) New.
2437 (rdpid): Make 32bit only.
2439 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
2442 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
2443 an IDENTIFIER_NODE for FUNCTION_TYPE's.
2445 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2448 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
2450 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
2452 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2454 PR rtl-optimization/84071
2455 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
2456 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
2458 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2460 * config/arc/arc.c (arc_handle_aux_attribute): New function.
2461 (arc_attribute_table): Add 'aux' attribute.
2462 (arc_in_small_data_p): Consider aux like variables.
2463 (arc_is_aux_reg_p): New function.
2464 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
2465 (arc_get_aux_arg): New function.
2466 (prepare_move_operands): Handle aux-register access.
2467 (arc_handle_aux_attribute): New function.
2468 * doc/extend.texi (ARC Variable attributes): Add subsection.
2470 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2472 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
2473 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
2474 (arc_attribute_table): Add 'uncached' attribute.
2475 (arc_print_operand): Print '.di' flag for uncached memory
2477 (arc_in_small_data_p): Do not consider for small data the uncached
2479 (arc_is_uncached_mem_p): New function.
2480 * config/arc/predicates.md (compact_store_memory_operand): Check
2481 for uncached memory accesses.
2482 (nonvol_nonimm_operand): Likewise.
2483 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
2485 2018-01-31 Jakub Jelinek <jakub@redhat.com>
2488 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
2489 falign-loops=): Add Optimization flag.
2491 2018-01-30 Jeff Law <law@redhat.com>
2494 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
2495 INT_REGISTERS_SAVED. Check it prior to calling
2496 get_scratch_register_on_entry.
2497 (ix86_adjust_stack_and_probe): Similarly.
2498 (ix86_emit_probe_stack_range): Similarly.
2499 (ix86_expand_prologue): Corresponding changes.
2501 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2504 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
2505 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
2507 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
2510 * lra-constraints.c (curr_insn_transform): Process AND in the
2513 2018-01-30 Jakub Jelinek <jakub@redhat.com>
2515 PR rtl-optimization/83986
2516 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
2517 dependence against last_pending_memory_flush in addition to
2520 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
2522 PR tree-optimization/81611
2523 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
2526 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2529 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
2532 2018-01-30 Richard Biener <rguenther@suse.de>
2533 Jakub Jelinek <jakub@redhat.com>
2535 PR tree-optimization/84111
2536 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
2537 inner loops added during recursion, as they don't have up-to-date
2540 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2543 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
2544 (can_inline_edge_by_limits_p): ... here.
2545 (can_early_inline_edge_p, check_callers,
2546 update_caller_keys, update_callee_keys, recursive_inlining,
2547 add_new_edges_to_heap, speculation_useful_p,
2548 inline_small_functions,
2549 inline_small_functions, flatten_function,
2550 inline_to_all_callers_1): Update.
2552 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2554 * profile-count.c (profile_count::combine_with_ipa_count): Handle
2557 2018-01-30 Richard Biener <rguenther@suse.de>
2559 PR tree-optimization/83008
2560 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
2561 invariant and constant vector uses in stmts when they need
2564 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2567 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
2568 * configure: Regenerate.
2570 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2572 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
2574 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
2575 Use gen_rtx_REG rather than gen_lowpart.
2577 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2579 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
2580 rather than 0 when creating partial subregs.
2582 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2584 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
2587 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
2590 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
2591 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
2592 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
2593 flags. This restores the settings used before the 2017-07-24.
2594 Turning off pre increment/decrement/modify allows IVOPTS to
2595 optimize DF/SF loops where the index is an int.
2597 2018-01-29 Richard Biener <rguenther@suse.de>
2598 Kelvin Nilsen <kelvin@gcc.gnu.org>
2601 * tree-vect-stmts.c (vectorizable_call): Don't call
2602 targetm.vectorize_builtin_md_vectorized_function if callee is
2605 2018-01-22 Carl Love <cel@us.ibm.com>
2607 * doc/extend.tex: Fix typo in second arg in
2608 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
2610 2018-01-29 Richard Biener <rguenther@suse.de>
2612 PR tree-optimization/84086
2613 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
2614 (flush_ssaname_freelist): When SSA names were released reset
2615 the SCEV hash table.
2617 2018-01-29 Richard Biener <rguenther@suse.de>
2619 PR tree-optimization/84057
2620 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
2621 removed paths when removing edges.
2623 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
2625 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
2626 -mfunction-return=@var{choice}.
2628 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
2631 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
2633 (layout::print_source_line): Likewise.
2634 (test_get_line_width_without_trailing_whitespace): Add test cases.
2636 2018-01-27 Jakub Jelinek <jakub@redhat.com>
2639 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
2642 2018-01-26 Jim Wilson <jimw@sifive.com>
2644 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
2646 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
2649 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2651 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
2652 and CMP + SUB-immediate -> SUBS.
2654 2018-01-26 Martin Sebor <msebor@redhat.com>
2656 PR tree-optimization/83896
2657 * tree-ssa-strlen.c (get_string_len): Rename...
2658 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
2659 Avoid assuming length is constant.
2660 (handle_char_store): Use HOST_WIDE_INT for string length.
2662 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
2665 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
2666 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
2668 2018-01-26 Richard Biener <rguenther@suse.de>
2670 PR rtl-optimization/84003
2671 * dse.c (record_store): Only record redundant stores when
2672 the earlier store aliases at least all accesses the later one does.
2674 2018-01-26 Jakub Jelinek <jakub@redhat.com>
2676 PR rtl-optimization/83985
2677 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
2678 REG_CFA_RESTORE insns.
2679 (delete_unmarked_insns): Don't ignore separate shrink wrapping
2680 REG_CFA_RESTORE insns here.
2683 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
2684 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
2686 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2688 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
2689 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
2690 (arc_init): Likewise.
2691 (arc_override_options): Likewise.
2692 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
2694 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
2696 * config/arc/arc.h (TARGET_DBNZ): Define.
2697 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
2698 properly set the tune attribute.
2699 (dbnz): Use TARGET_DBNZ guard.
2700 * config/arc/arc.opt (mtune): Add core3 option.
2702 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2704 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
2705 recognize new pic like addresses.
2706 (arc_delegitimize_address): Clean up.
2708 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2710 * config/arc/arc-arches.def: Option mrf16 valid for all
2712 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
2713 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
2714 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
2715 * config/arc/arc-tables.opt: Regenerate.
2716 * config/arc/arc.c (arc_conditional_register_usage): Handle
2717 reduced register file case.
2718 (arc_file_start): Set must have build attributes.
2719 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
2721 * config/arc/arc.opt (mrf16): Add new option.
2722 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
2723 * config/arc/genmultilib.awk: Handle new mrf16 option.
2724 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
2725 * config/arc/t-multilib: Regenerate.
2726 * doc/invoke.texi (ARC Options): Document mrf16 option.
2728 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2730 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
2731 * config/arc/arc.c (arc_handle_secure_attribute): New function.
2732 (arc_attribute_table): Add 'secure_call' attribute.
2733 (arc_print_operand): Print secure call operand.
2734 (arc_function_ok_for_sibcall): Don't optimize tail calls when
2736 (arc_is_secure_call_p): New function. * config/arc/arc.md
2737 (call_i): Add support for sjli instruction.
2738 (call_value_i): Likewise.
2739 * config/arc/constraints.md (Csc): New constraint.
2741 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2742 John Eric Martin <John.Martin@emmicro-us.com>
2744 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
2745 * config/arc/arc.c (_arc_jli_section): New struct.
2746 (arc_jli_section): New type.
2747 (rc_jli_sections): New static variable.
2748 (arc_handle_jli_attribute): New function.
2749 (arc_attribute_table): Add jli_always and jli_fixed attribute.
2750 (arc_file_end): New function.
2751 (TARGET_ASM_FILE_END): Define.
2752 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
2753 (arc_add_jli_section): New function.
2754 (jli_call_scan): Likewise.
2755 (arc_reorg): Call jli_call_scan.
2756 (arc_output_addsi): Remove 'S' from printing asm operand.
2757 (arc_is_jli_call_p): New function.
2758 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
2760 (movhi_insn): Likewise.
2761 (movsi_insn): Likewise.
2762 (movsi_set_cc_insn): Likewise.
2763 (loadqi_update): Likewise.
2764 (load_zeroextendqisi_update): Likewise.
2765 (load_signextendqisi_update): Likewise.
2766 (loadhi_update): Likewise.
2767 (load_zeroextendhisi_update): Likewise.
2768 (load_signextendhisi_update): Likewise.
2769 (loadsi_update): Likewise.
2770 (loadsf_update): Likewise.
2771 (movsicc_insn): Likewise.
2772 (bset_insn): Likewise.
2773 (bxor_insn): Likewise.
2774 (bclr_insn): Likewise.
2775 (bmsk_insn): Likewise.
2776 (bicsi3_insn): Likewise.
2777 (cmpsi_cc_c_insn): Likewise.
2778 (movsi_ne): Likewise.
2779 (movsi_cond_exec): Likewise.
2780 (clrsbsi2): Likewise.
2791 (call_i): Remove 'S' asm letter, add jli instruction.
2792 (call_value_i): Likewise.
2793 * config/arc/arc.op (mjli-always): New option.
2794 * config/arc/constraints.md (Cji): New constraint.
2795 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
2797 (subsf3_fpx): Likewise.
2798 (mulsf3_fpx): Likewise.
2799 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
2801 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
2803 * doc/invoke.texi (ARC): Document mjli-always option.
2805 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
2807 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
2808 avoid addition with 0 and use incw and decw where possible.
2810 2018-01-26 Richard Biener <rguenther@suse.de>
2812 PR tree-optimization/81082
2813 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
2814 association if it requires casting to unsigned.
2815 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
2816 from fold_plusminus_mult_expr to catch important cases late when
2817 range info is available.
2819 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2821 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
2822 * configure.ac (hidden_linkonce): New test.
2823 * configure: Regenerate.
2824 * config.in: Regenerate.
2826 2018-01-26 Julia Koval <julia.koval@intel.com>
2828 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
2829 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
2830 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
2831 _mm_mask_bitshuffle_epi64_mask): Fix type.
2832 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
2833 USI_FTYPE_V4DI_V4DI_USI): Remove.
2834 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
2835 __builtin_ia32_vpshufbitqmb256_mask,
2836 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
2837 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
2838 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
2840 2018-01-26 Alan Modra <amodra@gmail.com>
2843 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
2844 UNSPEC_VBPERMQ. Sort other unspecs.
2846 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
2848 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
2850 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
2853 * predict.c (drop_profile): Do not push/pop cfun; update also
2855 (handle_missing_profiles): Fix logic looking for zero profiles.
2857 2018-01-25 Jakub Jelinek <jakub@redhat.com>
2860 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
2861 on functions with #pragma omp declare simd or functions with simd
2863 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
2864 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
2865 Remove trailing \n from warning_at calls.
2867 2018-01-25 Tom de Vries <tom@codesourcery.com>
2870 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2871 for neutered workers.
2873 2018-01-24 Joseph Myers <joseph@codesourcery.com>
2876 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
2877 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
2879 2018-01-24 Jeff Law <law@redhat.com>
2882 * i386.c (get_probe_interval): Move to earlier point.
2883 (ix86_compute_frame_layout): If -fstack-clash-protection and
2884 the frame is larger than the probe interval, then use pushes
2885 to save registers rather than reg->mem moves.
2886 (ix86_expand_prologue): Remove conditional for int_registers_saved
2889 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
2892 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
2893 min/max for never referenced object.
2895 2018-01-24 Jakub Jelinek <jakub@redhat.com>
2898 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
2900 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
2901 attributes from DECL_ATTRIBUTES (decl) without affecting
2902 DECL_ATTRIBUTES (current_function_decl).
2903 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
2904 functions with non-NULL DECL_ABSTRACT_ORIGIN.
2906 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
2908 PR tree-optimization/83979
2909 * fold-const.c (fold_comparison): Use constant_boolean_node
2910 instead of boolean_{true,false}_node.
2912 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
2914 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
2917 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2919 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
2920 Simplify the clause that sets the length attribute.
2921 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
2922 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
2923 clause that sets the length attribute.
2924 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
2926 2018-01-24 Tom de Vries <tom@codesourcery.com>
2929 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
2930 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
2931 Add strict parameter.
2932 (prevent_branch_around_nothing): Insert dummy insn between branch to
2933 label and label with no ptx insn inbetween.
2934 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
2936 2018-01-24 Tom de Vries <tom@codesourcery.com>
2939 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2940 for neutered threads in warp.
2941 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
2943 2018-01-24 Richard Biener <rguenther@suse.de>
2945 PR tree-optimization/83176
2946 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
2949 2018-01-24 Richard Biener <rguenther@suse.de>
2951 PR tree-optimization/82819
2952 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
2953 code generating pluses that are no-ops in the target precision.
2955 2018-01-24 Richard Biener <rguenther@suse.de>
2958 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
2960 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2962 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
2963 to merge probabilities.
2964 * predict.c (probably_never_executed): Also mark as cold functions
2965 with global 0 profile and guessed local profile.
2966 * profile-count.c (profile_probability::combine_with_count): New
2968 * profile-count.h (profile_probability::operator*,
2969 profile_probability::operator*=, profile_probability::operator/,
2970 profile_probability::operator/=): Reduce precision to adjusted
2971 and set value to guessed on contradictory divisions.
2972 (profile_probability::combine_with_freq): Remove.
2973 (profile_probability::combine_wiht_count): Declare.
2974 (profile_count::force_nonzero):: Set to adjusted.
2975 (profile_count::probability_in):: Set quality to adjusted.
2976 * tree-ssa-tail-merge.c (replace_block_by): Use
2979 2018-01-23 Andrew Waterman <andrew@sifive.com>
2980 Jim Wilson <jimw@sifive.com>
2982 * config/riscv/riscv.c (riscv_stack_boundary): New.
2983 (riscv_option_override): Set riscv_stack_boundary. Handle
2984 riscv_preferred_stack_boundary_arg.
2985 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
2986 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
2987 (STACK_BOUNDARY): Set to riscv_stack_boundary.
2988 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
2989 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
2990 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
2992 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
2995 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
2996 of struct ix86_frame.
2997 (ix86_expand_epilogue): Likewise. Add a local variable for
2998 the reg_save_offset field in struct ix86_frame.
3000 2018-01-23 Bin Cheng <bin.cheng@arm.com>
3002 PR tree-optimization/82604
3003 * tree-loop-distribution.c (enum partition_kind): New enum item
3004 PKIND_PARTIAL_MEMSET.
3005 (partition_builtin_p): Support above new enum item.
3006 (generate_code_for_partition): Ditto.
3007 (compute_access_range): Differentiate cases that equality can be
3008 proven at all loops, the innermost loops or no loops.
3009 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
3010 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
3011 (finalize_partitions, distribute_loop): Don't fuse partition of
3012 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
3013 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
3016 2018-01-23 Martin Liska <mliska@suse.cz>
3018 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
3019 order to ignore the predictor.
3020 (PRED_POLYMORPHIC_CALL): Likewise.
3021 (PRED_RECURSIVE_CALL): Likewise.
3023 2018-01-23 Martin Liska <mliska@suse.cz>
3025 * tree-profile.c (tree_profiling): Print function header to
3026 aware reader which function we are working on.
3027 * value-prof.c (gimple_find_values_to_profile): Do not print
3028 not interesting value histograms.
3030 2018-01-23 Martin Liska <mliska@suse.cz>
3032 * profile-count.h (enum profile_quality): Add
3033 profile_uninitialized as the first value. Do not number values
3034 as they are zero based.
3035 (profile_count::verify): Update sanity check.
3036 (profile_probability::verify): Likewise.
3038 2018-01-23 Nathan Sidwell <nathan@acm.org>
3040 * doc/invoke.texi (ffor-scope): Deprecate.
3042 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3044 PR tree-optimization/83510
3045 * domwalk.c (set_all_edges_as_executable): New function.
3046 (dom_walker::dom_walker): Convert bool param
3047 "skip_unreachable_blocks" to enum reachability. Move setup of
3048 edge flags to set_all_edges_as_executable and only do it when
3049 reachability is REACHABLE_BLOCKS.
3050 * domwalk.h (enum dom_walker::reachability): New enum.
3051 (dom_walker::dom_walker): Convert bool param
3052 "skip_unreachable_blocks" to enum reachability.
3053 (set_all_edges_as_executable): New decl.
3054 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
3055 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
3057 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
3058 but converting true to REACHABLE_BLOCKS.
3059 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
3061 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
3062 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
3063 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
3065 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
3066 if check_all_array_refs will be called.
3068 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3070 * tree.c (selftest::test_location_wrappers): Add more test
3073 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3075 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
3076 (selftest::test_bit_in_range): Likewise.
3078 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3081 * doc/sourcebuild.texi (vect_float): Say that the selector
3082 only describes the situation when -funsafe-math-optimizations is on.
3083 (vect_float_strict): Document.
3085 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3087 PR tree-optimization/83965
3088 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3089 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
3090 instead of checking only for a reduction.
3091 (vect_recog_widen_sum_pattern): Likewise.
3093 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3095 * predict.c (probably_never_executed): Only use precise profile info.
3096 (compute_function_frequency): Skip after inlining hack since we now
3097 have quality checking.
3099 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3101 * profile-count.h (profile_probability::very_unlikely,
3102 profile_probability::unlikely, profile_probability::even): Set
3103 precision to guessed.
3105 2018-01-23 Richard Biener <rguenther@suse.de>
3107 PR tree-optimization/83963
3108 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
3109 Properly terminate dominator walk when crossing the exit edge not
3110 when visiting its source block.
3112 2018-01-23 Jakub Jelinek <jakub@redhat.com>
3115 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
3116 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
3118 2018-01-22 Jakub Jelinek <jakub@redhat.com>
3120 PR tree-optimization/83957
3121 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
3122 semicolon after for body surrounded by braces.
3124 PR tree-optimization/83081
3125 * profile-count.h (profile_probability::split): New method.
3126 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
3127 Use profile_probability::split.
3128 (do_compare_rtx_and_jump): Fix adjustment of probabilities
3129 when splitting a single conditional jump into 2.
3131 2018-01-22 David Malcolm <dmalcolm@redhat.com>
3133 PR tree-optimization/69452
3134 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
3137 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3139 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
3140 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
3141 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
3143 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3145 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
3146 * config/rl78/rl78.md (movdi): New define_expand.
3147 * config/rl78/rl78.c (rl78_split_movdi): New function.
3149 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
3152 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
3154 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
3155 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
3156 128-bit to produce an UNSPEC move to get the double word with the
3157 signbit and then a shift directly to do signbit.
3158 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
3159 implementation with a new version that just does either a direct
3160 move or a regular move. Move memory interface to separate insns.
3161 Move insns so they are next to the expander.
3162 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
3163 with signbit move. Split big and little endian case.
3164 (signbit<mode>2_dm_mem_le): Likewise.
3165 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
3166 (signbit<mode>2_dm2): Likewise.
3168 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3170 * config/rl78/rl78.md (anddi3): New define_expand.
3172 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3174 * config/rl78/rl78.md (umindi3): New define_expand.
3176 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3178 * config/rl78/rl78.md (smindi3): New define_expand.
3180 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3182 * config/rl78/rl78.md (smaxdi3): New define_expand.
3184 2018-01-22 Carl Love <cel@us.ibm.com>
3186 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
3187 LVX_V1TI): Add macro expansion.
3188 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
3189 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
3190 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
3191 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
3192 Change check to determine if the instruction is a byte reversing
3193 entry. Fix typo in comment.
3194 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
3195 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
3196 Add def_builtin calls for new builtins.
3197 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
3198 Add define_insn expansion.
3200 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3202 * config/rl78/rl78.md (umaxdi3): New define_expand.
3204 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3206 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
3207 for non-QImode registers.
3209 2018-01-22 Richard Biener <rguenther@suse.de>
3211 PR tree-optimization/83963
3212 * graphite-scop-detection.c (scop_detection::get_sese): Delay
3213 including the loop exit block.
3214 (scop_detection::merge_sese): Likewise.
3215 (scop_detection::add_scop): Do it here instead.
3217 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3219 * doc/sourcebuild.texi (arm_softfloat): Document.
3221 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
3224 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
3225 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
3226 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
3228 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3229 David Edelsohn <dje.gcc@gmail.com>
3232 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3233 Change "crset eq" to "crset 2".
3234 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3235 (*call_indirect_aix<mode>_nospec): Likewise.
3236 (*call_value_indirect_aix<mode>_nospec): Likewise.
3237 (*call_indirect_elfv2<mode>_nospec): Likewise.
3238 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
3239 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
3240 change assembly output from . to $.
3241 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3242 (indirect_jump<mode>_nospec): Change assembly output from . to $.
3243 (*tablejump<mode>_internal1_nospec): Likewise.
3245 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
3248 * config/sh/sh_optimize_sett_clrt.cc:
3249 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
3251 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
3253 PR tree-optimization/83940
3254 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
3255 offset_dt to vect_constant_def rather than vect_unknown_def_type.
3256 (vect_check_load_store_mask): Add a mask_dt_out parameter and
3257 use it to pass back the definition type.
3258 (vect_check_store_rhs): Likewise rhs_dt_out.
3259 (vect_build_gather_load_calls): Add a mask_dt argument and use
3260 it instead of a call to vect_is_simple_use.
3261 (vectorizable_store): Update calls to vect_check_load_store_mask
3262 and vect_check_store_rhs. Use the dt returned by the latter instead
3263 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
3264 instead of calls to vect_is_simple_use. Pass the scalar rather
3265 than the vector operand to vect_is_simple_use when handling
3266 second and subsequent copies of an rhs value.
3267 (vectorizable_load): Update calls to vect_check_load_store_mask
3268 and vect_build_gather_load_calls. Use the cached mask_dt and
3269 gs_info.offset_dt instead of calls to vect_is_simple_use.
3271 2018-01-20 Jakub Jelinek <jakub@redhat.com>
3274 * tree-emutls.c: Include gimplify.h.
3275 (lower_emutls_2): New function.
3276 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
3277 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
3278 it before further processing.
3281 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
3282 UINTVAL (trueop1) instead of INTVAL (op1).
3284 2018-01-19 Jakub Jelinek <jakub@redhat.com>
3288 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
3289 INCOMING_FRAME_SP_OFFSET if not defined.
3290 (scan_trace): Add ENTRY argument. If true and
3291 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
3292 emit a note to adjust the CFA offset.
3293 (create_cfi_notes): Adjust scan_trace callers.
3294 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
3295 INCOMING_FRAME_SP_OFFSET in the CIE.
3296 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
3297 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
3299 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
3300 * doc/tm.texi: Regenerated.
3302 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3304 PR rtl-optimization/83147
3305 * lra-constraints.c (remove_inheritance_pseudos): Use
3306 lra_substitute_pseudo_within_insn.
3308 2018-01-19 Tom de Vries <tom@codesourcery.com>
3309 Cesar Philippidis <cesar@codesourcery.com>
3312 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
3314 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
3317 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
3318 spaces for function labels.
3320 2018-01-19 Martin Liska <mliska@suse.cz>
3322 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
3323 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
3324 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
3325 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
3326 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
3327 (PRED_CONST_RETURN): Change from 69 to 65.
3328 (PRED_NULL_RETURN): Change from 91 to 71.
3329 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
3330 (PRED_LOOP_GUARD): Change from 66 to 73.
3332 2018-01-19 Martin Liska <mliska@suse.cz>
3334 * predict.c (predict_insn_def): Add new assert.
3335 (struct branch_predictor): Change type to signed integer.
3336 (test_prediction_value_range): Amend test to cover
3338 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
3339 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
3340 (PRED_LOOP_ITERATIONS_MAX): Likewise.
3341 (PRED_LOOP_IV_COMPARE): Likewise.
3342 * predict.h (PROB_UNINITIALIZED): Define new constant.
3344 2018-01-19 Martin Liska <mliska@suse.cz>
3346 * predict.c (dump_prediction): Add new format for
3347 analyze_brprob.py script which is enabled with -details
3349 * profile-count.h (precise_p): New function.
3351 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3353 PR tree-optimization/83922
3354 * tree-vect-loop.c (vect_verify_full_masking): Return false if
3355 there are no statements that need masking.
3356 (vect_active_double_reduction_p): New function.
3357 (vect_analyze_loop_operations): Use it when handling phis that
3358 are not in the loop header.
3360 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3362 PR tree-optimization/83914
3363 * tree-vect-loop.c (vectorizable_induction): Don't convert
3364 init_expr or apply the peeling adjustment for inductions
3365 that are nested within the vectorized loop.
3367 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3369 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
3372 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3376 * function.h (gimplify_parameters): Add gimple_seq * argument.
3377 * function.c: Include gimple.h and options.h.
3378 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
3379 for the added local temporaries if needed.
3380 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
3381 if there are any parameter cleanups, wrap whole body into a
3382 try/finally with the cleanups.
3384 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
3387 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
3388 Use GET_MODE_CLASS for scalar floating point.
3390 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3394 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3395 Fix call of call_cgraph_insertion_hooks.
3397 2018-01-18 Martin Sebor <msebor@redhat.com>
3399 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
3401 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3404 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
3407 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
3410 * common.opt: (-ffile-prefix-map): New option.
3411 * opts.c (common_handle_option): Defer it.
3412 * opts-global.c (handle_common_deferred_options): Handle it.
3413 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
3414 * file-prefix-map.h: New file.
3415 (remap_debug_filename, add_debug_prefix_map): ...here.
3416 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
3417 * final.c (debug_prefix_map, add_debug_prefix_map
3418 remap_debug_filename): Move to...
3419 * file-prefix-map.c: New file.
3420 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
3421 generalize, get rid of alloca(), use strrchr() instead of strchr().
3422 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
3423 Implement in terms of add_prefix_map().
3424 (remap_macro_filename, remap_debug_filename): Implement in term of
3426 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
3427 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
3428 * dbxout.c: Include file-prefix-map.h.
3429 * varasm.c: Likewise.
3430 * vmsdbgout.c: Likewise.
3431 * xcoffout.c: Likewise.
3432 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
3433 * doc/cppopts.texi (-fmacro-prefix-map): Document.
3434 * doc/invoke.texi (-ffile-prefix-map): Document.
3435 (-fdebug-prefix-map): Update description.
3437 2018-01-18 Martin Liska <mliska@suse.cz>
3439 * config/i386/i386.c (indirect_thunk_name): Document that also
3441 (output_indirect_thunk): Document why both instructions
3442 (pause and lfence) are generated.
3444 2018-01-18 Richard Biener <rguenther@suse.de>
3446 PR tree-optimization/83887
3447 * graphite-scop-detection.c
3448 (scop_detection::get_nearest_dom_with_single_entry): Remove.
3449 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
3450 (scop_detection::merge_sese): Re-implement with a flood-fill
3451 algorithm that properly finds a SESE region if it exists.
3453 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3456 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
3457 pointer_diff optimizations use view_convert instead of convert.
3459 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3461 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3462 Generate different code for -mno-speculate-indirect-jumps.
3463 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3464 (*call_indirect_aix<mode>): Disable for
3465 -mno-speculate-indirect-jumps.
3466 (*call_indirect_aix<mode>_nospec): New define_insn.
3467 (*call_value_indirect_aix<mode>): Disable for
3468 -mno-speculate-indirect-jumps.
3469 (*call_value_indirect_aix<mode>_nospec): New define_insn.
3470 (*sibcall_nonlocal_sysv<mode>): Generate different code for
3471 -mno-speculate-indirect-jumps.
3472 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3474 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3476 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
3477 long double type, set the flags for noting the default long double
3478 type, even if we don't pass or return a long double type.
3480 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
3483 * ipa-inline.c (flatten_function): Do not overwrite final inlining
3486 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
3488 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
3489 support for merge[hl].
3490 (fold_mergehl_helper): New helper function.
3491 (tree-vector-builder.h): New #include for tree_vector_builder usage.
3492 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
3493 (altivec_vmrglw_direct): Add xxmrglw insn.
3495 2018-01-17 Andrew Waterman <andrew@sifive.com>
3497 * config/riscv/riscv.c (riscv_conditional_register_usage): If
3498 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
3500 2018-01-17 David Malcolm <dmalcolm@redhat.com>
3503 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
3504 call the lto_location_cache before reading the
3505 DECL_SOURCE_LOCATION of the types.
3507 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3508 Richard Sandiford <richard.sandiford@linaro.org>
3510 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
3511 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
3512 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
3513 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
3514 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
3516 * config/aarch64/constraints.md (aarch64_movti_operand):
3518 * config/aarch64/predicates.md (Uti): Add new constraint.
3520 2018-01-17 Carl Love <cel@us.ibm.com>
3522 * config/rs6000/vsx.md (define_expand xl_len_r,
3523 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
3524 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
3526 (define_expand, define_insn): Move the shift left from the
3527 define_insn to the define_expand for lxvl and stxvl instructions.
3528 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
3529 and XL_LEN_R definitions to PURE.
3531 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
3533 * config/i386/i386.c (indirect_thunk_name): Declare regno
3534 as unsigned int. Compare regno with INVALID_REGNUM.
3535 (output_indirect_thunk): Ditto.
3536 (output_indirect_thunk_function): Ditto.
3537 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
3538 in the call to output_indirect_thunk_function.
3540 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
3543 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
3544 rather than the size of inner_type to determine the stack slot size
3545 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
3547 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
3550 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
3553 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
3555 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
3556 endian Linux systems to optionally enable multilibs for selecting
3557 the long double type if the user configured an explicit type.
3558 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
3559 have no long double multilibs if not defined.
3560 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
3561 warn if the user used -mabi={ieee,ibm}longdouble and we built
3562 multilibs for long double.
3563 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
3564 appropriate multilib option.
3565 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
3567 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
3568 for building long double multilibs.
3569 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
3571 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
3573 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
3576 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
3578 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
3581 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
3584 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
3587 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3589 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
3590 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
3592 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3594 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
3595 different rtl trees depending on TARGET_64BIT.
3596 (rs6000_gen_lvx): Likewise.
3598 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3600 * config/visium/visium.md (nop): Tweak comment.
3601 (hazard_nop): Likewise.
3603 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3605 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
3606 -mspeculate-indirect-jumps.
3607 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
3608 for -mno-speculate-indirect-jumps.
3609 (*call_indirect_elfv2<mode>_nospec): New define_insn.
3610 (*call_value_indirect_elfv2<mode>): Disable for
3611 -mno-speculate-indirect-jumps.
3612 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
3613 (indirect_jump): Emit different RTL for
3614 -mno-speculate-indirect-jumps.
3615 (*indirect_jump<mode>): Disable for
3616 -mno-speculate-indirect-jumps.
3617 (*indirect_jump<mode>_nospec): New define_insn.
3618 (tablejump): Emit different RTL for
3619 -mno-speculate-indirect-jumps.
3620 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
3621 (tablejumpsi_nospec): New define_expand.
3622 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
3623 (tablejumpdi_nospec): New define_expand.
3624 (*tablejump<mode>_internal1): Disable for
3625 -mno-speculate-indirect-jumps.
3626 (*tablejump<mode>_internal1_nospec): New define_insn.
3627 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
3630 2018-01-16 Artyom Skrobov tyomitch@gmail.com
3632 * caller-save.c (insert_save): Drop unnecessary parameter. All
3635 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3636 Richard Biener <rguenth@suse.de>
3639 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
3640 return early, inline manually is_gimple_sizepos. Make sure if we
3641 call gimplify_expr we don't end up with a gimple constant.
3642 * tree.c (variably_modified_type_p): Don't return true for
3643 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
3644 * gimplify.h (is_gimple_sizepos): Remove.
3646 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3648 PR tree-optimization/83857
3649 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
3650 vectorizable_live_operation for pure SLP statements.
3651 (vectorizable_live_operation): Handle PHIs.
3653 2018-01-16 Richard Biener <rguenther@suse.de>
3655 PR tree-optimization/83867
3656 * tree-vect-stmts.c (vect_transform_stmt): Precompute
3657 nested_in_vect_loop_p since the scalar stmt may get invalidated.
3659 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3662 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
3663 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
3664 If off is not INTEGER_CST, issue a may not be aligned warning
3665 rather than isn't aligned. Use isn%'t rather than isn't.
3666 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
3668 <case MULT_EXPR>: Improve the case when bottom and one of the
3669 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
3670 operand, in that case check if the other operand is multiple of
3671 bottom divided by the INTEGER_CST operand.
3673 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3676 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
3677 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
3678 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
3679 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
3680 * config/pa/pa.c (pa_function_arg_advance): Likewise.
3681 (pa_function_arg, pa_arg_partial_bytes): Likewise.
3682 (pa_function_arg_size): New function.
3684 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3686 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
3687 in a separate statement.
3689 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3691 PR tree-optimization/83847
3692 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
3693 group gathers and scatters.
3695 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3697 PR rtl-optimization/86620
3698 * params.def (max-sched-ready-insns): Bump minimum value to 1.
3700 PR rtl-optimization/83213
3701 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
3702 to last if both are JUMP_INSNs.
3704 PR tree-optimization/83843
3705 * gimple-ssa-store-merging.c
3706 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
3707 store_immediate_info for bswap/nop orig_stores.
3709 2018-01-15 Andrew Waterman <andrew@sifive.com>
3711 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
3713 <UDIV>: Increase cost if !TARGET_DIV.
3715 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
3717 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
3718 (define_attr "cr_logical_3op"): New.
3719 (cceq_ior_compare): Adjust.
3720 (cceq_ior_compare_complement): Adjust.
3721 (*cceq_rev_compare): Adjust.
3722 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
3723 (is_cracked_insn): Adjust.
3724 (insn_must_be_first_in_group): Adjust.
3725 * config/rs6000/40x.md: Adjust.
3726 * config/rs6000/440.md: Adjust.
3727 * config/rs6000/476.md: Adjust.
3728 * config/rs6000/601.md: Adjust.
3729 * config/rs6000/603.md: Adjust.
3730 * config/rs6000/6xx.md: Adjust.
3731 * config/rs6000/7450.md: Adjust.
3732 * config/rs6000/7xx.md: Adjust.
3733 * config/rs6000/8540.md: Adjust.
3734 * config/rs6000/cell.md: Adjust.
3735 * config/rs6000/e300c2c3.md: Adjust.
3736 * config/rs6000/e500mc.md: Adjust.
3737 * config/rs6000/e500mc64.md: Adjust.
3738 * config/rs6000/e5500.md: Adjust.
3739 * config/rs6000/e6500.md: Adjust.
3740 * config/rs6000/mpc.md: Adjust.
3741 * config/rs6000/power4.md: Adjust.
3742 * config/rs6000/power5.md: Adjust.
3743 * config/rs6000/power6.md: Adjust.
3744 * config/rs6000/power7.md: Adjust.
3745 * config/rs6000/power8.md: Adjust.
3746 * config/rs6000/power9.md: Adjust.
3747 * config/rs6000/rs64.md: Adjust.
3748 * config/rs6000/titan.md: Adjust.
3750 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3752 * config/i386/predicates.md (indirect_branch_operand): Rewrite
3753 ix86_indirect_branch_register logic.
3755 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3757 * config/i386/constraints.md (Bs): Update
3758 ix86_indirect_branch_register check. Don't check
3759 ix86_indirect_branch_register with GOT_memory_operand.
3761 * config/i386/predicates.md (GOT_memory_operand): Don't check
3762 ix86_indirect_branch_register here.
3763 (GOT32_symbol_operand): Likewise.
3765 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3767 * config/i386/predicates.md (constant_call_address_operand):
3768 Rewrite ix86_indirect_branch_register logic.
3769 (sibcall_insn_operand): Likewise.
3771 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3773 * config/i386/constraints.md (Bs): Replace
3774 ix86_indirect_branch_thunk_register with
3775 ix86_indirect_branch_register.
3777 * config/i386/i386.md (indirect_jump): Likewise.
3778 (tablejump): Likewise.
3779 (*sibcall_memory): Likewise.
3780 (*sibcall_value_memory): Likewise.
3781 Peepholes of indirect call and jump via memory: Likewise.
3782 * config/i386/i386.opt: Likewise.
3783 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3784 (GOT_memory_operand): Likewise.
3785 (call_insn_operand): Likewise.
3786 (sibcall_insn_operand): Likewise.
3787 (GOT32_symbol_operand): Likewise.
3789 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3792 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
3793 type rather than type addr's type points to.
3794 (expand_omp_atomic_mutex): Likewise.
3795 (expand_omp_atomic): Likewise.
3797 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3800 * config/i386/i386.c (output_indirect_thunk_function): Use
3801 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
3802 for __x86_return_thunk.
3804 2018-01-15 Richard Biener <rguenther@suse.de>
3807 * expmed.c (extract_bit_field_1): Fix typo.
3809 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3812 * config/arm/iterators.md (VF): New mode iterator.
3813 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
3814 Remove integer-related logic from pattern.
3815 (neon_vabd<mode>_3): Likewise.
3817 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3820 * common.opt (fstrict-overflow): No longer an alias.
3821 (fwrapv-pointer): New option.
3822 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
3823 also for pointer types based on flag_wrapv_pointer.
3824 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
3825 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
3826 opts->x_flag_wrapv got set.
3827 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
3828 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
3829 POINTER_TYPE_OVERFLOW_UNDEFINED.
3830 * match.pd: Likewise in address comparison pattern.
3831 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
3833 2018-01-15 Richard Biener <rguenther@suse.de>
3836 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
3837 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
3838 Reset type names to their identifier if their TYPE_DECL doesn't
3839 have linkage (and thus is used for ODR and devirt).
3840 (save_debug_info_for_decl): Remove.
3841 (save_debug_info_for_type): Likewise.
3842 (add_tree_to_fld_list): Adjust.
3843 * tree-pretty-print.c (dump_generic_node): Make dumping of
3844 type names more robust.
3846 2018-01-15 Richard Biener <rguenther@suse.de>
3848 * BASE-VER: Bump to 8.0.1.
3850 2018-01-14 Martin Sebor <msebor@redhat.com>
3853 * builtins.c (check_access): Avoid warning when the no-warning bit
3856 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
3858 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
3859 * ira-color (allocno_hard_regs_compare): Likewise.
3861 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
3864 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
3865 Use .pushsection/.popsection.
3867 2018-01-14 Martin Sebor <msebor@redhat.com>
3870 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
3872 2018-01-14 Jakub Jelinek <jakub@redhat.com>
3874 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
3875 entry from extra_headers.
3876 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
3877 extra_headers, make the list bitwise identical to the i?86-*-* one.
3879 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3881 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
3882 -mcmodel=large with -mindirect-branch=thunk,
3883 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
3884 -mfunction-return=thunk-extern.
3885 * doc/invoke.texi: Document -mcmodel=large is incompatible with
3886 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
3887 -mfunction-return=thunk and -mfunction-return=thunk-extern.
3889 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3891 * config/i386/i386.c (print_reg): Print the name of the full
3892 integer register without '%'.
3893 (ix86_print_operand): Handle 'V'.
3894 * doc/extend.texi: Document 'V' modifier.
3896 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3898 * config/i386/constraints.md (Bs): Disallow memory operand for
3899 -mindirect-branch-register.
3901 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3902 (GOT_memory_operand): Likewise.
3903 (call_insn_operand): Likewise.
3904 (sibcall_insn_operand): Likewise.
3905 (GOT32_symbol_operand): Likewise.
3906 * config/i386/i386.md (indirect_jump): Call convert_memory_address
3907 for -mindirect-branch-register.
3908 (tablejump): Likewise.
3909 (*sibcall_memory): Likewise.
3910 (*sibcall_value_memory): Likewise.
3911 Disallow peepholes of indirect call and jump via memory for
3912 -mindirect-branch-register.
3913 (*call_pop): Replace m with Bw.
3914 (*call_value_pop): Likewise.
3915 (*sibcall_pop_memory): Replace m with Bs.
3916 * config/i386/i386.opt (mindirect-branch-register): New option.
3917 * doc/invoke.texi: Document -mindirect-branch-register option.
3919 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3921 * config/i386/i386-protos.h (ix86_output_function_return): New.
3922 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
3923 set function_return_type.
3924 (indirect_thunk_name): Add ret_p to indicate thunk for function
3926 (output_indirect_thunk_function): Pass false to
3927 indirect_thunk_name.
3928 (ix86_output_indirect_branch_via_reg): Likewise.
3929 (ix86_output_indirect_branch_via_push): Likewise.
3930 (output_indirect_thunk_function): Create alias for function
3931 return thunk if regno < 0.
3932 (ix86_output_function_return): New function.
3933 (ix86_handle_fndecl_attribute): Handle function_return.
3934 (ix86_attribute_table): Add function_return.
3935 * config/i386/i386.h (machine_function): Add
3936 function_return_type.
3937 * config/i386/i386.md (simple_return_internal): Use
3938 ix86_output_function_return.
3939 (simple_return_internal_long): Likewise.
3940 * config/i386/i386.opt (mfunction-return=): New option.
3941 (indirect_branch): Mention -mfunction-return=.
3942 * doc/extend.texi: Document function_return function attribute.
3943 * doc/invoke.texi: Document -mfunction-return= option.
3945 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3947 * config/i386/i386-opts.h (indirect_branch): New.
3948 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
3949 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
3950 with local indirect jump when converting indirect call and jump.
3951 (ix86_set_indirect_branch_type): New.
3952 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
3953 (indirectlabelno): New.
3954 (indirect_thunk_needed): Likewise.
3955 (indirect_thunk_bnd_needed): Likewise.
3956 (indirect_thunks_used): Likewise.
3957 (indirect_thunks_bnd_used): Likewise.
3958 (INDIRECT_LABEL): Likewise.
3959 (indirect_thunk_name): Likewise.
3960 (output_indirect_thunk): Likewise.
3961 (output_indirect_thunk_function): Likewise.
3962 (ix86_output_indirect_branch_via_reg): Likewise.
3963 (ix86_output_indirect_branch_via_push): Likewise.
3964 (ix86_output_indirect_branch): Likewise.
3965 (ix86_output_indirect_jmp): Likewise.
3966 (ix86_code_end): Call output_indirect_thunk_function if needed.
3967 (ix86_output_call_insn): Call ix86_output_indirect_branch if
3969 (ix86_handle_fndecl_attribute): Handle indirect_branch.
3970 (ix86_attribute_table): Add indirect_branch.
3971 * config/i386/i386.h (machine_function): Add indirect_branch_type
3972 and has_local_indirect_jump.
3973 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
3975 (tablejump): Likewise.
3976 (*indirect_jump): Use ix86_output_indirect_jmp.
3977 (*tablejump_1): Likewise.
3978 (simple_return_indirect_internal): Likewise.
3979 * config/i386/i386.opt (mindirect-branch=): New option.
3980 (indirect_branch): New.
3983 (thunk-inline): Likewise.
3984 (thunk-extern): Likewise.
3985 * doc/extend.texi: Document indirect_branch function attribute.
3986 * doc/invoke.texi: Document -mindirect-branch= option.
3988 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
3991 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
3993 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
3995 * ipa-inline.c (want_inline_small_function_p): Return false if
3996 inlining has already failed with CIF_FINAL_ERROR.
3997 (update_caller_keys): Call want_inline_small_function_p before
3999 (update_callee_keys): Likewise.
4001 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4003 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
4005 (rs6000_quadword_masked_address_p): Likewise.
4006 (quad_aligned_load_p): Likewise.
4007 (quad_aligned_store_p): Likewise.
4008 (const_load_sequence_p): Add comment to describe the outer-most loop.
4009 (mimic_memory_attributes_and_flags): New function.
4010 (rs6000_gen_stvx): Likewise.
4011 (replace_swapped_aligned_store): Likewise.
4012 (rs6000_gen_lvx): Likewise.
4013 (replace_swapped_aligned_load): Likewise.
4014 (replace_swapped_load_constant): Capitalize argument name in
4015 comment describing this function.
4016 (rs6000_analyze_swaps): Add a third pass to search for vector loads
4017 and stores that access quad-word aligned addresses and replace
4018 with stvx or lvx instructions when appropriate.
4019 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
4020 New function prototype.
4021 (rs6000_quadword_masked_address_p): Likewise.
4022 (rs6000_gen_lvx): Likewise.
4023 (rs6000_gen_stvx): Likewise.
4024 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
4025 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
4026 when memory address is aligned.
4027 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
4028 this split to select lvx instruction when memory address is aligned.
4029 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
4030 instruction when memory address is aligned.
4031 (*vsx_le_perm_load_v16qi): Likewise.
4032 (four unnamed splitters): Modify to select the stvx instruction
4033 when memory is aligned.
4035 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
4037 * predict.c (determine_unlikely_bbs): Handle correctly BBs
4038 which appears in the queue multiple times.
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 * tree-vectorizer.h (vec_lower_bound): New structure.
4045 (_loop_vec_info): Add check_nonzero and lower_bounds.
4046 (LOOP_VINFO_CHECK_NONZERO): New macro.
4047 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
4048 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
4049 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
4050 fields. Make seg_len the distance travelled, not including the
4052 (dr_direction_indicator): Declare.
4053 (dr_zero_step_indicator): Likewise.
4054 (dr_known_forward_stride_p): Likewise.
4055 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
4057 (runtime_alias_check_p): Allow runtime alias checks with
4059 (operator ==): Compare access_size and align.
4060 (prune_runtime_alias_test_list): Rework for new distinction between
4061 the access_size and seg_len.
4062 (create_intersect_range_checks_index): Likewise. Cope with polynomial
4064 (get_segment_min_max): New function.
4065 (create_intersect_range_checks): Use it.
4066 (dr_step_indicator): New function.
4067 (dr_direction_indicator): Likewise.
4068 (dr_zero_step_indicator): Likewise.
4069 (dr_known_forward_stride_p): Likewise.
4070 * tree-loop-distribution.c (data_ref_segment_size): Return
4071 DR_STEP * (niters - 1).
4072 (compute_alias_check_pairs): Update call to the dr_with_seg_len
4074 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
4075 (vect_preserves_scalar_order_p): New function, split out from...
4076 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
4077 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
4078 (vect_vfa_access_size): New function.
4079 (vect_vfa_align): Likewise.
4080 (vect_compile_time_alias): Take access_size_a and access_b arguments.
4081 (dump_lower_bound): New function.
4082 (vect_check_lower_bound): Likewise.
4083 (vect_small_gap_p): Likewise.
4084 (vectorizable_with_step_bound_p): Likewise.
4085 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
4086 depencies if the vectorization factor is 1. Convert the checks
4087 for nonzero steps into checks on the bounds of DR_STEP. Try using
4088 a bunds check for variable steps if the minimum required step is
4089 relatively small. Update calls to the dr_with_seg_len
4090 constructor and to vect_compile_time_alias.
4091 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
4093 (vect_loop_versioning): Call it.
4094 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
4096 (vect_estimate_min_profitable_iters): Account for any bounds checks.
4098 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4099 Alan Hayward <alan.hayward@arm.com>
4100 David Sherwood <david.sherwood@arm.com>
4102 * doc/sourcebuild.texi (vect_scatter_store): Document.
4103 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
4105 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
4107 * genopinit.c (main): Add supports_vec_scatter_store and
4108 supports_vec_scatter_store_cached to target_optabs.
4109 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
4110 IFN_MASK_SCATTER_STORE.
4111 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
4113 * internal-fn.h (internal_store_fn_p): Declare.
4114 (internal_fn_stored_value_index): Likewise.
4115 * internal-fn.c (scatter_store_direct): New macro.
4116 (expand_scatter_store_optab_fn): New function.
4117 (direct_scatter_store_optab_supported_p): New macro.
4118 (internal_store_fn_p): New function.
4119 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
4120 IFN_MASK_SCATTER_STORE.
4121 (internal_fn_mask_index): Likewise.
4122 (internal_fn_stored_value_index): New function.
4123 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
4125 * optabs-query.h (supports_vec_scatter_store_p): Declare.
4126 * optabs-query.c (supports_vec_scatter_store_p): New function.
4127 * tree-vectorizer.h (vect_get_store_rhs): Declare.
4128 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
4129 true for scatter stores.
4130 (vect_gather_scatter_fn_p): Handle scatter stores too.
4131 (vect_check_gather_scatter): Consider using scatter stores if
4132 supports_vec_scatter_store_p.
4133 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
4135 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4136 internal_fn_stored_value_index.
4137 (check_load_store_masking): Handle scatter stores too.
4138 (vect_get_store_rhs): Make public.
4139 (vectorizable_call): Use internal_store_fn_p.
4140 (vectorizable_store): Handle scatter store internal functions.
4141 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
4142 when deciding whether the end of the group has been reached.
4143 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
4144 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
4145 (mask_scatter_store<mode>): New insns.
4147 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4148 Alan Hayward <alan.hayward@arm.com>
4149 David Sherwood <david.sherwood@arm.com>
4151 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
4152 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
4153 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
4155 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
4156 Use vect_truncate_gather_scatter_offset if we can't treat the
4157 operation as a normal gather load or scatter store.
4158 (get_group_load_store_type): Take the gather_scatter_info
4159 as argument. Try using a gather load or scatter store for
4160 single-element groups.
4161 (get_load_store_type): Update calls to get_group_load_store_type
4162 and vect_use_strided_gather_scatters_p.
4164 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4165 Alan Hayward <alan.hayward@arm.com>
4166 David Sherwood <david.sherwood@arm.com>
4168 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
4169 optional tree argument.
4170 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
4172 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
4173 but continue to use the current value as a fallback.
4174 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
4175 to compare the updates.
4176 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
4177 (get_load_store_type): Use it when handling a strided access.
4178 (vect_get_strided_load_store_ops): New function.
4179 (vect_get_data_ptr_increment): Likewise.
4180 (vectorizable_load): Handle strided gather loads. Always pass
4181 a step to vect_create_data_ref_ptr and bump_vector_ptr.
4183 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4184 Alan Hayward <alan.hayward@arm.com>
4185 David Sherwood <david.sherwood@arm.com>
4187 * doc/md.texi (gather_load@var{m}): Document.
4188 (mask_gather_load@var{m}): Likewise.
4189 * genopinit.c (main): Add supports_vec_gather_load and
4190 supports_vec_gather_load_cached to target_optabs.
4191 * optabs-tree.c (init_tree_optimization_optabs): Use
4192 ggc_cleared_alloc to allocate target_optabs.
4193 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
4194 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
4196 * internal-fn.h (internal_load_fn_p): Declare.
4197 (internal_gather_scatter_fn_p): Likewise.
4198 (internal_fn_mask_index): Likewise.
4199 (internal_gather_scatter_fn_supported_p): Likewise.
4200 * internal-fn.c (gather_load_direct): New macro.
4201 (expand_gather_load_optab_fn): New function.
4202 (direct_gather_load_optab_supported_p): New macro.
4203 (direct_internal_fn_optab): New function.
4204 (internal_load_fn_p): Likewise.
4205 (internal_gather_scatter_fn_p): Likewise.
4206 (internal_fn_mask_index): Likewise.
4207 (internal_gather_scatter_fn_supported_p): Likewise.
4208 * optabs-query.c (supports_at_least_one_mode_p): New function.
4209 (supports_vec_gather_load_p): Likewise.
4210 * optabs-query.h (supports_vec_gather_load_p): Declare.
4211 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
4212 and memory_type field.
4213 (NUM_PATTERNS): Bump to 15.
4214 * tree-vect-data-refs.c: Include internal-fn.h.
4215 (vect_gather_scatter_fn_p): New function.
4216 (vect_describe_gather_scatter_call): Likewise.
4217 (vect_check_gather_scatter): Try using internal functions for
4218 gather loads. Recognize existing calls to a gather load function.
4219 (vect_analyze_data_refs): Consider using gather loads if
4220 supports_vec_gather_load_p.
4221 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
4222 (vect_get_gather_scatter_offset_type): Likewise.
4223 (vect_convert_mask_for_vectype): Likewise.
4224 (vect_add_conversion_to_patterm): Likewise.
4225 (vect_try_gather_scatter_pattern): Likewise.
4226 (vect_recog_gather_scatter_pattern): New pattern recognizer.
4227 (vect_vect_recog_func_ptrs): Add it.
4228 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4229 internal_fn_mask_index and internal_gather_scatter_fn_p.
4230 (check_load_store_masking): Take the gather_scatter_info as an
4231 argument and handle gather loads.
4232 (vect_get_gather_scatter_ops): New function.
4233 (vectorizable_call): Check internal_load_fn_p.
4234 (vectorizable_load): Likewise. Handle gather load internal
4236 (vectorizable_store): Update call to check_load_store_masking.
4237 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
4238 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
4239 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
4240 (aarch64_gather_scale_operand_d): New predicates.
4241 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
4242 (mask_gather_load<mode>): New insns.
4244 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4245 Alan Hayward <alan.hayward@arm.com>
4246 David Sherwood <david.sherwood@arm.com>
4248 * optabs.def (fold_left_plus_optab): New optab.
4249 * doc/md.texi (fold_left_plus_@var{m}): Document.
4250 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
4251 * internal-fn.c (fold_left_direct): Define.
4252 (expand_fold_left_optab_fn): Likewise.
4253 (direct_fold_left_optab_supported_p): Likewise.
4254 * fold-const-call.c (fold_const_fold_left): New function.
4255 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
4256 * tree-parloops.c (valid_reduction_p): New function.
4257 (gather_scalar_reductions): Use it.
4258 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
4259 (vect_finish_replace_stmt): Declare.
4260 * tree-vect-loop.c (fold_left_reduction_fn): New function.
4261 (needs_fold_left_reduction_p): New function, split out from...
4262 (vect_is_simple_reduction): ...here. Accept reductions that
4263 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
4264 (vect_force_simple_reduction): Also store the reduction type in
4265 the assignment's STMT_VINFO_REDUC_TYPE.
4266 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
4267 (merge_with_identity): New function.
4268 (vect_expand_fold_left): Likewise.
4269 (vectorize_fold_left_reduction): Likewise.
4270 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
4271 scalar phi in place for it. Check for target support and reject
4272 cases that would reassociate the operation. Defer the transform
4273 phase to vectorize_fold_left_reduction.
4274 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
4275 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
4276 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
4278 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4280 * tree-if-conv.c (predicate_mem_writes): Remove redundant
4281 call to ifc_temp_var.
4283 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4284 Alan Hayward <alan.hayward@arm.com>
4285 David Sherwood <david.sherwood@arm.com>
4287 * target.def (legitimize_address_displacement): Take the original
4288 offset as a poly_int.
4289 * targhooks.h (default_legitimize_address_displacement): Update
4291 * targhooks.c (default_legitimize_address_displacement): Likewise.
4292 * doc/tm.texi: Regenerate.
4293 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
4294 as an argument, moving assert of ad->disp == ad->disp_term to...
4295 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
4296 Try calling targetm.legitimize_address_displacement before expanding
4297 the address rather than afterwards, and adjust for the new interface.
4298 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
4299 Match the new hook interface. Handle SVE addresses.
4300 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
4303 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4305 * Makefile.in (OBJS): Add early-remat.o.
4306 * target.def (select_early_remat_modes): New hook.
4307 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
4308 * doc/tm.texi: Regenerate.
4309 * targhooks.h (default_select_early_remat_modes): Declare.
4310 * targhooks.c (default_select_early_remat_modes): New function.
4311 * timevar.def (TV_EARLY_REMAT): New timevar.
4312 * passes.def (pass_early_remat): New pass.
4313 * tree-pass.h (make_pass_early_remat): Declare.
4314 * early-remat.c: New file.
4315 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
4317 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
4319 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4320 Alan Hayward <alan.hayward@arm.com>
4321 David Sherwood <david.sherwood@arm.com>
4323 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
4324 vfm1 with a bound_epilog parameter.
4325 (vect_do_peeling): Update calls accordingly, and move the prologue
4326 call earlier in the function. Treat the base bound_epilog as 0 for
4327 fully-masked loops and retain vf - 1 for other loops. Add 1 to
4328 this base when peeling for gaps.
4329 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
4330 with fully-masked loops.
4331 (vect_estimate_min_profitable_iters): Handle the single peeled
4332 iteration in that case.
4334 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4335 Alan Hayward <alan.hayward@arm.com>
4336 David Sherwood <david.sherwood@arm.com>
4338 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
4339 single-element interleaving even if the size is not a power of 2.
4340 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
4341 accesses for single-element interleaving if the group size is
4344 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4345 Alan Hayward <alan.hayward@arm.com>
4346 David Sherwood <david.sherwood@arm.com>
4348 * doc/md.texi (fold_extract_last_@var{m}): Document.
4349 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
4350 * optabs.def (fold_extract_last_optab): New optab.
4351 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
4352 * internal-fn.c (fold_extract_direct): New macro.
4353 (expand_fold_extract_optab_fn): Likewise.
4354 (direct_fold_extract_optab_supported_p): Likewise.
4355 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
4356 * tree-vect-loop.c (vect_model_reduction_cost): Handle
4357 EXTRACT_LAST_REDUCTION.
4358 (get_initial_def_for_reduction): Do not create an initial vector
4359 for EXTRACT_LAST_REDUCTION reductions.
4360 (vectorizable_reduction): Leave the scalar phi in place for
4361 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
4362 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
4363 epilogue code for EXTRACT_LAST_REDUCTION and defer the
4364 transform phase to vectorizable_condition.
4365 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
4367 (vect_finish_stmt_generation): ...here.
4368 (vect_finish_replace_stmt): New function.
4369 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
4370 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
4372 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
4374 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4375 Alan Hayward <alan.hayward@arm.com>
4376 David Sherwood <david.sherwood@arm.com>
4378 * doc/md.texi (extract_last_@var{m}): Document.
4379 * optabs.def (extract_last_optab): New optab.
4380 * internal-fn.def (EXTRACT_LAST): New internal function.
4381 * internal-fn.c (cond_unary_direct): New macro.
4382 (expand_cond_unary_optab_fn): Likewise.
4383 (direct_cond_unary_optab_supported_p): Likewise.
4384 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
4385 loops using EXTRACT_LAST.
4386 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
4387 (extract_last_<mode>): ...this optab.
4388 (vec_extract<mode><Vel>): Update accordingly.
4390 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4391 Alan Hayward <alan.hayward@arm.com>
4392 David Sherwood <david.sherwood@arm.com>
4394 * target.def (empty_mask_is_expensive): New hook.
4395 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
4396 * doc/tm.texi: Regenerate.
4397 * targhooks.h (default_empty_mask_is_expensive): Declare.
4398 * targhooks.c (default_empty_mask_is_expensive): New function.
4399 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
4400 if the target says that empty masks are expensive.
4401 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
4403 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
4405 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4406 Alan Hayward <alan.hayward@arm.com>
4407 David Sherwood <david.sherwood@arm.com>
4409 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
4410 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
4411 (vect_use_loop_mask_for_alignment_p): New function.
4412 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
4413 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
4414 niters_skip argument. Make sure that the first niters_skip elements
4415 of the first iteration are inactive.
4416 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
4417 Update call to vect_set_loop_masks_directly.
4418 (get_misalign_in_elems): New function, split out from...
4419 (vect_gen_prolog_loop_niters): ...here.
4420 (vect_update_init_of_dr): Take a code argument that specifies whether
4421 the adjustment should be added or subtracted.
4422 (vect_update_init_of_drs): Likewise.
4423 (vect_prepare_for_masked_peels): New function.
4424 (vect_do_peeling): Skip prologue peeling if we're using a mask
4425 instead. Update call to vect_update_inits_of_drs.
4426 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4428 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
4429 alignment. Do not include the number of peeled iterations in
4430 the minimum threshold in that case.
4431 (vectorizable_induction): Adjust the start value down by
4432 LOOP_VINFO_MASK_SKIP_NITERS iterations.
4433 (vect_transform_loop): Call vect_prepare_for_masked_peels.
4434 Take the number of skipped iterations into account when calculating
4436 * tree-vect-stmts.c (vect_gen_while_not): New function.
4438 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4439 Alan Hayward <alan.hayward@arm.com>
4440 David Sherwood <david.sherwood@arm.com>
4442 * doc/sourcebuild.texi (vect_fully_masked): Document.
4443 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
4445 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
4447 (vect_analyze_loop_2): ...here. Don't check the vectorization
4448 factor against the number of loop iterations if the loop is
4451 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4452 Alan Hayward <alan.hayward@arm.com>
4453 David Sherwood <david.sherwood@arm.com>
4455 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
4456 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
4457 (dump_groups): Update accordingly.
4458 (iv_use::mem_type): New member variable.
4459 (address_p): New function.
4460 (record_use): Add a mem_type argument and initialize the new
4462 (record_group_use): Add a mem_type argument. Use address_p.
4463 Remove obsolete null checks of base_object. Update call to record_use.
4464 (find_interesting_uses_op): Update call to record_group_use.
4465 (find_interesting_uses_cond): Likewise.
4466 (find_interesting_uses_address): Likewise.
4467 (get_mem_type_for_internal_fn): New function.
4468 (find_address_like_use): Likewise.
4469 (find_interesting_uses_stmt): Try find_address_like_use before
4470 calling find_interesting_uses_op.
4471 (addr_offset_valid_p): Use the iv mem_type field as the type
4472 of the addressed memory.
4473 (add_autoinc_candidates): Likewise.
4474 (get_address_cost): Likewise.
4475 (split_small_address_groups_p): Use address_p.
4476 (split_address_groups): Likewise.
4477 (add_iv_candidate_for_use): Likewise.
4478 (autoinc_possible_for_pair): Likewise.
4479 (rewrite_groups): Likewise.
4480 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
4481 (determine_group_iv_cost): Update after split of USE_ADDRESS.
4482 (get_alias_ptr_type_for_ptr_address): New function.
4483 (rewrite_use_address): Rewrite address uses in calls that were
4484 identified by find_address_like_use.
4486 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4487 Alan Hayward <alan.hayward@arm.com>
4488 David Sherwood <david.sherwood@arm.com>
4490 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
4492 * gimple-expr.h (is_gimple_addressable: Likewise.
4493 * gimple-expr.c (is_gimple_address): Likewise.
4494 * internal-fn.c (expand_call_mem_ref): New function.
4495 (expand_mask_load_optab_fn): Use it.
4496 (expand_mask_store_optab_fn): Likewise.
4498 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4499 Alan Hayward <alan.hayward@arm.com>
4500 David Sherwood <david.sherwood@arm.com>
4502 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
4503 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
4504 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
4505 (cond_umax@var{mode}): Document.
4506 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
4507 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
4508 (cond_umin_optab, cond_umax_optab): New optabs.
4509 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
4510 (COND_IOR, COND_XOR): New internal functions.
4511 * internal-fn.h (get_conditional_internal_fn): Declare.
4512 * internal-fn.c (cond_binary_direct): New macro.
4513 (expand_cond_binary_optab_fn): Likewise.
4514 (direct_cond_binary_optab_supported_p): Likewise.
4515 (get_conditional_internal_fn): New function.
4516 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
4517 Cope with reduction statements that are vectorized as calls rather
4519 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
4520 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
4521 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
4522 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4523 (UNSPEC_COND_EOR): New unspecs.
4524 (optab): Add mappings for them.
4525 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
4526 (sve_int_op, sve_fp_op): New int attributes.
4528 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4529 Alan Hayward <alan.hayward@arm.com>
4530 David Sherwood <david.sherwood@arm.com>
4532 * optabs.def (while_ult_optab): New optab.
4533 * doc/md.texi (while_ult@var{m}@var{n}): Document.
4534 * internal-fn.def (WHILE_ULT): New internal function.
4535 * internal-fn.h (direct_internal_fn_supported_p): New override
4536 that takes two types as argument.
4537 * internal-fn.c (while_direct): New macro.
4538 (expand_while_optab_fn): New function.
4539 (convert_optab_supported_p): Likewise.
4540 (direct_while_optab_supported_p): New macro.
4541 * wide-int.h (wi::udiv_ceil): New function.
4542 * tree-vectorizer.h (rgroup_masks): New structure.
4543 (vec_loop_masks): New typedef.
4544 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
4546 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
4547 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
4548 (vect_max_vf): New function.
4549 (slpeel_make_loop_iterate_ntimes): Delete.
4550 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
4551 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
4552 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
4553 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
4554 internal-fn.h, stor-layout.h and optabs-query.h.
4555 (vect_set_loop_mask): New function.
4556 (add_preheader_seq): Likewise.
4557 (add_header_seq): Likewise.
4558 (interleave_supported_p): Likewise.
4559 (vect_maybe_permute_loop_masks): Likewise.
4560 (vect_set_loop_masks_directly): Likewise.
4561 (vect_set_loop_condition_masked): Likewise.
4562 (vect_set_loop_condition_unmasked): New function, split out from
4563 slpeel_make_loop_iterate_ntimes.
4564 (slpeel_make_loop_iterate_ntimes): Rename to..
4565 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
4566 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
4567 (vect_do_peeling): Update call accordingly.
4568 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
4570 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4571 mask_compare_type, can_fully_mask_p and fully_masked_p.
4572 (release_vec_loop_masks): New function.
4573 (_loop_vec_info): Use it to free the loop masks.
4574 (can_produce_all_loop_masks_p): New function.
4575 (vect_get_max_nscalars_per_iter): Likewise.
4576 (vect_verify_full_masking): Likewise.
4577 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
4578 retries, and free the mask rgroups before retrying. Check loop-wide
4579 reasons for disallowing fully-masked loops. Make the final decision
4580 about whether use a fully-masked loop or not.
4581 (vect_estimate_min_profitable_iters): Do not assume that peeling
4582 for the number of iterations will be needed for fully-masked loops.
4583 (vectorizable_reduction): Disable fully-masked loops.
4584 (vectorizable_live_operation): Likewise.
4585 (vect_halve_mask_nunits): New function.
4586 (vect_double_mask_nunits): Likewise.
4587 (vect_record_loop_mask): Likewise.
4588 (vect_get_loop_mask): Likewise.
4589 (vect_transform_loop): Handle the case in which the final loop
4590 iteration might handle a partial vector. Call vect_set_loop_condition
4591 instead of slpeel_make_loop_iterate_ntimes.
4592 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
4593 (check_load_store_masking): New function.
4594 (prepare_load_store_mask): Likewise.
4595 (vectorizable_store): Handle fully-masked loops.
4596 (vectorizable_load): Likewise.
4597 (supportable_widening_operation): Use vect_halve_mask_nunits for
4599 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
4600 (vect_gen_while): New function.
4601 * config/aarch64/aarch64.md (umax<mode>3): New expander.
4602 (aarch64_uqdec<mode>): New insn.
4604 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4605 Alan Hayward <alan.hayward@arm.com>
4606 David Sherwood <david.sherwood@arm.com>
4608 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
4609 (reduc_xor_scal_optab): New optabs.
4610 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
4611 (reduc_xor_scal_@var{m}): Document.
4612 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
4613 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
4615 * fold-const-call.c (fold_const_call): Handle them.
4616 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
4617 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
4618 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
4619 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
4620 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
4621 (UNSPEC_XORV): New unspecs.
4622 (optab): Add entries for them.
4623 (BITWISEV): New int iterator.
4624 (bit_reduc_op): New int attributes.
4626 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4627 Alan Hayward <alan.hayward@arm.com>
4628 David Sherwood <david.sherwood@arm.com>
4630 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
4631 * internal-fn.def (VEC_SHL_INSERT): New internal function.
4632 * optabs.def (vec_shl_insert_optab): New optab.
4633 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
4634 (duplicate_and_interleave): Likewise.
4635 * tree-vect-loop.c: Include internal-fn.h.
4636 (neutral_op_for_slp_reduction): New function, split out from
4637 get_initial_defs_for_reduction.
4638 (get_initial_def_for_reduction): Handle option 2 for variable-length
4639 vectors by loading the neutral value into a vector and then shifting
4640 the initial value into element 0.
4641 (get_initial_defs_for_reduction): Replace the code argument with
4642 the neutral value calculated by neutral_op_for_slp_reduction.
4643 Use gimple_build_vector for constant-length vectors.
4644 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
4645 but the first group_size elements have a neutral value.
4646 Use duplicate_and_interleave otherwise.
4647 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
4648 Update call to get_initial_defs_for_reduction. Handle SLP
4649 reductions for variable-length vectors by creating one vector
4650 result for each scalar result, with the elements associated
4651 with other scalar results stubbed out with the neutral value.
4652 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
4653 Require IFN_VEC_SHL_INSERT for double reductions on
4654 variable-length vectors, or SLP reductions that have
4655 a neutral value. Require can_duplicate_and_interleave_p
4656 support for variable-length unchained SLP reductions if there
4657 is no neutral value, such as for MIN/MAX reductions. Also require
4658 the number of vector elements to be a multiple of the number of
4659 SLP statements when doing variable-length unchained SLP reductions.
4660 Update call to vect_create_epilog_for_reduction.
4661 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
4662 and remove initial values.
4663 (duplicate_and_interleave): Make public.
4664 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
4665 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
4667 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4668 Alan Hayward <alan.hayward@arm.com>
4669 David Sherwood <david.sherwood@arm.com>
4671 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
4672 (can_duplicate_and_interleave_p): New function.
4673 (vect_get_and_check_slp_defs): Take the vector of statements
4674 rather than just the current one. Remove excess parentheses.
4675 Restriction rejectinon of vect_constant_def and vect_external_def
4676 for variable-length vectors to boolean types, or types for which
4677 can_duplicate_and_interleave_p is false.
4678 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
4679 (duplicate_and_interleave): New function.
4680 (vect_get_constant_vectors): Use gimple_build_vector for
4681 constant-length vectors and suitable variable-length constant
4682 vectors. Use duplicate_and_interleave for other variable-length
4683 vectors. Don't defer the update when inserting new statements.
4685 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4686 Alan Hayward <alan.hayward@arm.com>
4687 David Sherwood <david.sherwood@arm.com>
4689 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
4690 min_profitable_iters doesn't go negative.
4692 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4693 Alan Hayward <alan.hayward@arm.com>
4694 David Sherwood <david.sherwood@arm.com>
4696 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
4697 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
4698 * optabs.def (vec_mask_load_lanes_optab): New optab.
4699 (vec_mask_store_lanes_optab): Likewise.
4700 * internal-fn.def (MASK_LOAD_LANES): New internal function.
4701 (MASK_STORE_LANES): Likewise.
4702 * internal-fn.c (mask_load_lanes_direct): New macro.
4703 (mask_store_lanes_direct): Likewise.
4704 (expand_mask_load_optab_fn): Handle masked operations.
4705 (expand_mask_load_lanes_optab_fn): New macro.
4706 (expand_mask_store_optab_fn): Handle masked operations.
4707 (expand_mask_store_lanes_optab_fn): New macro.
4708 (direct_mask_load_lanes_optab_supported_p): Likewise.
4709 (direct_mask_store_lanes_optab_supported_p): Likewise.
4710 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
4712 (vect_load_lanes_supported): Likewise.
4713 * tree-vect-data-refs.c (strip_conversion): New function.
4714 (can_group_stmts_p): Likewise.
4715 (vect_analyze_data_ref_accesses): Use it instead of checking
4716 for a pair of assignments.
4717 (vect_store_lanes_supported): Take a masked_p parameter.
4718 (vect_load_lanes_supported): Likewise.
4719 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
4720 vect_store_lanes_supported and vect_load_lanes_supported.
4721 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4722 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
4723 parameter. Don't allow gaps for masked accesses.
4724 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
4725 and vect_load_lanes_supported.
4726 (get_load_store_type): Take a masked_p parameter and update
4727 call to get_group_load_store_type.
4728 (vectorizable_store): Update call to get_load_store_type.
4729 Handle IFN_MASK_STORE_LANES.
4730 (vectorizable_load): Update call to get_load_store_type.
4731 Handle IFN_MASK_LOAD_LANES.
4733 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4734 Alan Hayward <alan.hayward@arm.com>
4735 David Sherwood <david.sherwood@arm.com>
4737 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
4739 * config/aarch64/aarch64-protos.h
4740 (aarch64_sve_struct_memory_operand_p): Declare.
4741 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
4742 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
4743 (VPRED, vpred): Handle SVE structure modes.
4744 * config/aarch64/constraints.md (Utx): New constraint.
4745 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
4746 (aarch64_sve_struct_nonimmediate_operand): New predicates.
4747 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
4748 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
4749 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
4750 structure modes. Split into pieces after RA.
4751 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
4752 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
4754 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
4755 SVE structure modes.
4756 (aarch64_classify_address): Likewise.
4757 (sizetochar): Move earlier in file.
4758 (aarch64_print_operand): Handle SVE register lists.
4759 (aarch64_array_mode): New function.
4760 (aarch64_sve_struct_memory_operand_p): Likewise.
4761 (TARGET_ARRAY_MODE): Redefine.
4763 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4764 Alan Hayward <alan.hayward@arm.com>
4765 David Sherwood <david.sherwood@arm.com>
4767 * target.def (array_mode): New target hook.
4768 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
4769 * doc/tm.texi: Regenerate.
4770 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
4771 * hooks.c (hook_optmode_mode_uhwi_none): New function.
4772 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
4774 * stor-layout.c (mode_for_array): Likewise. Support polynomial
4777 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4778 Alan Hayward <alan.hayward@arm.com>
4779 David Sherwood <david.sherwood@arm.com>
4781 * fold-const.c (fold_binary_loc): Check the argument types
4782 rather than the result type when testing for a vector operation.
4784 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4786 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
4787 * doc/tm.texi: Regenerate.
4789 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4790 Alan Hayward <alan.hayward@arm.com>
4791 David Sherwood <david.sherwood@arm.com>
4793 * doc/invoke.texi (-msve-vector-bits=): Document new option.
4794 (sve): Document new AArch64 extension.
4795 * doc/md.texi (w): Extend the description of the AArch64
4796 constraint to include SVE vectors.
4797 (Upl, Upa): Document new AArch64 predicate constraints.
4798 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
4800 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
4801 (msve-vector-bits=): New option.
4802 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
4803 SVE when these are disabled.
4804 (sve): New extension.
4805 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
4806 modes. Adjust their number of units based on aarch64_sve_vg.
4807 (MAX_BITSIZE_MODE_ANY_MODE): Define.
4808 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
4809 aarch64_addr_query_type.
4810 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
4811 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
4812 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
4813 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
4814 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
4815 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
4816 (aarch64_simd_imm_zero_p): Delete.
4817 (aarch64_check_zero_based_sve_index_immediate): Declare.
4818 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4819 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4820 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4821 (aarch64_sve_float_mul_immediate_p): Likewise.
4822 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4824 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
4825 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
4826 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
4827 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
4828 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
4829 (aarch64_regmode_natural_size): Likewise.
4830 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
4831 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
4833 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
4834 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
4835 for VG and the SVE predicate registers.
4836 (V_ALIASES): Add a "z"-prefixed alias.
4837 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
4838 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
4839 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
4840 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
4841 (REG_CLASS_NAMES): Add entries for them.
4842 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
4843 and the predicate registers.
4844 (aarch64_sve_vg): Declare.
4845 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
4846 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
4847 (REGMODE_NATURAL_SIZE): Define.
4848 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
4850 * config/aarch64/aarch64.c: Include cfgrtl.h.
4851 (simd_immediate_info): Add a constructor for series vectors,
4852 and an associated step field.
4853 (aarch64_sve_vg): New variable.
4854 (aarch64_dbx_register_number): Handle VG and the predicate registers.
4855 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
4856 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
4857 (VEC_ANY_DATA, VEC_STRUCT): New constants.
4858 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
4859 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
4860 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
4861 (aarch64_get_mask_mode): New functions.
4862 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
4863 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
4864 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
4865 predicate modes and predicate registers. Explicitly restrict
4866 GPRs to modes of 16 bytes or smaller. Only allow FP registers
4867 to store a vector mode if it is recognized by
4868 aarch64_classify_vector_mode.
4869 (aarch64_regmode_natural_size): New function.
4870 (aarch64_hard_regno_caller_save_mode): Return the original mode
4872 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
4873 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
4874 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
4875 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
4877 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
4878 does not overlap dest if the function is frame-related. Handle
4880 (aarch64_split_add_offset): New function.
4881 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
4882 them aarch64_add_offset.
4883 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
4884 and update call to aarch64_sub_sp.
4885 (aarch64_add_cfa_expression): New function.
4886 (aarch64_expand_prologue): Pass extra temporary registers to the
4887 functions above. Handle the case in which we need to emit new
4888 DW_CFA_expressions for registers that were originally saved
4889 relative to the stack pointer, but now have to be expressed
4890 relative to the frame pointer.
4891 (aarch64_output_mi_thunk): Pass extra temporary registers to the
4893 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
4894 IP0 and IP1 values for SVE frames.
4895 (aarch64_expand_vec_series): New function.
4896 (aarch64_expand_sve_widened_duplicate): Likewise.
4897 (aarch64_expand_sve_const_vector): Likewise.
4898 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
4899 Handle SVE constants. Use emit_move_insn to move a force_const_mem
4900 into the register, rather than emitting a SET directly.
4901 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
4902 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
4903 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
4904 (offset_9bit_signed_scaled_p): New functions.
4905 (aarch64_replicate_bitmask_imm): New function.
4906 (aarch64_bitmask_imm): Use it.
4907 (aarch64_cannot_force_const_mem): Reject expressions involving
4908 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
4909 (aarch64_classify_index): Handle SVE indices, by requiring
4910 a plain register index with a scale that matches the element size.
4911 (aarch64_classify_address): Handle SVE addresses. Assert that
4912 the mode of the address is VOIDmode or an integer mode.
4913 Update call to aarch64_classify_symbol.
4914 (aarch64_classify_symbolic_expression): Update call to
4915 aarch64_classify_symbol.
4916 (aarch64_const_vec_all_in_range_p): New function.
4917 (aarch64_print_vector_float_operand): Likewise.
4918 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
4919 "vN" for FP registers with SVE modes. Handle (const ...) vectors
4920 and the FP immediates 1.0 and 0.5.
4921 (aarch64_print_address_internal): Handle SVE addresses.
4922 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
4923 (aarch64_regno_regclass): Handle predicate registers.
4924 (aarch64_secondary_reload): Handle big-endian reloads of SVE
4926 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
4927 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
4928 (aarch64_convert_sve_vector_bits): New function.
4929 (aarch64_override_options): Use it to handle -msve-vector-bits=.
4930 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4932 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
4933 Handle SVE vector and predicate modes. Accept VL-based constants
4934 that need only one temporary register, and VL offsets that require
4935 no temporary registers.
4936 (aarch64_conditional_register_usage): Mark the predicate registers
4937 as fixed if SVE isn't available.
4938 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
4939 Return true for SVE vector and predicate modes.
4940 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
4941 rather than an unsigned int. Handle SVE modes.
4942 (aarch64_preferred_simd_mode): Update call accordingly. Handle
4944 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
4946 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4947 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4948 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4949 (aarch64_sve_float_mul_immediate_p): New functions.
4950 (aarch64_sve_valid_immediate): New function.
4951 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
4952 Explicitly reject structure modes. Check for INDEX constants.
4953 Handle PTRUE and PFALSE constants.
4954 (aarch64_check_zero_based_sve_index_immediate): New function.
4955 (aarch64_simd_imm_zero_p): Delete.
4956 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
4957 vector modes. Accept constants in the range of CNT[BHWD].
4958 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
4959 ask for an Advanced SIMD mode.
4960 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
4961 (aarch64_simd_vector_alignment): Handle SVE predicates.
4962 (aarch64_vectorize_preferred_vector_alignment): New function.
4963 (aarch64_simd_vector_alignment_reachable): Use it instead of
4965 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
4966 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
4968 (MAX_VECT_LEN): Delete.
4969 (expand_vec_perm_d): Add a vec_flags field.
4970 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
4971 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
4972 (aarch64_evpc_ext): Don't apply a big-endian lane correction
4974 (aarch64_evpc_rev): Rename to...
4975 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
4976 (aarch64_evpc_rev_global): New function.
4977 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
4978 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
4980 (aarch64_evpc_sve_tbl): New function.
4981 (aarch64_expand_vec_perm_const_1): Update after rename of
4982 aarch64_evpc_rev. Handle SVE permutes too, trying
4983 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
4984 than aarch64_evpc_tbl.
4985 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
4986 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
4987 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
4988 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
4989 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
4990 (aarch64_expand_sve_vcond): New functions.
4991 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
4992 of aarch64_vector_mode_p.
4993 (aarch64_dwarf_poly_indeterminate_value): New function.
4994 (aarch64_compute_pressure_classes): Likewise.
4995 (aarch64_can_change_mode_class): Likewise.
4996 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
4997 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
4998 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
4999 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
5000 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
5001 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
5002 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
5003 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
5005 (Dn, Dl, Dr): Accept const as well as const_vector.
5006 (Dz): Likewise. Compare against CONST0_RTX.
5007 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
5008 of "vector" where appropriate.
5009 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
5010 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
5011 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
5012 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
5013 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
5014 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
5015 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
5016 (v_int_equiv): Extend to SVE modes.
5017 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
5019 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
5020 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
5021 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
5022 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
5023 (SVE_COND_FP_CMP): New int iterators.
5024 (perm_hilo): Handle the new unpack unspecs.
5025 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
5027 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
5028 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
5029 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
5030 (aarch64_equality_operator, aarch64_constant_vector_operand)
5031 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
5032 (aarch64_sve_nonimmediate_operand): Likewise.
5033 (aarch64_sve_general_operand): Likewise.
5034 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
5035 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
5036 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
5037 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
5038 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
5039 (aarch64_sve_float_arith_immediate): Likewise.
5040 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
5041 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
5042 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
5043 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
5044 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
5045 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
5046 (aarch64_sve_float_arith_operand): Likewise.
5047 (aarch64_sve_float_arith_with_sub_operand): Likewise.
5048 (aarch64_sve_float_mul_operand): Likewise.
5049 (aarch64_sve_vec_perm_operand): Likewise.
5050 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
5051 (aarch64_mov_operand): Accept const_poly_int and const_vector.
5052 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
5053 as well as const_vector.
5054 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
5055 in file. Use CONST0_RTX and CONSTM1_RTX.
5056 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
5057 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
5058 Use aarch64_simd_imm_zero.
5059 * config/aarch64/aarch64-sve.md: New file.
5060 * config/aarch64/aarch64.md: Include it.
5061 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
5062 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
5063 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
5064 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
5065 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
5066 (sve): New attribute.
5067 (enabled): Disable instructions with the sve attribute unless
5069 (movqi, movhi): Pass CONST_POLY_INT operaneds through
5070 aarch64_expand_mov_immediate.
5071 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
5072 CNT[BHSD] immediates.
5073 (movti): Split CONST_POLY_INT moves into two halves.
5074 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
5075 Split additions that need a temporary here if the destination
5076 is the stack pointer.
5077 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
5078 (*add<mode>3_poly_1): New instruction.
5079 (set_clobber_cc): New expander.
5081 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5083 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
5084 parameter and use it instead of GET_MODE_SIZE (innermode). Use
5085 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
5086 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
5087 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
5088 Change innermode from fixed_mode_size to machine_mode.
5089 (simplify_subreg): Update call accordingly. Handle a constant-sized
5090 subreg of a variable-length CONST_VECTOR.
5092 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5093 Alan Hayward <alan.hayward@arm.com>
5094 David Sherwood <david.sherwood@arm.com>
5096 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
5097 (add_offset_to_base): New function, split out from...
5098 (create_mem_ref): ...here. When handling a scale other than 1,
5099 check first whether the address is valid without the offset.
5100 Add it into the base if so, leaving the index and scale as-is.
5102 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5105 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
5106 fold_for_warn before checking if arg2 is INTEGER_CST.
5108 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
5110 * config/rs6000/predicates.md (load_multiple_operation): Delete.
5111 (store_multiple_operation): Delete.
5112 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
5113 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
5114 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
5115 guarded by TARGET_STRING.
5116 (rs6000_output_load_multiple): Delete.
5117 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
5118 OPTION_MASK_STRING / TARGET_STRING handling.
5119 (print_operand) <'N', 'O'>: Add comment that these are unused now.
5120 (const rs6000_opt_masks) <"string">: Change mask to 0.
5121 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
5122 (MASK_STRING): Delete.
5123 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
5125 (load_multiple): Delete.
5132 (store_multiple): Delete.
5139 (movmemsi_8reg): Delete.
5140 (corresponding unnamed define_insn): Delete.
5141 (movmemsi_6reg): Delete.
5142 (corresponding unnamed define_insn): Delete.
5143 (movmemsi_4reg): Delete.
5144 (corresponding unnamed define_insn): Delete.
5145 (movmemsi_2reg): Delete.
5146 (corresponding unnamed define_insn): Delete.
5147 (movmemsi_1reg): Delete.
5148 (corresponding unnamed define_insn): Delete.
5149 * config/rs6000/rs6000.opt (mno-string): New.
5150 (mstring): Replace by deprecation warning stub.
5151 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
5153 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5155 * regrename.c (regrename_do_replace): If replacing the same
5156 reg multiple times, try to reuse last created gen_raw_REG.
5159 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
5160 main to workaround a bug in GDB.
5162 2018-01-12 Tom de Vries <tom@codesourcery.com>
5165 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
5167 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
5169 PR rtl-optimization/80481
5170 * ira-color.c (get_cap_member): New function.
5171 (allocnos_conflict_by_live_ranges_p): Use it.
5172 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
5173 (setup_slot_coalesced_allocno_live_ranges): Ditto.
5175 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
5178 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
5179 (*saddl_se_1): Ditto.
5181 (*ssubl_se_1): Ditto.
5183 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5185 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
5186 rather than wi::to_widest for DR_INITs.
5187 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
5188 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
5189 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
5191 (vect_analyze_group_access_1): Note that here.
5193 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5195 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
5196 polynomial type sizes.
5198 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5200 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
5201 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
5202 (gimple_add_tmp_var): Likewise.
5204 2018-01-12 Martin Liska <mliska@suse.cz>
5206 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
5207 (gimple_alloc_sizes): Likewise.
5208 (dump_gimple_statistics): Use PRIu64 in printf format.
5209 * gimple.h: Change uint64_t to int.
5211 2018-01-12 Martin Liska <mliska@suse.cz>
5213 * tree-core.h: Use uint64_t instead of int.
5214 * tree.c (tree_node_counts): Likewise.
5215 (tree_node_sizes): Likewise.
5216 (dump_tree_statistics): Use PRIu64 in printf format.
5218 2018-01-12 Martin Liska <mliska@suse.cz>
5220 * Makefile.in: As qsort_chk is implemented in vec.c, add
5221 vec.o to linkage of gencfn-macros.
5222 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
5223 passing the info to record_node_allocation_statistics.
5224 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
5226 * ggc-common.c (struct ggc_usage): Add operator== and use
5227 it in operator< and compare function.
5228 * mem-stats.h (struct mem_usage): Likewise.
5229 * vec.c (struct vec_usage): Remove operator< and compare
5230 function. Can be simply inherited.
5232 2018-01-12 Martin Jambor <mjambor@suse.cz>
5235 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
5236 * tree-ssa-math-opts.c: Include domwalk.h.
5237 (convert_mult_to_fma_1): New function.
5238 (fma_transformation_info): New type.
5239 (fma_deferring_state): Likewise.
5240 (cancel_fma_deferring): New function.
5241 (result_of_phi): Likewise.
5242 (last_fma_candidate_feeds_initial_phi): Likewise.
5243 (convert_mult_to_fma): Added deferring logic, split actual
5244 transformation to convert_mult_to_fma_1.
5245 (math_opts_dom_walker): New type.
5246 (math_opts_dom_walker::after_dom_children): New method, body moved
5247 here from pass_optimize_widening_mul::execute, added deferring logic
5249 (pass_optimize_widening_mul::execute): Moved most of code to
5250 math_opts_dom_walker::after_dom_children.
5251 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
5252 * config/i386/i386.c (ix86_option_override_internal): Added
5253 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
5255 2018-01-12 Richard Biener <rguenther@suse.de>
5258 * dwarf2out.c (gen_variable_die): Do not reset old_die for
5259 inline instance vars.
5261 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
5264 * config/rx/rx.c (rx_is_restricted_memory_address):
5267 2018-01-12 Richard Biener <rguenther@suse.de>
5269 PR tree-optimization/80846
5270 * target.def (split_reduction): New target hook.
5271 * targhooks.c (default_split_reduction): New function.
5272 * targhooks.h (default_split_reduction): Declare.
5273 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
5274 target requests first reduce vectors by combining low and high
5276 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
5277 (get_vectype_for_scalar_type_and_size): Export.
5278 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
5279 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
5280 * doc/tm.texi: Regenerate.
5281 * config/i386/i386.c (ix86_split_reduction): Implement
5282 TARGET_VECTORIZE_SPLIT_REDUCTION.
5284 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5287 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
5288 in PIC mode except for TARGET_VXWORKS_RTP.
5289 * config/sparc/sparc.c: Include cfgrtl.h.
5290 (TARGET_INIT_PIC_REG): Define.
5291 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
5292 (sparc_pic_register_p): New predicate.
5293 (sparc_legitimate_address_p): Use it.
5294 (sparc_legitimize_pic_address): Likewise.
5295 (sparc_delegitimize_address): Likewise.
5296 (sparc_mode_dependent_address_p): Likewise.
5297 (gen_load_pcrel_sym): Remove 4th parameter.
5298 (load_got_register): Adjust call to above. Remove obsolete stuff.
5299 (sparc_expand_prologue): Do not call load_got_register here.
5300 (sparc_flat_expand_prologue): Likewise.
5301 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
5302 (sparc_use_pseudo_pic_reg): New function.
5303 (sparc_init_pic_reg): Likewise.
5304 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5305 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
5307 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
5309 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
5310 Add item for branch_cost.
5312 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5314 PR rtl-optimization/83565
5315 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
5316 not extend the result to a larger mode for rotate operations.
5317 (num_sign_bit_copies1): Likewise.
5319 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5322 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
5324 Use values-Xc.o for -pedantic.
5325 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
5327 2018-01-12 Martin Liska <mliska@suse.cz>
5330 * ipa-devirt.c (final_warning_record::grow_type_warnings):
5332 (possible_polymorphic_call_targets): Use it.
5333 (ipa_devirt): Likewise.
5335 2018-01-12 Martin Liska <mliska@suse.cz>
5337 * profile-count.h (enum profile_quality): Use 0 as invalid
5338 enum value of profile_quality.
5340 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
5342 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
5343 -mext-string options.
5345 2018-01-12 Richard Biener <rguenther@suse.de>
5347 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
5348 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
5349 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
5351 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
5353 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5355 * configure.ac (--with-long-double-format): Add support for the
5356 configuration option to change the default long double format on
5358 * config.gcc (powerpc*-linux*-*): Likewise.
5359 * configure: Regenerate.
5360 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
5361 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
5362 used without modification.
5364 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5366 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
5367 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
5368 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
5369 MISC_BUILTIN_SPEC_BARRIER.
5370 (rs6000_init_builtins): Likewise.
5371 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
5373 (speculation_barrier): New define_insn.
5374 * doc/extend.texi: Document __builtin_speculation_barrier.
5376 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5379 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
5380 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
5381 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
5383 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
5384 integral modes instead of "ss" and "sd".
5385 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
5386 vectors with 32-bit and 64-bit elements.
5387 (vecdupssescalarmodesuffix): New mode attribute.
5388 (vec_dup<mode>): Use it.
5390 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
5393 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
5394 frame if argument is passed on stack.
5396 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5399 * ree.c (combine_reaching_defs): Optimize also
5400 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
5401 reg2=any_extend(exp); reg1=reg2;, formatting fix.
5403 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5406 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
5408 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5411 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
5412 after they are computed.
5414 2018-01-11 Bin Cheng <bin.cheng@arm.com>
5416 PR tree-optimization/83695
5417 * gimple-loop-linterchange.cc
5418 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
5419 reset cached scev information after interchange.
5420 (pass_linterchange::execute): Remove call to scev_reset_htab.
5422 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5424 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
5425 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
5426 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
5427 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
5428 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
5429 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
5430 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
5431 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
5432 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
5433 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
5434 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
5435 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
5436 (V_lane_reg): Likewise.
5437 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
5439 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
5440 (vfmal_lane_low<mode>_intrinsic,
5441 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
5442 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
5443 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
5444 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
5445 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
5446 vfmsl_lane_high<mode>_intrinsic): New define_insns.
5448 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5450 * config/arm/arm-cpus.in (fp16fml): New feature.
5451 (ALL_SIMD): Add fp16fml.
5452 (armv8.2-a): Add fp16fml as an option.
5453 (armv8.3-a): Likewise.
5454 (armv8.4-a): Add fp16fml as part of fp16.
5455 * config/arm/arm.h (TARGET_FP16FML): Define.
5456 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
5458 * config/arm/arm-modes.def (V2HF): Define.
5459 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
5460 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
5461 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
5462 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
5463 vfmsl_low, vfmsl_high): New set of builtins.
5464 * config/arm/iterators.md (PLUSMINUS): New code iterator.
5465 (vfml_op): New code attribute.
5466 (VFMLHALVES): New int iterator.
5467 (VFML, VFMLSEL): New mode attributes.
5468 (V_reg): Define mapping for V2HF.
5469 (V_hi, V_lo): New mode attributes.
5470 (VF_constraint): Likewise.
5471 (vfml_half, vfml_half_selector): New int attributes.
5472 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
5474 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
5475 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
5477 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
5478 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
5479 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
5480 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
5482 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
5483 Document new effective target and option set.
5485 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5487 * config/arm/arm-cpus.in (armv8_4): New feature.
5488 (ARMv8_4a): New fgroup.
5489 (armv8.4-a): New arch.
5490 * config/arm/arm-tables.opt: Regenerate.
5491 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
5492 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
5493 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
5494 Add matching rules for -march=armv8.4-a and extensions.
5495 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
5497 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
5500 * config/rx/rx.md (BW): New mode attribute.
5501 (sync_lock_test_and_setsi): Add mode suffix to insn output.
5503 2018-01-11 Richard Biener <rguenther@suse.de>
5505 PR tree-optimization/83435
5506 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
5507 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
5508 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
5510 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5511 Alan Hayward <alan.hayward@arm.com>
5512 David Sherwood <david.sherwood@arm.com>
5514 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
5516 (aarch64_classify_address): Initialize it. Track polynomial offsets.
5517 (aarch64_print_address_internal): Use it to check for a zero offset.
5519 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5520 Alan Hayward <alan.hayward@arm.com>
5521 David Sherwood <david.sherwood@arm.com>
5523 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
5524 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
5525 Return a poly_int64 rather than a HOST_WIDE_INT.
5526 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
5527 rather than a HOST_WIDE_INT.
5528 * config/aarch64/aarch64.h (aarch64_frame): Protect with
5529 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
5530 hard_fp_offset, frame_size, initial_adjust, callee_offset and
5531 final_offset from HOST_WIDE_INT to poly_int64.
5532 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
5533 to_constant when getting the number of units in an Advanced SIMD
5535 (aarch64_builtin_vectorized_function): Check for a constant number
5537 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
5539 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
5540 attribute instead of GET_MODE_NUNITS.
5541 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
5542 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
5543 GET_MODE_SIZE for fixed-size registers.
5544 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
5545 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
5546 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
5547 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
5548 (aarch64_print_operand, aarch64_print_address_internal)
5549 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
5550 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
5551 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
5552 Handle polynomial GET_MODE_SIZE.
5553 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
5554 wider than SImode without modification.
5555 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
5556 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
5557 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
5558 passing and returning SVE modes.
5559 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
5560 rather than GEN_INT.
5561 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
5562 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
5563 (aarch64_allocate_and_probe_stack_space): Likewise.
5564 (aarch64_layout_frame): Cope with polynomial offsets.
5565 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
5566 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
5568 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
5569 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
5570 poly_int64 rather than a HOST_WIDE_INT.
5571 (aarch64_get_separate_components, aarch64_process_components)
5572 (aarch64_expand_prologue, aarch64_expand_epilogue)
5573 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
5574 (aarch64_anchor_offset): New function, split out from...
5575 (aarch64_legitimize_address): ...here.
5576 (aarch64_builtin_vectorization_cost): Handle polynomial
5577 TYPE_VECTOR_SUBPARTS.
5578 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
5580 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
5581 number of elements from the PARALLEL rather than the mode.
5582 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
5583 rather than GET_MODE_BITSIZE.
5584 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
5585 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
5586 (aarch64_expand_vec_perm_const_1): Handle polynomial
5587 d->perm.length () and d->perm elements.
5588 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
5589 Apply to_constant to d->perm elements.
5590 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
5591 polynomial CONST_VECTOR_NUNITS.
5592 (aarch64_move_pointer): Take amount as a poly_int64 rather
5594 (aarch64_progress_pointer): Avoid temporary variable.
5595 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
5596 the mode attribute instead of GET_MODE.
5598 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5599 Alan Hayward <alan.hayward@arm.com>
5600 David Sherwood <david.sherwood@arm.com>
5602 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
5603 x exists before using it.
5604 (aarch64_add_constant_internal): Rename to...
5605 (aarch64_add_offset_1): ...this. Replace regnum with separate
5606 src and dest rtxes. Handle the case in which they're different,
5607 including when the offset is zero. Replace scratchreg with an rtx.
5608 Use 2 additions if there is no spare register into which we can
5609 move a 16-bit constant.
5610 (aarch64_add_constant): Delete.
5611 (aarch64_add_offset): Replace reg with separate src and dest
5612 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
5613 Use aarch64_add_offset_1.
5614 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
5615 an rtx rather than an int. Take the delta as a poly_int64
5616 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
5617 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
5618 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
5619 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
5620 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
5622 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
5623 aarch64_add_constant.
5625 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5627 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
5628 Use scalar_float_mode.
5630 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5632 * config/aarch64/aarch64-simd.md
5633 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
5634 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
5635 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
5636 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
5637 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
5638 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
5639 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
5640 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
5641 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
5642 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
5644 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5647 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
5648 targ_options->x_arm_arch_string is non NULL.
5650 2018-01-11 Tamar Christina <tamar.christina@arm.com>
5652 * config/aarch64/aarch64.h
5653 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
5655 2018-01-11 Sudakshina Das <sudi.das@arm.com>
5658 * expmed.c (emit_store_flag_force): Swap if const op0
5659 and change VOIDmode to mode of op0.
5661 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5663 PR rtl-optimization/83761
5664 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
5665 than bytes to mode_for_size.
5667 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5670 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
5671 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
5674 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5677 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
5678 when in layout mode.
5679 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
5680 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
5683 2018-01-10 Michael Collison <michael.collison@arm.com>
5685 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
5686 * config/aarch64/aarch64-option-extension.def: Add
5687 AARCH64_OPT_EXTENSION of 'fp16fml'.
5688 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5689 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
5690 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
5691 * config/aarch64/constraints.md (Ui7): New constraint.
5692 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
5693 (VFMLA_SEL_W): Ditto.
5696 (VFMLA16_LOW): New int iterator.
5697 (VFMLA16_HIGH): Ditto.
5698 (UNSPEC_FMLAL): New unspec.
5699 (UNSPEC_FMLSL): Ditto.
5700 (UNSPEC_FMLAL2): Ditto.
5701 (UNSPEC_FMLSL2): Ditto.
5702 (f16mac): New code attribute.
5703 * config/aarch64/aarch64-simd-builtins.def
5704 (aarch64_fmlal_lowv2sf): Ditto.
5705 (aarch64_fmlsl_lowv2sf): Ditto.
5706 (aarch64_fmlalq_lowv4sf): Ditto.
5707 (aarch64_fmlslq_lowv4sf): Ditto.
5708 (aarch64_fmlal_highv2sf): Ditto.
5709 (aarch64_fmlsl_highv2sf): Ditto.
5710 (aarch64_fmlalq_highv4sf): Ditto.
5711 (aarch64_fmlslq_highv4sf): Ditto.
5712 (aarch64_fmlal_lane_lowv2sf): Ditto.
5713 (aarch64_fmlsl_lane_lowv2sf): Ditto.
5714 (aarch64_fmlal_laneq_lowv2sf): Ditto.
5715 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
5716 (aarch64_fmlalq_lane_lowv4sf): Ditto.
5717 (aarch64_fmlsl_lane_lowv4sf): Ditto.
5718 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
5719 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
5720 (aarch64_fmlal_lane_highv2sf): Ditto.
5721 (aarch64_fmlsl_lane_highv2sf): Ditto.
5722 (aarch64_fmlal_laneq_highv2sf): Ditto.
5723 (aarch64_fmlsl_laneq_highv2sf): Ditto.
5724 (aarch64_fmlalq_lane_highv4sf): Ditto.
5725 (aarch64_fmlsl_lane_highv4sf): Ditto.
5726 (aarch64_fmlalq_laneq_highv4sf): Ditto.
5727 (aarch64_fmlsl_laneq_highv4sf): Ditto.
5728 * config/aarch64/aarch64-simd.md:
5729 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
5730 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5731 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
5732 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5733 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
5734 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
5735 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
5736 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
5737 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
5738 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
5739 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
5740 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
5741 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
5742 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
5743 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
5744 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
5745 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
5746 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
5747 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
5748 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
5749 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
5750 (vfmlsl_low_u32): Ditto.
5751 (vfmlalq_low_u32): Ditto.
5752 (vfmlslq_low_u32): Ditto.
5753 (vfmlal_high_u32): Ditto.
5754 (vfmlsl_high_u32): Ditto.
5755 (vfmlalq_high_u32): Ditto.
5756 (vfmlslq_high_u32): Ditto.
5757 (vfmlal_lane_low_u32): Ditto.
5758 (vfmlsl_lane_low_u32): Ditto.
5759 (vfmlal_laneq_low_u32): Ditto.
5760 (vfmlsl_laneq_low_u32): Ditto.
5761 (vfmlalq_lane_low_u32): Ditto.
5762 (vfmlslq_lane_low_u32): Ditto.
5763 (vfmlalq_laneq_low_u32): Ditto.
5764 (vfmlslq_laneq_low_u32): Ditto.
5765 (vfmlal_lane_high_u32): Ditto.
5766 (vfmlsl_lane_high_u32): Ditto.
5767 (vfmlal_laneq_high_u32): Ditto.
5768 (vfmlsl_laneq_high_u32): Ditto.
5769 (vfmlalq_lane_high_u32): Ditto.
5770 (vfmlslq_lane_high_u32): Ditto.
5771 (vfmlalq_laneq_high_u32): Ditto.
5772 (vfmlslq_laneq_high_u32): Ditto.
5773 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
5774 (AARCH64_FL_FOR_ARCH8_4): New.
5775 (AARCH64_ISA_F16FML): New ISA flag.
5776 (TARGET_F16FML): New feature flag for fp16fml.
5777 (doc/invoke.texi): Document new fp16fml option.
5779 2018-01-10 Michael Collison <michael.collison@arm.com>
5781 * config/aarch64/aarch64-builtins.c:
5782 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
5783 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5784 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
5785 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
5786 (AARCH64_ISA_SHA3): New ISA flag.
5787 (TARGET_SHA3): New feature flag for sha3.
5788 * config/aarch64/iterators.md (sha512_op): New int attribute.
5789 (CRYPTO_SHA512): New int iterator.
5790 (UNSPEC_SHA512H): New unspec.
5791 (UNSPEC_SHA512H2): Ditto.
5792 (UNSPEC_SHA512SU0): Ditto.
5793 (UNSPEC_SHA512SU1): Ditto.
5794 * config/aarch64/aarch64-simd-builtins.def
5795 (aarch64_crypto_sha512hqv2di): New builtin.
5796 (aarch64_crypto_sha512h2qv2di): Ditto.
5797 (aarch64_crypto_sha512su0qv2di): Ditto.
5798 (aarch64_crypto_sha512su1qv2di): Ditto.
5799 (aarch64_eor3qv8hi): Ditto.
5800 (aarch64_rax1qv2di): Ditto.
5801 (aarch64_xarqv2di): Ditto.
5802 (aarch64_bcaxqv8hi): Ditto.
5803 * config/aarch64/aarch64-simd.md:
5804 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
5805 (aarch64_crypto_sha512su0qv2di): Ditto.
5806 (aarch64_crypto_sha512su1qv2di): Ditto.
5807 (aarch64_eor3qv8hi): Ditto.
5808 (aarch64_rax1qv2di): Ditto.
5809 (aarch64_xarqv2di): Ditto.
5810 (aarch64_bcaxqv8hi): Ditto.
5811 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
5812 (vsha512h2q_u64): Ditto.
5813 (vsha512su0q_u64): Ditto.
5814 (vsha512su1q_u64): Ditto.
5815 (veor3q_u16): Ditto.
5816 (vrax1q_u64): Ditto.
5818 (vbcaxq_u16): Ditto.
5819 * config/arm/types.md (crypto_sha512): New type attribute.
5820 (crypto_sha3): Ditto.
5821 (doc/invoke.texi): Document new sha3 option.
5823 2018-01-10 Michael Collison <michael.collison@arm.com>
5825 * config/aarch64/aarch64-builtins.c:
5826 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
5827 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5828 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
5829 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
5830 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
5831 (AARCH64_ISA_SM4): New ISA flag.
5832 (TARGET_SM4): New feature flag for sm4.
5833 * config/aarch64/aarch64-simd-builtins.def
5834 (aarch64_sm3ss1qv4si): Ditto.
5835 (aarch64_sm3tt1aq4si): Ditto.
5836 (aarch64_sm3tt1bq4si): Ditto.
5837 (aarch64_sm3tt2aq4si): Ditto.
5838 (aarch64_sm3tt2bq4si): Ditto.
5839 (aarch64_sm3partw1qv4si): Ditto.
5840 (aarch64_sm3partw2qv4si): Ditto.
5841 (aarch64_sm4eqv4si): Ditto.
5842 (aarch64_sm4ekeyqv4si): Ditto.
5843 * config/aarch64/aarch64-simd.md:
5844 (aarch64_sm3ss1qv4si): Ditto.
5845 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
5846 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
5847 (aarch64_sm4eqv4si): Ditto.
5848 (aarch64_sm4ekeyqv4si): Ditto.
5849 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
5850 (sm3part_op): Ditto.
5851 (CRYPTO_SM3TT): Ditto.
5852 (CRYPTO_SM3PART): Ditto.
5853 (UNSPEC_SM3SS1): New unspec.
5854 (UNSPEC_SM3TT1A): Ditto.
5855 (UNSPEC_SM3TT1B): Ditto.
5856 (UNSPEC_SM3TT2A): Ditto.
5857 (UNSPEC_SM3TT2B): Ditto.
5858 (UNSPEC_SM3PARTW1): Ditto.
5859 (UNSPEC_SM3PARTW2): Ditto.
5860 (UNSPEC_SM4E): Ditto.
5861 (UNSPEC_SM4EKEY): Ditto.
5862 * config/aarch64/constraints.md (Ui2): New constraint.
5863 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
5864 * config/arm/types.md (crypto_sm3): New type attribute.
5865 (crypto_sm4): Ditto.
5866 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
5867 (vsm3tt1aq_u32): Ditto.
5868 (vsm3tt1bq_u32): Ditto.
5869 (vsm3tt2aq_u32): Ditto.
5870 (vsm3tt2bq_u32): Ditto.
5871 (vsm3partw1q_u32): Ditto.
5872 (vsm3partw2q_u32): Ditto.
5873 (vsm4eq_u32): Ditto.
5874 (vsm4ekeyq_u32): Ditto.
5875 (doc/invoke.texi): Document new sm4 option.
5877 2018-01-10 Michael Collison <michael.collison@arm.com>
5879 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
5880 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
5881 (AARCH64_FL_FOR_ARCH8_4): New.
5882 (AARCH64_FL_V8_4): New flag.
5883 (doc/invoke.texi): Document new armv8.4-a option.
5885 2018-01-10 Michael Collison <michael.collison@arm.com>
5887 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5888 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
5889 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
5890 * config/aarch64/aarch64-option-extension.def: Add
5891 AARCH64_OPT_EXTENSION of 'sha2'.
5892 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
5893 (crypto): Disable sha2 and aes if crypto disabled.
5894 (crypto): Enable aes and sha2 if enabled.
5895 (simd): Disable sha2 and aes if simd disabled.
5896 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
5898 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
5899 (TARGET_SHA2): New feature flag for sha2.
5900 (TARGET_AES): New feature flag for aes.
5901 * config/aarch64/aarch64-simd.md:
5902 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
5903 conditional on TARGET_AES.
5904 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
5905 (aarch64_crypto_sha1hsi): Make pattern conditional
5907 (aarch64_crypto_sha1hv4si): Ditto.
5908 (aarch64_be_crypto_sha1hv4si): Ditto.
5909 (aarch64_crypto_sha1su1v4si): Ditto.
5910 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
5911 (aarch64_crypto_sha1su0v4si): Ditto.
5912 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
5913 (aarch64_crypto_sha256su0v4si): Ditto.
5914 (aarch64_crypto_sha256su1v4si): Ditto.
5915 (doc/invoke.texi): Document new aes and sha2 options.
5917 2018-01-10 Martin Sebor <msebor@redhat.com>
5919 PR tree-optimization/83781
5920 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
5923 2018-01-11 Martin Sebor <msebor@gmail.com>
5924 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5926 PR tree-optimization/83501
5927 PR tree-optimization/81703
5929 * tree-ssa-strlen.c (get_string_cst): Rename...
5930 (get_string_len): ...to this. Handle global constants.
5931 (handle_char_store): Adjust.
5933 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
5934 Jim Wilson <jimw@sifive.com>
5936 * config/riscv/riscv-protos.h (riscv_output_return): New.
5937 * config/riscv/riscv.c (struct machine_function): New naked_p field.
5938 (riscv_attribute_table, riscv_output_return),
5939 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
5940 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
5941 (riscv_compute_frame_info): Only compute frame->mask if not a naked
5943 (riscv_expand_prologue): Add early return for naked function.
5944 (riscv_expand_epilogue): Likewise.
5945 (riscv_function_ok_for_sibcall): Return false for naked function.
5946 (riscv_set_current_function): New.
5947 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
5948 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
5949 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
5950 * doc/extend.texi (RISC-V Function Attributes): New.
5952 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
5954 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
5955 check for 128-bit long double before checking TCmode.
5956 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
5957 128-bit long doubles before checking TFmode or TCmode.
5958 (FLOAT128_IBM_P): Likewise.
5960 2018-01-10 Martin Sebor <msebor@redhat.com>
5962 PR tree-optimization/83671
5963 * builtins.c (c_strlen): Unconditionally return zero for the empty
5965 Use -Warray-bounds for warnings.
5966 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
5967 for non-constant array indices with COMPONENT_REF, arrays of
5968 arrays, and pointers to arrays.
5969 (gimple_fold_builtin_strlen): Determine and set length range for
5970 non-constant character arrays.
5972 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
5975 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
5978 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
5980 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
5982 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5985 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
5986 VECTOR_MEM_ALTIVEC_OR_VSX_P.
5987 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
5988 indexed_or_indirect_operand predicate.
5989 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
5990 (*vsx_le_perm_load_v8hi): Likewise.
5991 (*vsx_le_perm_load_v16qi): Likewise.
5992 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
5993 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
5994 (*vsx_le_perm_store_v8hi): Likewise.
5995 (*vsx_le_perm_store_v16qi): Likewise.
5996 (eight unnamed splitters): Likewise.
5998 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6000 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
6001 * config/rs6000/emmintrin.h: Likewise.
6002 * config/rs6000/mmintrin.h: Likewise.
6003 * config/rs6000/xmmintrin.h: Likewise.
6005 2018-01-10 David Malcolm <dmalcolm@redhat.com>
6008 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
6010 * tree.c (tree_nop_conversion): Return true for location wrapper
6012 (maybe_wrap_with_location): New function.
6013 (selftest::check_strip_nops): New function.
6014 (selftest::test_location_wrappers): New function.
6015 (selftest::tree_c_tests): Call it.
6016 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
6017 (maybe_wrap_with_location): New decl.
6018 (EXPR_LOCATION_WRAPPER_P): New macro.
6019 (location_wrapper_p): New inline function.
6020 (tree_strip_any_location_wrapper): New inline function.
6022 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
6025 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
6026 stack_realign_offset for the largest alignment of stack slot
6028 (ix86_find_max_used_stack_alignment): New function.
6029 (ix86_finalize_stack_frame_flags): Use it. Set
6030 max_used_stack_alignment if we don't realign stack.
6031 * config/i386/i386.h (machine_function): Add
6032 max_used_stack_alignment.
6034 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
6036 * config/arm/arm.opt (-mbranch-cost): New option.
6037 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
6040 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
6043 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
6044 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
6046 2018-01-10 Richard Biener <rguenther@suse.de>
6049 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
6050 early out so it also covers the case where we have a non-NULL
6053 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6055 PR tree-optimization/83753
6056 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
6057 for non-strided grouped accesses if the number of elements is 1.
6059 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6062 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
6063 * i386.h (TARGET_USE_GATHER): Define.
6064 * x86-tune.def (X86_TUNE_USE_GATHER): New.
6066 2018-01-10 Martin Liska <mliska@suse.cz>
6069 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
6070 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
6072 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
6073 CLEANUP_NO_PARTITIONING is not set.
6075 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6077 * doc/rtl.texi: Remove documentation of (const ...) wrappers
6078 for vectors, as a partial revert of r254296.
6079 * rtl.h (const_vec_p): Delete.
6080 (const_vec_duplicate_p): Don't test for vector CONSTs.
6081 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
6082 * expmed.c (make_tree): Likewise.
6085 * common.md (E, F): Use CONSTANT_P instead of checking for
6087 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
6088 checking for CONST_VECTOR.
6090 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6093 * predict.c (force_edge_cold): Handle in more sane way edges
6096 2018-01-09 Carl Love <cel@us.ibm.com>
6098 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
6100 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
6101 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
6102 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
6103 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
6104 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
6105 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
6106 * config/rs6000/rs6000-protos.h: Add extern defition for
6107 rs6000_generate_float2_double_code.
6108 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
6110 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
6111 (float2_v2df): Add define_expand.
6113 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
6116 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
6117 op_mode in the force_to_mode call.
6119 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6121 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
6122 instead of checking each element individually.
6123 (aarch64_evpc_uzp): Likewise.
6124 (aarch64_evpc_zip): Likewise.
6125 (aarch64_evpc_ext): Likewise.
6126 (aarch64_evpc_rev): Likewise.
6127 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
6128 instead of checking each element individually. Return true without
6130 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
6131 whether all selected elements come from the same input, instead of
6132 checking each element individually. Remove calls to gen_rtx_REG,
6133 start_sequence and end_sequence and instead assert that no rtl is
6136 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6138 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
6139 order of HIGH and CONST checks.
6141 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6143 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
6144 if the destination isn't an SSA_NAME.
6146 2018-01-09 Richard Biener <rguenther@suse.de>
6148 PR tree-optimization/83668
6149 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
6151 (canonicalize_loop_form): ... here, renamed from ...
6152 (canonicalize_loop_closed_ssa_form): ... this and amended to
6153 swap successor edges for loop exit blocks to make us use
6154 the RPO order we need for initial schedule generation.
6156 2018-01-09 Joseph Myers <joseph@codesourcery.com>
6158 PR tree-optimization/64811
6159 * match.pd: When optimizing comparisons with Inf, avoid
6160 introducing or losing exceptions from comparisons with NaN.
6162 2018-01-09 Martin Liska <mliska@suse.cz>
6165 * asan.c (shadow_mem_size): Add gcc_assert.
6167 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
6169 Don't save registers in main().
6172 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
6173 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
6174 * config/avr/avr.c (avr_set_current_function): Don't error if
6175 naked, OS_task or OS_main are specified at the same time.
6176 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
6178 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
6180 * common/config/avr/avr-common.c (avr_option_optimization_table):
6181 Switch on -mmain-is-OS_task for optimizing compilations.
6183 2018-01-09 Richard Biener <rguenther@suse.de>
6185 PR tree-optimization/83572
6186 * graphite.c: Include cfganal.h.
6187 (graphite_transform_loops): Connect infinite loops to exit
6188 and remove fake edges at the end.
6190 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6192 * ipa-inline.c (edge_badness): Revert accidental checkin.
6194 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6197 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
6198 symbols; not inline clones.
6200 2018-01-09 Jakub Jelinek <jakub@redhat.com>
6203 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
6204 hard registers. Formatting fixes.
6206 PR preprocessor/83722
6207 * gcc.c (try_generate_repro): Pass
6208 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
6209 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
6212 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
6213 Kito Cheng <kito.cheng@gmail.com>
6215 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
6216 (riscv_leaf_function_p): Delete.
6217 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
6219 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6221 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
6223 (do_ifelse): New function.
6224 (do_isel): New function.
6225 (do_sub3): New function.
6226 (do_add3): New function.
6227 (do_load_mask_compare): New function.
6228 (do_overlap_load_compare): New function.
6229 (expand_compare_loop): New function.
6230 (expand_block_compare): Call expand_compare_loop() when appropriate.
6231 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
6233 (-mblock-compare-inline-loop-limit): New option.
6235 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6238 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
6239 Reverse order of second and third operands in first alternative.
6240 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
6241 of first and second elements in UNSPEC_VPERMR vector.
6242 (altivec_expand_vec_perm_le): Likewise.
6244 2018-01-08 Jeff Law <law@redhat.com>
6246 PR rtl-optimizatin/81308
6247 * tree-switch-conversion.c (cfg_altered): New file scoped static.
6248 (process_switch): If group_case_labels makes a change, then set
6250 (pass_convert_switch::execute): If a switch is converted, then
6251 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
6253 PR rtl-optimization/81308
6254 * recog.c (split_all_insns): Conditionally cleanup the CFG after
6257 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
6259 PR target/83663 - Revert r255946
6260 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
6261 generation for cases where splatting a value is not useful.
6262 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
6263 across a vec_duplicate and a paradoxical subreg forming a vector
6264 mode to a vec_concat.
6266 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6268 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
6269 -march=armv8.3-a variants.
6270 * config/arm/t-multilib: Likewise.
6271 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
6273 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6275 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
6277 (cceq_ior_compare_complement): Give it a name so I can use it, and
6278 change boolean_or_operator predicate to boolean_operator so it can
6279 be used to generate a crand.
6280 (eqne): New code iterator.
6281 (bd/bd_neg): New code_attrs.
6282 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
6283 a single define_insn.
6284 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
6285 decrement (bdnzt/bdnzf/bdzt/bdzf).
6286 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
6287 with the new names of the branch decrement patterns, and added the
6288 names of the branch decrement conditional patterns.
6290 2018-01-08 Richard Biener <rguenther@suse.de>
6292 PR tree-optimization/83563
6293 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
6296 2018-01-08 Richard Biener <rguenther@suse.de>
6299 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
6301 2018-01-08 Richard Biener <rguenther@suse.de>
6303 PR tree-optimization/83685
6304 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
6305 references to abnormals.
6307 2018-01-08 Richard Biener <rguenther@suse.de>
6310 * dwarf2out.c (output_indirect_strings): Handle empty
6311 skeleton_debug_str_hash.
6312 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
6314 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6316 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
6317 (emit_store_direct): Likewise.
6318 (arc_trampoline_adjust_address): Likewise.
6319 (arc_asm_trampoline_template): New function.
6320 (arc_initialize_trampoline): Use asm_trampoline_template.
6321 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
6322 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
6323 * config/arc/arc.md (flush_icache): Delete pattern.
6325 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6327 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
6328 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
6331 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6334 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
6335 by not USED_FOR_TARGET.
6336 (make_pass_resolve_sw_modes): Likewise.
6338 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6340 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
6343 2018-01-08 Richard Biener <rguenther@suse.de>
6346 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
6348 2018-01-08 Richard Biener <rguenther@suse.de>
6351 * match.pd ((t * 2) / 2) -> t): Add missing :c.
6353 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
6356 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
6357 basic blocks with a small number of successors.
6358 (convert_control_dep_chain_into_preds): Improve handling of
6360 (dump_predicates): Split apart into...
6361 (dump_pred_chain): ...here...
6362 (dump_pred_info): ...and here.
6363 (can_one_predicate_be_invalidated_p): Add debugging printfs.
6364 (can_chain_union_be_invalidated_p): Improve check for invalidation
6366 (uninit_uses_cannot_happen): Avoid unnecessary if
6367 convert_control_dep_chain_into_preds yielded nothing.
6369 2018-01-06 Martin Sebor <msebor@redhat.com>
6371 PR tree-optimization/83640
6372 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
6373 subtracting negative offset from size.
6374 (builtin_access::overlap): Adjust offset bounds of the access to fall
6375 within the size of the object if possible.
6377 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
6379 PR rtl-optimization/83699
6380 * expmed.c (extract_bit_field_1): Restrict the vector usage of
6381 extract_bit_field_as_subreg to cases in which the extracted
6382 value is also a vector.
6384 * lra-constraints.c (process_alt_operands): Test for the equivalence
6385 substitutions when detecting a possible reload cycle.
6387 2018-01-06 Jakub Jelinek <jakub@redhat.com>
6390 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
6391 by default if flag_selective_schedling{,2}. Formatting fixes.
6393 PR rtl-optimization/83682
6394 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
6395 if it has non-VECTOR_MODE element mode.
6396 (vec_duplicate_p): Likewise.
6399 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
6400 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
6402 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6405 * config/i386/i386-builtin.def
6406 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
6407 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
6408 Require also OPTION_MASK_ISA_AVX512F in addition to
6409 OPTION_MASK_ISA_GFNI.
6410 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
6411 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
6412 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
6413 to OPTION_MASK_ISA_GFNI.
6414 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
6415 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
6416 OPTION_MASK_ISA_AVX512BW.
6417 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
6418 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
6419 addition to OPTION_MASK_ISA_GFNI.
6420 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
6421 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
6422 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
6423 to OPTION_MASK_ISA_GFNI.
6424 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
6425 a requirement for all ISAs rather than any of them with a few
6427 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
6429 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
6430 bitmasks to be enabled with 3 exceptions, instead of requiring any
6431 enabled ISA with lots of exceptions.
6432 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
6433 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
6434 Change avx512bw in isa attribute to avx512f.
6435 * config/i386/sgxintrin.h: Add license boilerplate.
6436 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
6437 to __AVX512F__ and __AVX512VL to __AVX512VL__.
6438 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
6439 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
6441 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
6442 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
6443 temporarily sse2 rather than sse if not enabled already.
6446 * config/i386/sse.md (VI248_VLBW): Rename to ...
6447 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
6448 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
6449 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
6450 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
6451 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
6452 mode iterator instead of VI248_VLBW.
6454 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
6456 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
6457 (record_modified): Skip clobbers; add debug output.
6458 (param_change_prob): Use sreal frequencies.
6460 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6462 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
6463 punt for user-aligned variables.
6465 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6467 * tree-chrec.c (chrec_contains_symbols): Return true for
6470 2018-01-05 Sudakshina Das <sudi.das@arm.com>
6473 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
6474 of (x|y) == x for BICS pattern.
6476 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6478 PR tree-optimization/83605
6479 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
6480 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
6483 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
6485 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
6486 * config/epiphany/rtems.h: New file.
6488 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6489 Uros Bizjak <ubizjak@gmail.com>
6492 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
6493 QIreg_operand instead of register_operand predicate.
6494 * config/i386/i386.c (ix86_rop_should_change_byte_p,
6495 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
6496 comments instead of -fmitigate[-_]rop.
6498 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6501 * cgraphunit.c (symbol_table::compile): Switch to text_section
6502 before calling assembly_start debug hook.
6503 * run-rtl-passes.c (run_rtl_passes): Likewise.
6506 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6508 * tree-vrp.c (extract_range_from_binary_expr_1): Check
6509 range_int_cst_p rather than !symbolic_range_p before calling
6510 extract_range_from_multiplicative_op_1.
6512 2018-01-04 Jeff Law <law@redhat.com>
6514 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
6515 redundant test in assertion.
6517 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6519 * doc/rtl.texi: Document machine_mode wrapper classes.
6521 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6523 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
6526 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6528 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
6529 the VEC_PERM_EXPR fold to fail.
6531 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6534 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
6535 to switched_sections.
6537 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6540 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
6543 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
6546 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
6547 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
6549 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6552 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
6553 is BLKmode and bitpos not zero or mode change is needed.
6555 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6558 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
6561 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
6564 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
6565 instead of MULT rtx. Update all corresponding splitters.
6567 (*ssub<modesuffix>): Ditto.
6569 (*cmp_sadd_di): Update split patterns.
6570 (*cmp_sadd_si): Ditto.
6571 (*cmp_sadd_sidi): Ditto.
6572 (*cmp_ssub_di): Ditto.
6573 (*cmp_ssub_si): Ditto.
6574 (*cmp_ssub_sidi): Ditto.
6575 * config/alpha/predicates.md (const23_operand): New predicate.
6576 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
6577 Look for ASHIFT, not MULT inner operand.
6578 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
6580 2018-01-04 Martin Liska <mliska@suse.cz>
6582 PR gcov-profile/83669
6583 * gcov.c (output_intermediate_file): Add version to intermediate
6585 * doc/gcov.texi: Document new field 'version' in intermediate
6586 file format. Fix location of '-k' option of gcov command.
6588 2018-01-04 Martin Liska <mliska@suse.cz>
6591 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
6593 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6595 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
6597 2018-01-03 Martin Sebor <msebor@redhat.com>
6599 PR tree-optimization/83655
6600 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
6601 checking calls with invalid arguments.
6603 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6605 * tree-vect-stmts.c (vect_get_store_rhs): New function.
6606 (vectorizable_mask_load_store): Delete.
6607 (vectorizable_call): Return false for masked loads and stores.
6608 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
6609 instead of gimple_assign_rhs1.
6610 (vectorizable_load): Handle IFN_MASK_LOAD.
6611 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
6613 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6615 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
6617 (vectorizable_mask_load_store): ...here.
6618 (vectorizable_load): ...and here.
6620 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6622 * tree-vect-stmts.c (vect_build_all_ones_mask)
6623 (vect_build_zero_merge_argument): New functions, split out from...
6624 (vectorizable_load): ...here.
6626 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6628 * tree-vect-stmts.c (vect_check_store_rhs): New function,
6630 (vectorizable_mask_load_store): ...here.
6631 (vectorizable_store): ...and here.
6633 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6635 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
6637 (vectorizable_mask_load_store): ...here.
6639 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6641 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
6642 (vect_model_store_cost): Take a vec_load_store_type instead of a
6644 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
6645 (vect_model_store_cost): Take a vec_load_store_type instead of a
6647 (vectorizable_mask_load_store): Update accordingly.
6648 (vectorizable_store): Likewise.
6649 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
6651 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6653 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
6654 IFN_MASK_LOAD calls here rather than...
6655 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
6657 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6658 Alan Hayward <alan.hayward@arm.com>
6659 David Sherwood <david.sherwood@arm.com>
6661 * expmed.c (extract_bit_field_1): For vector extracts,
6662 fall back to extract_bit_field_as_subreg if vec_extract
6665 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6666 Alan Hayward <alan.hayward@arm.com>
6667 David Sherwood <david.sherwood@arm.com>
6669 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
6670 they are variable or constant sized.
6671 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
6672 slots for constant-sized data.
6674 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6675 Alan Hayward <alan.hayward@arm.com>
6676 David Sherwood <david.sherwood@arm.com>
6678 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
6679 handling COND_EXPRs with boolean comparisons, try to find a better
6680 basis for the mask type than the boolean itself.
6682 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6684 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
6685 is calculated and how it can be overridden.
6686 * genmodes.c (max_bitsize_mode_any_mode): New variable.
6687 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
6689 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
6692 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6693 Alan Hayward <alan.hayward@arm.com>
6694 David Sherwood <david.sherwood@arm.com>
6696 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
6697 Remove the mode argument.
6698 (aarch64_simd_valid_immediate): Remove the mode and inverse
6700 * config/aarch64/iterators.md (bitsize): New iterator.
6701 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
6702 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
6703 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
6704 aarch64_simd_valid_immediate.
6705 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
6706 (aarch64_reg_or_bic_imm): Likewise.
6707 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
6708 with an insn_type enum and msl with a modifier_type enum.
6709 Replace element_width with a scalar_mode. Change the shift
6710 to unsigned int. Add constructors for scalar_float_mode and
6711 scalar_int_mode elements.
6712 (aarch64_vect_float_const_representable_p): Delete.
6713 (aarch64_can_const_movi_rtx_p)
6714 (aarch64_simd_scalar_immediate_valid_for_move)
6715 (aarch64_simd_make_constant): Update call to
6716 aarch64_simd_valid_immediate.
6717 (aarch64_advsimd_valid_immediate_hs): New function.
6718 (aarch64_advsimd_valid_immediate): Likewise.
6719 (aarch64_simd_valid_immediate): Remove mode and inverse
6720 arguments. Rewrite to use the above. Use const_vec_duplicate_p
6721 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
6722 and aarch64_float_const_representable_p on the result.
6723 (aarch64_output_simd_mov_immediate): Remove mode argument.
6724 Update call to aarch64_simd_valid_immediate and use of
6725 simd_immediate_info.
6726 (aarch64_output_scalar_simd_mov_immediate): Update call
6729 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6730 Alan Hayward <alan.hayward@arm.com>
6731 David Sherwood <david.sherwood@arm.com>
6733 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
6734 (mode_nunits): Likewise CONST_MODE_NUNITS.
6735 * machmode.def (ADJUST_NUNITS): Document.
6736 * genmodes.c (mode_data::need_nunits_adj): New field.
6737 (blank_mode): Update accordingly.
6738 (adj_nunits): New variable.
6739 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
6741 (emit_mode_size_inline): Set need_bytesize_adj for all modes
6742 listed in adj_nunits.
6743 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
6744 listed in adj_nunits. Don't emit case statements for such modes.
6745 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
6746 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
6747 nothing if adj_nunits is nonnull.
6748 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
6749 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
6750 (emit_mode_fbit): Update use of print_maybe_const_decl.
6751 (emit_move_size): Likewise. Treat the array as non-const
6753 (emit_mode_adjustments): Handle adj_nunits.
6755 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6757 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
6758 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
6759 (VECTOR_MODES): Use it.
6760 (make_vector_modes): Take the prefix as an argument.
6762 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6763 Alan Hayward <alan.hayward@arm.com>
6764 David Sherwood <david.sherwood@arm.com>
6766 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
6767 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
6768 for MODE_VECTOR_BOOL.
6769 * machmode.def (VECTOR_BOOL_MODE): Document.
6770 * genmodes.c (VECTOR_BOOL_MODE): New macro.
6771 (make_vector_bool_mode): New function.
6772 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
6774 * lto-streamer-in.c (lto_input_mode_table): Likewise.
6775 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
6777 * stor-layout.c (int_mode_for_mode): Likewise.
6778 * tree.c (build_vector_type_for_mode): Likewise.
6779 * varasm.c (output_constant_pool_2): Likewise.
6780 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
6781 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
6782 for MODE_VECTOR_BOOL.
6783 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
6784 of mode class checks.
6785 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
6786 instead of a list of mode class checks.
6787 (expand_vector_scalar_condition): Likewise.
6788 (type_for_widest_vector_mode): Handle BImode as an inner mode.
6790 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6791 Alan Hayward <alan.hayward@arm.com>
6792 David Sherwood <david.sherwood@arm.com>
6794 * machmode.h (mode_size): Change from unsigned short to
6796 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
6797 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6798 or if measurement_type is not polynomial.
6799 (fixed_size_mode::includes_p): Check for constant-sized modes.
6800 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
6801 return a poly_uint16 rather than an unsigned short.
6802 (emit_mode_size): Change the type of mode_size from unsigned short
6803 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
6804 (emit_mode_adjustments): Cope with polynomial vector sizes.
6805 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6807 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6809 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
6810 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
6811 * caller-save.c (setup_save_areas): Likewise.
6812 (replace_reg_with_saved_mem): Likewise.
6813 * calls.c (emit_library_call_value_1): Likewise.
6814 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
6815 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
6816 (gen_lowpart_for_combine): Likewise.
6817 * convert.c (convert_to_integer_1): Likewise.
6818 * cse.c (equiv_constant, cse_insn): Likewise.
6819 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
6820 (cselib_subst_to_values): Likewise.
6821 * dce.c (word_dce_process_block): Likewise.
6822 * df-problems.c (df_word_lr_mark_ref): Likewise.
6823 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
6824 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
6825 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
6826 (rtl_for_decl_location): Likewise.
6827 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
6828 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
6829 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
6830 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
6831 (expand_expr_real_1): Likewise.
6832 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
6833 (pad_below): Likewise.
6834 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6835 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
6836 * ira.c (get_subreg_tracking_sizes): Likewise.
6837 * ira-build.c (ira_create_allocno_objects): Likewise.
6838 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
6839 (ira_sort_regnos_for_alter_reg): Likewise.
6840 * ira-costs.c (record_operand_costs): Likewise.
6841 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
6842 (resolve_simple_move): Likewise.
6843 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
6844 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
6845 (lra_constraints): Likewise.
6846 (CONST_POOL_OK_P): Reject variable-sized modes.
6847 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
6848 (add_pseudo_to_slot, lra_spill): Likewise.
6849 * omp-low.c (omp_clause_aligned_alignment): Likewise.
6850 * optabs-query.c (get_best_extraction_insn): Likewise.
6851 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6852 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
6853 (expand_mult_highpart, valid_multiword_target_p): Likewise.
6854 * recog.c (offsettable_address_addr_space_p): Likewise.
6855 * regcprop.c (maybe_mode_change): Likewise.
6856 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
6857 * regrename.c (build_def_use): Likewise.
6858 * regstat.c (dump_reg_info): Likewise.
6859 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
6860 (find_reloads, find_reloads_subreg_address): Likewise.
6861 * reload1.c (eliminate_regs_1): Likewise.
6862 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
6863 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
6864 (simplify_binary_operation_1, simplify_subreg): Likewise.
6865 * targhooks.c (default_function_arg_padding): Likewise.
6866 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
6867 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
6868 (verify_gimple_assign_ternary): Likewise.
6869 * tree-inline.c (estimate_move_cost): Likewise.
6870 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6871 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
6872 (get_address_cost_ainc): Likewise.
6873 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
6874 (vect_supportable_dr_alignment): Likewise.
6875 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6876 (vectorizable_reduction): Likewise.
6877 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
6878 (vectorizable_operation, vectorizable_load): Likewise.
6879 * tree.c (build_same_sized_truth_vector_type): Likewise.
6880 * valtrack.c (cleanup_auto_inc_dec): Likewise.
6881 * var-tracking.c (emit_note_insn_var_location): Likewise.
6882 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
6883 (ADDR_VEC_ALIGN): Likewise.
6885 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6886 Alan Hayward <alan.hayward@arm.com>
6887 David Sherwood <david.sherwood@arm.com>
6889 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
6891 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6892 or if measurement_type is polynomial.
6893 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
6894 * combine.c (make_extraction): Likewise.
6895 * dse.c (find_shift_sequence): Likewise.
6896 * dwarf2out.c (mem_loc_descriptor): Likewise.
6897 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
6898 (extract_bit_field, extract_low_bits): Likewise.
6899 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
6900 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
6901 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
6902 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
6903 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6904 * reload.c (find_reloads): Likewise.
6905 * reload1.c (alter_reg): Likewise.
6906 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
6907 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
6908 * tree-if-conv.c (predicate_mem_writes): Likewise.
6909 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
6910 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
6911 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
6912 * valtrack.c (dead_debug_insert_temp): Likewise.
6913 * varasm.c (mergeable_constant_section): Likewise.
6914 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
6916 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6917 Alan Hayward <alan.hayward@arm.com>
6918 David Sherwood <david.sherwood@arm.com>
6920 * expr.c (expand_assignment): Cope with polynomial mode sizes
6921 when assigning to a CONCAT.
6923 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6924 Alan Hayward <alan.hayward@arm.com>
6925 David Sherwood <david.sherwood@arm.com>
6927 * machmode.h (mode_precision): Change from unsigned short to
6929 (mode_to_precision): Return a poly_uint16 rather than an unsigned
6931 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
6932 or if measurement_type is not polynomial.
6933 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
6934 in which the mode is already known to be a scalar_int_mode.
6935 * genmodes.c (emit_mode_precision): Change the type of mode_precision
6936 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
6938 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6939 for GET_MODE_PRECISION.
6940 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6941 for GET_MODE_PRECISION.
6942 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
6944 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
6945 (expand_field_assignment, make_extraction): Likewise.
6946 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
6947 (get_last_value): Likewise.
6948 * convert.c (convert_to_integer_1): Likewise.
6949 * cse.c (cse_insn): Likewise.
6950 * expr.c (expand_expr_real_1): Likewise.
6951 * lra-constraints.c (simplify_operand_subreg): Likewise.
6952 * optabs-query.c (can_atomic_load_p): Likewise.
6953 * optabs.c (expand_atomic_load): Likewise.
6954 (expand_atomic_store): Likewise.
6955 * ree.c (combine_reaching_defs): Likewise.
6956 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
6957 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
6958 * tree.h (type_has_mode_precision_p): Likewise.
6959 * ubsan.c (instrument_si_overflow): Likewise.
6961 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6962 Alan Hayward <alan.hayward@arm.com>
6963 David Sherwood <david.sherwood@arm.com>
6965 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
6966 polynomial numbers of units.
6967 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
6968 (valid_vector_subparts_p): New function.
6969 (build_vector_type): Remove temporary shim and take the number
6970 of units as a poly_uint64 rather than an int.
6971 (build_opaque_vector_type): Take the number of units as a
6972 poly_uint64 rather than an int.
6973 * tree.c (build_vector_from_ctor): Handle polynomial
6974 TYPE_VECTOR_SUBPARTS.
6975 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
6976 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
6977 (build_vector_from_val): If the number of units is variable,
6978 use build_vec_duplicate_cst for constant operands and
6979 VEC_DUPLICATE_EXPR otherwise.
6980 (make_vector_type): Remove temporary is_constant ().
6981 (build_vector_type, build_opaque_vector_type): Take the number of
6982 units as a poly_uint64 rather than an int.
6983 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
6985 * cfgexpand.c (expand_debug_expr): Likewise.
6986 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
6987 (store_constructor, expand_expr_real_1): Likewise.
6988 (const_scalar_mask_from_tree): Likewise.
6989 * fold-const-call.c (fold_const_reduction): Likewise.
6990 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
6991 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
6992 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
6993 (fold_relational_const): Likewise.
6994 (native_interpret_vector): Likewise. Change the size from an
6995 int to an unsigned int.
6996 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
6997 TYPE_VECTOR_SUBPARTS.
6998 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
6999 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
7000 duplicating a non-constant operand into a variable-length vector.
7001 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
7002 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
7003 * ipa-icf.c (sem_variable::equals): Likewise.
7004 * match.pd: Likewise.
7005 * omp-simd-clone.c (simd_clone_subparts): Likewise.
7006 * print-tree.c (print_node): Likewise.
7007 * stor-layout.c (layout_type): Likewise.
7008 * targhooks.c (default_builtin_vectorization_cost): Likewise.
7009 * tree-cfg.c (verify_gimple_comparison): Likewise.
7010 (verify_gimple_assign_binary): Likewise.
7011 (verify_gimple_assign_ternary): Likewise.
7012 (verify_gimple_assign_single): Likewise.
7013 * tree-pretty-print.c (dump_generic_node): Likewise.
7014 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7015 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
7016 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
7017 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
7018 (vect_shift_permute_load_chain): Likewise.
7019 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
7020 (expand_vector_condition, optimize_vector_constructor): Likewise.
7021 (lower_vec_perm, get_compute_type): Likewise.
7022 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7023 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
7024 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
7025 (vect_recog_mask_conversion_pattern): Likewise.
7026 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
7027 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
7028 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7029 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
7030 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
7031 (vectorizable_shift, vectorizable_operation, vectorizable_store)
7032 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
7033 (supportable_widening_operation): Likewise.
7034 (supportable_narrowing_operation): Likewise.
7035 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
7037 * varasm.c (output_constant): Likewise.
7039 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7040 Alan Hayward <alan.hayward@arm.com>
7041 David Sherwood <david.sherwood@arm.com>
7043 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
7044 so that both the length == 3 and length != 3 cases set up their
7045 own permute vectors. Add comments explaining why we know the
7046 number of elements is constant.
7047 (vect_permute_load_chain): Likewise.
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 * machmode.h (mode_nunits): Change from unsigned char to
7055 (ONLY_FIXED_SIZE_MODES): New macro.
7056 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
7057 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
7058 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
7060 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
7061 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
7062 or if measurement_type is not polynomial.
7063 * genmodes.c (ZERO_COEFFS): New macro.
7064 (emit_mode_nunits_inline): Make mode_nunits_inline return a
7066 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
7067 Use ZERO_COEFFS when emitting initializers.
7068 * data-streamer.h (bp_pack_poly_value): New function.
7069 (bp_unpack_poly_value): Likewise.
7070 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7071 for GET_MODE_NUNITS.
7072 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7073 for GET_MODE_NUNITS.
7074 * tree.c (make_vector_type): Remove temporary shim and make
7075 the real function take the number of units as a poly_uint64
7077 (build_vector_type_for_mode): Handle polynomial nunits.
7078 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
7079 * emit-rtl.c (const_vec_series_p_1): Likewise.
7080 (gen_rtx_CONST_VECTOR): Likewise.
7081 * fold-const.c (test_vec_duplicate_folding): Likewise.
7082 * genrecog.c (validate_pattern): Likewise.
7083 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
7084 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7085 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
7086 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
7087 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
7088 * rtlanal.c (subreg_get_info): Likewise.
7089 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7090 (vect_grouped_load_supported): Likewise.
7091 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
7092 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
7093 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
7094 (simplify_const_unary_operation, simplify_binary_operation_1)
7095 (simplify_const_binary_operation, simplify_ternary_operation)
7096 (test_vector_ops_duplicate, test_vector_ops): Likewise.
7097 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
7098 instead of CONST_VECTOR_NUNITS.
7099 * varasm.c (output_constant_pool_2): Likewise.
7100 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
7101 explicit-encoded elements in the XVEC for variable-length vectors.
7103 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7105 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
7107 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7108 Alan Hayward <alan.hayward@arm.com>
7109 David Sherwood <david.sherwood@arm.com>
7111 * coretypes.h (fixed_size_mode): Declare.
7112 (fixed_size_mode_pod): New typedef.
7113 * builtins.h (target_builtins::x_apply_args_mode)
7114 (target_builtins::x_apply_result_mode): Change type to
7115 fixed_size_mode_pod.
7116 * builtins.c (apply_args_size, apply_result_size, result_vector)
7117 (expand_builtin_apply_args_1, expand_builtin_apply)
7118 (expand_builtin_return): Update accordingly.
7120 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7122 * cse.c (hash_rtx_cb): Hash only the encoded elements.
7123 * cselib.c (cselib_hash_rtx): Likewise.
7124 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
7125 CONST_VECTOR encoding.
7127 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7128 Jeff Law <law@redhat.com>
7131 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
7132 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
7133 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
7134 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
7137 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
7138 explicitly probe *sp in a noreturn function if there were any callee
7139 register saves or frame pointer is needed.
7141 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7144 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
7145 BLKmode for ternary, binary or unary expressions.
7148 * var-tracking.c (delete_vta_debug_insn): New inline function.
7149 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
7150 insns from get_insns () to NULL instead of each bb separately.
7151 Use delete_vta_debug_insn. No longer static.
7152 (vt_debug_insns_local, variable_tracking_main_1): Adjust
7153 delete_vta_debug_insns callers.
7154 * rtl.h (delete_vta_debug_insns): Declare.
7155 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
7156 instead of variable_tracking_main.
7158 2018-01-03 Martin Sebor <msebor@redhat.com>
7160 PR tree-optimization/83603
7161 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
7162 arguments past the endof the argument list in functions declared
7163 without a prototype.
7164 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7165 Avoid checking when arguments are null.
7167 2018-01-03 Martin Sebor <msebor@redhat.com>
7170 * doc/extend.texi (attribute const): Fix a typo.
7171 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
7172 issuing -Wsuggest-attribute for void functions.
7174 2018-01-03 Martin Sebor <msebor@redhat.com>
7176 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
7177 offset_int::from instead of wide_int::to_shwi.
7178 (maybe_diag_overlap): Remove assertion.
7179 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
7180 * gimple-ssa-sprintf.c (format_directive): Same.
7181 (parse_directive): Same.
7182 (sprintf_dom_walker::compute_format_length): Same.
7183 (try_substitute_return_value): Same.
7185 2018-01-03 Jeff Law <law@redhat.com>
7188 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
7189 non-constant residual for zero at runtime and avoid probing in
7190 that case. Reorganize code for trailing problem to mirror handling
7193 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7195 PR tree-optimization/83501
7196 * tree-ssa-strlen.c (get_string_cst): New.
7197 (handle_char_store): Call get_string_cst.
7199 2018-01-03 Martin Liska <mliska@suse.cz>
7201 PR tree-optimization/83593
7202 * tree-ssa-strlen.c: Include tree-cfg.h.
7203 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
7204 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
7205 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
7207 (strlen_dom_walker::before_dom_children): Call
7208 gimple_purge_dead_eh_edges. Dump tranformation with details
7210 (strlen_dom_walker::before_dom_children): Update call by adding
7211 new argument cleanup_eh.
7212 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
7214 2018-01-03 Martin Liska <mliska@suse.cz>
7217 * cif-code.def (VARIADIC_THUNK): New enum value.
7218 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
7221 2018-01-03 Jan Beulich <jbeulich@suse.com>
7223 * sse.md (mov<mode>_internal): Tighten condition for when to use
7224 vmovdqu<ssescalarsize> for TI and OI modes.
7226 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7228 Update copyright years.
7230 2018-01-03 Martin Liska <mliska@suse.cz>
7233 * ipa-visibility.c (function_and_variable_visibility): Skip
7234 functions with noipa attribure.
7236 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7238 * gcc.c (process_command): Update copyright notice dates.
7239 * gcov-dump.c (print_version): Ditto.
7240 * gcov.c (print_version): Ditto.
7241 * gcov-tool.c (print_version): Ditto.
7242 * gengtype.c (create_file): Ditto.
7243 * doc/cpp.texi: Bump @copying's copyright year.
7244 * doc/cppinternals.texi: Ditto.
7245 * doc/gcc.texi: Ditto.
7246 * doc/gccint.texi: Ditto.
7247 * doc/gcov.texi: Ditto.
7248 * doc/install.texi: Ditto.
7249 * doc/invoke.texi: Ditto.
7251 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7253 * vector-builder.h (vector_builder::m_full_nelts): Change from
7254 unsigned int to poly_uint64.
7255 (vector_builder::full_nelts): Update prototype accordingly.
7256 (vector_builder::new_vector): Likewise.
7257 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
7258 (vector_builder::operator ==): Likewise.
7259 (vector_builder::finalize): Likewise.
7260 * int-vector-builder.h (int_vector_builder::int_vector_builder):
7261 Take the number of elements as a poly_uint64 rather than an
7263 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
7264 from unsigned int to poly_uint64.
7265 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
7266 (vec_perm_indices::new_vector): Likewise.
7267 (vec_perm_indices::length): Likewise.
7268 (vec_perm_indices::nelts_per_input): Likewise.
7269 (vec_perm_indices::input_nelts): Likewise.
7270 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
7271 number of elements per input as a poly_uint64 rather than an
7272 unsigned int. Use the original encoding for variable-length
7273 vectors, rather than clamping each individual element.
7274 For the second and subsequent elements in each pattern,
7275 clamp the step and base before clamping their sum.
7276 (vec_perm_indices::series_p): Handle polynomial element counts.
7277 (vec_perm_indices::all_in_range_p): Likewise.
7278 (vec_perm_indices_to_tree): Likewise.
7279 (vec_perm_indices_to_rtx): Likewise.
7280 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
7281 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
7282 (tree_vector_builder::new_binary_operation): Handle polynomial
7283 element counts. Return false if we need to know the number
7284 of elements at compile time.
7285 * fold-const.c (fold_vec_perm): Punt if the number of elements
7286 isn't known at compile time.
7288 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7290 * vec-perm-indices.h (vec_perm_builder): Change element type
7291 from HOST_WIDE_INT to poly_int64.
7292 (vec_perm_indices::element_type): Update accordingly.
7293 (vec_perm_indices::clamp): Handle polynomial element_types.
7294 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7295 (vec_perm_indices::all_in_range_p): Likewise.
7296 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
7298 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
7299 polynomial vec_perm_indices element types.
7300 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
7301 * fold-const.c (fold_vec_perm): Likewise.
7302 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
7303 * tree-vect-generic.c (lower_vec_perm): Likewise.
7304 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7305 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
7306 element type to HOST_WIDE_INT.
7308 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7309 Alan Hayward <alan.hayward@arm.com>
7310 David Sherwood <david.sherwood@arm.com>
7312 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
7313 rather than an int. Use plus_constant.
7314 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
7315 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
7317 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7318 Alan Hayward <alan.hayward@arm.com>
7319 David Sherwood <david.sherwood@arm.com>
7321 * calls.c (emit_call_1, expand_call): Change struct_value_size from
7322 a HOST_WIDE_INT to a poly_int64.
7324 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7325 Alan Hayward <alan.hayward@arm.com>
7326 David Sherwood <david.sherwood@arm.com>
7328 * calls.c (load_register_parameters): Cope with polynomial
7329 mode sizes. Require a constant size for BLKmode parameters
7330 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
7331 forces a parameter to be padded at the lsb end in order to
7332 fill a complete number of words, require the parameter size
7333 to be ordered wrt UNITS_PER_WORD.
7335 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7336 Alan Hayward <alan.hayward@arm.com>
7337 David Sherwood <david.sherwood@arm.com>
7339 * reload1.c (spill_stack_slot_width): Change element type
7340 from unsigned int to poly_uint64_pod.
7341 (alter_reg): Treat mode sizes as polynomial.
7343 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7344 Alan Hayward <alan.hayward@arm.com>
7345 David Sherwood <david.sherwood@arm.com>
7347 * reload.c (complex_word_subreg_p): New function.
7348 (reload_inner_reg_of_subreg, push_reload): Use it.
7350 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7351 Alan Hayward <alan.hayward@arm.com>
7352 David Sherwood <david.sherwood@arm.com>
7354 * lra-constraints.c (process_alt_operands): Reject matched
7355 operands whose sizes aren't ordered.
7356 (match_reload): Refer to this check here.
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 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
7363 that the mode size is in the set {1, 2, 4, 8, 16}.
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 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
7370 Use plus_constant instead of gen_rtx_PLUS.
7372 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7373 Alan Hayward <alan.hayward@arm.com>
7374 David Sherwood <david.sherwood@arm.com>
7376 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
7377 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
7378 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
7379 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
7380 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
7381 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
7382 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
7383 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
7384 * config/i386/i386.c (ix86_push_rounding): ...this new function.
7385 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
7387 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
7388 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
7389 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
7390 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
7391 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
7392 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
7393 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
7394 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
7395 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
7396 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
7398 * expr.c (emit_move_resolve_push): Treat the input and result
7399 of PUSH_ROUNDING as a poly_int64.
7400 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
7401 (emit_push_insn): Likewise.
7402 * lra-eliminations.c (mark_not_eliminable): Likewise.
7403 * recog.c (push_operand): Likewise.
7404 * reload1.c (elimination_effects): Likewise.
7405 * rtlanal.c (nonzero_bits1): Likewise.
7406 * calls.c (store_one_arg): Likewise. Require the padding to be
7407 known at compile time.
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 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
7414 Use plus_constant instead of gen_rtx_PLUS.
7416 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7417 Alan Hayward <alan.hayward@arm.com>
7418 David Sherwood <david.sherwood@arm.com>
7420 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
7423 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7424 Alan Hayward <alan.hayward@arm.com>
7425 David Sherwood <david.sherwood@arm.com>
7427 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
7428 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
7429 via stack temporaries. Treat the mode size as polynomial too.
7431 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7432 Alan Hayward <alan.hayward@arm.com>
7433 David Sherwood <david.sherwood@arm.com>
7435 * expr.c (expand_expr_real_2): When handling conversions involving
7436 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
7437 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
7438 as a poly_uint64 too.
7440 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7441 Alan Hayward <alan.hayward@arm.com>
7442 David Sherwood <david.sherwood@arm.com>
7444 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
7446 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7447 Alan Hayward <alan.hayward@arm.com>
7448 David Sherwood <david.sherwood@arm.com>
7450 * combine.c (can_change_dest_mode): Handle polynomial
7451 REGMODE_NATURAL_SIZE.
7452 * expmed.c (store_bit_field_1): Likewise.
7453 * expr.c (store_constructor): Likewise.
7454 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
7455 and polynomial REGMODE_NATURAL_SIZE.
7456 (gen_lowpart_common): Likewise.
7457 * reginfo.c (record_subregs_of_mode): Likewise.
7458 * rtlanal.c (read_modify_subreg_p): Likewise.
7460 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7461 Alan Hayward <alan.hayward@arm.com>
7462 David Sherwood <david.sherwood@arm.com>
7464 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
7465 numbers of elements.
7467 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7468 Alan Hayward <alan.hayward@arm.com>
7469 David Sherwood <david.sherwood@arm.com>
7471 * match.pd: Cope with polynomial numbers of vector elements.
7473 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7474 Alan Hayward <alan.hayward@arm.com>
7475 David Sherwood <david.sherwood@arm.com>
7477 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
7478 in a POINTER_PLUS_EXPR.
7480 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7481 Alan Hayward <alan.hayward@arm.com>
7482 David Sherwood <david.sherwood@arm.com>
7484 * omp-simd-clone.c (simd_clone_subparts): New function.
7485 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
7486 (ipa_simd_modify_function_body): Likewise.
7488 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7489 Alan Hayward <alan.hayward@arm.com>
7490 David Sherwood <david.sherwood@arm.com>
7492 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
7493 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
7494 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
7495 (expand_vector_condition, vector_element): Likewise.
7496 (subparts_gt): New function.
7497 (get_compute_type): Use subparts_gt.
7498 (count_type_subparts): Delete.
7499 (expand_vector_operations_1): Use subparts_gt instead of
7500 count_type_subparts.
7502 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7503 Alan Hayward <alan.hayward@arm.com>
7504 David Sherwood <david.sherwood@arm.com>
7506 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
7507 (vect_compile_time_alias): ...this new function. Do the calculation
7508 on poly_ints rather than trees.
7509 (vect_prune_runtime_alias_test_list): Update call accordingly.
7511 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7512 Alan Hayward <alan.hayward@arm.com>
7513 David Sherwood <david.sherwood@arm.com>
7515 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
7517 (vect_schedule_slp_instance): Likewise.
7519 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7520 Alan Hayward <alan.hayward@arm.com>
7521 David Sherwood <david.sherwood@arm.com>
7523 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
7524 constant and extern definitions for variable-length vectors.
7525 (vect_get_constant_vectors): Note that the number of units
7526 is known to be constant.
7528 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7529 Alan Hayward <alan.hayward@arm.com>
7530 David Sherwood <david.sherwood@arm.com>
7532 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
7533 of units as polynomial. Choose between WIDE and NARROW based
7536 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7537 Alan Hayward <alan.hayward@arm.com>
7538 David Sherwood <david.sherwood@arm.com>
7540 * tree-vect-stmts.c (simd_clone_subparts): New function.
7541 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
7543 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7544 Alan Hayward <alan.hayward@arm.com>
7545 David Sherwood <david.sherwood@arm.com>
7547 * tree-vect-stmts.c (vectorizable_call): Treat the number of
7548 vectors as polynomial. Use build_index_vector for
7551 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7552 Alan Hayward <alan.hayward@arm.com>
7553 David Sherwood <david.sherwood@arm.com>
7555 * tree-vect-stmts.c (get_load_store_type): Treat the number of
7556 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
7557 for variable-length vectors.
7558 (vectorizable_mask_load_store): Treat the number of units as
7559 polynomial, asserting that it is constant if the condition has
7560 already been enforced.
7561 (vectorizable_store, vectorizable_load): Likewise.
7563 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7564 Alan Hayward <alan.hayward@arm.com>
7565 David Sherwood <david.sherwood@arm.com>
7567 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
7568 of units as polynomial. Punt if we can't tell at compile time
7569 which vector contains the final result.
7571 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7572 Alan Hayward <alan.hayward@arm.com>
7573 David Sherwood <david.sherwood@arm.com>
7575 * tree-vect-loop.c (vectorizable_induction): Treat the number
7576 of units as polynomial. Punt on SLP inductions. Use an integer
7577 VEC_SERIES_EXPR for variable-length integer reductions. Use a
7578 cast of such a series for variable-length floating-point
7581 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7582 Alan Hayward <alan.hayward@arm.com>
7583 David Sherwood <david.sherwood@arm.com>
7585 * tree.h (build_index_vector): Declare.
7586 * tree.c (build_index_vector): New function.
7587 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
7588 of units as polynomial, forcibly converting it to a constant if
7589 vectorizable_reduction has already enforced the condition.
7590 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
7591 to create a {1,2,3,...} vector.
7592 (vectorizable_reduction): Treat the number of units as polynomial.
7593 Choose vectype_in based on the largest scalar element size rather
7594 than the smallest number of units. Enforce the restrictions
7597 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7598 Alan Hayward <alan.hayward@arm.com>
7599 David Sherwood <david.sherwood@arm.com>
7601 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
7602 number of units as polynomial.
7604 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7605 Alan Hayward <alan.hayward@arm.com>
7606 David Sherwood <david.sherwood@arm.com>
7608 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
7609 * target.def (autovectorize_vector_sizes): Return the vector sizes
7610 by pointer, using vector_sizes rather than a bitmask.
7611 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
7612 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
7613 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
7615 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
7616 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
7617 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
7618 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
7619 * omp-general.c (omp_max_vf): Likewise.
7620 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7621 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
7622 * tree-vect-loop.c (vect_analyze_loop): Likewise.
7623 * tree-vect-slp.c (vect_slp_bb): Likewise.
7624 * doc/tm.texi: Regenerate.
7625 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
7627 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
7628 the vector size as a poly_uint64 rather than an unsigned int.
7629 (current_vector_size): Change from an unsigned int to a poly_uint64.
7630 (get_vectype_for_scalar_type): Update accordingly.
7631 * tree.h (build_truth_vector_type): Take the size and number of
7632 units as a poly_uint64 rather than an unsigned int.
7633 (build_vector_type): Add a temporary overload that takes
7634 the number of units as a poly_uint64 rather than an unsigned int.
7635 * tree.c (make_vector_type): Likewise.
7636 (build_truth_vector_type): Take the number of units as a poly_uint64
7637 rather than an unsigned int.
7639 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7640 Alan Hayward <alan.hayward@arm.com>
7641 David Sherwood <david.sherwood@arm.com>
7643 * target.def (get_mask_mode): Take the number of units and length
7644 as poly_uint64s rather than unsigned ints.
7645 * targhooks.h (default_get_mask_mode): Update accordingly.
7646 * targhooks.c (default_get_mask_mode): Likewise.
7647 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
7648 * doc/tm.texi: Regenerate.
7650 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7651 Alan Hayward <alan.hayward@arm.com>
7652 David Sherwood <david.sherwood@arm.com>
7654 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
7655 * omp-general.c (omp_max_vf): Likewise.
7656 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
7657 (expand_omp_simd): Handle polynomial safelen.
7658 * omp-low.c (omplow_simd_context): Add a default constructor.
7659 (omplow_simd_context::max_vf): Change from int to poly_uint64.
7660 (lower_rec_simd_input_clauses): Update accordingly.
7661 (lower_rec_input_clauses): Likewise.
7663 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7664 Alan Hayward <alan.hayward@arm.com>
7665 David Sherwood <david.sherwood@arm.com>
7667 * tree-vectorizer.h (vect_nunits_for_cost): New function.
7668 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
7669 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
7670 (vect_analyze_slp_cost): Likewise.
7671 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
7672 (vect_model_load_cost): Likewise.
7674 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7675 Alan Hayward <alan.hayward@arm.com>
7676 David Sherwood <david.sherwood@arm.com>
7678 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
7679 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
7680 from an unsigned int * to a poly_uint64_pod *.
7681 (calculate_unrolling_factor): New function.
7682 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
7684 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7685 Alan Hayward <alan.hayward@arm.com>
7686 David Sherwood <david.sherwood@arm.com>
7688 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
7689 from an unsigned int to a poly_uint64.
7690 (_loop_vec_info::slp_unrolling_factor): Likewise.
7691 (_loop_vec_info::vectorization_factor): Change from an int
7693 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
7694 (vect_get_num_vectors): New function.
7695 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
7696 (vect_get_num_copies): Use vect_get_num_vectors.
7697 (vect_analyze_data_ref_dependences): Change max_vf from an int *
7698 to an unsigned int *.
7699 (vect_analyze_data_refs): Change min_vf from an int * to a
7701 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7702 than an unsigned HOST_WIDE_INT.
7703 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7704 (vect_analyze_data_ref_dependence): Change max_vf from an int *
7705 to an unsigned int *.
7706 (vect_analyze_data_ref_dependences): Likewise.
7707 (vect_compute_data_ref_alignment): Handle polynomial vf.
7708 (vect_enhance_data_refs_alignment): Likewise.
7709 (vect_prune_runtime_alias_test_list): Likewise.
7710 (vect_shift_permute_load_chain): Likewise.
7711 (vect_supportable_dr_alignment): Likewise.
7712 (dependence_distance_ge_vf): Take the vectorization factor as a
7713 poly_uint64 rather than an unsigned HOST_WIDE_INT.
7714 (vect_analyze_data_refs): Change min_vf from an int * to a
7716 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
7717 vfm1 as a poly_uint64 rather than an int. Make the same change
7718 for the returned bound_scalar.
7719 (vect_gen_vector_loop_niters): Handle polynomial vf.
7720 (vect_do_peeling): Likewise. Update call to
7721 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
7722 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
7724 * tree-vect-loop.c (vect_determine_vectorization_factor)
7725 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
7726 (vect_get_known_peeling_cost): Likewise.
7727 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
7728 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
7729 (vect_transform_loop): Likewise. Use the lowest possible VF when
7730 updating the upper bounds of the loop.
7731 (vect_min_worthwhile_factor): Make static. Return an unsigned int
7733 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
7734 polynomial unroll factors.
7735 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
7736 (vect_make_slp_decision): Likewise.
7737 (vect_supported_load_permutation_p): Likewise, and polynomial
7739 (vect_analyze_slp_cost): Handle polynomial vf.
7740 (vect_slp_analyze_node_operations): Likewise.
7741 (vect_slp_analyze_bb_1): Likewise.
7742 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7743 than an unsigned HOST_WIDE_INT.
7744 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
7745 (vectorizable_load): Handle polynomial vf.
7746 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
7748 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
7750 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7751 Alan Hayward <alan.hayward@arm.com>
7752 David Sherwood <david.sherwood@arm.com>
7754 * match.pd: Handle bit operations involving three constants
7755 and try to fold one pair.
7757 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7759 * tree-vect-loop-manip.c: Include gimple-fold.h.
7760 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
7761 niters_maybe_zero parameters. Handle other cases besides a step of 1.
7762 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
7763 Add a path that uses a step of VF instead of 1, but disable it
7765 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
7766 and niters_no_overflow parameters. Update calls to
7767 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
7768 Create a new SSA name if the latter choses to use a ste other
7769 than zero, and return it via niters_vector_mult_vf_var.
7770 * tree-vect-loop.c (vect_transform_loop): Update calls to
7771 vect_do_peeling, vect_gen_vector_loop_niters and
7772 slpeel_make_loop_iterate_ntimes.
7773 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
7774 (vect_gen_vector_loop_niters): Update declarations after above changes.
7776 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
7778 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
7779 128-bit round to integer instructions.
7780 (ceil<mode>2): Likewise.
7781 (btrunc<mode>2): Likewise.
7782 (round<mode>2): Likewise.
7784 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7786 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
7787 unaligned VSX load/store on P8/P9.
7788 (expand_block_clear): Allow the use of unaligned VSX
7789 load/store on P8/P9.
7791 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7793 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
7795 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
7796 swap associated with both a load and a store.
7798 2018-01-02 Andrew Waterman <andrew@sifive.com>
7800 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
7801 * config/riscv/riscv.md (clear_cache): Use it.
7803 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
7805 * web.c: Remove out-of-date comment.
7807 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7809 * expr.c (fixup_args_size_notes): Check that any existing
7810 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
7811 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
7812 (emit_single_push_insn): ...here.
7814 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7816 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
7817 (const_vector_encoded_nelts): New function.
7818 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
7819 (const_vector_int_elt, const_vector_elt): Declare.
7820 * emit-rtl.c (const_vector_int_elt_1): New function.
7821 (const_vector_elt): Likewise.
7822 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
7823 of CONST_VECTOR_ELT.
7825 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7827 * expr.c: Include rtx-vector-builder.h.
7828 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
7829 directly on the tree encoding.
7830 (const_vector_from_tree): Likewise.
7831 * optabs.c: Include rtx-vector-builder.h.
7832 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
7833 sequence of "u" values.
7834 * vec-perm-indices.c: Include rtx-vector-builder.h.
7835 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
7836 directly on the vec_perm_indices encoding.
7838 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7840 * doc/rtl.texi (const_vector): Describe new encoding scheme.
7841 * Makefile.in (OBJS): Add rtx-vector-builder.o.
7842 * rtx-vector-builder.h: New file.
7843 * rtx-vector-builder.c: Likewise.
7844 * rtl.h (rtx_def::u2): Add a const_vector field.
7845 (CONST_VECTOR_NPATTERNS): New macro.
7846 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
7847 (CONST_VECTOR_DUPLICATE_P): Likewise.
7848 (CONST_VECTOR_STEPPED_P): Likewise.
7849 (CONST_VECTOR_ENCODED_ELT): Likewise.
7850 (const_vec_duplicate_p): Check for a duplicated vector encoding.
7851 (unwrap_const_vec_duplicate): Likewise.
7852 (const_vec_series_p): Check for a non-duplicated vector encoding.
7853 Say that the function only returns true for integer vectors.
7854 * emit-rtl.c: Include rtx-vector-builder.h.
7855 (gen_const_vec_duplicate_1): Delete.
7856 (gen_const_vector): Call gen_const_vec_duplicate instead of
7857 gen_const_vec_duplicate_1.
7858 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
7859 (gen_const_vec_duplicate): Use rtx_vector_builder.
7860 (gen_const_vec_series): Likewise.
7861 (gen_rtx_CONST_VECTOR): Likewise.
7862 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
7863 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7864 Build a new vector rather than modifying a CONST_VECTOR in-place.
7865 (handle_special_swappables): Update call accordingly.
7866 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
7867 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7868 Build a new vector rather than modifying a CONST_VECTOR in-place.
7869 (handle_special_swappables): Update call accordingly.
7871 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7873 * simplify-rtx.c (simplify_const_binary_operation): Use
7874 CONST_VECTOR_ELT instead of XVECEXP.
7876 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7878 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
7879 the selector elements to be different from the data elements
7880 if the selector is a VECTOR_CST.
7881 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
7882 ssizetype for the selector.
7884 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7886 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
7887 before testing each element individually.
7888 * tree-vect-generic.c (lower_vec_perm): Likewise.
7890 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7892 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
7893 * selftest-run-tests.c (selftest::run_tests): Call it.
7894 * vector-builder.h (vector_builder::operator ==): New function.
7895 (vector_builder::operator !=): Likewise.
7896 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
7897 (vec_perm_indices::all_from_input_p): New function.
7898 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7899 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
7900 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
7901 instead of reading the VECTOR_CST directly. Detect whether both
7902 vector inputs are the same before constructing the vec_perm_indices,
7903 and update the number of inputs argument accordingly. Use the
7904 utility functions added above. Only construct sel2 if we need to.
7906 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7908 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
7909 the broadcast of the low byte.
7910 (expand_mult_highpart): Use an explicit encoding for the permutes.
7911 * optabs-query.c (can_mult_highpart_p): Likewise.
7912 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
7913 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7914 (vectorizable_bswap): Likewise.
7915 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
7916 explicit encoding for the power-of-2 permutes.
7917 (vect_permute_store_chain): Likewise.
7918 (vect_grouped_load_supported): Likewise.
7919 (vect_permute_load_chain): Likewise.
7921 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7923 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
7924 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
7925 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
7926 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7927 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
7928 (vect_gen_perm_mask_any): Likewise.
7930 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7932 * int-vector-builder.h: New file.
7933 * vec-perm-indices.h: Include int-vector-builder.h.
7934 (vec_perm_indices): Redefine as an int_vector_builder.
7935 (auto_vec_perm_indices): Delete.
7936 (vec_perm_builder): Redefine as a stand-alone class.
7937 (vec_perm_indices::vec_perm_indices): New function.
7938 (vec_perm_indices::clamp): Likewise.
7939 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
7940 (vec_perm_indices::new_vector): New function.
7941 (vec_perm_indices::new_expanded_vector): Update for new
7942 vec_perm_indices class.
7943 (vec_perm_indices::rotate_inputs): New function.
7944 (vec_perm_indices::all_in_range_p): Operate directly on the
7945 encoded form, without computing elided elements.
7946 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
7947 encoding. Update for new vec_perm_indices class.
7948 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
7949 the given vec_perm_builder.
7950 (expand_vec_perm_var): Update vec_perm_builder constructor.
7951 (expand_mult_highpart): Use vec_perm_builder instead of
7952 auto_vec_perm_indices.
7953 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
7954 vec_perm_indices instead of auto_vec_perm_indices. Use a single
7955 or double series encoding as appropriate.
7956 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
7957 vec_perm_indices instead of auto_vec_perm_indices.
7958 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7959 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7960 (vect_permute_store_chain): Likewise.
7961 (vect_grouped_load_supported): Likewise.
7962 (vect_permute_load_chain): Likewise.
7963 (vect_shift_permute_load_chain): Likewise.
7964 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
7965 (vect_transform_slp_perm_load): Likewise.
7966 (vect_schedule_slp_instance): Likewise.
7967 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7968 (vectorizable_mask_load_store): Likewise.
7969 (vectorizable_bswap): Likewise.
7970 (vectorizable_store): Likewise.
7971 (vectorizable_load): Likewise.
7972 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
7973 vec_perm_indices instead of auto_vec_perm_indices. Use
7974 tree_to_vec_perm_builder to read the vector from a tree.
7975 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
7976 vec_perm_builder instead of a vec_perm_indices.
7977 (have_whole_vector_shift): Use vec_perm_builder and
7978 vec_perm_indices instead of auto_vec_perm_indices. Leave the
7979 truncation to calc_vec_perm_mask_for_shift.
7980 (vect_create_epilog_for_reduction): Likewise.
7981 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
7982 from auto_vec_perm_indices to vec_perm_indices.
7983 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7984 instead of changing individual elements.
7985 (aarch64_vectorize_vec_perm_const): Use new_vector to install
7986 the vector in d.perm.
7987 * config/arm/arm.c (expand_vec_perm_d::perm): Change
7988 from auto_vec_perm_indices to vec_perm_indices.
7989 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7990 instead of changing individual elements.
7991 (arm_vectorize_vec_perm_const): Use new_vector to install
7992 the vector in d.perm.
7993 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
7994 Update vec_perm_builder constructor.
7995 (rs6000_expand_interleave): Likewise.
7996 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
7997 (rs6000_expand_interleave): Likewise.
7999 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8001 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
8002 to qimode could truncate the indices.
8003 * optabs.c (expand_vec_perm_var): Likewise.
8005 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8007 * Makefile.in (OBJS): Add vec-perm-indices.o.
8008 * vec-perm-indices.h: New file.
8009 * vec-perm-indices.c: Likewise.
8010 * target.h (vec_perm_indices): Replace with a forward class
8012 (auto_vec_perm_indices): Move to vec-perm-indices.h.
8013 * optabs.h: Include vec-perm-indices.h.
8014 (expand_vec_perm): Delete.
8015 (selector_fits_mode_p, expand_vec_perm_var): Declare.
8016 (expand_vec_perm_const): Declare.
8017 * target.def (vec_perm_const_ok): Replace with...
8018 (vec_perm_const): ...this new hook.
8019 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
8020 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
8021 * doc/tm.texi: Regenerate.
8022 * optabs.def (vec_perm_const): Delete.
8023 * doc/md.texi (vec_perm_const): Likewise.
8024 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
8025 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
8026 expand_vec_perm for constant permutation vectors. Assert that
8027 the mode of variable permutation vectors is the integer equivalent
8028 of the mode that is being permuted.
8029 * optabs-query.h (selector_fits_mode_p): Declare.
8030 * optabs-query.c: Include vec-perm-indices.h.
8031 (selector_fits_mode_p): New function.
8032 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
8033 is defined, instead of checking whether the vec_perm_const_optab
8034 exists. Use targetm.vectorize.vec_perm_const instead of
8035 targetm.vectorize.vec_perm_const_ok. Check whether the indices
8036 fit in the vector mode before using a variable permute.
8037 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
8038 vec_perm_indices instead of an rtx.
8039 (expand_vec_perm): Replace with...
8040 (expand_vec_perm_const): ...this new function. Take the selector
8041 as a vec_perm_indices rather than an rtx. Also take the mode of
8042 the selector. Update call to shift_amt_for_vec_perm_mask.
8043 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
8044 Use vec_perm_indices::new_expanded_vector to expand the original
8045 selector into bytes. Check whether the indices fit in the vector
8046 mode before using a variable permute.
8047 (expand_vec_perm_var): Make global.
8048 (expand_mult_highpart): Use expand_vec_perm_const.
8049 * fold-const.c: Includes vec-perm-indices.h.
8050 * tree-ssa-forwprop.c: Likewise.
8051 * tree-vect-data-refs.c: Likewise.
8052 * tree-vect-generic.c: Likewise.
8053 * tree-vect-loop.c: Likewise.
8054 * tree-vect-slp.c: Likewise.
8055 * tree-vect-stmts.c: Likewise.
8056 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
8058 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
8059 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
8060 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
8061 (aarch64_vectorize_vec_perm_const): ...this new function.
8062 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8063 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8064 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
8065 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
8066 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8067 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8068 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
8070 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
8071 check for NEON modes.
8072 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
8073 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
8074 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
8075 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
8077 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
8078 the old VEC_PERM_CONST conditions.
8079 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
8080 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
8081 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
8082 (ia64_vectorize_vec_perm_const_ok): Merge into...
8083 (ia64_vectorize_vec_perm_const): ...this new function.
8084 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
8085 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
8086 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
8087 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
8088 * config/mips/mips.c (mips_expand_vec_perm_const)
8089 (mips_vectorize_vec_perm_const_ok): Merge into...
8090 (mips_vectorize_vec_perm_const): ...this new function.
8091 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
8092 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
8093 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
8094 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
8095 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
8096 (rs6000_expand_vec_perm_const): Delete.
8097 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
8099 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8100 (altivec_expand_vec_perm_const_le): Take each operand individually.
8101 Operate on constant selectors rather than rtxes.
8102 (altivec_expand_vec_perm_const): Likewise. Update call to
8103 altivec_expand_vec_perm_const_le.
8104 (rs6000_expand_vec_perm_const): Delete.
8105 (rs6000_vectorize_vec_perm_const_ok): Delete.
8106 (rs6000_vectorize_vec_perm_const): New function.
8107 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8108 an element count and rtx array.
8109 (rs6000_expand_extract_even): Update call accordingly.
8110 (rs6000_expand_interleave): Likewise.
8111 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
8112 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
8113 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
8114 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
8115 (rs6000_expand_vec_perm_const): Delete.
8116 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8117 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8118 (altivec_expand_vec_perm_const_le): Take each operand individually.
8119 Operate on constant selectors rather than rtxes.
8120 (altivec_expand_vec_perm_const): Likewise. Update call to
8121 altivec_expand_vec_perm_const_le.
8122 (rs6000_expand_vec_perm_const): Delete.
8123 (rs6000_vectorize_vec_perm_const_ok): Delete.
8124 (rs6000_vectorize_vec_perm_const): New function. Remove stray
8125 reference to the SPE evmerge intructions.
8126 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8127 an element count and rtx array.
8128 (rs6000_expand_extract_even): Update call accordingly.
8129 (rs6000_expand_interleave): Likewise.
8130 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
8131 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
8133 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8135 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8137 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
8138 vector mode and that that mode matches the mode of the data
8140 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
8141 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
8142 directly using expand_vec_perm_1 when forcing selectors into
8144 (expand_vec_perm_var): New function, split out from expand_vec_perm.
8146 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8148 * optabs-query.h (can_vec_perm_p): Delete.
8149 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
8150 * optabs-query.c (can_vec_perm_p): Split into...
8151 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
8152 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
8153 particular selector is valid.
8154 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8155 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8156 (vect_grouped_load_supported): Likewise.
8157 (vect_shift_permute_load_chain): Likewise.
8158 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8159 (vect_transform_slp_perm_load): Likewise.
8160 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8161 (vectorizable_bswap): Likewise.
8162 (vect_gen_perm_mask_checked): Likewise.
8163 * fold-const.c (fold_ternary_loc): Likewise. Don't take
8164 implementations of variable permutation vectors into account
8165 when deciding which selector to use.
8166 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
8167 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
8168 with a false third argument.
8169 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
8170 to test whether the constant selector is valid and can_vec_perm_var_p
8171 to test whether a variable selector is valid.
8173 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8175 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
8176 * optabs-query.c (can_vec_perm_p): Likewise.
8177 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
8178 instead of vec_perm_indices.
8179 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
8180 (vect_gen_perm_mask_checked): Likewise,
8181 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
8182 (vect_gen_perm_mask_checked): Likewise,
8184 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8186 * optabs-query.h (qimode_for_vec_perm): Declare.
8187 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
8188 (qimode_for_vec_perm): ...this new function.
8189 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
8191 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8193 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
8194 does not have a conditional at the top.
8196 2018-01-02 Richard Biener <rguenther@suse.de>
8198 * ipa-inline.c (big_speedup_p): Fix expression.
8200 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8203 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
8206 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8210 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
8211 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
8212 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
8213 cond_taken_branch_cost 3->4.
8215 2018-01-01 Jakub Jelinek <jakub@redhat.com>
8217 PR tree-optimization/83581
8218 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
8219 TODO_cleanup_cfg if any changes have been made.
8222 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
8223 convert_modes if target mode has the right side, but different mode
8227 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
8228 last argument when extracting from CONCAT. If either from_real or
8229 from_imag is NULL, use expansion through memory. If result is not
8230 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
8231 the parts directly to inner mode, if even that fails, use expansion
8235 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
8236 check for bswap in mode rather than HImode and use that in expand_unop
8239 Copyright (C) 2018 Free Software Foundation, Inc.
8241 Copying and distribution of this file, with or without modification,
8242 are permitted in any medium without royalty provided the copyright
8243 notice and this notice are preserved.