1 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4 * combine.c (try_combine): Use reg_or_subregno instead of handling
5 just paradoxical SUBREGs and REGs.
7 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
9 * config/arc/arc.c (arc_finalize_pic): Remove function.
10 (arc_must_save_register): We use single base PIC register, remove
11 checks to save/restore the PIC register.
12 (arc_expand_prologue): Likewise.
13 * config/arc/arc-protos.h (arc_set_default_type_attributes):
15 (arc_verify_short): Likewise.
16 (arc_attr_type): Likewise.
17 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
18 (walk_stores): Likewise.
19 (arc_address_cost): Make it static.
20 (arc_verify_short): Likewise.
21 (branch_dest): Likewise.
22 (arc_attr_type): Likewise.
23 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
24 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
25 (arc_final_prescan_insn): Remove inserting the nops due to
26 hardware hazards. It is done in reorg step.
27 (insn_length_variant_t): Remove.
28 (insn_length_parameters_t): Likewise.
29 (arc_insn_length_parameters): Likewise.
30 (arc_get_insn_variants): Likewise.
31 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
33 2018-03-06 Jakub Jelinek <jakub@redhat.com>
36 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
39 PR tree-optimization/84687
40 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
42 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
44 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
46 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
47 Rename to ppc_speculation_barrier.
48 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
49 __builtin_ppc_speculation_barrier.
51 2018-03-05 Jakub Jelinek <jakub@redhat.com>
54 * combine.c (combine_simplify_rtx): Don't try to simplify if
55 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
58 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
60 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
61 to 32 bytes when compiling for POWER9.
63 2018-03-05 Jakub Jelinek <jakub@redhat.com>
66 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
67 regparm >= 3 with no arg reg available also for calls with
68 flag_force_indirect_call. Pass decl to ix86_function_regparm.
71 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
73 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
75 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
78 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
80 2018-03-05 Richard Biener <rguenther@suse.de>
82 PR tree-optimization/84486
83 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
84 When inserting a __builtin_assume_aligned call set the LHS
85 SSA name alignment info accordingly.
87 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
89 PR tree-optimization/84114
90 * config/aarch64/aarch64.c (aarch64_reassociation_width)
91 Avoid reassociation of FLOAT_MODE addition.
93 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
95 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
96 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
97 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
98 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
99 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
100 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
101 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
103 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
104 __builtin_ia32_wbinvd): New builtins.
105 (SPECIAL_ARGS2): New.
106 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
107 (SPECIAL_ARGS2): New.
108 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
109 (ix86_valid_target_attribute_inner_p): Ditto.
110 (ix86_init_mmx_sse_builtins): Add special_args2.
111 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
112 TARGET_WBNOINVD_P): New.
113 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
114 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
115 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
116 * config/i386/immintrin.h (_wbinvd): New intrinsic.
117 * config/i386/pconfigintrin.h: New file.
118 * config/i386/wbnoinvdintrin.h: Ditto.
119 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
120 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
122 2018-03-05 Richard Biener <rguenther@suse.de>
124 PR tree-optimization/84670
125 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
127 (BB_VISITED_WITH_VISITED_SUCCS): New define.
128 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
129 (compute_antic_aux): Only assert the number of values in ANTIC_IN
130 doesn't grow if all successors (recursively) were visited at least
133 2018-03-05 Richard Biener <rguenther@suse.de>
135 PR tree-optimization/84650
136 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
137 if executed in the loop pipeline.
139 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
141 * doc/configfiles.texi (Configuration Files): Move info about
142 conditionalizing $target-protos.h to...
143 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
144 differs from $target-protos.h.
146 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
147 Chung-Ju Wu <jasonwucj@gmail.com>
149 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
150 * config/nds32/nds32-multiple.md (setmemsi): Define.
151 * config/nds32/nds32-memory-manipulation.c
152 (nds32_gen_dup_4_byte_to_word_value): New.
153 (emit_setmem_word_loop): New.
154 (emit_setmem_byte_loop): New.
155 (nds32_expand_setmem_loop): New.
156 (nds32_expand_setmem_loop_v3m): New.
157 (nds32_expand_setmem_unroll): New.
158 (nds32_expand_setmem): New.
160 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
161 Chung-Ju Wu <jasonwucj@gmail.com>
163 * config/nds32/nds32-memory-manipulation.c
164 (nds32_emit_load_store): New.
165 (nds32_emit_post_inc_load_store): New.
166 (nds32_emit_mem_move): New.
167 (nds32_emit_mem_move_block): New.
168 (nds32_expand_movmemsi_loop_unknown_size): New.
169 (nds32_expand_movmemsi_loop_known_size): New.
170 (nds32_expand_movmemsi_loop): New.
171 (nds32_expand_movmemsi_unroll): New.
172 (nds32_expand_movmemqi): Rename ...
173 (nds32_expand_movmemsi): ... to this.
174 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
175 (movmemsi): ... to this.
176 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
177 (nds32_expand_movmemsi): ... to this.
179 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
180 Monk Chiang <sh.chiang04@gmail.com>
181 Chung-Ju Wu <jasonwucj@gmail.com>
183 * config/nds32/nds32-protos.h
184 (nds32_expand_load_multiple): New arguments.
185 (nds32_expand_store_multiple): Ditto.
186 (nds32_valid_multiple_load_store): Rename ...
187 (nds32_valid_multiple_load_store_p): ... to this.
188 * config/nds32/nds32-memory-manipulation.c
189 (nds32_expand_load_multiple): Refine implementation.
190 (nds32_expand_store_multiple): Ditto.
191 * config/nds32/nds32-multiple.md
192 (load_multiple): Update nds32_expand_load_multiple interface.
193 (store_multiple): Update nds32_expand_store_multiple interface.
194 * config/nds32/nds32-predicates.c
195 (nds32_valid_multiple_load_store): Rename ...
196 (nds32_valid_multiple_load_store_p): ... to this and refine
198 * config/nds32/predicates.md
199 (nds32_load_multiple_and_update_address_operation): New predicate.
200 (nds32_store_multiple_and_update_address_operation): New predicate.
202 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
203 Chung-Ju Wu <jasonwucj@gmail.com>
205 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
206 (combo): New attribute.
207 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
209 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
211 * config/nds32/nds32.opt: Change -mcmodel= default value.
213 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
214 Monk Chiang <sh.chiang04@gmail.com>
215 Chung-Ju Wu <jasonwucj@gmail.com>
217 * config/nds32/constants.md (unspec_element): New enum.
218 * config/nds32/constraints.md (Umw): New constraint.
219 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
220 * config/nds32/nds32-intrinsic.md: Likewise.
221 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
222 (nds32_valid_smw_lwm_base_p): New.
223 (nds32_output_smw_single_word): New.
224 (nds32_output_lmw_single_word): New.
225 (nds32_expand_unaligned_load): New.
226 (nds32_expand_unaligned_store): New.
227 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
228 (nds32_output_smw_single_word): Declare.
229 (nds32_output_lmw_single_word): Declare.
230 (nds32_expand_unaligned_load): Declare.
231 (nds32_expand_unaligned_store): Declare.
232 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
233 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
234 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
235 NDS32_BUILTIN_UASTORE_DW.
236 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
239 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
240 Kito Cheng <kito.cheng@gmail.com>
241 Chung-Ju Wu <jasonwucj@gmail.com>
243 * config/nds32/nds32-intrinsic.c
244 (nds32_expand_builtin_null_ftype_reg): Delete.
245 (nds32_expand_builtin_reg_ftype_imm): Ditto.
246 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
247 (nds32_read_argument): New.
248 (nds32_legitimize_target): Ditto.
249 (nds32_legitimize_argument): Ditto.
250 (nds32_check_constant_argument): Ditto.
251 (nds32_expand_unop_builtin): Ditto.
252 (nds32_expand_unopimm_builtin): Ditto.
253 (nds32_expand_binop_builtin): Ditto.
254 (nds32_builtin_decl_impl): Ditto.
255 (builtin_description): Ditto.
256 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
257 (nds32_init_builtins_impl): Ditto.
258 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
259 (nds32_builtin_decl): New.
260 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
261 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
263 2018-03-02 Jeff Law <law@redhat.com>
265 * reorg.c (stop_search_p): Handle DEBUG_INSN.
266 (redundant_insn, fill_simple_delay_slots): Likewise.
267 (fill_slots_from_thread): Likewise.
268 * resource.c (mark_referenced_resources): Likewise.
269 (mark_set_resources, find_dead_or_set_registers): Likewise.
271 2018-03-02 Jakub Jelinek <jakub@redhat.com>
273 * substring-locations.h (format_warning_va): Formatting fix for
275 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
277 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
278 * substring-locations.c: Include intl.h.
279 (format_warning_va): Turned into small wrapper around
280 format_warning_n_va, renamed to ...
281 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
282 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
284 (format_warning_at_substring_n): New function.
285 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
286 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
287 format_warning_at_substring with just a shorter name instead of
288 const function pointer.
289 (fmtwarn_n): New function.
290 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
291 appropriate, get rid of all the fmtstr temporaries, move conditionals
292 with G_() wrapped string literals directly into fmtwarn arguments,
293 cast dir.len to (int), formatting fixes.
295 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
297 * doc/invoke.texi: Remove "Cilk Plus" references.
299 2018-03-02 Jakub Jelinek <jakub@redhat.com>
300 Richard Biener <rguenther@suse.de>
303 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
304 for error or warning attributes if CALL_FROM_THUNK_P is set.
307 2018-03-02 Jakub Jelinek <jakub@redhat.com>
310 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
311 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
314 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
315 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
317 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
318 instead of -1U in last predictors element's probability member.
320 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
323 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
324 arguments if they are comparable.
326 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
328 PR tree-optimization/84634
329 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
330 masks and masked_loop_p with a single loop_masks, making sure it's
331 null for bb vectorization.
333 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
335 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
336 (vect_analyze_data_ref_access): Use loop->safe_len rather than
337 loop->force_vectorize to check whether there is no alias.
339 2018-03-02 Jakub Jelinek <jakub@redhat.com>
342 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
344 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
346 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
347 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
348 instead of a loop around prev_real_insn.
349 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
353 * config/i386/i386.c (ix86_print_operand): Use conditional
354 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
357 2018-03-02 Richard Biener <rguenther@suse.de>
359 PR tree-optimization/84427
360 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
361 (bitmap_set_subtract_values): Rewrite to handle multiple
364 (prune_clobbered_mems): Likewise.
365 (phi_translate): Take edge instead of pred/phiblock.
366 (phi_translate_1): Likewise.
367 (phi_translate_set): Likewise. Insert all translated
368 exprs for a value into the set, keeping possibly multiple
369 expressions per value.
370 (compute_antic_aux): Adjust for phi_translate changes.
371 When intersecting union the expressions and prune those
372 not in the final value set, keeping possibly multiple
373 expressions per value. Do not use value-insertion
374 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
375 all expressions. Add verification that the value-sets
376 only shrink during iteration.
377 (compute_partial_antic_aux): Adjust for the phi_translate changes.
378 (do_pre_regular_insertion): Likewise.
379 (do_pre_partial_partial_insertion): Likewise.
381 2018-03-02 Richard Biener <rguenther@suse.de>
384 * config/darwin.c (saved_debug_info_level): New static global.
385 (darwin_asm_lto_start): Disable debug info generation for LTO out.
386 (darwin_asm_lto_end): Restore debug info generation settings.
388 2018-03-01 Martin Liska <mliska@suse.cz>
391 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
394 2018-03-01 Richard Biener <rguenther@suse.de>
397 * dwarf2out.c (gen_variable_die): Properly handle late VLA
398 type annotation with LTO when debug was disabled at compile-time.
400 2018-03-01 Matthew Fortune <mfortune@gmail.com>
402 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
404 (mips_final_postscan_insn): Likewise.
406 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
408 PR rtl-optimization/84528
409 * alias.c (init_alias_target): Add commentary.
410 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
411 a unique base value if the frame pointer is not eliminated
412 to the stack pointer.
414 2018-03-01 Tom de Vries <tom@codesourcery.com>
416 PR rtl-optimization/83327
417 * lra-int.h (hard_regs_spilled_into): Declare.
418 * lra.c (hard_regs_spilled_into): Define.
419 (init_reg_info): Init hard_regs_spilled_into.
420 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
421 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
422 (process_bb_lives): Handle hard_regs_spilled_into.
423 (lra_create_live_ranges_1): Before doing liveness propagation, clear
424 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
426 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
428 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
429 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
430 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
431 * config/rs6000/aix72.h: New file.
433 2018-02-28 Jakub Jelinek <jakub@redhat.com>
435 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
436 instead of warning_at with conditional singular and plural messages
440 * stor-layout.c (update_alignment_for_field): For
441 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
442 && !DECL_PACKED (field), do the alignment update, just use
443 only desired_align instead of MAX (type_align, desired_align)
445 (place_field): Don't do known_align < desired_align handling
446 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
447 is non-NULL, instead do it after rli->prev_field handling and
448 only if not within a bitfield word. For DECL_PACKED (field)
449 use type_align of BITS_PER_UNIT.
451 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
453 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
454 superfluous parentheses and trailing spaces.
456 2018-02-28 Richard Biener <rguenther@suse.de>
458 PR tree-optimization/84584
459 * graphite-scop-detection.c (scop_detection::add_scop): Discard
460 SCoPs with fake exit edge.
462 2018-02-28 Martin Liska <mliska@suse.cz>
465 * timevar.c (timer::print): Fix format to properly print 100%
468 2018-02-28 Richard Biener <rguenther@suse.de>
471 * genmatch.c (capture_info::walk_match): Do not mark
472 captured expressions without operands as expr_p given
473 they act more like predicates and should be subject to
474 "lost tail" side-effect preserving.
476 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
478 PR rtl-optimization/81611
479 * auto-inc-dec.c (attempt_change): Move dead note from
480 mem_insn if it's the next use of regno
481 (find_address): Take address use of reg holding
482 non-incremented value. Add parm to limit search to the named
484 (merge_in_block): Attempt to use a mem insn that is the next
485 use of the original regno.
487 2018-02-27 Martin Sebor <msebor@redhat.com>
490 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
491 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
493 2018-02-27 Martin Sebor <msebor@redhat.com>
496 * diagnostic-core.h (warning_n, error_n, inform_n): Change
497 n argument to unsigned HOST_WIDE_INT.
498 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
499 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
500 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
501 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
503 2018-02-27 Richard Biener <rguenther@suse.de>
505 PR tree-optimization/84512
506 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
507 Do not use the estimate returned from record_stmt_cost for
508 the scalar iteration cost but sum properly using add_stmt_cost.
510 2018-02-27 Richard Biener <rguenther@suse.de>
512 PR tree-optimization/84466
513 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
514 Adjust last change to less strictly validate use operands.
516 2018-02-27 Martin Liska <mliska@suse.cz>
518 PR gcov-profile/84548
519 * gcov.c (process_file): Allow partial overlap and consider it
520 also as group functions.
521 (output_lines): Properly calculate range of lines for a group.
523 2018-02-27 Martin Liska <mliska@suse.cz>
525 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
526 'ggc' suffixes. Change first column width.
527 (timer::print): Fix formatting of the column.
529 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
531 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
532 preserve inline entry blocks for the sake of debug inline
533 entry point markers alone.
534 (remove_unused_locals): Suggest in comments a better place to
535 force the preservation of inline entry blocks that are
536 otherwise unused, but do not preserve them.
538 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
540 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
542 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
545 * config/i386/constraints.md (Bs): Replace
546 ix86_indirect_branch_register with
547 TARGET_INDIRECT_BRANCH_REGISTER.
549 * config/i386/i386.md (indirect_jump): Likewise.
550 (tablejump): Likewise.
551 (*sibcall_memory): Likewise.
552 (*sibcall_value_memory): Likewise.
553 Peepholes of indirect call and jump via memory: Likewise.
554 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
555 (*sibcall_value_GOT_32): Likewise.
556 * config/i386/predicates.md (indirect_branch_operand): Likewise.
557 (GOT_memory_operand): Likewise.
558 (call_insn_operand): Likewise.
559 (sibcall_insn_operand): Likewise.
560 (GOT32_symbol_operand): Likewise.
561 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
563 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
565 PR rtl-optimization/83496
566 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
567 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
568 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
569 redundant insn, if any.
570 (relax_delay_slots): Likewise.
571 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
573 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
575 PR tree-optimization/83965
576 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
577 that grouped statements are part of a reduction chain. Return
578 true if the statement is not marked as a reduction itself but
580 (vect_recog_dot_prod_pattern): Don't check whether the statement
581 is part of a group here.
582 (vect_recog_sad_pattern): Likewise.
583 (vect_recog_widen_sum_pattern): Likewise.
585 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
588 * final.c (rest_of_clean_state): Also look for calls inside sequences.
590 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
593 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
595 (ix86_output_indirect_function_return): New prototype.
596 (ix86_split_simple_return_pop_internal): Likewise.
597 * config/i386/i386.c (indirect_return_via_cx): New.
598 (indirect_return_via_cx_bnd): Likewise.
599 (indirect_thunk_name): Handle return va CX_REG.
600 (output_indirect_thunk_function): Create alias for
601 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
602 (ix86_output_indirect_jmp): Remove the bool argument.
603 (ix86_output_indirect_function_return): New function.
604 (ix86_split_simple_return_pop_internal): Likewise.
605 * config/i386/i386.md (*indirect_jump): Don't pass false
606 to ix86_output_indirect_jmp.
607 (*tablejump_1): Likewise.
608 (simple_return_pop_internal): Change it to define_insn_and_split.
609 Call ix86_split_simple_return_pop_internal to split it for
611 (simple_return_indirect_internal): Call
612 ix86_output_indirect_function_return instead of
613 ix86_output_indirect_jmp.
615 2018-02-26 Jakub Jelinek <jakub@redhat.com>
618 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
619 memset and value initialization afterwards.
621 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
623 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
625 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
628 * common/config/aarch64/aarch64-common.c
629 (aarch_option_optimization_table[]): Switch
630 off fomit-frame-pointer
632 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
633 Chung-Ju Wu <jasonwucj@gmail.com>
635 * config/nds32/nds32-multiple.md (load_multiple): Disallow
637 (store_multiple): Ditto.
639 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
641 * config.gcc: Add --with-cpu support for nds32 target.
642 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
643 * config/nds32/nds32.opt: Add -mcpu= option.
645 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
647 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
648 isel=yes): Warn for these deprecated options.
650 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
652 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
653 ISA_2_5_MASKS_EMBEDDED.
655 2018-02-23 Jakub Jelinek <jakub@redhat.com>
657 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
658 p->max as pointers rather than using iterative_hash_expr.
660 2018-02-23 Carl Love <cel@us.ibm.com>
662 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
663 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
665 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
666 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
667 P8V_BUILTIN_VEC_VUNSIGNED2.
669 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
672 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
673 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
675 * lra-constraints.c (curr_insn_transform): Set up
676 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
678 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
679 * lra-eliminations.c (spill_pseudos): Ditto.
680 (process_insn_for_elimination): Ditto.
681 * lra-lives.c (reg_early_clobber_p): Use the new macros.
682 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
683 LRA_NON_CLOBBERED_ALT.
685 2018-02-22 Martin Sebor <msebor@redhat.com>
687 PR tree-optimization/84480
688 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
689 to maybe_diag_stxncpy_trunc. Call it.
690 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
691 from gimple_fold_builtin_strcpy. Print inlining stack.
692 (handle_builtin_stxncpy): Print inlining stack.
693 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
695 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
698 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
699 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
700 and -fcheck-pointer-bounds are used together.
701 (indirect_thunk_prefix): New enum.
702 (indirect_thunk_need_prefix): New function.
703 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
704 "_nt" instead of "_bnd" for NOTRACK prefix.
705 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
706 (output_indirect_thunk_function): Likewise.
708 (ix86_code_end): Update output_indirect_thunk_function calls.
709 (ix86_output_indirect_branch_via_reg): Replace
710 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
711 (ix86_output_indirect_branch_via_push): Likewise.
712 (ix86_output_function_return): Likewise.
713 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
714 incompatible with -fcf-protection=branch and
715 -fcheck-pointer-bounds.
717 2018-02-22 Steve Ellcey <sellcey@cavium.com>
720 * config/aarch64/aarch64.c (aarch64_print_address_internal):
721 Change gcc_assert call to output_operand_lossage.
723 2018-02-22 Steve Ellcey <sellcey@cavium.com>
725 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
727 2018-02-22 DJ Delorie <dj@redhat.com>
728 Sebastian Perta <sebastian.perta@renesas.com>
729 Oleg Endo <olegendo@gcc.gnu.org>
731 * config/rx/rx.c (rx_rtx_costs): New function.
732 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
734 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
736 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
738 2018-02-22 Martin Liska <mliska@suse.cz>
741 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
742 Add "native" as a possible value.
744 2018-02-22 Martin Liska <mliska@suse.cz>
747 * config/i386/i386.c (ix86_option_override_internal):
748 Add "native" as a possible value for -march and -mtune.
750 2018-02-22 Jakub Jelinek <jakub@redhat.com>
753 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
754 to all type variants.
756 PR tree-optimization/84503
757 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
758 width as info->bitpos + info->bitsize - start.
759 (merged_store_group::merge_overlapping): Simplify width computation.
760 (check_no_overlap): New function.
761 (imm_store_chain_info::try_coalesce_bswap): Compute expected
762 start + width and last_order of the group, fail if check_no_overlap
764 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
765 to group if check_no_overlap fails.
767 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
769 * config/rs6000/altivec.md: Delete contraint arguments to
770 define_expand, define_split, and define_peephole2, and in
771 define_insn_and_split if always unused.
772 * config/rs6000/darwin.md: Ditto.
773 * config/rs6000/dfp.md: Ditto.
774 * config/rs6000/rs6000.md: Ditto.
775 * config/rs6000/sync.md: Ditto.
776 * config/rs6000/vector.md: Ditto.
777 * config/rs6000/vsx.md: Ditto.
779 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
781 * config/rs6000/altivec.md: Write output control strings as braced
782 blocks instead of double-quoted strings.
783 * config/rs6000/darwin.md: Ditto.
784 * config/rs6000/rs6000.md: Ditto.
785 * config/rs6000/vector.md: Ditto.
786 * config/rs6000/vsx.md: Ditto.
788 2018-02-21 Jason Merrill <jason@redhat.com>
790 PR c++/84314 - ICE with templates and fastcall attribute.
791 * attribs.c (build_type_attribute_qual_variant): Remove assert.
793 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
795 * ipa-cp.c (determine_versionability): Fix comment typos.
797 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
800 * ipa-cp.c (determine_versionability): Do not version functions caling
803 2018-02-21 Martin Liska <mliska@suse.cz>
806 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
807 Add "native" as a possible value.
808 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
809 the macro when native cpu detection is available.
811 2018-02-21 Martin Liska <mliska@suse.cz>
814 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
815 Add "native" as a possible value.
816 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
817 when native cpu detection is available.
819 2018-02-21 Jakub Jelinek <jakub@redhat.com>
820 Martin Sebor <msebor@redhat.com>
822 PR tree-optimization/84478
823 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
825 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
826 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
827 support which is conservatively correct, for 2 only stay conservative
828 for maxlen. Formatting and comment capitalization fixes. Add STRICT
829 argument to the 2 argument get_range_strlen, adjust 6 arg
830 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
832 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
833 (gimple_fold_builtin_strlen): Pass true as last argument to
836 2018-02-20 Martin Sebor <msebor@redhat.com>
839 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
840 (builtin_memref::set_base_and_offset): Same. Handle inner references.
841 (builtin_memref::builtin_memref): Factor out parts into
842 set_base_and_offset and call it.
844 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
847 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
848 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
849 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
850 search at the associated MODE_INT.
852 2018-02-20 Jeff Law <law@redhat.com>
855 PR tree-optimization/81592
857 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
858 for range data rather than using global data.
859 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
860 range data rather than using global data.
861 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
862 pass it to children as needed.
863 (struct directive::fmtresult): Similarly.
864 (struct directive::set_width): Similarly.
865 (struct directive::set_precision): Similarly.
866 (format_integer, format_directive, parse_directive): Similarly.
867 (format_none): Accept unnamed vr_values parameter.
868 (format_percent, format_floating, format_character): Similarly.
869 (format_string, format_plain): Similarly.
870 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
871 the EVRP range analyzer for range data rather than using global data.
872 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
873 gimple-ssa-evrp-analyze.h
874 (class sprintf_dom_walker): Add after_dom_children member function.
875 Add evrp_range_analyzer member.
876 (sprintf_dom_walker::before_dom_children): Call into the EVRP
877 range analyzer as needed.
878 (sprintf_dom_walker::after_dom_children): New member function.
879 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
881 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
882 (evrp_range_analyzer::pop_to_marker): Likewise.
884 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
886 PR tree-optimization/84419
887 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
888 with the required type if its current type is compatible but
891 2018-02-20 Jakub Jelinek <jakub@redhat.com>
894 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
897 2018-02-20 Martin Liska <mliska@suse.cz>
900 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
901 possible values if we don't have a hint.
903 2018-02-20 Martin Liska <mliska@suse.cz>
907 * final.c (shorten_branches): Build align_tab array with one
909 * opts.c (finish_options): Add alignment option limit check.
910 (MAX_CODE_ALIGN): Likewise.
911 (MAX_CODE_ALIGN_VALUE): Likewise.
912 * doc/invoke.texi: Document maximum allowed option value for
913 all -falign-* options.
915 2018-02-19 Jakub Jelinek <jakub@redhat.com>
918 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
919 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
920 * var-tracking.c (emit_note_insn_var_location): Remove all references
921 to NOTE_INSN_CALL_ARG_LOCATION.
922 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
923 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
924 Use copy_rtx_if_shared.
925 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
926 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
927 (dwarf2out_var_location): Remove handling of
928 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
930 * final.c (final_scan_insn): Remove all references to
931 NOTE_INSN_CALL_ARG_LOCATION.
932 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
933 before dumping final insns.
934 * except.c (emit_note_eh_region_end): Remove all references to
935 NOTE_INSN_CALL_ARG_LOCATION.
936 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
937 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
938 * config/arc/arc.c (hwloop_optimize): Likewise.
939 * config/arm/arm.c (create_fix_barrier): Likewise.
940 * config/s390/s390.c (s390_chunkify_start): Likewise.
941 * config/sh/sh.c (find_barrier): Likewise.
942 * config/i386/i386.c (rest_of_insert_endbranch,
943 ix86_seh_fixup_eh_fallthru): Likewise.
944 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
945 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
946 * config/frv/frv.c (frv_function_prologue): Likewise.
947 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
949 (note_outside_basic_block_p): Remove all references to
950 NOTE_INSN_CALL_ARG_LOCATION.
951 * gengtype.c (adjust_field_rtx_def): Likewise.
952 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
954 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
955 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
958 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
961 PR tree-optimization/84452
962 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
963 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
966 2018-02-19 Martin Liska <mliska@suse.cz>
969 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
971 2018-02-19 Martin Liska <mliska@suse.cz>
972 Richard Sandiford <richard.sandiford@linaro.org>
974 PR tree-optimization/82491
975 * gimple-fold.c (get_base_constructor): Make earlier bail out
978 2018-02-19 Carl Love <cel@us.ibm.com>
980 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
981 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
983 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
986 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
988 * config/rl78/rl78.md (movdf): New define expand.
990 2018-02-19 Martin Liska <mliska@suse.cz>
993 * doc/invoke.texi: Fix typo.
994 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
996 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
998 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
999 handle rs6000_single_float and rs6000_double_float specially for
1002 2018-02-16 Jeff Law <law@redhat.com>
1004 * config/rx/rx.c (add_pop_cfi_notes): New function.;
1007 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1010 * ipa-inline.c (inline_small_functions): Fix a typo.
1012 2018-02-16 Nathan Sidwell <nathan@acm.org>
1014 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
1016 2018-02-16 Carl Love <cel@us.ibm.com>
1018 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
1019 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
1020 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
1021 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
1022 expansion to P8V_BUILTIN_VEC_FLOAT2.
1024 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
1026 PR rtl-optimization/70023
1027 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
1028 src_regno into account.
1030 2018-02-16 Carl Love <cel@us.ibm.com>
1032 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
1033 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
1034 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
1035 * config/rs6000/rs6000.c: Remove case statements for
1036 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
1037 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
1038 and P9V_BUILTIN_VEC_VINSERT4B.
1039 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
1040 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
1041 * config/rs6000/vsx.md:
1042 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
1045 2018-02-16 Carl Love <cel@us.ibm.com>
1047 * config/rs6000/altivec.h: Add builtin names vec_extract4b
1049 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
1051 * config/rs6000/rs6000-c.c: Add the definitions for
1052 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
1053 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
1054 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
1055 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
1056 definition for insert4b and define insn *insert3b_internal.
1057 * doc/extend.texi: Add documentation for vec_extract4b.
1059 2018-02-16 Nathan Sidwell <nathan@acm.org>
1061 * doc/extend.texi (Backwards Compatibility): Mention friend
1062 injection. Note for-scope is deprecated.
1063 * doc/invoke.texi (-ffriend-injection): Deprecate.
1065 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
1067 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
1068 that moved to I2, also allow destinations that are a paradoxical
1069 subreg (instead of a normal reg).
1071 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
1074 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
1077 2018-02-16 Richard Biener <rguenther@suse.de>
1079 PR tree-optimization/84037
1080 PR tree-optimization/84016
1082 * config/i386/i386.c (ix86_builtin_vectorization_cost):
1083 Adjust vec_construct for the fact we need additional higher latency
1084 128bit inserts for AVX256 and AVX512 vector builds.
1085 (ix86_add_stmt_cost): Scale vector construction cost for
1088 2018-02-16 Richard Biener <rguenther@suse.de>
1090 PR tree-optimization/84417
1091 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
1092 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
1093 (non_rewritable_lvalue_p): Likewise, use poly-ints.
1095 2018-02-16 Martin Liska <mliska@suse.cz>
1098 * internal-fn.def (ASAN_CHECK): Set proper flags.
1099 (ASAN_MARK): Likewise.
1101 2018-02-16 Julia Koval <julia.koval@intel.com>
1103 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
1104 from PTA_CANNONLAKE.
1106 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1109 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
1110 Use ++iter rather than iter++ for std::list iterators.
1111 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
1112 defer deleting them until all nodes in the forest are processed. Do
1113 free even leaf nodes. Change to_process into auto_vec.
1116 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
1117 * vec.h (vec_default_construct): Use memset instead of placement new
1118 if BROKEN_VALUE_INITIALIZATION is defined.
1119 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
1120 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
1123 PR rtl-optimization/83723
1124 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
1125 * lra.c (lra_substitute_pseudo): Likewise. If true, use
1126 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
1128 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
1130 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
1132 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
1134 PR rtl-optimization/81443
1135 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
1136 from inner REGs to paradoxical SUBREGs.
1138 2018-02-16 Richard Biener <rguenther@suse.de>
1140 PR tree-optimization/84399
1141 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1142 For operands we can analyze at their definition make sure we can
1143 analyze them at each use as well.
1145 2018-02-16 Richard Biener <rguenther@suse.de>
1147 PR tree-optimization/84190
1148 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
1149 volatile accesses if the decl isn't volatile.
1151 2018-02-15 Jason Merrill <jason@redhat.com>
1153 PR c++/84314 - ICE with templates and fastcall attribute.
1154 * attribs.c (build_type_attribute_qual_variant): Don't clobber
1155 TYPE_CANONICAL on an existing type.
1157 2018-02-15 Jakub Jelinek <jakub@redhat.com>
1159 PR tree-optimization/84383
1160 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
1161 dstoff nor call operand_equal_p if dstbase is NULL.
1163 PR tree-optimization/84334
1164 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
1165 also a CONSTANT_CLASS_P, punt.
1167 2018-02-14 Jim Wilson <jimw@sifive.com>
1169 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
1170 first SMALL_OPERAND check. New local min_second_step. Move assert
1171 to where locals are set. Add TARGET_RVC support.
1172 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
1174 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
1176 * doc/invoke.texi: Correct -Wformat-overflow code sample.
1178 2018-02-14 Martin Sebor <msebor@redhat.com>
1180 PR tree-optimization/83698
1181 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
1182 arrays constrain the offset range to their bounds.
1183 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
1184 (builtin_access::overlap): Avoid setting the size of overlap if it's
1186 (maybe_diag_overlap): Also consider arrays when deciding what values
1187 of offsets to include in diagnostics.
1189 2018-02-14 Martin Sebor <msebor@redhat.com>
1192 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
1193 that correspond to the kind of a declaration.
1195 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
1198 * config/pa/pa.md: Load address of PIC label using the linkage table
1199 if the label is nonlocal.
1201 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
1203 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
1204 warning message if user requests -maltivec=be.
1205 * doc/invoke.texi: Document deprecation of -maltivec=be.
1207 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
1210 * config/rs6000/rs6000-c.c: Update definitions for
1211 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
1212 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
1214 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1217 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
1218 add _get_ssp intrinsics. Remove argument from
1219 __builtin_ia32_rdssp[d|q].
1220 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
1221 * config/i386/i386-builtin.def: Remove argument from
1222 __builtin_ia32_rdssp[d|q].
1223 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
1224 ix86_expand_special_args_builtin for _rdssp[d|q].
1225 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
1226 Clear register before usage.
1227 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
1228 Add documentation for new _get_ssp and _inc_ssp intrinsics.
1230 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
1232 PR tree-optimization/84357
1233 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
1234 operand 1 of an ARRAY_REF too.
1236 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
1239 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
1240 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1242 (set_of_reg): New struct.
1243 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
1244 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
1245 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1247 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
1248 Split into bitclr, bitset, bitinvert patterns if appropriate.
1249 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
1250 use rx_fuse_in_memory_bitop.
1251 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
1252 to named insn, correct maximum insn length.
1254 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
1257 * machmode.def: Define a complex mode for PARTIAL_INT.
1258 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
1260 * doc/rtl.texi: Document CSPImode.
1261 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
1263 (msp430_hard_regno_nregs_with_padding): Likewise.
1265 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
1268 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
1270 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
1272 PR rtl-optimization/84169
1273 * combine.c (try_combine): New variable split_i2i3. Set it to true if
1274 we generated a parallel as new i3 and we split that to new i2 and i3
1275 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
1276 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
1277 those to i2, not i1. Partially rewrite this scan code.
1279 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1282 * stor-layout.c (place_field): For variable length fields, adjust
1283 offset_align afterwards not just based on the field's alignment,
1284 but also on the size.
1287 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
1288 of exps and logs in the use_exp2 case.
1290 2018-02-13 Jeff Law <law@redhat.com>
1292 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
1295 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
1298 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
1302 * common.opt (gas-loc-support, gas-locview-support): New.
1303 (ginline-points, ginternal-reset-location-views): New.
1304 * doc/invoke.texi: Document them. Use @itemx where intended.
1305 (gvariable-location-views): Adjust.
1306 * target.def (reset_location_view): New.
1307 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
1308 (TARGET_RESET_LOCATION_VIEW): New.
1309 * doc/tm.texi: Rebuilt.
1310 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
1311 (dwarf2out_default_as_locview_support): New.
1312 (output_asm_line_debug_info): Use option variables.
1313 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
1314 (output_loc_list): Likewise.
1315 (add_high_low_attributes): Check option variables.
1316 Don't output entry view attribute in strict mode.
1317 (gen_inlined_subroutine_die): Check option variables.
1318 (dwarf2out_inline_entry): Likewise.
1319 (init_sections_and_labels): Likewise.
1320 (dwarf2out_early_finish): Likewise.
1321 (maybe_reset_location_view): New, from...
1322 (dwarf2out_var_location): ... here. Call it.
1323 * debug.h (dwarf2out_default_as_loc_support): Declare.
1324 (dwarf2out_default_as_locview_support): Declare.
1325 * hooks.c (hook_int_rtx_insn_0): New.
1326 * hooks.h (hook_int_rtx_insn_0): Declare.
1327 * toplev.c (process_options): Take -gas-loc-support and
1328 -gas-locview-support from dwarf2out. Enable
1329 -gvariable-location-views by default only with locview
1330 assembler support. Enable -ginternal-reset-location-views by
1331 default only if the target defines the corresponding hook.
1332 Enable -ginline-points by default if location views are
1333 enabled; force it disabled if statement frontiers are
1335 * tree-inline.c (expand_call_inline): Check option variables.
1336 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1338 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
1340 PR tree-optimization/84321
1341 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
1342 handling. Also check whether the anti-range contains any values
1343 that satisfy the mask; switch to a VR_RANGE if not.
1345 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
1348 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
1350 2018-02-13 Martin Jambor <mjambor@suse.cz>
1353 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
1354 of call statements, also set location of a load to a temporary.
1356 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
1358 * config/rl78/rl78.c (add_vector_labels): New function.
1359 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
1360 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
1361 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
1362 which checks that no arguments are passed.
1363 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
1364 * doc/extend.texi: Documentation for the new attribute.
1366 2018-02-13 Andreas Schwab <schwab@suse.de>
1368 * config/riscv/linux.h (CPP_SPEC): Define.
1370 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1373 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
1374 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
1375 OPTION_MASK_ISA_AES as first argument to def_builtin_const
1376 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
1377 instead of OPTION_MASK_ISA_PCLMUL as first argument to
1378 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
1379 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
1380 temporarily for AES and PCLMUL builtins.
1382 PR tree-optimization/84339
1383 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
1384 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
1388 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
1389 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
1390 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
1392 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
1394 * omp-simd-clone.h: New file.
1395 * omp-simd-clone.c: Include omp-simd-clone.h.
1396 (expand_simd_clones): No longer static.
1397 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
1398 cgraph.h and omp-simd-clone.h.
1399 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
1400 (vect_recog_widen_shift_pattern): Formatting fix.
1401 (vect_pattern_recog_1): Don't check optab for calls.
1404 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
1405 operands[2] into a REG before using gen_lowpart on it.
1407 2018-02-12 Jeff Law <law@redhat.com>
1410 * config/sh/sh.c (find_barrier): Consider a sibling call
1413 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
1414 successfully back substituting a reg.
1416 2018-02-12 Richard Biener <rguenther@suse.de>
1418 PR tree-optimization/84037
1419 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
1420 parameter, move visited init to caller.
1421 (vect_slp_analyze_operations): Separate cost from validity
1422 check, initialize visited once for all instances.
1423 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
1425 * tree-vect-stmts.c (vect_model_simple_cost): Make early
1427 (vect_model_promotion_demotion_cost): Likewise.
1428 (vectorizable_bswap): Guard cost modeling with !slp_node
1429 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
1431 (vectorizable_call): Likewise.
1432 (vectorizable_conversion): Likewise.
1433 (vectorizable_assignment): Likewise.
1434 (vectorizable_shift): Likewise.
1435 (vectorizable_operation): Likewise.
1436 (vectorizable_store): Likewise.
1437 (vectorizable_load): Likewise.
1438 (vectorizable_condition): Likewise.
1439 (vectorizable_comparison): Likewise.
1441 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
1444 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
1445 (ASAN_MARK): Fix fnspec to account for return value, change pointer
1446 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
1448 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1451 * params.def (inline-min-speedup): Increase from 8 to 15.
1452 (max-inline-insns-auto): Decrease from 40 to 30.
1453 * ipa-split.c (consider_split): Add some buffer for function to
1454 be considered inlining candidate.
1455 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
1458 2018-02-12 Richard Biener <rguenther@suse.de>
1460 PR tree-optimization/84037
1461 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
1462 matched stmts if we cannot swap the non-matched ones.
1464 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1466 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
1467 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
1468 _mm_maskz_scalef_round_ss): New intrinsics.
1469 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
1470 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
1471 __builtin_ia32_scalefss_round): Remove.
1472 (__builtin_ia32_scalefsd_mask_round,
1473 __builtin_ia32_scalefss_mask_round): New intrinsics.
1474 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
1475 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
1476 ((match_operand:VF_128 2 "<round_nimm_predicate>"
1477 "<round_constraint>")): Changed to ...
1478 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
1479 "<round_scalar_constraint>")): ... this.
1480 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
1481 %0, %1, %2<round_op3>}"): Changed to ...
1482 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
1483 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
1484 %2<round_scalar_mask_op3>}"): ... this.
1485 * config/i386/subst.md (round_scalar_nimm_predicate): New.
1487 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1489 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
1490 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
1491 (_mm_maskz_sqrt_round_ss): New intrinsics.
1492 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
1493 (__builtin_ia32_sqrtsd_mask_round)
1494 (__builtin_ia32_sqrtss_mask_round): New builtins.
1495 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
1496 (__builtin_ia32_sqrtss_round): Remove.
1497 (__builtin_ia32_sqrtsd_mask_round)
1498 (__builtin_ia32_sqrtss_mask_round): New builtins.
1499 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
1500 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
1501 ((match_operand:VF_128 1 "vector_operand"
1502 "xBm,<round_constraint>")): Changed to ...
1503 ((match_operand:VF_128 1 "vector_operand"
1504 "xBm,<round_scalar_constraint>")): ... this.
1505 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
1506 %0, %2, %<iptr>1<round_op3>}): Changed to ...
1507 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
1508 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
1509 %<iptr>1<round_scalar_mask_op3>}): ... this.
1510 ((set_attr "prefix" "<round_prefix>")): Changed to ...
1511 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
1513 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
1516 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
1517 Cast vec_cmpeq result to correct type.
1518 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
1519 Cast vec_cmpgt result to correct type.
1521 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
1523 * final.c (final_scan_insn_1): Renamed from...
1524 (final_scan_insn): ... this. New wrapper, to recover
1525 seen from the outermost call in recursive ones.
1526 * config/sparc/sparc.c (output_return): Drop seen from call.
1527 (output_sibcall): Likewise.
1528 * config/visium/visium.c (output_branch): Likewise.
1530 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
1532 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
1535 2018-02-10 Alan Modra <amodra@gmail.com>
1538 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
1539 Specify LR as an input.
1541 2018-02-10 Jakub Jelinek <jakub@redhat.com>
1544 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
1545 remove_member_access_dummy_vars): New functions.
1546 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
1547 lower_omp_1, execute_lower_omp): Use them.
1549 PR rtl-optimization/84308
1550 * shrink-wrap.c (spread_components): Release todo vector.
1552 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
1554 PR rtl-optimization/57193
1555 * ira-color.c (struct allocno_color_data): Add member
1556 conflict_allocno_hard_prefs.
1557 (update_conflict_allocno_hard_prefs): New.
1558 (bucket_allocno_compare_func): Add a preference based on
1559 conflict_allocno_hard_prefs.
1560 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
1561 (color_allocnos): Remove a dead code. Initiate
1562 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
1564 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1567 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
1568 constraint from =wa to wa. Avoid a subreg on the output operand,
1569 instead use a pseudo and subreg it in a move.
1570 (p9_xxbrd_<mode>): Changed to ...
1571 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
1572 (p9_xxbrd_v2df): New expander.
1573 (p9_xxbrw_<mode>): Changed to ...
1574 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
1575 (p9_xxbrw_v4sf): New expander.
1577 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1579 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
1581 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
1584 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
1585 multiply in 32-bit mode.
1586 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
1587 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
1590 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1592 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
1593 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
1594 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
1595 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
1597 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
1600 * dwarf2out.c (is_trivial_indirect_ref): New function.
1601 (dwarf2out_late_global_decl): Do not generate a location
1602 attribute for variables that have a non-trivial DECL_VALUE_EXPR
1603 and that are not defined in the current unit.
1605 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1607 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
1608 instead of a libcall for UNORDERED.
1610 2018-02-09 Tamar Christina <tamar.christina@arm.com>
1613 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
1614 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
1616 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1619 * config/s390/s390.c (s390_set_current_function): Invoke
1620 s390_indirect_branch_settings also if fndecl didn't change.
1622 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1624 * config/rs6000/rs6000.md (blockage): Set length to zero.
1626 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1628 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
1630 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1633 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
1634 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
1638 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
1639 PARALLEL incoming that failed vt_get_decl_and_offset check.
1642 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
1643 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
1645 (get_variable_section): For decls in named .bss* sections pass true as
1646 second argument to bss_initializer_p.
1648 2018-02-09 Marek Polacek <polacek@redhat.com>
1649 Jakub Jelinek <jakub@redhat.com>
1652 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
1653 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
1654 Sync some changes from cxx_fold_indirect_ref.
1656 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1658 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
1660 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
1661 (BLOCK_INLINE_ENTRY_LABEL): New.
1662 (dwarf2out_var_location): Disregard inline entry markers.
1663 (inline_entry_data): New struct.
1664 (inline_entry_data_hasher): New hashtable type.
1665 (inline_entry_data_hasher::hash): New.
1666 (inline_entry_data_hasher::equal): New.
1667 (inline_entry_data_table): New variable.
1668 (add_high_low_attributes): Add DW_AT_entry_pc and
1669 DW_AT_GNU_entry_view attributes if a pending entry is found
1670 in inline_entry_data_table. Add old entry_pc attribute only
1671 if debug nonbinding markers are disabled.
1672 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
1673 markers are enabled.
1674 (block_within_block_p, dwarf2out_inline_entry): New.
1675 (dwarf2out_finish): Check that no entries remained in
1676 inline_entry_data_table.
1677 * final.c (reemit_insn_block_notes): Handle inline entry notes.
1678 (final_scan_insn, notice_source_line): Likewise.
1679 (rest_of_clean_state): Skip inline entry markers.
1680 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
1682 * gimple.c (gimple_build_debug_inline_entry): New.
1683 * gimple.h (enum gimple_debug_subcode): Add
1684 GIMPLE_DEBUG_INLINE_ENTRY.
1685 (gimple_build_debug_inline_entry): Declare.
1686 (gimple_debug_inline_entry_p): New.
1687 (gimple_debug_nonbind_marker_p): Adjust.
1688 * insn-notes.def (INLINE_ENTRY): New.
1689 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
1690 inline entry marker notes.
1691 (print_insn): Likewise.
1692 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
1693 (INSN_DEBUG_MARKER_KIND): Likewise.
1694 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
1695 * tree-inline.c (expand_call_inline): Build and insert
1696 debug_inline_entry stmt.
1697 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
1698 inline entry blocks early, if nonbind markers are enabled.
1699 (dump_scope_block): Dump fragment info.
1700 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
1701 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
1702 (gimple_build_debug_inline_entry): New.
1703 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
1704 Enable/disable inline entry points too.
1705 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
1706 (DEBUG_INSN): Describe inline entry markers.
1708 * common.opt (gvariable-location-views): New.
1709 (gvariable-location-views=incompat5): New.
1710 * config.in: Rebuilt.
1711 * configure: Rebuilt.
1712 * configure.ac: Test assembler for view support.
1713 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
1714 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
1715 * dwarf2out.c (var_loc_view): New typedef.
1716 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
1717 (dwarf2out_locviews_in_attribute): New.
1718 (dwarf2out_locviews_in_loclist): New.
1719 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
1720 (enum dw_line_info_opcode): Add LI_adv_address.
1721 (struct dw_line_info_table): Add view.
1722 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
1723 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
1724 (zero_view_p): New variable.
1725 (ZERO_VIEW_P): New macro.
1726 (output_asm_line_debug_info): New.
1727 (struct var_loc_node): Add view.
1728 (add_AT_view_list, AT_loc_list): New.
1729 (add_var_loc_to_decl): Add view param. Test it against last.
1730 (new_loc_list): Add view params. Record them.
1731 (AT_loc_list_ptr): Handle loc and view lists.
1732 (view_list_to_loc_list_val_node): New.
1733 (print_dw_val): Handle dw_val_class_view_list.
1734 (size_of_die): Likewise.
1735 (value_format): Likewise.
1736 (loc_list_has_views): New.
1737 (gen_llsym): Set vl_symbol too.
1738 (maybe_gen_llsym, skip_loc_list_entry): New.
1739 (dwarf2out_maybe_output_loclist_view_pair): New.
1740 (output_loc_list): Output view list or entries too.
1741 (output_view_list_offset): New.
1742 (output_die): Handle dw_val_class_view_list.
1743 (output_dwarf_version): New.
1744 (output_compilation_unit_header): Use it.
1745 (output_skeleton_debug_sections): Likewise.
1746 (output_rnglists, output_line_info): Likewise.
1747 (output_pubnames, output_aranges): Update version comments.
1748 (output_one_line_info_table): Output view numbers in asm comments.
1749 (dw_loc_list): Determine current endview, pass it to new_loc_list.
1750 Call maybe_gen_llsym.
1751 (loc_list_from_tree_1): Adjust.
1752 (add_AT_location_description): Create view list attribute if
1753 needed, check it's absent otherwise.
1754 (convert_cfa_to_fb_loc_list): Adjust.
1755 (maybe_emit_file): Call output_asm_line_debug_info for test.
1756 (dwarf2out_var_location): Reset views as needed. Precompute
1757 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
1758 attribute. Set view.
1759 (new_line_info_table): Reset next view.
1760 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
1761 (dwarf2out_source_line): Likewise. Output view resets and labels to
1762 the assembler, or select appropriate line info opcodes.
1763 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
1764 (optimize_string_length): Catch it. Adjust.
1765 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
1766 dw_val_class_view_list, and remove it if no longer needed.
1767 (hash_loc_list): Hash view numbers.
1768 (loc_list_hasher::equal): Compare them.
1769 (optimize_location_lists): Check whether a view list symbol is
1770 needed, and whether the locview attribute is present, and
1771 whether they match. Remove the locview attribute if no longer
1773 (index_location_lists): Call skip_loc_list_entry for test.
1774 (dwarf2out_finish): Call output_asm_line_debug_info for test.
1775 Use output_dwarf_version.
1776 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
1777 (struct dw_val_node): Add val_view_list.
1778 * final.c (SEEN_NEXT_VIEW): New.
1779 (set_next_view_needed): New.
1780 (clear_next_view_needed): New.
1781 (maybe_output_next_view): New.
1782 (final_start_function): Rename to...
1783 (final_start_function_1): ... this. Take pointer to FIRST,
1784 add SEEN parameter. Emit param bindings in the initial view.
1785 (final_start_function): Reintroduce SEEN-less interface.
1786 (final): Rename to...
1787 (final_1): ... this. Take SEEN parameter. Output final pending
1788 next view at the end.
1789 (final): Reintroduce seen-less interface.
1790 (final_scan_insn): Output pending next view before switching
1791 sections or ending a block. Mark the next view as needed when
1792 outputting variable locations. Notify debug backend of section
1793 changes, and of location view changes.
1794 (rest_of_handle_final): Adjust.
1795 * toplev.c (process_options): Autodetect value for debug variable
1796 location views option. Warn on incompat5 without -gdwarf-5.
1797 * doc/invoke.texi (gvariable-location-views): New.
1798 (gvariable-location-views=incompat5): New.
1799 (gno-variable-location-views): New.
1801 2018-02-08 David Malcolm <dmalcolm@redhat.com>
1803 PR tree-optimization/84136
1804 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
1805 that the result of find_edge is non-NULL.
1807 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1810 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
1811 storing integer register in SImode. Fix cost of 256 and 512
1812 byte aligned SSE register store.
1814 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1816 * config/i386/i386.c (ix86_multiplication_cost): Fix
1817 multiplication cost for TARGET_AVX512DQ.
1819 2018-02-08 Marek Polacek <polacek@redhat.com>
1821 PR tree-optimization/84238
1822 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
1825 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1827 PR tree-optimization/84265
1828 * tree-vect-stmts.c (vectorizable_store): Don't treat
1829 VMAT_CONTIGUOUS accesses as grouped.
1830 (vectorizable_load): Likewise.
1832 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1834 PR tree-optimization/81635
1835 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
1836 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
1837 (test_round_for_mask): New functions.
1838 (wide_int_cc_tests): Call test_round_for_mask.
1839 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
1840 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
1841 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
1842 range returned by get_range_info.
1844 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1847 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
1848 * symtab.c: Include builtins.h
1849 (symtab_node::output_to_lto_symbol_table_p): Move here
1850 from lto-streamer-out.c:output_symbol_p.
1851 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
1852 (output_symbol_p): Move all logic to symtab.c
1853 (produce_symtab): Update.
1855 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1857 * config/s390/s390-opts.h (enum indirect_branch): Define.
1858 * config/s390/s390-protos.h (s390_return_addr_from_memory)
1859 (s390_indirect_branch_via_thunk)
1860 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
1861 (enum s390_indirect_branch_type): Define.
1862 * config/s390/s390.c (struct s390_frame_layout, struct
1863 machine_function): Remove.
1864 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
1865 (indirect_branch_table_label_no, indirect_branch_table_name):
1867 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
1868 (enum s390_indirect_branch_option): Define.
1869 (s390_return_addr_from_memory): New function.
1870 (s390_handle_string_attribute): New function.
1871 (s390_attribute_table): Add new attribute handler.
1872 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
1873 (s390_indirect_branch_via_thunk): New function.
1874 (s390_indirect_branch_via_inline_thunk): New function.
1875 (s390_function_ok_for_sibcall): When jumping via thunk disallow
1876 sibling call optimization for non z10 compiles.
1877 (s390_emit_call): Force indirect branch target to be a single
1878 register. Add r1 clobber for non-z10 compiles.
1879 (s390_emit_epilogue): Emit return jump via return_use expander.
1880 (s390_reorg): Handle JUMP_INSNs as execute targets.
1881 (s390_option_override_internal): Perform validity checks for the
1882 new command line options.
1883 (s390_indirect_branch_attrvalue): New function.
1884 (s390_indirect_branch_settings): New function.
1885 (s390_set_current_function): Invoke s390_indirect_branch_settings.
1886 (s390_output_indirect_thunk_function): New function.
1887 (s390_code_end): Implement target hook.
1888 (s390_case_values_threshold): Implement target hook.
1889 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
1891 * config/s390/s390.h (struct s390_frame_layout)
1892 (struct machine_function): Move here from s390.c.
1893 (TARGET_INDIRECT_BRANCH_NOBP_RET)
1894 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
1895 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
1896 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
1897 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
1898 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
1899 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
1900 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
1901 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
1902 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
1903 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
1904 (mnemonic attribute): Add values which aren't recognized
1906 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
1907 pattern for branch conversion. Fix mnemonic attribute.
1908 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
1909 indirect branch via thunk if requested.
1910 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
1911 ("*indirect_jump"): Disable for branch conversion using out of
1913 ("indirect_jump_via_thunk<mode>_z10")
1914 ("indirect_jump_via_thunk<mode>")
1915 ("indirect_jump_via_inlinethunk<mode>_z10")
1916 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
1917 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
1918 ("casesi_jump_via_inlinethunk<mode>_z10")
1919 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
1920 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
1921 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
1922 ("*indirect2_jump"): Disable for branch conversion.
1923 ("casesi_jump"): Turn into expander and expand patterns for branch
1925 ("return_use"): New expander.
1926 ("*return"): Emit return via thunk and rename it to ...
1927 ("*return<mode>"): ... this one.
1928 * config/s390/s390.opt: Add new options and and enum for the
1931 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1933 * lra-constraints.c (match_reload): Unconditionally use
1934 gen_lowpart_SUBREG, rather than selecting between that
1935 and equivalent gen_rtx_SUBREG code.
1937 2018-02-08 Richard Biener <rguenther@suse.de>
1939 PR tree-optimization/84233
1940 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
1941 changed flag instead of boguously re-using phi_inserted.
1943 2018-02-08 Martin Jambor <mjambor@suse.cz>
1945 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
1946 static local variables.
1948 2018-02-08 Richard Biener <rguenther@suse.de>
1950 PR tree-optimization/84278
1951 * tree-vect-stmts.c (vectorizable_store): When looking for
1952 smaller vector types to perform grouped strided loads/stores
1953 make sure the mode is supported by the target.
1954 (vectorizable_load): Likewise.
1956 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1958 * config/aarch64/aarch64.c (aarch64_components_for_bb):
1959 Increase LDP/STP opportunities by adding adjacent callee-saves.
1961 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1963 PR rtl-optimization/84068
1964 PR rtl-optimization/83459
1965 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
1967 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
1969 PR tree-optimization/84224
1970 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
1971 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
1974 2018-02-07 Iain Sandoe <iain@codesourcery.com>
1977 * config/rs6000/altivec.md (*restore_world): Remove LR use.
1978 * config/rs6000/predicates.md (restore_world_operation): Adjust op
1979 count, remove one USE.
1981 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
1983 * doc/install.texi (Configuration): Document the
1984 --with-long-double-format={ibm,ieee} PowerPC configuration
1988 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
1989 Convert from define_expand to be define_insn_and_split. Rework
1990 float/double/_Float128 conversions to QI/HI/SImode to work with
1991 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
1992 conversions to QI/HImode types did a store and then a load to
1993 truncate the value. For conversions to VSX registers, don't split
1994 the insn, instead emit the code directly. Use the code iterator
1995 any_fix to combine signed and unsigned conversions.
1996 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
1997 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
1998 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
1999 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
2000 (fix_<mode>di2_hw): Likewise.
2001 (fixuns_<mode>di2_hw): Likewise.
2002 (fix_<mode>si2_hw): Likewise.
2003 (fixuns_<mode>si2_hw): Likewise.
2004 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
2005 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2006 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
2007 fix<uns>_trunc<SFDF:mode>si2_p8.
2008 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
2010 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
2011 (fix<uns>_<mode>_mem): Likewise.
2012 (fctiw<u>z_<mode>_mem): Likewise.
2013 (fix<uns>_<mode>_mem): Likewise.
2014 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
2015 the register allocator from doing a direct move to the GPRs to do
2016 a store, and instead use the ISA 3.0 store byte/half-word from
2017 vector register instruction. For IEEE 128-bit floating point,
2018 also optimize stores of 32-bit ints.
2019 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
2021 2018-02-07 Alan Hayward <alan.hayward@arm.com>
2023 * genextract.c (push_pathstr_operand): New function to support
2025 (walk_rtx): Call push_pathstr_operand.
2026 (print_path): Support [a-zA-Z].
2028 2018-02-07 Richard Biener <rguenther@suse.de>
2030 PR tree-optimization/84037
2031 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
2032 (cse_and_gimplify_to_preheader): Declare.
2033 (vect_get_place_in_interleaving_chain): Likewise.
2034 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2036 (_loop_vec_info::~_loop_vec_info): Delete it.
2037 (cse_and_gimplify_to_preheader): New function.
2038 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
2039 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
2040 (vectorizable_load): Likewise. For grouped stores always base
2041 the IV on the first element.
2042 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
2043 condition before gimplifying.
2045 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2047 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
2048 *DIV_EXPR and *MOD_EXPR.
2050 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
2053 * config/i386/i386.c (ix86_option_override_internal): Mask out
2054 the CF_SET bit when checking -fcf-protection.
2056 2018-02-07 Tom de Vries <tom@codesourcery.com>
2059 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
2062 2018-02-07 Richard Biener <rguenther@suse.de>
2064 PR tree-optimization/84204
2065 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2068 PR tree-optimization/84205
2069 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
2070 special-case isl_ast_op_zdiv_r.
2072 PR tree-optimization/84223
2073 * graphite-scop-detection.c (gather_bbs::before_dom_children):
2074 Only add conditions from within the region.
2075 (gather_bbs::after_dom_children): Adjust.
2077 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
2080 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
2081 * config/avr/avr.md: Only post-reload split REG-REG moves if
2082 either register is REGERAL_REG_P.
2084 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2086 PR tree-optimization/84235
2087 * tree-ssa-scopedtables.c
2088 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
2089 if the subtraction is performed in floating point type where NaNs are
2090 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
2091 build 1. Formatting fix.
2093 2018-02-06 Jakub Jelinek <jakub@redhat.com>
2096 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
2097 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
2098 and skip it regardless of bb boundaries. Use CALL_P macro,
2099 don't test INSN_P (insn) together with CALL_P or JUMP_P check
2100 unnecessarily, formatting fix.
2102 2018-02-06 Michael Collison <michael.collison@arm.com>
2104 * config/arm/thumb2.md:
2105 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
2106 (*thumb_mov_notscc): Ditto.
2108 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2111 * config/rs6000/rs6000.md (su code attribute): Use "u" for
2112 unsigned_fix, not "s".
2114 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2116 * configure.ac (gcc_fn_eh_frame_ro): New function.
2117 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
2118 correct .eh_frame permissions.
2119 * configure: Regenerate.
2121 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
2123 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
2126 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2128 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2129 Display warning message for -mno-speculate-indirect-jumps.
2131 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
2133 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
2135 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
2137 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
2139 PR tree-optimization/84225
2140 * tree-eh.c (find_trapping_overflow): Only call
2141 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
2143 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2146 * config/i386/i386.c: Reimplement the check of possible options
2147 -mibt/-mshstk conbination. Change error messages.
2148 * doc/invoke.texi: Fix a typo: remove extra '='.
2150 2018-02-06 Marek Polacek <polacek@redhat.com>
2152 PR tree-optimization/84228
2153 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
2155 2018-02-06 Tamar Christina <tamar.christina@arm.com>
2158 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
2159 emitted arch directives.
2160 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
2161 __ARM_FEATURE_COPROC before changing architectures.
2163 2018-02-06 Richard Biener <rguenther@suse.de>
2165 * config/i386/i386.c (print_reg): Fix typo.
2166 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
2168 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
2170 * configure: Regenerate.
2172 2018-02-05 Martin Sebor <msebor@redhat.com>
2174 PR tree-optimization/83369
2175 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
2178 2018-02-05 Martin Liska <mliska@suse.cz>
2180 * doc/invoke.texi: Cherry-pick upstream r323995.
2182 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
2184 * ira.c (ira_init_register_move_cost): Adjust comment.
2186 2018-02-05 Martin Liska <mliska@suse.cz>
2188 PR gcov-profile/84137
2189 * doc/gcov.texi: Fix typo in documentation.
2191 2018-02-05 Martin Liska <mliska@suse.cz>
2193 PR gcov-profile/83879
2194 * doc/gcov.texi: Document necessity of --dynamic-list-data when
2195 using dlopen functionality.
2197 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
2199 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
2200 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
2201 _mm_maskz_range_ss, _mm_mask_range_round_ss,
2202 _mm_maskz_range_round_ss): New intrinsics.
2203 (__builtin_ia32_rangesd128_round)
2204 (__builtin_ia32_rangess128_round): Remove.
2205 (__builtin_ia32_rangesd128_mask_round,
2206 __builtin_ia32_rangess128_mask_round): New builtins.
2207 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
2208 __builtin_ia32_rangess128_round): Remove.
2209 (__builtin_ia32_rangesd128_mask_round,
2210 __builtin_ia32_rangess128_mask_round): New builtins.
2211 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
2212 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
2213 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
2214 "<round_saeonly_constraint>")): Changed to ...
2215 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
2216 "<round_saeonly_scalar_constraint>")): ... this.
2217 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
2218 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
2219 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
2220 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
2221 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
2223 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
2225 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
2227 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
2228 Remove all values except native, 8540 and 8548.
2230 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
2232 * config/i386/i386.c (ix86_output_function_return): Pass
2233 INVALID_REGNUM, instead of -1, as invalid register number to
2234 indirect_thunk_name and output_indirect_thunk.
2236 2018-02-02 Julia Koval <julia.koval@intel.com>
2238 * config.gcc: Add -march=icelake.
2239 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
2240 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
2241 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
2242 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
2243 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
2244 (processor_target_table): Add icelake.
2245 (ix86_option_override_internal): Handle new PTAs.
2246 (get_builtin_code_for_version): Handle icelake.
2247 (M_INTEL_COREI7_ICELAKE): New.
2248 (fold_builtin_cpu): Handle icelake.
2249 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
2250 * doc/invoke.texi: Add -march=icelake.
2252 2018-02-02 Julia Koval <julia.koval@intel.com>
2254 * config/i386/i386.c (ix86_option_override_internal): Change flags type
2255 to wide_int_bitmask.
2256 * wide-int-bitmask.h: New.
2258 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2261 * config/i386/i386.md: Replace Pmode with word_mode in
2262 builtin_setjmp_setup and builtin_longjmp to support x32.
2264 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
2268 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
2270 (rs6000_supported_cpu_names): New static variable.
2271 (linux_cpu_translation_table): Likewise.
2272 (elf_platform) <cpu>: Define new static variable and use it.
2273 Translate kernel AT_PLATFORM name to canonical name if needed.
2274 Error if platform name is unknown.
2276 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
2279 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
2281 2018-02-01 Jeff Law <law@redhat.com>
2284 * config/i386/i386.c (release_scratch_register_on_entry): Add new
2285 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
2286 the scratch if RELEASE_VIA_POP is false.
2287 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
2288 If we have to save a temporary register, decrement SIZE appropriately.
2289 Pass new arguments to release_scratch_register_on_entry.
2290 (ix86_adjust_stack_and_probe): Likewise.
2291 (ix86_emit_probe_stack_range): Pass new arguments to
2292 release_scratch_register_on_entry.
2294 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
2296 PR rtl-optimization/84157
2297 * combine.c (change_zero_ext): Use REG_P predicate in
2298 front of HARD_REGISTER_P predicate.
2300 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
2302 * config/avr/avr.c (avr_option_override): Move disabling of
2303 -fdelete-null-pointer-checks to...
2304 * common/config/avr/avr-common.c (avr_option_optimization_table):
2307 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2309 PR tree-optimization/81635
2310 * tree-data-ref.c (split_constant_offset_1): For types that
2311 wrap on overflow, try to use range info to prove that wrapping
2314 2018-02-01 Renlin Li <renlin.li@arm.com>
2317 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
2319 (aarch64_register_move_cost): Likewise.
2320 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
2322 (REG_CLASS_NAMES): Likewise.
2323 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
2324 TAILCALL_ADDR_REGS. Remove IP registers.
2325 * config/aarch64/aarch64.md (Ucs): Update register constraint.
2327 2018-02-01 Richard Biener <rguenther@suse.de>
2329 * domwalk.h (dom_walker::dom_walker): Add additional constructor
2330 for specifying RPO order and allow NULL for that.
2331 * domwalk.c (dom_walker::dom_walker): Likewise.
2332 (dom_walker::walk): Handle NULL RPO order.
2333 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
2335 (rewrite_update_dom_walker): Likewise.
2336 (mark_def_dom_walker): Likewise.
2338 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2340 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
2341 (aarch64_maybe_expand_sve_subreg_move): Declare.
2342 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
2343 * config/aarch64/predicates.md (aarch64_any_register_operand): New
2345 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
2346 that are semantically a reverse operation.
2347 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
2348 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
2349 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
2351 (aarch64_can_change_mode_class): For big-endian, forbid changes
2352 between two SVE modes if they have different element sizes.
2354 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2356 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
2357 the TImode handling for big-endian targets.
2359 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2361 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
2362 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
2364 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
2365 Remove BSWAP handing for big-endian targets and use the form of
2366 LD1RQ appropariate for the mode.
2368 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2370 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
2371 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
2374 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2377 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
2378 check for operands that need to go through aarch64_sve_reload_be.
2380 2018-02-01 Jakub Jelinek <jakub@redhat.com>
2382 PR tree-optimization/81661
2383 PR tree-optimization/84117
2384 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
2385 * tree-eh.c: Include gimplify.h.
2386 (find_trapping_overflow, replace_trapping_overflow,
2387 rewrite_to_non_trapping_overflow): New functions.
2388 * tree-vect-loop.c: Include tree-eh.h.
2389 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
2390 * tree-data-ref.c: Include tree-eh.h.
2391 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
2393 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
2395 PR rtl-optimization/84123
2396 * combine.c (change_zero_ext): Check if hard register satisfies
2397 can_change_dest_mode before calling gen_lowpart_SUBREG.
2399 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
2402 * ira.c (ira_init_register_move_cost): Remove assert.
2404 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2406 PR rtl-optimization/84071
2407 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
2408 * doc/tm.texi: Regenerate.
2410 2018-01-31 Richard Biener <rguenther@suse.de>
2412 PR tree-optimization/84132
2413 * tree-data-ref.c (analyze_miv_subscript): Properly
2414 check whether evolution_function_is_affine_multivariate_p
2415 before calling gcd_of_steps_may_divide_p.
2417 2018-01-31 Julia Koval <julia.koval@intel.com>
2420 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
2421 * config/i386/i386.md (rdpid_rex64) New.
2422 (rdpid): Make 32bit only.
2424 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
2427 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
2428 an IDENTIFIER_NODE for FUNCTION_TYPE's.
2430 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2433 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
2435 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
2437 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2439 PR rtl-optimization/84071
2440 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
2441 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
2443 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2445 * config/arc/arc.c (arc_handle_aux_attribute): New function.
2446 (arc_attribute_table): Add 'aux' attribute.
2447 (arc_in_small_data_p): Consider aux like variables.
2448 (arc_is_aux_reg_p): New function.
2449 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
2450 (arc_get_aux_arg): New function.
2451 (prepare_move_operands): Handle aux-register access.
2452 (arc_handle_aux_attribute): New function.
2453 * doc/extend.texi (ARC Variable attributes): Add subsection.
2455 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2457 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
2458 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
2459 (arc_attribute_table): Add 'uncached' attribute.
2460 (arc_print_operand): Print '.di' flag for uncached memory
2462 (arc_in_small_data_p): Do not consider for small data the uncached
2464 (arc_is_uncached_mem_p): New function.
2465 * config/arc/predicates.md (compact_store_memory_operand): Check
2466 for uncached memory accesses.
2467 (nonvol_nonimm_operand): Likewise.
2468 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
2470 2018-01-31 Jakub Jelinek <jakub@redhat.com>
2473 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
2474 falign-loops=): Add Optimization flag.
2476 2018-01-30 Jeff Law <law@redhat.com>
2479 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
2480 INT_REGISTERS_SAVED. Check it prior to calling
2481 get_scratch_register_on_entry.
2482 (ix86_adjust_stack_and_probe): Similarly.
2483 (ix86_emit_probe_stack_range): Similarly.
2484 (ix86_expand_prologue): Corresponding changes.
2486 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2489 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
2490 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
2492 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
2495 * lra-constraints.c (curr_insn_transform): Process AND in the
2498 2018-01-30 Jakub Jelinek <jakub@redhat.com>
2500 PR rtl-optimization/83986
2501 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
2502 dependence against last_pending_memory_flush in addition to
2505 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
2507 PR tree-optimization/81611
2508 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
2511 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2514 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
2517 2018-01-30 Richard Biener <rguenther@suse.de>
2518 Jakub Jelinek <jakub@redhat.com>
2520 PR tree-optimization/84111
2521 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
2522 inner loops added during recursion, as they don't have up-to-date
2525 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2528 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
2529 (can_inline_edge_by_limits_p): ... here.
2530 (can_early_inline_edge_p, check_callers,
2531 update_caller_keys, update_callee_keys, recursive_inlining,
2532 add_new_edges_to_heap, speculation_useful_p,
2533 inline_small_functions,
2534 inline_small_functions, flatten_function,
2535 inline_to_all_callers_1): Update.
2537 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2539 * profile-count.c (profile_count::combine_with_ipa_count): Handle
2542 2018-01-30 Richard Biener <rguenther@suse.de>
2544 PR tree-optimization/83008
2545 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
2546 invariant and constant vector uses in stmts when they need
2549 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2552 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
2553 * configure: Regenerate.
2555 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2557 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
2559 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
2560 Use gen_rtx_REG rather than gen_lowpart.
2562 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2564 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
2565 rather than 0 when creating partial subregs.
2567 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2569 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
2572 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
2575 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
2576 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
2577 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
2578 flags. This restores the settings used before the 2017-07-24.
2579 Turning off pre increment/decrement/modify allows IVOPTS to
2580 optimize DF/SF loops where the index is an int.
2582 2018-01-29 Richard Biener <rguenther@suse.de>
2583 Kelvin Nilsen <kelvin@gcc.gnu.org>
2586 * tree-vect-stmts.c (vectorizable_call): Don't call
2587 targetm.vectorize_builtin_md_vectorized_function if callee is
2590 2018-01-22 Carl Love <cel@us.ibm.com>
2592 * doc/extend.tex: Fix typo in second arg in
2593 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
2595 2018-01-29 Richard Biener <rguenther@suse.de>
2597 PR tree-optimization/84086
2598 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
2599 (flush_ssaname_freelist): When SSA names were released reset
2600 the SCEV hash table.
2602 2018-01-29 Richard Biener <rguenther@suse.de>
2604 PR tree-optimization/84057
2605 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
2606 removed paths when removing edges.
2608 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
2610 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
2611 -mfunction-return=@var{choice}.
2613 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
2616 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
2618 (layout::print_source_line): Likewise.
2619 (test_get_line_width_without_trailing_whitespace): Add test cases.
2621 2018-01-27 Jakub Jelinek <jakub@redhat.com>
2624 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
2627 2018-01-26 Jim Wilson <jimw@sifive.com>
2629 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
2631 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
2634 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2636 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
2637 and CMP + SUB-immediate -> SUBS.
2639 2018-01-26 Martin Sebor <msebor@redhat.com>
2641 PR tree-optimization/83896
2642 * tree-ssa-strlen.c (get_string_len): Rename...
2643 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
2644 Avoid assuming length is constant.
2645 (handle_char_store): Use HOST_WIDE_INT for string length.
2647 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
2650 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
2651 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
2653 2018-01-26 Richard Biener <rguenther@suse.de>
2655 PR rtl-optimization/84003
2656 * dse.c (record_store): Only record redundant stores when
2657 the earlier store aliases at least all accesses the later one does.
2659 2018-01-26 Jakub Jelinek <jakub@redhat.com>
2661 PR rtl-optimization/83985
2662 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
2663 REG_CFA_RESTORE insns.
2664 (delete_unmarked_insns): Don't ignore separate shrink wrapping
2665 REG_CFA_RESTORE insns here.
2668 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
2669 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
2671 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2673 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
2674 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
2675 (arc_init): Likewise.
2676 (arc_override_options): Likewise.
2677 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
2679 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
2681 * config/arc/arc.h (TARGET_DBNZ): Define.
2682 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
2683 properly set the tune attribute.
2684 (dbnz): Use TARGET_DBNZ guard.
2685 * config/arc/arc.opt (mtune): Add core3 option.
2687 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2689 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
2690 recognize new pic like addresses.
2691 (arc_delegitimize_address): Clean up.
2693 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2695 * config/arc/arc-arches.def: Option mrf16 valid for all
2697 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
2698 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
2699 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
2700 * config/arc/arc-tables.opt: Regenerate.
2701 * config/arc/arc.c (arc_conditional_register_usage): Handle
2702 reduced register file case.
2703 (arc_file_start): Set must have build attributes.
2704 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
2706 * config/arc/arc.opt (mrf16): Add new option.
2707 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
2708 * config/arc/genmultilib.awk: Handle new mrf16 option.
2709 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
2710 * config/arc/t-multilib: Regenerate.
2711 * doc/invoke.texi (ARC Options): Document mrf16 option.
2713 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2715 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
2716 * config/arc/arc.c (arc_handle_secure_attribute): New function.
2717 (arc_attribute_table): Add 'secure_call' attribute.
2718 (arc_print_operand): Print secure call operand.
2719 (arc_function_ok_for_sibcall): Don't optimize tail calls when
2721 (arc_is_secure_call_p): New function. * config/arc/arc.md
2722 (call_i): Add support for sjli instruction.
2723 (call_value_i): Likewise.
2724 * config/arc/constraints.md (Csc): New constraint.
2726 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2727 John Eric Martin <John.Martin@emmicro-us.com>
2729 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
2730 * config/arc/arc.c (_arc_jli_section): New struct.
2731 (arc_jli_section): New type.
2732 (rc_jli_sections): New static variable.
2733 (arc_handle_jli_attribute): New function.
2734 (arc_attribute_table): Add jli_always and jli_fixed attribute.
2735 (arc_file_end): New function.
2736 (TARGET_ASM_FILE_END): Define.
2737 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
2738 (arc_add_jli_section): New function.
2739 (jli_call_scan): Likewise.
2740 (arc_reorg): Call jli_call_scan.
2741 (arc_output_addsi): Remove 'S' from printing asm operand.
2742 (arc_is_jli_call_p): New function.
2743 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
2745 (movhi_insn): Likewise.
2746 (movsi_insn): Likewise.
2747 (movsi_set_cc_insn): Likewise.
2748 (loadqi_update): Likewise.
2749 (load_zeroextendqisi_update): Likewise.
2750 (load_signextendqisi_update): Likewise.
2751 (loadhi_update): Likewise.
2752 (load_zeroextendhisi_update): Likewise.
2753 (load_signextendhisi_update): Likewise.
2754 (loadsi_update): Likewise.
2755 (loadsf_update): Likewise.
2756 (movsicc_insn): Likewise.
2757 (bset_insn): Likewise.
2758 (bxor_insn): Likewise.
2759 (bclr_insn): Likewise.
2760 (bmsk_insn): Likewise.
2761 (bicsi3_insn): Likewise.
2762 (cmpsi_cc_c_insn): Likewise.
2763 (movsi_ne): Likewise.
2764 (movsi_cond_exec): Likewise.
2765 (clrsbsi2): Likewise.
2776 (call_i): Remove 'S' asm letter, add jli instruction.
2777 (call_value_i): Likewise.
2778 * config/arc/arc.op (mjli-always): New option.
2779 * config/arc/constraints.md (Cji): New constraint.
2780 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
2782 (subsf3_fpx): Likewise.
2783 (mulsf3_fpx): Likewise.
2784 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
2786 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
2788 * doc/invoke.texi (ARC): Document mjli-always option.
2790 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
2792 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
2793 avoid addition with 0 and use incw and decw where possible.
2795 2018-01-26 Richard Biener <rguenther@suse.de>
2797 PR tree-optimization/81082
2798 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
2799 association if it requires casting to unsigned.
2800 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
2801 from fold_plusminus_mult_expr to catch important cases late when
2802 range info is available.
2804 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2806 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
2807 * configure.ac (hidden_linkonce): New test.
2808 * configure: Regenerate.
2809 * config.in: Regenerate.
2811 2018-01-26 Julia Koval <julia.koval@intel.com>
2813 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
2814 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
2815 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
2816 _mm_mask_bitshuffle_epi64_mask): Fix type.
2817 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
2818 USI_FTYPE_V4DI_V4DI_USI): Remove.
2819 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
2820 __builtin_ia32_vpshufbitqmb256_mask,
2821 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
2822 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
2823 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
2825 2018-01-26 Alan Modra <amodra@gmail.com>
2828 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
2829 UNSPEC_VBPERMQ. Sort other unspecs.
2831 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
2833 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
2835 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
2838 * predict.c (drop_profile): Do not push/pop cfun; update also
2840 (handle_missing_profiles): Fix logic looking for zero profiles.
2842 2018-01-25 Jakub Jelinek <jakub@redhat.com>
2845 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
2846 on functions with #pragma omp declare simd or functions with simd
2848 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
2849 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
2850 Remove trailing \n from warning_at calls.
2852 2018-01-25 Tom de Vries <tom@codesourcery.com>
2855 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2856 for neutered workers.
2858 2018-01-24 Joseph Myers <joseph@codesourcery.com>
2861 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
2862 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
2864 2018-01-24 Jeff Law <law@redhat.com>
2867 * i386.c (get_probe_interval): Move to earlier point.
2868 (ix86_compute_frame_layout): If -fstack-clash-protection and
2869 the frame is larger than the probe interval, then use pushes
2870 to save registers rather than reg->mem moves.
2871 (ix86_expand_prologue): Remove conditional for int_registers_saved
2874 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
2877 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
2878 min/max for never referenced object.
2880 2018-01-24 Jakub Jelinek <jakub@redhat.com>
2883 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
2885 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
2886 attributes from DECL_ATTRIBUTES (decl) without affecting
2887 DECL_ATTRIBUTES (current_function_decl).
2888 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
2889 functions with non-NULL DECL_ABSTRACT_ORIGIN.
2891 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
2893 PR tree-optimization/83979
2894 * fold-const.c (fold_comparison): Use constant_boolean_node
2895 instead of boolean_{true,false}_node.
2897 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
2899 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
2902 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2904 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
2905 Simplify the clause that sets the length attribute.
2906 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
2907 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
2908 clause that sets the length attribute.
2909 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
2911 2018-01-24 Tom de Vries <tom@codesourcery.com>
2914 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
2915 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
2916 Add strict parameter.
2917 (prevent_branch_around_nothing): Insert dummy insn between branch to
2918 label and label with no ptx insn inbetween.
2919 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
2921 2018-01-24 Tom de Vries <tom@codesourcery.com>
2924 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2925 for neutered threads in warp.
2926 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
2928 2018-01-24 Richard Biener <rguenther@suse.de>
2930 PR tree-optimization/83176
2931 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
2934 2018-01-24 Richard Biener <rguenther@suse.de>
2936 PR tree-optimization/82819
2937 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
2938 code generating pluses that are no-ops in the target precision.
2940 2018-01-24 Richard Biener <rguenther@suse.de>
2943 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
2945 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2947 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
2948 to merge probabilities.
2949 * predict.c (probably_never_executed): Also mark as cold functions
2950 with global 0 profile and guessed local profile.
2951 * profile-count.c (profile_probability::combine_with_count): New
2953 * profile-count.h (profile_probability::operator*,
2954 profile_probability::operator*=, profile_probability::operator/,
2955 profile_probability::operator/=): Reduce precision to adjusted
2956 and set value to guessed on contradictory divisions.
2957 (profile_probability::combine_with_freq): Remove.
2958 (profile_probability::combine_wiht_count): Declare.
2959 (profile_count::force_nonzero):: Set to adjusted.
2960 (profile_count::probability_in):: Set quality to adjusted.
2961 * tree-ssa-tail-merge.c (replace_block_by): Use
2964 2018-01-23 Andrew Waterman <andrew@sifive.com>
2965 Jim Wilson <jimw@sifive.com>
2967 * config/riscv/riscv.c (riscv_stack_boundary): New.
2968 (riscv_option_override): Set riscv_stack_boundary. Handle
2969 riscv_preferred_stack_boundary_arg.
2970 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
2971 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
2972 (STACK_BOUNDARY): Set to riscv_stack_boundary.
2973 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
2974 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
2975 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
2977 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
2980 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
2981 of struct ix86_frame.
2982 (ix86_expand_epilogue): Likewise. Add a local variable for
2983 the reg_save_offset field in struct ix86_frame.
2985 2018-01-23 Bin Cheng <bin.cheng@arm.com>
2987 PR tree-optimization/82604
2988 * tree-loop-distribution.c (enum partition_kind): New enum item
2989 PKIND_PARTIAL_MEMSET.
2990 (partition_builtin_p): Support above new enum item.
2991 (generate_code_for_partition): Ditto.
2992 (compute_access_range): Differentiate cases that equality can be
2993 proven at all loops, the innermost loops or no loops.
2994 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
2995 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
2996 (finalize_partitions, distribute_loop): Don't fuse partition of
2997 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
2998 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
3001 2018-01-23 Martin Liska <mliska@suse.cz>
3003 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
3004 order to ignore the predictor.
3005 (PRED_POLYMORPHIC_CALL): Likewise.
3006 (PRED_RECURSIVE_CALL): Likewise.
3008 2018-01-23 Martin Liska <mliska@suse.cz>
3010 * tree-profile.c (tree_profiling): Print function header to
3011 aware reader which function we are working on.
3012 * value-prof.c (gimple_find_values_to_profile): Do not print
3013 not interesting value histograms.
3015 2018-01-23 Martin Liska <mliska@suse.cz>
3017 * profile-count.h (enum profile_quality): Add
3018 profile_uninitialized as the first value. Do not number values
3019 as they are zero based.
3020 (profile_count::verify): Update sanity check.
3021 (profile_probability::verify): Likewise.
3023 2018-01-23 Nathan Sidwell <nathan@acm.org>
3025 * doc/invoke.texi (ffor-scope): Deprecate.
3027 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3029 PR tree-optimization/83510
3030 * domwalk.c (set_all_edges_as_executable): New function.
3031 (dom_walker::dom_walker): Convert bool param
3032 "skip_unreachable_blocks" to enum reachability. Move setup of
3033 edge flags to set_all_edges_as_executable and only do it when
3034 reachability is REACHABLE_BLOCKS.
3035 * domwalk.h (enum dom_walker::reachability): New enum.
3036 (dom_walker::dom_walker): Convert bool param
3037 "skip_unreachable_blocks" to enum reachability.
3038 (set_all_edges_as_executable): New decl.
3039 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
3040 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
3042 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
3043 but converting true to REACHABLE_BLOCKS.
3044 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
3046 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
3047 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
3048 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
3050 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
3051 if check_all_array_refs will be called.
3053 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3055 * tree.c (selftest::test_location_wrappers): Add more test
3058 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3060 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
3061 (selftest::test_bit_in_range): Likewise.
3063 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3066 * doc/sourcebuild.texi (vect_float): Say that the selector
3067 only describes the situation when -funsafe-math-optimizations is on.
3068 (vect_float_strict): Document.
3070 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3072 PR tree-optimization/83965
3073 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3074 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
3075 instead of checking only for a reduction.
3076 (vect_recog_widen_sum_pattern): Likewise.
3078 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3080 * predict.c (probably_never_executed): Only use precise profile info.
3081 (compute_function_frequency): Skip after inlining hack since we now
3082 have quality checking.
3084 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3086 * profile-count.h (profile_probability::very_unlikely,
3087 profile_probability::unlikely, profile_probability::even): Set
3088 precision to guessed.
3090 2018-01-23 Richard Biener <rguenther@suse.de>
3092 PR tree-optimization/83963
3093 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
3094 Properly terminate dominator walk when crossing the exit edge not
3095 when visiting its source block.
3097 2018-01-23 Jakub Jelinek <jakub@redhat.com>
3100 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
3101 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
3103 2018-01-22 Jakub Jelinek <jakub@redhat.com>
3105 PR tree-optimization/83957
3106 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
3107 semicolon after for body surrounded by braces.
3109 PR tree-optimization/83081
3110 * profile-count.h (profile_probability::split): New method.
3111 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
3112 Use profile_probability::split.
3113 (do_compare_rtx_and_jump): Fix adjustment of probabilities
3114 when splitting a single conditional jump into 2.
3116 2018-01-22 David Malcolm <dmalcolm@redhat.com>
3118 PR tree-optimization/69452
3119 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
3122 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3124 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
3125 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
3126 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
3128 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3130 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
3131 * config/rl78/rl78.md (movdi): New define_expand.
3132 * config/rl78/rl78.c (rl78_split_movdi): New function.
3134 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
3137 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
3139 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
3140 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
3141 128-bit to produce an UNSPEC move to get the double word with the
3142 signbit and then a shift directly to do signbit.
3143 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
3144 implementation with a new version that just does either a direct
3145 move or a regular move. Move memory interface to separate insns.
3146 Move insns so they are next to the expander.
3147 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
3148 with signbit move. Split big and little endian case.
3149 (signbit<mode>2_dm_mem_le): Likewise.
3150 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
3151 (signbit<mode>2_dm2): Likewise.
3153 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3155 * config/rl78/rl78.md (anddi3): New define_expand.
3157 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3159 * config/rl78/rl78.md (umindi3): New define_expand.
3161 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3163 * config/rl78/rl78.md (smindi3): New define_expand.
3165 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3167 * config/rl78/rl78.md (smaxdi3): New define_expand.
3169 2018-01-22 Carl Love <cel@us.ibm.com>
3171 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
3172 LVX_V1TI): Add macro expansion.
3173 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
3174 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
3175 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
3176 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
3177 Change check to determine if the instruction is a byte reversing
3178 entry. Fix typo in comment.
3179 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
3180 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
3181 Add def_builtin calls for new builtins.
3182 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
3183 Add define_insn expansion.
3185 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3187 * config/rl78/rl78.md (umaxdi3): New define_expand.
3189 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3191 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
3192 for non-QImode registers.
3194 2018-01-22 Richard Biener <rguenther@suse.de>
3196 PR tree-optimization/83963
3197 * graphite-scop-detection.c (scop_detection::get_sese): Delay
3198 including the loop exit block.
3199 (scop_detection::merge_sese): Likewise.
3200 (scop_detection::add_scop): Do it here instead.
3202 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3204 * doc/sourcebuild.texi (arm_softfloat): Document.
3206 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
3209 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
3210 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
3211 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
3213 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3214 David Edelsohn <dje.gcc@gmail.com>
3217 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3218 Change "crset eq" to "crset 2".
3219 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3220 (*call_indirect_aix<mode>_nospec): Likewise.
3221 (*call_value_indirect_aix<mode>_nospec): Likewise.
3222 (*call_indirect_elfv2<mode>_nospec): Likewise.
3223 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
3224 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
3225 change assembly output from . to $.
3226 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3227 (indirect_jump<mode>_nospec): Change assembly output from . to $.
3228 (*tablejump<mode>_internal1_nospec): Likewise.
3230 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
3233 * config/sh/sh_optimize_sett_clrt.cc:
3234 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
3236 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
3238 PR tree-optimization/83940
3239 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
3240 offset_dt to vect_constant_def rather than vect_unknown_def_type.
3241 (vect_check_load_store_mask): Add a mask_dt_out parameter and
3242 use it to pass back the definition type.
3243 (vect_check_store_rhs): Likewise rhs_dt_out.
3244 (vect_build_gather_load_calls): Add a mask_dt argument and use
3245 it instead of a call to vect_is_simple_use.
3246 (vectorizable_store): Update calls to vect_check_load_store_mask
3247 and vect_check_store_rhs. Use the dt returned by the latter instead
3248 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
3249 instead of calls to vect_is_simple_use. Pass the scalar rather
3250 than the vector operand to vect_is_simple_use when handling
3251 second and subsequent copies of an rhs value.
3252 (vectorizable_load): Update calls to vect_check_load_store_mask
3253 and vect_build_gather_load_calls. Use the cached mask_dt and
3254 gs_info.offset_dt instead of calls to vect_is_simple_use.
3256 2018-01-20 Jakub Jelinek <jakub@redhat.com>
3259 * tree-emutls.c: Include gimplify.h.
3260 (lower_emutls_2): New function.
3261 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
3262 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
3263 it before further processing.
3266 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
3267 UINTVAL (trueop1) instead of INTVAL (op1).
3269 2018-01-19 Jakub Jelinek <jakub@redhat.com>
3273 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
3274 INCOMING_FRAME_SP_OFFSET if not defined.
3275 (scan_trace): Add ENTRY argument. If true and
3276 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
3277 emit a note to adjust the CFA offset.
3278 (create_cfi_notes): Adjust scan_trace callers.
3279 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
3280 INCOMING_FRAME_SP_OFFSET in the CIE.
3281 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
3282 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
3284 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
3285 * doc/tm.texi: Regenerated.
3287 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3289 PR rtl-optimization/83147
3290 * lra-constraints.c (remove_inheritance_pseudos): Use
3291 lra_substitute_pseudo_within_insn.
3293 2018-01-19 Tom de Vries <tom@codesourcery.com>
3294 Cesar Philippidis <cesar@codesourcery.com>
3297 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
3299 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
3302 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
3303 spaces for function labels.
3305 2018-01-19 Martin Liska <mliska@suse.cz>
3307 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
3308 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
3309 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
3310 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
3311 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
3312 (PRED_CONST_RETURN): Change from 69 to 65.
3313 (PRED_NULL_RETURN): Change from 91 to 71.
3314 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
3315 (PRED_LOOP_GUARD): Change from 66 to 73.
3317 2018-01-19 Martin Liska <mliska@suse.cz>
3319 * predict.c (predict_insn_def): Add new assert.
3320 (struct branch_predictor): Change type to signed integer.
3321 (test_prediction_value_range): Amend test to cover
3323 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
3324 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
3325 (PRED_LOOP_ITERATIONS_MAX): Likewise.
3326 (PRED_LOOP_IV_COMPARE): Likewise.
3327 * predict.h (PROB_UNINITIALIZED): Define new constant.
3329 2018-01-19 Martin Liska <mliska@suse.cz>
3331 * predict.c (dump_prediction): Add new format for
3332 analyze_brprob.py script which is enabled with -details
3334 * profile-count.h (precise_p): New function.
3336 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3338 PR tree-optimization/83922
3339 * tree-vect-loop.c (vect_verify_full_masking): Return false if
3340 there are no statements that need masking.
3341 (vect_active_double_reduction_p): New function.
3342 (vect_analyze_loop_operations): Use it when handling phis that
3343 are not in the loop header.
3345 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3347 PR tree-optimization/83914
3348 * tree-vect-loop.c (vectorizable_induction): Don't convert
3349 init_expr or apply the peeling adjustment for inductions
3350 that are nested within the vectorized loop.
3352 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3354 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
3357 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3361 * function.h (gimplify_parameters): Add gimple_seq * argument.
3362 * function.c: Include gimple.h and options.h.
3363 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
3364 for the added local temporaries if needed.
3365 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
3366 if there are any parameter cleanups, wrap whole body into a
3367 try/finally with the cleanups.
3369 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
3372 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
3373 Use GET_MODE_CLASS for scalar floating point.
3375 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3379 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3380 Fix call of call_cgraph_insertion_hooks.
3382 2018-01-18 Martin Sebor <msebor@redhat.com>
3384 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
3386 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3389 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
3392 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
3395 * common.opt: (-ffile-prefix-map): New option.
3396 * opts.c (common_handle_option): Defer it.
3397 * opts-global.c (handle_common_deferred_options): Handle it.
3398 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
3399 * file-prefix-map.h: New file.
3400 (remap_debug_filename, add_debug_prefix_map): ...here.
3401 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
3402 * final.c (debug_prefix_map, add_debug_prefix_map
3403 remap_debug_filename): Move to...
3404 * file-prefix-map.c: New file.
3405 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
3406 generalize, get rid of alloca(), use strrchr() instead of strchr().
3407 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
3408 Implement in terms of add_prefix_map().
3409 (remap_macro_filename, remap_debug_filename): Implement in term of
3411 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
3412 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
3413 * dbxout.c: Include file-prefix-map.h.
3414 * varasm.c: Likewise.
3415 * vmsdbgout.c: Likewise.
3416 * xcoffout.c: Likewise.
3417 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
3418 * doc/cppopts.texi (-fmacro-prefix-map): Document.
3419 * doc/invoke.texi (-ffile-prefix-map): Document.
3420 (-fdebug-prefix-map): Update description.
3422 2018-01-18 Martin Liska <mliska@suse.cz>
3424 * config/i386/i386.c (indirect_thunk_name): Document that also
3426 (output_indirect_thunk): Document why both instructions
3427 (pause and lfence) are generated.
3429 2018-01-18 Richard Biener <rguenther@suse.de>
3431 PR tree-optimization/83887
3432 * graphite-scop-detection.c
3433 (scop_detection::get_nearest_dom_with_single_entry): Remove.
3434 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
3435 (scop_detection::merge_sese): Re-implement with a flood-fill
3436 algorithm that properly finds a SESE region if it exists.
3438 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3441 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
3442 pointer_diff optimizations use view_convert instead of convert.
3444 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3446 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3447 Generate different code for -mno-speculate-indirect-jumps.
3448 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3449 (*call_indirect_aix<mode>): Disable for
3450 -mno-speculate-indirect-jumps.
3451 (*call_indirect_aix<mode>_nospec): New define_insn.
3452 (*call_value_indirect_aix<mode>): Disable for
3453 -mno-speculate-indirect-jumps.
3454 (*call_value_indirect_aix<mode>_nospec): New define_insn.
3455 (*sibcall_nonlocal_sysv<mode>): Generate different code for
3456 -mno-speculate-indirect-jumps.
3457 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3459 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3461 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
3462 long double type, set the flags for noting the default long double
3463 type, even if we don't pass or return a long double type.
3465 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
3468 * ipa-inline.c (flatten_function): Do not overwrite final inlining
3471 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
3473 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
3474 support for merge[hl].
3475 (fold_mergehl_helper): New helper function.
3476 (tree-vector-builder.h): New #include for tree_vector_builder usage.
3477 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
3478 (altivec_vmrglw_direct): Add xxmrglw insn.
3480 2018-01-17 Andrew Waterman <andrew@sifive.com>
3482 * config/riscv/riscv.c (riscv_conditional_register_usage): If
3483 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
3485 2018-01-17 David Malcolm <dmalcolm@redhat.com>
3488 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
3489 call the lto_location_cache before reading the
3490 DECL_SOURCE_LOCATION of the types.
3492 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3493 Richard Sandiford <richard.sandiford@linaro.org>
3495 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
3496 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
3497 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
3498 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
3499 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
3501 * config/aarch64/constraints.md (aarch64_movti_operand):
3503 * config/aarch64/predicates.md (Uti): Add new constraint.
3505 2018-01-17 Carl Love <cel@us.ibm.com>
3507 * config/rs6000/vsx.md (define_expand xl_len_r,
3508 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
3509 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
3511 (define_expand, define_insn): Move the shift left from the
3512 define_insn to the define_expand for lxvl and stxvl instructions.
3513 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
3514 and XL_LEN_R definitions to PURE.
3516 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
3518 * config/i386/i386.c (indirect_thunk_name): Declare regno
3519 as unsigned int. Compare regno with INVALID_REGNUM.
3520 (output_indirect_thunk): Ditto.
3521 (output_indirect_thunk_function): Ditto.
3522 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
3523 in the call to output_indirect_thunk_function.
3525 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
3528 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
3529 rather than the size of inner_type to determine the stack slot size
3530 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
3532 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
3535 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
3538 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
3540 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
3541 endian Linux systems to optionally enable multilibs for selecting
3542 the long double type if the user configured an explicit type.
3543 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
3544 have no long double multilibs if not defined.
3545 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
3546 warn if the user used -mabi={ieee,ibm}longdouble and we built
3547 multilibs for long double.
3548 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
3549 appropriate multilib option.
3550 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
3552 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
3553 for building long double multilibs.
3554 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
3556 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
3558 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
3561 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
3563 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
3566 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
3569 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
3572 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3574 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
3575 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
3577 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3579 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
3580 different rtl trees depending on TARGET_64BIT.
3581 (rs6000_gen_lvx): Likewise.
3583 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3585 * config/visium/visium.md (nop): Tweak comment.
3586 (hazard_nop): Likewise.
3588 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3590 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
3591 -mspeculate-indirect-jumps.
3592 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
3593 for -mno-speculate-indirect-jumps.
3594 (*call_indirect_elfv2<mode>_nospec): New define_insn.
3595 (*call_value_indirect_elfv2<mode>): Disable for
3596 -mno-speculate-indirect-jumps.
3597 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
3598 (indirect_jump): Emit different RTL for
3599 -mno-speculate-indirect-jumps.
3600 (*indirect_jump<mode>): Disable for
3601 -mno-speculate-indirect-jumps.
3602 (*indirect_jump<mode>_nospec): New define_insn.
3603 (tablejump): Emit different RTL for
3604 -mno-speculate-indirect-jumps.
3605 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
3606 (tablejumpsi_nospec): New define_expand.
3607 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
3608 (tablejumpdi_nospec): New define_expand.
3609 (*tablejump<mode>_internal1): Disable for
3610 -mno-speculate-indirect-jumps.
3611 (*tablejump<mode>_internal1_nospec): New define_insn.
3612 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
3615 2018-01-16 Artyom Skrobov tyomitch@gmail.com
3617 * caller-save.c (insert_save): Drop unnecessary parameter. All
3620 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3621 Richard Biener <rguenth@suse.de>
3624 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
3625 return early, inline manually is_gimple_sizepos. Make sure if we
3626 call gimplify_expr we don't end up with a gimple constant.
3627 * tree.c (variably_modified_type_p): Don't return true for
3628 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
3629 * gimplify.h (is_gimple_sizepos): Remove.
3631 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3633 PR tree-optimization/83857
3634 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
3635 vectorizable_live_operation for pure SLP statements.
3636 (vectorizable_live_operation): Handle PHIs.
3638 2018-01-16 Richard Biener <rguenther@suse.de>
3640 PR tree-optimization/83867
3641 * tree-vect-stmts.c (vect_transform_stmt): Precompute
3642 nested_in_vect_loop_p since the scalar stmt may get invalidated.
3644 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3647 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
3648 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
3649 If off is not INTEGER_CST, issue a may not be aligned warning
3650 rather than isn't aligned. Use isn%'t rather than isn't.
3651 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
3653 <case MULT_EXPR>: Improve the case when bottom and one of the
3654 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
3655 operand, in that case check if the other operand is multiple of
3656 bottom divided by the INTEGER_CST operand.
3658 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3661 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
3662 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
3663 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
3664 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
3665 * config/pa/pa.c (pa_function_arg_advance): Likewise.
3666 (pa_function_arg, pa_arg_partial_bytes): Likewise.
3667 (pa_function_arg_size): New function.
3669 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3671 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
3672 in a separate statement.
3674 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3676 PR tree-optimization/83847
3677 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
3678 group gathers and scatters.
3680 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3682 PR rtl-optimization/86620
3683 * params.def (max-sched-ready-insns): Bump minimum value to 1.
3685 PR rtl-optimization/83213
3686 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
3687 to last if both are JUMP_INSNs.
3689 PR tree-optimization/83843
3690 * gimple-ssa-store-merging.c
3691 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
3692 store_immediate_info for bswap/nop orig_stores.
3694 2018-01-15 Andrew Waterman <andrew@sifive.com>
3696 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
3698 <UDIV>: Increase cost if !TARGET_DIV.
3700 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
3702 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
3703 (define_attr "cr_logical_3op"): New.
3704 (cceq_ior_compare): Adjust.
3705 (cceq_ior_compare_complement): Adjust.
3706 (*cceq_rev_compare): Adjust.
3707 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
3708 (is_cracked_insn): Adjust.
3709 (insn_must_be_first_in_group): Adjust.
3710 * config/rs6000/40x.md: Adjust.
3711 * config/rs6000/440.md: Adjust.
3712 * config/rs6000/476.md: Adjust.
3713 * config/rs6000/601.md: Adjust.
3714 * config/rs6000/603.md: Adjust.
3715 * config/rs6000/6xx.md: Adjust.
3716 * config/rs6000/7450.md: Adjust.
3717 * config/rs6000/7xx.md: Adjust.
3718 * config/rs6000/8540.md: Adjust.
3719 * config/rs6000/cell.md: Adjust.
3720 * config/rs6000/e300c2c3.md: Adjust.
3721 * config/rs6000/e500mc.md: Adjust.
3722 * config/rs6000/e500mc64.md: Adjust.
3723 * config/rs6000/e5500.md: Adjust.
3724 * config/rs6000/e6500.md: Adjust.
3725 * config/rs6000/mpc.md: Adjust.
3726 * config/rs6000/power4.md: Adjust.
3727 * config/rs6000/power5.md: Adjust.
3728 * config/rs6000/power6.md: Adjust.
3729 * config/rs6000/power7.md: Adjust.
3730 * config/rs6000/power8.md: Adjust.
3731 * config/rs6000/power9.md: Adjust.
3732 * config/rs6000/rs64.md: Adjust.
3733 * config/rs6000/titan.md: Adjust.
3735 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3737 * config/i386/predicates.md (indirect_branch_operand): Rewrite
3738 ix86_indirect_branch_register logic.
3740 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3742 * config/i386/constraints.md (Bs): Update
3743 ix86_indirect_branch_register check. Don't check
3744 ix86_indirect_branch_register with GOT_memory_operand.
3746 * config/i386/predicates.md (GOT_memory_operand): Don't check
3747 ix86_indirect_branch_register here.
3748 (GOT32_symbol_operand): Likewise.
3750 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3752 * config/i386/predicates.md (constant_call_address_operand):
3753 Rewrite ix86_indirect_branch_register logic.
3754 (sibcall_insn_operand): Likewise.
3756 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3758 * config/i386/constraints.md (Bs): Replace
3759 ix86_indirect_branch_thunk_register with
3760 ix86_indirect_branch_register.
3762 * config/i386/i386.md (indirect_jump): Likewise.
3763 (tablejump): Likewise.
3764 (*sibcall_memory): Likewise.
3765 (*sibcall_value_memory): Likewise.
3766 Peepholes of indirect call and jump via memory: Likewise.
3767 * config/i386/i386.opt: Likewise.
3768 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3769 (GOT_memory_operand): Likewise.
3770 (call_insn_operand): Likewise.
3771 (sibcall_insn_operand): Likewise.
3772 (GOT32_symbol_operand): Likewise.
3774 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3777 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
3778 type rather than type addr's type points to.
3779 (expand_omp_atomic_mutex): Likewise.
3780 (expand_omp_atomic): Likewise.
3782 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3785 * config/i386/i386.c (output_indirect_thunk_function): Use
3786 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
3787 for __x86_return_thunk.
3789 2018-01-15 Richard Biener <rguenther@suse.de>
3792 * expmed.c (extract_bit_field_1): Fix typo.
3794 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3797 * config/arm/iterators.md (VF): New mode iterator.
3798 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
3799 Remove integer-related logic from pattern.
3800 (neon_vabd<mode>_3): Likewise.
3802 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3805 * common.opt (fstrict-overflow): No longer an alias.
3806 (fwrapv-pointer): New option.
3807 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
3808 also for pointer types based on flag_wrapv_pointer.
3809 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
3810 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
3811 opts->x_flag_wrapv got set.
3812 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
3813 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
3814 POINTER_TYPE_OVERFLOW_UNDEFINED.
3815 * match.pd: Likewise in address comparison pattern.
3816 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
3818 2018-01-15 Richard Biener <rguenther@suse.de>
3821 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
3822 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
3823 Reset type names to their identifier if their TYPE_DECL doesn't
3824 have linkage (and thus is used for ODR and devirt).
3825 (save_debug_info_for_decl): Remove.
3826 (save_debug_info_for_type): Likewise.
3827 (add_tree_to_fld_list): Adjust.
3828 * tree-pretty-print.c (dump_generic_node): Make dumping of
3829 type names more robust.
3831 2018-01-15 Richard Biener <rguenther@suse.de>
3833 * BASE-VER: Bump to 8.0.1.
3835 2018-01-14 Martin Sebor <msebor@redhat.com>
3838 * builtins.c (check_access): Avoid warning when the no-warning bit
3841 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
3843 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
3844 * ira-color (allocno_hard_regs_compare): Likewise.
3846 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
3849 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
3850 Use .pushsection/.popsection.
3852 2018-01-14 Martin Sebor <msebor@redhat.com>
3855 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
3857 2018-01-14 Jakub Jelinek <jakub@redhat.com>
3859 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
3860 entry from extra_headers.
3861 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
3862 extra_headers, make the list bitwise identical to the i?86-*-* one.
3864 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3866 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
3867 -mcmodel=large with -mindirect-branch=thunk,
3868 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
3869 -mfunction-return=thunk-extern.
3870 * doc/invoke.texi: Document -mcmodel=large is incompatible with
3871 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
3872 -mfunction-return=thunk and -mfunction-return=thunk-extern.
3874 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3876 * config/i386/i386.c (print_reg): Print the name of the full
3877 integer register without '%'.
3878 (ix86_print_operand): Handle 'V'.
3879 * doc/extend.texi: Document 'V' modifier.
3881 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3883 * config/i386/constraints.md (Bs): Disallow memory operand for
3884 -mindirect-branch-register.
3886 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3887 (GOT_memory_operand): Likewise.
3888 (call_insn_operand): Likewise.
3889 (sibcall_insn_operand): Likewise.
3890 (GOT32_symbol_operand): Likewise.
3891 * config/i386/i386.md (indirect_jump): Call convert_memory_address
3892 for -mindirect-branch-register.
3893 (tablejump): Likewise.
3894 (*sibcall_memory): Likewise.
3895 (*sibcall_value_memory): Likewise.
3896 Disallow peepholes of indirect call and jump via memory for
3897 -mindirect-branch-register.
3898 (*call_pop): Replace m with Bw.
3899 (*call_value_pop): Likewise.
3900 (*sibcall_pop_memory): Replace m with Bs.
3901 * config/i386/i386.opt (mindirect-branch-register): New option.
3902 * doc/invoke.texi: Document -mindirect-branch-register option.
3904 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3906 * config/i386/i386-protos.h (ix86_output_function_return): New.
3907 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
3908 set function_return_type.
3909 (indirect_thunk_name): Add ret_p to indicate thunk for function
3911 (output_indirect_thunk_function): Pass false to
3912 indirect_thunk_name.
3913 (ix86_output_indirect_branch_via_reg): Likewise.
3914 (ix86_output_indirect_branch_via_push): Likewise.
3915 (output_indirect_thunk_function): Create alias for function
3916 return thunk if regno < 0.
3917 (ix86_output_function_return): New function.
3918 (ix86_handle_fndecl_attribute): Handle function_return.
3919 (ix86_attribute_table): Add function_return.
3920 * config/i386/i386.h (machine_function): Add
3921 function_return_type.
3922 * config/i386/i386.md (simple_return_internal): Use
3923 ix86_output_function_return.
3924 (simple_return_internal_long): Likewise.
3925 * config/i386/i386.opt (mfunction-return=): New option.
3926 (indirect_branch): Mention -mfunction-return=.
3927 * doc/extend.texi: Document function_return function attribute.
3928 * doc/invoke.texi: Document -mfunction-return= option.
3930 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3932 * config/i386/i386-opts.h (indirect_branch): New.
3933 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
3934 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
3935 with local indirect jump when converting indirect call and jump.
3936 (ix86_set_indirect_branch_type): New.
3937 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
3938 (indirectlabelno): New.
3939 (indirect_thunk_needed): Likewise.
3940 (indirect_thunk_bnd_needed): Likewise.
3941 (indirect_thunks_used): Likewise.
3942 (indirect_thunks_bnd_used): Likewise.
3943 (INDIRECT_LABEL): Likewise.
3944 (indirect_thunk_name): Likewise.
3945 (output_indirect_thunk): Likewise.
3946 (output_indirect_thunk_function): Likewise.
3947 (ix86_output_indirect_branch_via_reg): Likewise.
3948 (ix86_output_indirect_branch_via_push): Likewise.
3949 (ix86_output_indirect_branch): Likewise.
3950 (ix86_output_indirect_jmp): Likewise.
3951 (ix86_code_end): Call output_indirect_thunk_function if needed.
3952 (ix86_output_call_insn): Call ix86_output_indirect_branch if
3954 (ix86_handle_fndecl_attribute): Handle indirect_branch.
3955 (ix86_attribute_table): Add indirect_branch.
3956 * config/i386/i386.h (machine_function): Add indirect_branch_type
3957 and has_local_indirect_jump.
3958 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
3960 (tablejump): Likewise.
3961 (*indirect_jump): Use ix86_output_indirect_jmp.
3962 (*tablejump_1): Likewise.
3963 (simple_return_indirect_internal): Likewise.
3964 * config/i386/i386.opt (mindirect-branch=): New option.
3965 (indirect_branch): New.
3968 (thunk-inline): Likewise.
3969 (thunk-extern): Likewise.
3970 * doc/extend.texi: Document indirect_branch function attribute.
3971 * doc/invoke.texi: Document -mindirect-branch= option.
3973 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
3976 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
3978 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
3980 * ipa-inline.c (want_inline_small_function_p): Return false if
3981 inlining has already failed with CIF_FINAL_ERROR.
3982 (update_caller_keys): Call want_inline_small_function_p before
3984 (update_callee_keys): Likewise.
3986 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3988 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
3990 (rs6000_quadword_masked_address_p): Likewise.
3991 (quad_aligned_load_p): Likewise.
3992 (quad_aligned_store_p): Likewise.
3993 (const_load_sequence_p): Add comment to describe the outer-most loop.
3994 (mimic_memory_attributes_and_flags): New function.
3995 (rs6000_gen_stvx): Likewise.
3996 (replace_swapped_aligned_store): Likewise.
3997 (rs6000_gen_lvx): Likewise.
3998 (replace_swapped_aligned_load): Likewise.
3999 (replace_swapped_load_constant): Capitalize argument name in
4000 comment describing this function.
4001 (rs6000_analyze_swaps): Add a third pass to search for vector loads
4002 and stores that access quad-word aligned addresses and replace
4003 with stvx or lvx instructions when appropriate.
4004 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
4005 New function prototype.
4006 (rs6000_quadword_masked_address_p): Likewise.
4007 (rs6000_gen_lvx): Likewise.
4008 (rs6000_gen_stvx): Likewise.
4009 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
4010 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
4011 when memory address is aligned.
4012 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
4013 this split to select lvx instruction when memory address is aligned.
4014 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
4015 instruction when memory address is aligned.
4016 (*vsx_le_perm_load_v16qi): Likewise.
4017 (four unnamed splitters): Modify to select the stvx instruction
4018 when memory is aligned.
4020 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
4022 * predict.c (determine_unlikely_bbs): Handle correctly BBs
4023 which appears in the queue multiple times.
4025 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4026 Alan Hayward <alan.hayward@arm.com>
4027 David Sherwood <david.sherwood@arm.com>
4029 * tree-vectorizer.h (vec_lower_bound): New structure.
4030 (_loop_vec_info): Add check_nonzero and lower_bounds.
4031 (LOOP_VINFO_CHECK_NONZERO): New macro.
4032 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
4033 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
4034 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
4035 fields. Make seg_len the distance travelled, not including the
4037 (dr_direction_indicator): Declare.
4038 (dr_zero_step_indicator): Likewise.
4039 (dr_known_forward_stride_p): Likewise.
4040 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
4042 (runtime_alias_check_p): Allow runtime alias checks with
4044 (operator ==): Compare access_size and align.
4045 (prune_runtime_alias_test_list): Rework for new distinction between
4046 the access_size and seg_len.
4047 (create_intersect_range_checks_index): Likewise. Cope with polynomial
4049 (get_segment_min_max): New function.
4050 (create_intersect_range_checks): Use it.
4051 (dr_step_indicator): New function.
4052 (dr_direction_indicator): Likewise.
4053 (dr_zero_step_indicator): Likewise.
4054 (dr_known_forward_stride_p): Likewise.
4055 * tree-loop-distribution.c (data_ref_segment_size): Return
4056 DR_STEP * (niters - 1).
4057 (compute_alias_check_pairs): Update call to the dr_with_seg_len
4059 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
4060 (vect_preserves_scalar_order_p): New function, split out from...
4061 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
4062 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
4063 (vect_vfa_access_size): New function.
4064 (vect_vfa_align): Likewise.
4065 (vect_compile_time_alias): Take access_size_a and access_b arguments.
4066 (dump_lower_bound): New function.
4067 (vect_check_lower_bound): Likewise.
4068 (vect_small_gap_p): Likewise.
4069 (vectorizable_with_step_bound_p): Likewise.
4070 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
4071 depencies if the vectorization factor is 1. Convert the checks
4072 for nonzero steps into checks on the bounds of DR_STEP. Try using
4073 a bunds check for variable steps if the minimum required step is
4074 relatively small. Update calls to the dr_with_seg_len
4075 constructor and to vect_compile_time_alias.
4076 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
4078 (vect_loop_versioning): Call it.
4079 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
4081 (vect_estimate_min_profitable_iters): Account for any bounds checks.
4083 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4084 Alan Hayward <alan.hayward@arm.com>
4085 David Sherwood <david.sherwood@arm.com>
4087 * doc/sourcebuild.texi (vect_scatter_store): Document.
4088 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
4090 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
4092 * genopinit.c (main): Add supports_vec_scatter_store and
4093 supports_vec_scatter_store_cached to target_optabs.
4094 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
4095 IFN_MASK_SCATTER_STORE.
4096 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
4098 * internal-fn.h (internal_store_fn_p): Declare.
4099 (internal_fn_stored_value_index): Likewise.
4100 * internal-fn.c (scatter_store_direct): New macro.
4101 (expand_scatter_store_optab_fn): New function.
4102 (direct_scatter_store_optab_supported_p): New macro.
4103 (internal_store_fn_p): New function.
4104 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
4105 IFN_MASK_SCATTER_STORE.
4106 (internal_fn_mask_index): Likewise.
4107 (internal_fn_stored_value_index): New function.
4108 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
4110 * optabs-query.h (supports_vec_scatter_store_p): Declare.
4111 * optabs-query.c (supports_vec_scatter_store_p): New function.
4112 * tree-vectorizer.h (vect_get_store_rhs): Declare.
4113 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
4114 true for scatter stores.
4115 (vect_gather_scatter_fn_p): Handle scatter stores too.
4116 (vect_check_gather_scatter): Consider using scatter stores if
4117 supports_vec_scatter_store_p.
4118 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
4120 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4121 internal_fn_stored_value_index.
4122 (check_load_store_masking): Handle scatter stores too.
4123 (vect_get_store_rhs): Make public.
4124 (vectorizable_call): Use internal_store_fn_p.
4125 (vectorizable_store): Handle scatter store internal functions.
4126 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
4127 when deciding whether the end of the group has been reached.
4128 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
4129 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
4130 (mask_scatter_store<mode>): New insns.
4132 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4133 Alan Hayward <alan.hayward@arm.com>
4134 David Sherwood <david.sherwood@arm.com>
4136 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
4137 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
4138 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
4140 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
4141 Use vect_truncate_gather_scatter_offset if we can't treat the
4142 operation as a normal gather load or scatter store.
4143 (get_group_load_store_type): Take the gather_scatter_info
4144 as argument. Try using a gather load or scatter store for
4145 single-element groups.
4146 (get_load_store_type): Update calls to get_group_load_store_type
4147 and vect_use_strided_gather_scatters_p.
4149 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4150 Alan Hayward <alan.hayward@arm.com>
4151 David Sherwood <david.sherwood@arm.com>
4153 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
4154 optional tree argument.
4155 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
4157 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
4158 but continue to use the current value as a fallback.
4159 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
4160 to compare the updates.
4161 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
4162 (get_load_store_type): Use it when handling a strided access.
4163 (vect_get_strided_load_store_ops): New function.
4164 (vect_get_data_ptr_increment): Likewise.
4165 (vectorizable_load): Handle strided gather loads. Always pass
4166 a step to vect_create_data_ref_ptr and bump_vector_ptr.
4168 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4169 Alan Hayward <alan.hayward@arm.com>
4170 David Sherwood <david.sherwood@arm.com>
4172 * doc/md.texi (gather_load@var{m}): Document.
4173 (mask_gather_load@var{m}): Likewise.
4174 * genopinit.c (main): Add supports_vec_gather_load and
4175 supports_vec_gather_load_cached to target_optabs.
4176 * optabs-tree.c (init_tree_optimization_optabs): Use
4177 ggc_cleared_alloc to allocate target_optabs.
4178 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
4179 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
4181 * internal-fn.h (internal_load_fn_p): Declare.
4182 (internal_gather_scatter_fn_p): Likewise.
4183 (internal_fn_mask_index): Likewise.
4184 (internal_gather_scatter_fn_supported_p): Likewise.
4185 * internal-fn.c (gather_load_direct): New macro.
4186 (expand_gather_load_optab_fn): New function.
4187 (direct_gather_load_optab_supported_p): New macro.
4188 (direct_internal_fn_optab): New function.
4189 (internal_load_fn_p): Likewise.
4190 (internal_gather_scatter_fn_p): Likewise.
4191 (internal_fn_mask_index): Likewise.
4192 (internal_gather_scatter_fn_supported_p): Likewise.
4193 * optabs-query.c (supports_at_least_one_mode_p): New function.
4194 (supports_vec_gather_load_p): Likewise.
4195 * optabs-query.h (supports_vec_gather_load_p): Declare.
4196 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
4197 and memory_type field.
4198 (NUM_PATTERNS): Bump to 15.
4199 * tree-vect-data-refs.c: Include internal-fn.h.
4200 (vect_gather_scatter_fn_p): New function.
4201 (vect_describe_gather_scatter_call): Likewise.
4202 (vect_check_gather_scatter): Try using internal functions for
4203 gather loads. Recognize existing calls to a gather load function.
4204 (vect_analyze_data_refs): Consider using gather loads if
4205 supports_vec_gather_load_p.
4206 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
4207 (vect_get_gather_scatter_offset_type): Likewise.
4208 (vect_convert_mask_for_vectype): Likewise.
4209 (vect_add_conversion_to_patterm): Likewise.
4210 (vect_try_gather_scatter_pattern): Likewise.
4211 (vect_recog_gather_scatter_pattern): New pattern recognizer.
4212 (vect_vect_recog_func_ptrs): Add it.
4213 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4214 internal_fn_mask_index and internal_gather_scatter_fn_p.
4215 (check_load_store_masking): Take the gather_scatter_info as an
4216 argument and handle gather loads.
4217 (vect_get_gather_scatter_ops): New function.
4218 (vectorizable_call): Check internal_load_fn_p.
4219 (vectorizable_load): Likewise. Handle gather load internal
4221 (vectorizable_store): Update call to check_load_store_masking.
4222 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
4223 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
4224 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
4225 (aarch64_gather_scale_operand_d): New predicates.
4226 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
4227 (mask_gather_load<mode>): New insns.
4229 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4230 Alan Hayward <alan.hayward@arm.com>
4231 David Sherwood <david.sherwood@arm.com>
4233 * optabs.def (fold_left_plus_optab): New optab.
4234 * doc/md.texi (fold_left_plus_@var{m}): Document.
4235 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
4236 * internal-fn.c (fold_left_direct): Define.
4237 (expand_fold_left_optab_fn): Likewise.
4238 (direct_fold_left_optab_supported_p): Likewise.
4239 * fold-const-call.c (fold_const_fold_left): New function.
4240 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
4241 * tree-parloops.c (valid_reduction_p): New function.
4242 (gather_scalar_reductions): Use it.
4243 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
4244 (vect_finish_replace_stmt): Declare.
4245 * tree-vect-loop.c (fold_left_reduction_fn): New function.
4246 (needs_fold_left_reduction_p): New function, split out from...
4247 (vect_is_simple_reduction): ...here. Accept reductions that
4248 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
4249 (vect_force_simple_reduction): Also store the reduction type in
4250 the assignment's STMT_VINFO_REDUC_TYPE.
4251 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
4252 (merge_with_identity): New function.
4253 (vect_expand_fold_left): Likewise.
4254 (vectorize_fold_left_reduction): Likewise.
4255 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
4256 scalar phi in place for it. Check for target support and reject
4257 cases that would reassociate the operation. Defer the transform
4258 phase to vectorize_fold_left_reduction.
4259 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
4260 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
4261 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
4263 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4265 * tree-if-conv.c (predicate_mem_writes): Remove redundant
4266 call to ifc_temp_var.
4268 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4269 Alan Hayward <alan.hayward@arm.com>
4270 David Sherwood <david.sherwood@arm.com>
4272 * target.def (legitimize_address_displacement): Take the original
4273 offset as a poly_int.
4274 * targhooks.h (default_legitimize_address_displacement): Update
4276 * targhooks.c (default_legitimize_address_displacement): Likewise.
4277 * doc/tm.texi: Regenerate.
4278 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
4279 as an argument, moving assert of ad->disp == ad->disp_term to...
4280 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
4281 Try calling targetm.legitimize_address_displacement before expanding
4282 the address rather than afterwards, and adjust for the new interface.
4283 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
4284 Match the new hook interface. Handle SVE addresses.
4285 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
4288 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4290 * Makefile.in (OBJS): Add early-remat.o.
4291 * target.def (select_early_remat_modes): New hook.
4292 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
4293 * doc/tm.texi: Regenerate.
4294 * targhooks.h (default_select_early_remat_modes): Declare.
4295 * targhooks.c (default_select_early_remat_modes): New function.
4296 * timevar.def (TV_EARLY_REMAT): New timevar.
4297 * passes.def (pass_early_remat): New pass.
4298 * tree-pass.h (make_pass_early_remat): Declare.
4299 * early-remat.c: New file.
4300 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
4302 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
4304 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4305 Alan Hayward <alan.hayward@arm.com>
4306 David Sherwood <david.sherwood@arm.com>
4308 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
4309 vfm1 with a bound_epilog parameter.
4310 (vect_do_peeling): Update calls accordingly, and move the prologue
4311 call earlier in the function. Treat the base bound_epilog as 0 for
4312 fully-masked loops and retain vf - 1 for other loops. Add 1 to
4313 this base when peeling for gaps.
4314 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
4315 with fully-masked loops.
4316 (vect_estimate_min_profitable_iters): Handle the single peeled
4317 iteration in that case.
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-data-refs.c (vect_analyze_group_access_1): Allow
4324 single-element interleaving even if the size is not a power of 2.
4325 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
4326 accesses for single-element interleaving if the group size is
4329 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4330 Alan Hayward <alan.hayward@arm.com>
4331 David Sherwood <david.sherwood@arm.com>
4333 * doc/md.texi (fold_extract_last_@var{m}): Document.
4334 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
4335 * optabs.def (fold_extract_last_optab): New optab.
4336 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
4337 * internal-fn.c (fold_extract_direct): New macro.
4338 (expand_fold_extract_optab_fn): Likewise.
4339 (direct_fold_extract_optab_supported_p): Likewise.
4340 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
4341 * tree-vect-loop.c (vect_model_reduction_cost): Handle
4342 EXTRACT_LAST_REDUCTION.
4343 (get_initial_def_for_reduction): Do not create an initial vector
4344 for EXTRACT_LAST_REDUCTION reductions.
4345 (vectorizable_reduction): Leave the scalar phi in place for
4346 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
4347 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
4348 epilogue code for EXTRACT_LAST_REDUCTION and defer the
4349 transform phase to vectorizable_condition.
4350 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
4352 (vect_finish_stmt_generation): ...here.
4353 (vect_finish_replace_stmt): New function.
4354 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
4355 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
4357 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
4359 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4360 Alan Hayward <alan.hayward@arm.com>
4361 David Sherwood <david.sherwood@arm.com>
4363 * doc/md.texi (extract_last_@var{m}): Document.
4364 * optabs.def (extract_last_optab): New optab.
4365 * internal-fn.def (EXTRACT_LAST): New internal function.
4366 * internal-fn.c (cond_unary_direct): New macro.
4367 (expand_cond_unary_optab_fn): Likewise.
4368 (direct_cond_unary_optab_supported_p): Likewise.
4369 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
4370 loops using EXTRACT_LAST.
4371 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
4372 (extract_last_<mode>): ...this optab.
4373 (vec_extract<mode><Vel>): Update accordingly.
4375 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4376 Alan Hayward <alan.hayward@arm.com>
4377 David Sherwood <david.sherwood@arm.com>
4379 * target.def (empty_mask_is_expensive): New hook.
4380 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
4381 * doc/tm.texi: Regenerate.
4382 * targhooks.h (default_empty_mask_is_expensive): Declare.
4383 * targhooks.c (default_empty_mask_is_expensive): New function.
4384 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
4385 if the target says that empty masks are expensive.
4386 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
4388 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
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 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
4395 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
4396 (vect_use_loop_mask_for_alignment_p): New function.
4397 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
4398 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
4399 niters_skip argument. Make sure that the first niters_skip elements
4400 of the first iteration are inactive.
4401 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
4402 Update call to vect_set_loop_masks_directly.
4403 (get_misalign_in_elems): New function, split out from...
4404 (vect_gen_prolog_loop_niters): ...here.
4405 (vect_update_init_of_dr): Take a code argument that specifies whether
4406 the adjustment should be added or subtracted.
4407 (vect_update_init_of_drs): Likewise.
4408 (vect_prepare_for_masked_peels): New function.
4409 (vect_do_peeling): Skip prologue peeling if we're using a mask
4410 instead. Update call to vect_update_inits_of_drs.
4411 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4413 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
4414 alignment. Do not include the number of peeled iterations in
4415 the minimum threshold in that case.
4416 (vectorizable_induction): Adjust the start value down by
4417 LOOP_VINFO_MASK_SKIP_NITERS iterations.
4418 (vect_transform_loop): Call vect_prepare_for_masked_peels.
4419 Take the number of skipped iterations into account when calculating
4421 * tree-vect-stmts.c (vect_gen_while_not): New function.
4423 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4424 Alan Hayward <alan.hayward@arm.com>
4425 David Sherwood <david.sherwood@arm.com>
4427 * doc/sourcebuild.texi (vect_fully_masked): Document.
4428 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
4430 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
4432 (vect_analyze_loop_2): ...here. Don't check the vectorization
4433 factor against the number of loop iterations if the loop is
4436 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4437 Alan Hayward <alan.hayward@arm.com>
4438 David Sherwood <david.sherwood@arm.com>
4440 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
4441 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
4442 (dump_groups): Update accordingly.
4443 (iv_use::mem_type): New member variable.
4444 (address_p): New function.
4445 (record_use): Add a mem_type argument and initialize the new
4447 (record_group_use): Add a mem_type argument. Use address_p.
4448 Remove obsolete null checks of base_object. Update call to record_use.
4449 (find_interesting_uses_op): Update call to record_group_use.
4450 (find_interesting_uses_cond): Likewise.
4451 (find_interesting_uses_address): Likewise.
4452 (get_mem_type_for_internal_fn): New function.
4453 (find_address_like_use): Likewise.
4454 (find_interesting_uses_stmt): Try find_address_like_use before
4455 calling find_interesting_uses_op.
4456 (addr_offset_valid_p): Use the iv mem_type field as the type
4457 of the addressed memory.
4458 (add_autoinc_candidates): Likewise.
4459 (get_address_cost): Likewise.
4460 (split_small_address_groups_p): Use address_p.
4461 (split_address_groups): Likewise.
4462 (add_iv_candidate_for_use): Likewise.
4463 (autoinc_possible_for_pair): Likewise.
4464 (rewrite_groups): Likewise.
4465 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
4466 (determine_group_iv_cost): Update after split of USE_ADDRESS.
4467 (get_alias_ptr_type_for_ptr_address): New function.
4468 (rewrite_use_address): Rewrite address uses in calls that were
4469 identified by find_address_like_use.
4471 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4472 Alan Hayward <alan.hayward@arm.com>
4473 David Sherwood <david.sherwood@arm.com>
4475 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
4477 * gimple-expr.h (is_gimple_addressable: Likewise.
4478 * gimple-expr.c (is_gimple_address): Likewise.
4479 * internal-fn.c (expand_call_mem_ref): New function.
4480 (expand_mask_load_optab_fn): Use it.
4481 (expand_mask_store_optab_fn): Likewise.
4483 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4484 Alan Hayward <alan.hayward@arm.com>
4485 David Sherwood <david.sherwood@arm.com>
4487 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
4488 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
4489 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
4490 (cond_umax@var{mode}): Document.
4491 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
4492 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
4493 (cond_umin_optab, cond_umax_optab): New optabs.
4494 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
4495 (COND_IOR, COND_XOR): New internal functions.
4496 * internal-fn.h (get_conditional_internal_fn): Declare.
4497 * internal-fn.c (cond_binary_direct): New macro.
4498 (expand_cond_binary_optab_fn): Likewise.
4499 (direct_cond_binary_optab_supported_p): Likewise.
4500 (get_conditional_internal_fn): New function.
4501 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
4502 Cope with reduction statements that are vectorized as calls rather
4504 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
4505 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
4506 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
4507 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4508 (UNSPEC_COND_EOR): New unspecs.
4509 (optab): Add mappings for them.
4510 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
4511 (sve_int_op, sve_fp_op): New int attributes.
4513 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4514 Alan Hayward <alan.hayward@arm.com>
4515 David Sherwood <david.sherwood@arm.com>
4517 * optabs.def (while_ult_optab): New optab.
4518 * doc/md.texi (while_ult@var{m}@var{n}): Document.
4519 * internal-fn.def (WHILE_ULT): New internal function.
4520 * internal-fn.h (direct_internal_fn_supported_p): New override
4521 that takes two types as argument.
4522 * internal-fn.c (while_direct): New macro.
4523 (expand_while_optab_fn): New function.
4524 (convert_optab_supported_p): Likewise.
4525 (direct_while_optab_supported_p): New macro.
4526 * wide-int.h (wi::udiv_ceil): New function.
4527 * tree-vectorizer.h (rgroup_masks): New structure.
4528 (vec_loop_masks): New typedef.
4529 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
4531 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
4532 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
4533 (vect_max_vf): New function.
4534 (slpeel_make_loop_iterate_ntimes): Delete.
4535 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
4536 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
4537 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
4538 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
4539 internal-fn.h, stor-layout.h and optabs-query.h.
4540 (vect_set_loop_mask): New function.
4541 (add_preheader_seq): Likewise.
4542 (add_header_seq): Likewise.
4543 (interleave_supported_p): Likewise.
4544 (vect_maybe_permute_loop_masks): Likewise.
4545 (vect_set_loop_masks_directly): Likewise.
4546 (vect_set_loop_condition_masked): Likewise.
4547 (vect_set_loop_condition_unmasked): New function, split out from
4548 slpeel_make_loop_iterate_ntimes.
4549 (slpeel_make_loop_iterate_ntimes): Rename to..
4550 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
4551 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
4552 (vect_do_peeling): Update call accordingly.
4553 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
4555 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4556 mask_compare_type, can_fully_mask_p and fully_masked_p.
4557 (release_vec_loop_masks): New function.
4558 (_loop_vec_info): Use it to free the loop masks.
4559 (can_produce_all_loop_masks_p): New function.
4560 (vect_get_max_nscalars_per_iter): Likewise.
4561 (vect_verify_full_masking): Likewise.
4562 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
4563 retries, and free the mask rgroups before retrying. Check loop-wide
4564 reasons for disallowing fully-masked loops. Make the final decision
4565 about whether use a fully-masked loop or not.
4566 (vect_estimate_min_profitable_iters): Do not assume that peeling
4567 for the number of iterations will be needed for fully-masked loops.
4568 (vectorizable_reduction): Disable fully-masked loops.
4569 (vectorizable_live_operation): Likewise.
4570 (vect_halve_mask_nunits): New function.
4571 (vect_double_mask_nunits): Likewise.
4572 (vect_record_loop_mask): Likewise.
4573 (vect_get_loop_mask): Likewise.
4574 (vect_transform_loop): Handle the case in which the final loop
4575 iteration might handle a partial vector. Call vect_set_loop_condition
4576 instead of slpeel_make_loop_iterate_ntimes.
4577 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
4578 (check_load_store_masking): New function.
4579 (prepare_load_store_mask): Likewise.
4580 (vectorizable_store): Handle fully-masked loops.
4581 (vectorizable_load): Likewise.
4582 (supportable_widening_operation): Use vect_halve_mask_nunits for
4584 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
4585 (vect_gen_while): New function.
4586 * config/aarch64/aarch64.md (umax<mode>3): New expander.
4587 (aarch64_uqdec<mode>): New insn.
4589 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4590 Alan Hayward <alan.hayward@arm.com>
4591 David Sherwood <david.sherwood@arm.com>
4593 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
4594 (reduc_xor_scal_optab): New optabs.
4595 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
4596 (reduc_xor_scal_@var{m}): Document.
4597 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
4598 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
4600 * fold-const-call.c (fold_const_call): Handle them.
4601 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
4602 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
4603 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
4604 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
4605 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
4606 (UNSPEC_XORV): New unspecs.
4607 (optab): Add entries for them.
4608 (BITWISEV): New int iterator.
4609 (bit_reduc_op): New int attributes.
4611 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4612 Alan Hayward <alan.hayward@arm.com>
4613 David Sherwood <david.sherwood@arm.com>
4615 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
4616 * internal-fn.def (VEC_SHL_INSERT): New internal function.
4617 * optabs.def (vec_shl_insert_optab): New optab.
4618 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
4619 (duplicate_and_interleave): Likewise.
4620 * tree-vect-loop.c: Include internal-fn.h.
4621 (neutral_op_for_slp_reduction): New function, split out from
4622 get_initial_defs_for_reduction.
4623 (get_initial_def_for_reduction): Handle option 2 for variable-length
4624 vectors by loading the neutral value into a vector and then shifting
4625 the initial value into element 0.
4626 (get_initial_defs_for_reduction): Replace the code argument with
4627 the neutral value calculated by neutral_op_for_slp_reduction.
4628 Use gimple_build_vector for constant-length vectors.
4629 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
4630 but the first group_size elements have a neutral value.
4631 Use duplicate_and_interleave otherwise.
4632 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
4633 Update call to get_initial_defs_for_reduction. Handle SLP
4634 reductions for variable-length vectors by creating one vector
4635 result for each scalar result, with the elements associated
4636 with other scalar results stubbed out with the neutral value.
4637 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
4638 Require IFN_VEC_SHL_INSERT for double reductions on
4639 variable-length vectors, or SLP reductions that have
4640 a neutral value. Require can_duplicate_and_interleave_p
4641 support for variable-length unchained SLP reductions if there
4642 is no neutral value, such as for MIN/MAX reductions. Also require
4643 the number of vector elements to be a multiple of the number of
4644 SLP statements when doing variable-length unchained SLP reductions.
4645 Update call to vect_create_epilog_for_reduction.
4646 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
4647 and remove initial values.
4648 (duplicate_and_interleave): Make public.
4649 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
4650 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
4652 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4653 Alan Hayward <alan.hayward@arm.com>
4654 David Sherwood <david.sherwood@arm.com>
4656 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
4657 (can_duplicate_and_interleave_p): New function.
4658 (vect_get_and_check_slp_defs): Take the vector of statements
4659 rather than just the current one. Remove excess parentheses.
4660 Restriction rejectinon of vect_constant_def and vect_external_def
4661 for variable-length vectors to boolean types, or types for which
4662 can_duplicate_and_interleave_p is false.
4663 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
4664 (duplicate_and_interleave): New function.
4665 (vect_get_constant_vectors): Use gimple_build_vector for
4666 constant-length vectors and suitable variable-length constant
4667 vectors. Use duplicate_and_interleave for other variable-length
4668 vectors. Don't defer the update when inserting new statements.
4670 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4671 Alan Hayward <alan.hayward@arm.com>
4672 David Sherwood <david.sherwood@arm.com>
4674 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
4675 min_profitable_iters doesn't go negative.
4677 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4678 Alan Hayward <alan.hayward@arm.com>
4679 David Sherwood <david.sherwood@arm.com>
4681 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
4682 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
4683 * optabs.def (vec_mask_load_lanes_optab): New optab.
4684 (vec_mask_store_lanes_optab): Likewise.
4685 * internal-fn.def (MASK_LOAD_LANES): New internal function.
4686 (MASK_STORE_LANES): Likewise.
4687 * internal-fn.c (mask_load_lanes_direct): New macro.
4688 (mask_store_lanes_direct): Likewise.
4689 (expand_mask_load_optab_fn): Handle masked operations.
4690 (expand_mask_load_lanes_optab_fn): New macro.
4691 (expand_mask_store_optab_fn): Handle masked operations.
4692 (expand_mask_store_lanes_optab_fn): New macro.
4693 (direct_mask_load_lanes_optab_supported_p): Likewise.
4694 (direct_mask_store_lanes_optab_supported_p): Likewise.
4695 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
4697 (vect_load_lanes_supported): Likewise.
4698 * tree-vect-data-refs.c (strip_conversion): New function.
4699 (can_group_stmts_p): Likewise.
4700 (vect_analyze_data_ref_accesses): Use it instead of checking
4701 for a pair of assignments.
4702 (vect_store_lanes_supported): Take a masked_p parameter.
4703 (vect_load_lanes_supported): Likewise.
4704 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
4705 vect_store_lanes_supported and vect_load_lanes_supported.
4706 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4707 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
4708 parameter. Don't allow gaps for masked accesses.
4709 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
4710 and vect_load_lanes_supported.
4711 (get_load_store_type): Take a masked_p parameter and update
4712 call to get_group_load_store_type.
4713 (vectorizable_store): Update call to get_load_store_type.
4714 Handle IFN_MASK_STORE_LANES.
4715 (vectorizable_load): Update call to get_load_store_type.
4716 Handle IFN_MASK_LOAD_LANES.
4718 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4719 Alan Hayward <alan.hayward@arm.com>
4720 David Sherwood <david.sherwood@arm.com>
4722 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
4724 * config/aarch64/aarch64-protos.h
4725 (aarch64_sve_struct_memory_operand_p): Declare.
4726 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
4727 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
4728 (VPRED, vpred): Handle SVE structure modes.
4729 * config/aarch64/constraints.md (Utx): New constraint.
4730 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
4731 (aarch64_sve_struct_nonimmediate_operand): New predicates.
4732 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
4733 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
4734 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
4735 structure modes. Split into pieces after RA.
4736 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
4737 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
4739 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
4740 SVE structure modes.
4741 (aarch64_classify_address): Likewise.
4742 (sizetochar): Move earlier in file.
4743 (aarch64_print_operand): Handle SVE register lists.
4744 (aarch64_array_mode): New function.
4745 (aarch64_sve_struct_memory_operand_p): Likewise.
4746 (TARGET_ARRAY_MODE): Redefine.
4748 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4749 Alan Hayward <alan.hayward@arm.com>
4750 David Sherwood <david.sherwood@arm.com>
4752 * target.def (array_mode): New target hook.
4753 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
4754 * doc/tm.texi: Regenerate.
4755 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
4756 * hooks.c (hook_optmode_mode_uhwi_none): New function.
4757 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
4759 * stor-layout.c (mode_for_array): Likewise. Support polynomial
4762 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4763 Alan Hayward <alan.hayward@arm.com>
4764 David Sherwood <david.sherwood@arm.com>
4766 * fold-const.c (fold_binary_loc): Check the argument types
4767 rather than the result type when testing for a vector operation.
4769 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4771 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
4772 * doc/tm.texi: Regenerate.
4774 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4775 Alan Hayward <alan.hayward@arm.com>
4776 David Sherwood <david.sherwood@arm.com>
4778 * doc/invoke.texi (-msve-vector-bits=): Document new option.
4779 (sve): Document new AArch64 extension.
4780 * doc/md.texi (w): Extend the description of the AArch64
4781 constraint to include SVE vectors.
4782 (Upl, Upa): Document new AArch64 predicate constraints.
4783 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
4785 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
4786 (msve-vector-bits=): New option.
4787 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
4788 SVE when these are disabled.
4789 (sve): New extension.
4790 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
4791 modes. Adjust their number of units based on aarch64_sve_vg.
4792 (MAX_BITSIZE_MODE_ANY_MODE): Define.
4793 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
4794 aarch64_addr_query_type.
4795 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
4796 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
4797 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
4798 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
4799 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
4800 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
4801 (aarch64_simd_imm_zero_p): Delete.
4802 (aarch64_check_zero_based_sve_index_immediate): Declare.
4803 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4804 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4805 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4806 (aarch64_sve_float_mul_immediate_p): Likewise.
4807 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4809 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
4810 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
4811 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
4812 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
4813 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
4814 (aarch64_regmode_natural_size): Likewise.
4815 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
4816 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
4818 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
4819 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
4820 for VG and the SVE predicate registers.
4821 (V_ALIASES): Add a "z"-prefixed alias.
4822 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
4823 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
4824 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
4825 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
4826 (REG_CLASS_NAMES): Add entries for them.
4827 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
4828 and the predicate registers.
4829 (aarch64_sve_vg): Declare.
4830 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
4831 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
4832 (REGMODE_NATURAL_SIZE): Define.
4833 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
4835 * config/aarch64/aarch64.c: Include cfgrtl.h.
4836 (simd_immediate_info): Add a constructor for series vectors,
4837 and an associated step field.
4838 (aarch64_sve_vg): New variable.
4839 (aarch64_dbx_register_number): Handle VG and the predicate registers.
4840 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
4841 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
4842 (VEC_ANY_DATA, VEC_STRUCT): New constants.
4843 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
4844 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
4845 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
4846 (aarch64_get_mask_mode): New functions.
4847 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
4848 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
4849 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
4850 predicate modes and predicate registers. Explicitly restrict
4851 GPRs to modes of 16 bytes or smaller. Only allow FP registers
4852 to store a vector mode if it is recognized by
4853 aarch64_classify_vector_mode.
4854 (aarch64_regmode_natural_size): New function.
4855 (aarch64_hard_regno_caller_save_mode): Return the original mode
4857 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
4858 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
4859 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
4860 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
4862 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
4863 does not overlap dest if the function is frame-related. Handle
4865 (aarch64_split_add_offset): New function.
4866 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
4867 them aarch64_add_offset.
4868 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
4869 and update call to aarch64_sub_sp.
4870 (aarch64_add_cfa_expression): New function.
4871 (aarch64_expand_prologue): Pass extra temporary registers to the
4872 functions above. Handle the case in which we need to emit new
4873 DW_CFA_expressions for registers that were originally saved
4874 relative to the stack pointer, but now have to be expressed
4875 relative to the frame pointer.
4876 (aarch64_output_mi_thunk): Pass extra temporary registers to the
4878 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
4879 IP0 and IP1 values for SVE frames.
4880 (aarch64_expand_vec_series): New function.
4881 (aarch64_expand_sve_widened_duplicate): Likewise.
4882 (aarch64_expand_sve_const_vector): Likewise.
4883 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
4884 Handle SVE constants. Use emit_move_insn to move a force_const_mem
4885 into the register, rather than emitting a SET directly.
4886 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
4887 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
4888 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
4889 (offset_9bit_signed_scaled_p): New functions.
4890 (aarch64_replicate_bitmask_imm): New function.
4891 (aarch64_bitmask_imm): Use it.
4892 (aarch64_cannot_force_const_mem): Reject expressions involving
4893 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
4894 (aarch64_classify_index): Handle SVE indices, by requiring
4895 a plain register index with a scale that matches the element size.
4896 (aarch64_classify_address): Handle SVE addresses. Assert that
4897 the mode of the address is VOIDmode or an integer mode.
4898 Update call to aarch64_classify_symbol.
4899 (aarch64_classify_symbolic_expression): Update call to
4900 aarch64_classify_symbol.
4901 (aarch64_const_vec_all_in_range_p): New function.
4902 (aarch64_print_vector_float_operand): Likewise.
4903 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
4904 "vN" for FP registers with SVE modes. Handle (const ...) vectors
4905 and the FP immediates 1.0 and 0.5.
4906 (aarch64_print_address_internal): Handle SVE addresses.
4907 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
4908 (aarch64_regno_regclass): Handle predicate registers.
4909 (aarch64_secondary_reload): Handle big-endian reloads of SVE
4911 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
4912 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
4913 (aarch64_convert_sve_vector_bits): New function.
4914 (aarch64_override_options): Use it to handle -msve-vector-bits=.
4915 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4917 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
4918 Handle SVE vector and predicate modes. Accept VL-based constants
4919 that need only one temporary register, and VL offsets that require
4920 no temporary registers.
4921 (aarch64_conditional_register_usage): Mark the predicate registers
4922 as fixed if SVE isn't available.
4923 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
4924 Return true for SVE vector and predicate modes.
4925 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
4926 rather than an unsigned int. Handle SVE modes.
4927 (aarch64_preferred_simd_mode): Update call accordingly. Handle
4929 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
4931 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4932 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4933 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4934 (aarch64_sve_float_mul_immediate_p): New functions.
4935 (aarch64_sve_valid_immediate): New function.
4936 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
4937 Explicitly reject structure modes. Check for INDEX constants.
4938 Handle PTRUE and PFALSE constants.
4939 (aarch64_check_zero_based_sve_index_immediate): New function.
4940 (aarch64_simd_imm_zero_p): Delete.
4941 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
4942 vector modes. Accept constants in the range of CNT[BHWD].
4943 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
4944 ask for an Advanced SIMD mode.
4945 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
4946 (aarch64_simd_vector_alignment): Handle SVE predicates.
4947 (aarch64_vectorize_preferred_vector_alignment): New function.
4948 (aarch64_simd_vector_alignment_reachable): Use it instead of
4950 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
4951 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
4953 (MAX_VECT_LEN): Delete.
4954 (expand_vec_perm_d): Add a vec_flags field.
4955 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
4956 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
4957 (aarch64_evpc_ext): Don't apply a big-endian lane correction
4959 (aarch64_evpc_rev): Rename to...
4960 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
4961 (aarch64_evpc_rev_global): New function.
4962 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
4963 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
4965 (aarch64_evpc_sve_tbl): New function.
4966 (aarch64_expand_vec_perm_const_1): Update after rename of
4967 aarch64_evpc_rev. Handle SVE permutes too, trying
4968 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
4969 than aarch64_evpc_tbl.
4970 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
4971 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
4972 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
4973 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
4974 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
4975 (aarch64_expand_sve_vcond): New functions.
4976 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
4977 of aarch64_vector_mode_p.
4978 (aarch64_dwarf_poly_indeterminate_value): New function.
4979 (aarch64_compute_pressure_classes): Likewise.
4980 (aarch64_can_change_mode_class): Likewise.
4981 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
4982 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
4983 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
4984 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
4985 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
4986 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
4987 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
4988 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
4990 (Dn, Dl, Dr): Accept const as well as const_vector.
4991 (Dz): Likewise. Compare against CONST0_RTX.
4992 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
4993 of "vector" where appropriate.
4994 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
4995 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
4996 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
4997 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
4998 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
4999 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
5000 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
5001 (v_int_equiv): Extend to SVE modes.
5002 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
5004 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
5005 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
5006 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
5007 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
5008 (SVE_COND_FP_CMP): New int iterators.
5009 (perm_hilo): Handle the new unpack unspecs.
5010 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
5012 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
5013 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
5014 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
5015 (aarch64_equality_operator, aarch64_constant_vector_operand)
5016 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
5017 (aarch64_sve_nonimmediate_operand): Likewise.
5018 (aarch64_sve_general_operand): Likewise.
5019 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
5020 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
5021 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
5022 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
5023 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
5024 (aarch64_sve_float_arith_immediate): Likewise.
5025 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
5026 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
5027 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
5028 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
5029 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
5030 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
5031 (aarch64_sve_float_arith_operand): Likewise.
5032 (aarch64_sve_float_arith_with_sub_operand): Likewise.
5033 (aarch64_sve_float_mul_operand): Likewise.
5034 (aarch64_sve_vec_perm_operand): Likewise.
5035 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
5036 (aarch64_mov_operand): Accept const_poly_int and const_vector.
5037 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
5038 as well as const_vector.
5039 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
5040 in file. Use CONST0_RTX and CONSTM1_RTX.
5041 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
5042 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
5043 Use aarch64_simd_imm_zero.
5044 * config/aarch64/aarch64-sve.md: New file.
5045 * config/aarch64/aarch64.md: Include it.
5046 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
5047 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
5048 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
5049 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
5050 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
5051 (sve): New attribute.
5052 (enabled): Disable instructions with the sve attribute unless
5054 (movqi, movhi): Pass CONST_POLY_INT operaneds through
5055 aarch64_expand_mov_immediate.
5056 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
5057 CNT[BHSD] immediates.
5058 (movti): Split CONST_POLY_INT moves into two halves.
5059 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
5060 Split additions that need a temporary here if the destination
5061 is the stack pointer.
5062 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
5063 (*add<mode>3_poly_1): New instruction.
5064 (set_clobber_cc): New expander.
5066 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5068 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
5069 parameter and use it instead of GET_MODE_SIZE (innermode). Use
5070 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
5071 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
5072 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
5073 Change innermode from fixed_mode_size to machine_mode.
5074 (simplify_subreg): Update call accordingly. Handle a constant-sized
5075 subreg of a variable-length CONST_VECTOR.
5077 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5078 Alan Hayward <alan.hayward@arm.com>
5079 David Sherwood <david.sherwood@arm.com>
5081 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
5082 (add_offset_to_base): New function, split out from...
5083 (create_mem_ref): ...here. When handling a scale other than 1,
5084 check first whether the address is valid without the offset.
5085 Add it into the base if so, leaving the index and scale as-is.
5087 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5090 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
5091 fold_for_warn before checking if arg2 is INTEGER_CST.
5093 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
5095 * config/rs6000/predicates.md (load_multiple_operation): Delete.
5096 (store_multiple_operation): Delete.
5097 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
5098 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
5099 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
5100 guarded by TARGET_STRING.
5101 (rs6000_output_load_multiple): Delete.
5102 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
5103 OPTION_MASK_STRING / TARGET_STRING handling.
5104 (print_operand) <'N', 'O'>: Add comment that these are unused now.
5105 (const rs6000_opt_masks) <"string">: Change mask to 0.
5106 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
5107 (MASK_STRING): Delete.
5108 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
5110 (load_multiple): Delete.
5117 (store_multiple): Delete.
5124 (movmemsi_8reg): Delete.
5125 (corresponding unnamed define_insn): Delete.
5126 (movmemsi_6reg): Delete.
5127 (corresponding unnamed define_insn): Delete.
5128 (movmemsi_4reg): Delete.
5129 (corresponding unnamed define_insn): Delete.
5130 (movmemsi_2reg): Delete.
5131 (corresponding unnamed define_insn): Delete.
5132 (movmemsi_1reg): Delete.
5133 (corresponding unnamed define_insn): Delete.
5134 * config/rs6000/rs6000.opt (mno-string): New.
5135 (mstring): Replace by deprecation warning stub.
5136 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
5138 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5140 * regrename.c (regrename_do_replace): If replacing the same
5141 reg multiple times, try to reuse last created gen_raw_REG.
5144 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
5145 main to workaround a bug in GDB.
5147 2018-01-12 Tom de Vries <tom@codesourcery.com>
5150 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
5152 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
5154 PR rtl-optimization/80481
5155 * ira-color.c (get_cap_member): New function.
5156 (allocnos_conflict_by_live_ranges_p): Use it.
5157 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
5158 (setup_slot_coalesced_allocno_live_ranges): Ditto.
5160 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
5163 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
5164 (*saddl_se_1): Ditto.
5166 (*ssubl_se_1): Ditto.
5168 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5170 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
5171 rather than wi::to_widest for DR_INITs.
5172 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
5173 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
5174 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
5176 (vect_analyze_group_access_1): Note that here.
5178 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5180 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
5181 polynomial type sizes.
5183 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5185 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
5186 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
5187 (gimple_add_tmp_var): Likewise.
5189 2018-01-12 Martin Liska <mliska@suse.cz>
5191 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
5192 (gimple_alloc_sizes): Likewise.
5193 (dump_gimple_statistics): Use PRIu64 in printf format.
5194 * gimple.h: Change uint64_t to int.
5196 2018-01-12 Martin Liska <mliska@suse.cz>
5198 * tree-core.h: Use uint64_t instead of int.
5199 * tree.c (tree_node_counts): Likewise.
5200 (tree_node_sizes): Likewise.
5201 (dump_tree_statistics): Use PRIu64 in printf format.
5203 2018-01-12 Martin Liska <mliska@suse.cz>
5205 * Makefile.in: As qsort_chk is implemented in vec.c, add
5206 vec.o to linkage of gencfn-macros.
5207 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
5208 passing the info to record_node_allocation_statistics.
5209 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
5211 * ggc-common.c (struct ggc_usage): Add operator== and use
5212 it in operator< and compare function.
5213 * mem-stats.h (struct mem_usage): Likewise.
5214 * vec.c (struct vec_usage): Remove operator< and compare
5215 function. Can be simply inherited.
5217 2018-01-12 Martin Jambor <mjambor@suse.cz>
5220 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
5221 * tree-ssa-math-opts.c: Include domwalk.h.
5222 (convert_mult_to_fma_1): New function.
5223 (fma_transformation_info): New type.
5224 (fma_deferring_state): Likewise.
5225 (cancel_fma_deferring): New function.
5226 (result_of_phi): Likewise.
5227 (last_fma_candidate_feeds_initial_phi): Likewise.
5228 (convert_mult_to_fma): Added deferring logic, split actual
5229 transformation to convert_mult_to_fma_1.
5230 (math_opts_dom_walker): New type.
5231 (math_opts_dom_walker::after_dom_children): New method, body moved
5232 here from pass_optimize_widening_mul::execute, added deferring logic
5234 (pass_optimize_widening_mul::execute): Moved most of code to
5235 math_opts_dom_walker::after_dom_children.
5236 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
5237 * config/i386/i386.c (ix86_option_override_internal): Added
5238 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
5240 2018-01-12 Richard Biener <rguenther@suse.de>
5243 * dwarf2out.c (gen_variable_die): Do not reset old_die for
5244 inline instance vars.
5246 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
5249 * config/rx/rx.c (rx_is_restricted_memory_address):
5252 2018-01-12 Richard Biener <rguenther@suse.de>
5254 PR tree-optimization/80846
5255 * target.def (split_reduction): New target hook.
5256 * targhooks.c (default_split_reduction): New function.
5257 * targhooks.h (default_split_reduction): Declare.
5258 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
5259 target requests first reduce vectors by combining low and high
5261 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
5262 (get_vectype_for_scalar_type_and_size): Export.
5263 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
5264 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
5265 * doc/tm.texi: Regenerate.
5266 * config/i386/i386.c (ix86_split_reduction): Implement
5267 TARGET_VECTORIZE_SPLIT_REDUCTION.
5269 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5272 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
5273 in PIC mode except for TARGET_VXWORKS_RTP.
5274 * config/sparc/sparc.c: Include cfgrtl.h.
5275 (TARGET_INIT_PIC_REG): Define.
5276 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
5277 (sparc_pic_register_p): New predicate.
5278 (sparc_legitimate_address_p): Use it.
5279 (sparc_legitimize_pic_address): Likewise.
5280 (sparc_delegitimize_address): Likewise.
5281 (sparc_mode_dependent_address_p): Likewise.
5282 (gen_load_pcrel_sym): Remove 4th parameter.
5283 (load_got_register): Adjust call to above. Remove obsolete stuff.
5284 (sparc_expand_prologue): Do not call load_got_register here.
5285 (sparc_flat_expand_prologue): Likewise.
5286 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
5287 (sparc_use_pseudo_pic_reg): New function.
5288 (sparc_init_pic_reg): Likewise.
5289 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5290 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
5292 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
5294 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
5295 Add item for branch_cost.
5297 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5299 PR rtl-optimization/83565
5300 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
5301 not extend the result to a larger mode for rotate operations.
5302 (num_sign_bit_copies1): Likewise.
5304 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5307 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
5309 Use values-Xc.o for -pedantic.
5310 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
5312 2018-01-12 Martin Liska <mliska@suse.cz>
5315 * ipa-devirt.c (final_warning_record::grow_type_warnings):
5317 (possible_polymorphic_call_targets): Use it.
5318 (ipa_devirt): Likewise.
5320 2018-01-12 Martin Liska <mliska@suse.cz>
5322 * profile-count.h (enum profile_quality): Use 0 as invalid
5323 enum value of profile_quality.
5325 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
5327 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
5328 -mext-string options.
5330 2018-01-12 Richard Biener <rguenther@suse.de>
5332 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
5333 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
5334 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
5336 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
5338 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5340 * configure.ac (--with-long-double-format): Add support for the
5341 configuration option to change the default long double format on
5343 * config.gcc (powerpc*-linux*-*): Likewise.
5344 * configure: Regenerate.
5345 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
5346 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
5347 used without modification.
5349 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5351 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
5352 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
5353 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
5354 MISC_BUILTIN_SPEC_BARRIER.
5355 (rs6000_init_builtins): Likewise.
5356 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
5358 (speculation_barrier): New define_insn.
5359 * doc/extend.texi: Document __builtin_speculation_barrier.
5361 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5364 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
5365 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
5366 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
5368 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
5369 integral modes instead of "ss" and "sd".
5370 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
5371 vectors with 32-bit and 64-bit elements.
5372 (vecdupssescalarmodesuffix): New mode attribute.
5373 (vec_dup<mode>): Use it.
5375 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
5378 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
5379 frame if argument is passed on stack.
5381 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5384 * ree.c (combine_reaching_defs): Optimize also
5385 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
5386 reg2=any_extend(exp); reg1=reg2;, formatting fix.
5388 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5391 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
5393 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5396 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
5397 after they are computed.
5399 2018-01-11 Bin Cheng <bin.cheng@arm.com>
5401 PR tree-optimization/83695
5402 * gimple-loop-linterchange.cc
5403 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
5404 reset cached scev information after interchange.
5405 (pass_linterchange::execute): Remove call to scev_reset_htab.
5407 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5409 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
5410 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
5411 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
5412 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
5413 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
5414 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
5415 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
5416 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
5417 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
5418 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
5419 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
5420 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
5421 (V_lane_reg): Likewise.
5422 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
5424 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
5425 (vfmal_lane_low<mode>_intrinsic,
5426 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
5427 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
5428 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
5429 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
5430 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
5431 vfmsl_lane_high<mode>_intrinsic): New define_insns.
5433 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5435 * config/arm/arm-cpus.in (fp16fml): New feature.
5436 (ALL_SIMD): Add fp16fml.
5437 (armv8.2-a): Add fp16fml as an option.
5438 (armv8.3-a): Likewise.
5439 (armv8.4-a): Add fp16fml as part of fp16.
5440 * config/arm/arm.h (TARGET_FP16FML): Define.
5441 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
5443 * config/arm/arm-modes.def (V2HF): Define.
5444 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
5445 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
5446 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
5447 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
5448 vfmsl_low, vfmsl_high): New set of builtins.
5449 * config/arm/iterators.md (PLUSMINUS): New code iterator.
5450 (vfml_op): New code attribute.
5451 (VFMLHALVES): New int iterator.
5452 (VFML, VFMLSEL): New mode attributes.
5453 (V_reg): Define mapping for V2HF.
5454 (V_hi, V_lo): New mode attributes.
5455 (VF_constraint): Likewise.
5456 (vfml_half, vfml_half_selector): New int attributes.
5457 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
5459 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
5460 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
5462 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
5463 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
5464 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
5465 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
5467 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
5468 Document new effective target and option set.
5470 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5472 * config/arm/arm-cpus.in (armv8_4): New feature.
5473 (ARMv8_4a): New fgroup.
5474 (armv8.4-a): New arch.
5475 * config/arm/arm-tables.opt: Regenerate.
5476 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
5477 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
5478 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
5479 Add matching rules for -march=armv8.4-a and extensions.
5480 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
5482 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
5485 * config/rx/rx.md (BW): New mode attribute.
5486 (sync_lock_test_and_setsi): Add mode suffix to insn output.
5488 2018-01-11 Richard Biener <rguenther@suse.de>
5490 PR tree-optimization/83435
5491 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
5492 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
5493 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
5495 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5496 Alan Hayward <alan.hayward@arm.com>
5497 David Sherwood <david.sherwood@arm.com>
5499 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
5501 (aarch64_classify_address): Initialize it. Track polynomial offsets.
5502 (aarch64_print_address_internal): Use it to check for a zero offset.
5504 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5505 Alan Hayward <alan.hayward@arm.com>
5506 David Sherwood <david.sherwood@arm.com>
5508 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
5509 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
5510 Return a poly_int64 rather than a HOST_WIDE_INT.
5511 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
5512 rather than a HOST_WIDE_INT.
5513 * config/aarch64/aarch64.h (aarch64_frame): Protect with
5514 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
5515 hard_fp_offset, frame_size, initial_adjust, callee_offset and
5516 final_offset from HOST_WIDE_INT to poly_int64.
5517 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
5518 to_constant when getting the number of units in an Advanced SIMD
5520 (aarch64_builtin_vectorized_function): Check for a constant number
5522 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
5524 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
5525 attribute instead of GET_MODE_NUNITS.
5526 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
5527 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
5528 GET_MODE_SIZE for fixed-size registers.
5529 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
5530 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
5531 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
5532 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
5533 (aarch64_print_operand, aarch64_print_address_internal)
5534 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
5535 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
5536 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
5537 Handle polynomial GET_MODE_SIZE.
5538 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
5539 wider than SImode without modification.
5540 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
5541 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
5542 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
5543 passing and returning SVE modes.
5544 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
5545 rather than GEN_INT.
5546 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
5547 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
5548 (aarch64_allocate_and_probe_stack_space): Likewise.
5549 (aarch64_layout_frame): Cope with polynomial offsets.
5550 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
5551 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
5553 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
5554 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
5555 poly_int64 rather than a HOST_WIDE_INT.
5556 (aarch64_get_separate_components, aarch64_process_components)
5557 (aarch64_expand_prologue, aarch64_expand_epilogue)
5558 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
5559 (aarch64_anchor_offset): New function, split out from...
5560 (aarch64_legitimize_address): ...here.
5561 (aarch64_builtin_vectorization_cost): Handle polynomial
5562 TYPE_VECTOR_SUBPARTS.
5563 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
5565 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
5566 number of elements from the PARALLEL rather than the mode.
5567 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
5568 rather than GET_MODE_BITSIZE.
5569 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
5570 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
5571 (aarch64_expand_vec_perm_const_1): Handle polynomial
5572 d->perm.length () and d->perm elements.
5573 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
5574 Apply to_constant to d->perm elements.
5575 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
5576 polynomial CONST_VECTOR_NUNITS.
5577 (aarch64_move_pointer): Take amount as a poly_int64 rather
5579 (aarch64_progress_pointer): Avoid temporary variable.
5580 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
5581 the mode attribute instead of GET_MODE.
5583 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5584 Alan Hayward <alan.hayward@arm.com>
5585 David Sherwood <david.sherwood@arm.com>
5587 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
5588 x exists before using it.
5589 (aarch64_add_constant_internal): Rename to...
5590 (aarch64_add_offset_1): ...this. Replace regnum with separate
5591 src and dest rtxes. Handle the case in which they're different,
5592 including when the offset is zero. Replace scratchreg with an rtx.
5593 Use 2 additions if there is no spare register into which we can
5594 move a 16-bit constant.
5595 (aarch64_add_constant): Delete.
5596 (aarch64_add_offset): Replace reg with separate src and dest
5597 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
5598 Use aarch64_add_offset_1.
5599 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
5600 an rtx rather than an int. Take the delta as a poly_int64
5601 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
5602 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
5603 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
5604 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
5605 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
5607 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
5608 aarch64_add_constant.
5610 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5612 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
5613 Use scalar_float_mode.
5615 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5617 * config/aarch64/aarch64-simd.md
5618 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
5619 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
5620 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
5621 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
5622 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
5623 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
5624 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
5625 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
5626 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
5627 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
5629 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5632 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
5633 targ_options->x_arm_arch_string is non NULL.
5635 2018-01-11 Tamar Christina <tamar.christina@arm.com>
5637 * config/aarch64/aarch64.h
5638 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
5640 2018-01-11 Sudakshina Das <sudi.das@arm.com>
5643 * expmed.c (emit_store_flag_force): Swap if const op0
5644 and change VOIDmode to mode of op0.
5646 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5648 PR rtl-optimization/83761
5649 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
5650 than bytes to mode_for_size.
5652 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5655 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
5656 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
5659 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5662 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
5663 when in layout mode.
5664 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
5665 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
5668 2018-01-10 Michael Collison <michael.collison@arm.com>
5670 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
5671 * config/aarch64/aarch64-option-extension.def: Add
5672 AARCH64_OPT_EXTENSION of 'fp16fml'.
5673 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5674 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
5675 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
5676 * config/aarch64/constraints.md (Ui7): New constraint.
5677 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
5678 (VFMLA_SEL_W): Ditto.
5681 (VFMLA16_LOW): New int iterator.
5682 (VFMLA16_HIGH): Ditto.
5683 (UNSPEC_FMLAL): New unspec.
5684 (UNSPEC_FMLSL): Ditto.
5685 (UNSPEC_FMLAL2): Ditto.
5686 (UNSPEC_FMLSL2): Ditto.
5687 (f16mac): New code attribute.
5688 * config/aarch64/aarch64-simd-builtins.def
5689 (aarch64_fmlal_lowv2sf): Ditto.
5690 (aarch64_fmlsl_lowv2sf): Ditto.
5691 (aarch64_fmlalq_lowv4sf): Ditto.
5692 (aarch64_fmlslq_lowv4sf): Ditto.
5693 (aarch64_fmlal_highv2sf): Ditto.
5694 (aarch64_fmlsl_highv2sf): Ditto.
5695 (aarch64_fmlalq_highv4sf): Ditto.
5696 (aarch64_fmlslq_highv4sf): Ditto.
5697 (aarch64_fmlal_lane_lowv2sf): Ditto.
5698 (aarch64_fmlsl_lane_lowv2sf): Ditto.
5699 (aarch64_fmlal_laneq_lowv2sf): Ditto.
5700 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
5701 (aarch64_fmlalq_lane_lowv4sf): Ditto.
5702 (aarch64_fmlsl_lane_lowv4sf): Ditto.
5703 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
5704 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
5705 (aarch64_fmlal_lane_highv2sf): Ditto.
5706 (aarch64_fmlsl_lane_highv2sf): Ditto.
5707 (aarch64_fmlal_laneq_highv2sf): Ditto.
5708 (aarch64_fmlsl_laneq_highv2sf): Ditto.
5709 (aarch64_fmlalq_lane_highv4sf): Ditto.
5710 (aarch64_fmlsl_lane_highv4sf): Ditto.
5711 (aarch64_fmlalq_laneq_highv4sf): Ditto.
5712 (aarch64_fmlsl_laneq_highv4sf): Ditto.
5713 * config/aarch64/aarch64-simd.md:
5714 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
5715 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5716 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
5717 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5718 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
5719 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
5720 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
5721 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
5722 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
5723 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
5724 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
5725 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
5726 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
5727 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
5728 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
5729 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
5730 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
5731 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
5732 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
5733 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
5734 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
5735 (vfmlsl_low_u32): Ditto.
5736 (vfmlalq_low_u32): Ditto.
5737 (vfmlslq_low_u32): Ditto.
5738 (vfmlal_high_u32): Ditto.
5739 (vfmlsl_high_u32): Ditto.
5740 (vfmlalq_high_u32): Ditto.
5741 (vfmlslq_high_u32): Ditto.
5742 (vfmlal_lane_low_u32): Ditto.
5743 (vfmlsl_lane_low_u32): Ditto.
5744 (vfmlal_laneq_low_u32): Ditto.
5745 (vfmlsl_laneq_low_u32): Ditto.
5746 (vfmlalq_lane_low_u32): Ditto.
5747 (vfmlslq_lane_low_u32): Ditto.
5748 (vfmlalq_laneq_low_u32): Ditto.
5749 (vfmlslq_laneq_low_u32): Ditto.
5750 (vfmlal_lane_high_u32): Ditto.
5751 (vfmlsl_lane_high_u32): Ditto.
5752 (vfmlal_laneq_high_u32): Ditto.
5753 (vfmlsl_laneq_high_u32): Ditto.
5754 (vfmlalq_lane_high_u32): Ditto.
5755 (vfmlslq_lane_high_u32): Ditto.
5756 (vfmlalq_laneq_high_u32): Ditto.
5757 (vfmlslq_laneq_high_u32): Ditto.
5758 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
5759 (AARCH64_FL_FOR_ARCH8_4): New.
5760 (AARCH64_ISA_F16FML): New ISA flag.
5761 (TARGET_F16FML): New feature flag for fp16fml.
5762 (doc/invoke.texi): Document new fp16fml option.
5764 2018-01-10 Michael Collison <michael.collison@arm.com>
5766 * config/aarch64/aarch64-builtins.c:
5767 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
5768 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5769 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
5770 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
5771 (AARCH64_ISA_SHA3): New ISA flag.
5772 (TARGET_SHA3): New feature flag for sha3.
5773 * config/aarch64/iterators.md (sha512_op): New int attribute.
5774 (CRYPTO_SHA512): New int iterator.
5775 (UNSPEC_SHA512H): New unspec.
5776 (UNSPEC_SHA512H2): Ditto.
5777 (UNSPEC_SHA512SU0): Ditto.
5778 (UNSPEC_SHA512SU1): Ditto.
5779 * config/aarch64/aarch64-simd-builtins.def
5780 (aarch64_crypto_sha512hqv2di): New builtin.
5781 (aarch64_crypto_sha512h2qv2di): Ditto.
5782 (aarch64_crypto_sha512su0qv2di): Ditto.
5783 (aarch64_crypto_sha512su1qv2di): Ditto.
5784 (aarch64_eor3qv8hi): Ditto.
5785 (aarch64_rax1qv2di): Ditto.
5786 (aarch64_xarqv2di): Ditto.
5787 (aarch64_bcaxqv8hi): Ditto.
5788 * config/aarch64/aarch64-simd.md:
5789 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
5790 (aarch64_crypto_sha512su0qv2di): Ditto.
5791 (aarch64_crypto_sha512su1qv2di): Ditto.
5792 (aarch64_eor3qv8hi): Ditto.
5793 (aarch64_rax1qv2di): Ditto.
5794 (aarch64_xarqv2di): Ditto.
5795 (aarch64_bcaxqv8hi): Ditto.
5796 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
5797 (vsha512h2q_u64): Ditto.
5798 (vsha512su0q_u64): Ditto.
5799 (vsha512su1q_u64): Ditto.
5800 (veor3q_u16): Ditto.
5801 (vrax1q_u64): Ditto.
5803 (vbcaxq_u16): Ditto.
5804 * config/arm/types.md (crypto_sha512): New type attribute.
5805 (crypto_sha3): Ditto.
5806 (doc/invoke.texi): Document new sha3 option.
5808 2018-01-10 Michael Collison <michael.collison@arm.com>
5810 * config/aarch64/aarch64-builtins.c:
5811 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
5812 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5813 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
5814 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
5815 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
5816 (AARCH64_ISA_SM4): New ISA flag.
5817 (TARGET_SM4): New feature flag for sm4.
5818 * config/aarch64/aarch64-simd-builtins.def
5819 (aarch64_sm3ss1qv4si): Ditto.
5820 (aarch64_sm3tt1aq4si): Ditto.
5821 (aarch64_sm3tt1bq4si): Ditto.
5822 (aarch64_sm3tt2aq4si): Ditto.
5823 (aarch64_sm3tt2bq4si): Ditto.
5824 (aarch64_sm3partw1qv4si): Ditto.
5825 (aarch64_sm3partw2qv4si): Ditto.
5826 (aarch64_sm4eqv4si): Ditto.
5827 (aarch64_sm4ekeyqv4si): Ditto.
5828 * config/aarch64/aarch64-simd.md:
5829 (aarch64_sm3ss1qv4si): Ditto.
5830 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
5831 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
5832 (aarch64_sm4eqv4si): Ditto.
5833 (aarch64_sm4ekeyqv4si): Ditto.
5834 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
5835 (sm3part_op): Ditto.
5836 (CRYPTO_SM3TT): Ditto.
5837 (CRYPTO_SM3PART): Ditto.
5838 (UNSPEC_SM3SS1): New unspec.
5839 (UNSPEC_SM3TT1A): Ditto.
5840 (UNSPEC_SM3TT1B): Ditto.
5841 (UNSPEC_SM3TT2A): Ditto.
5842 (UNSPEC_SM3TT2B): Ditto.
5843 (UNSPEC_SM3PARTW1): Ditto.
5844 (UNSPEC_SM3PARTW2): Ditto.
5845 (UNSPEC_SM4E): Ditto.
5846 (UNSPEC_SM4EKEY): Ditto.
5847 * config/aarch64/constraints.md (Ui2): New constraint.
5848 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
5849 * config/arm/types.md (crypto_sm3): New type attribute.
5850 (crypto_sm4): Ditto.
5851 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
5852 (vsm3tt1aq_u32): Ditto.
5853 (vsm3tt1bq_u32): Ditto.
5854 (vsm3tt2aq_u32): Ditto.
5855 (vsm3tt2bq_u32): Ditto.
5856 (vsm3partw1q_u32): Ditto.
5857 (vsm3partw2q_u32): Ditto.
5858 (vsm4eq_u32): Ditto.
5859 (vsm4ekeyq_u32): Ditto.
5860 (doc/invoke.texi): Document new sm4 option.
5862 2018-01-10 Michael Collison <michael.collison@arm.com>
5864 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
5865 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
5866 (AARCH64_FL_FOR_ARCH8_4): New.
5867 (AARCH64_FL_V8_4): New flag.
5868 (doc/invoke.texi): Document new armv8.4-a option.
5870 2018-01-10 Michael Collison <michael.collison@arm.com>
5872 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5873 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
5874 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
5875 * config/aarch64/aarch64-option-extension.def: Add
5876 AARCH64_OPT_EXTENSION of 'sha2'.
5877 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
5878 (crypto): Disable sha2 and aes if crypto disabled.
5879 (crypto): Enable aes and sha2 if enabled.
5880 (simd): Disable sha2 and aes if simd disabled.
5881 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
5883 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
5884 (TARGET_SHA2): New feature flag for sha2.
5885 (TARGET_AES): New feature flag for aes.
5886 * config/aarch64/aarch64-simd.md:
5887 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
5888 conditional on TARGET_AES.
5889 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
5890 (aarch64_crypto_sha1hsi): Make pattern conditional
5892 (aarch64_crypto_sha1hv4si): Ditto.
5893 (aarch64_be_crypto_sha1hv4si): Ditto.
5894 (aarch64_crypto_sha1su1v4si): Ditto.
5895 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
5896 (aarch64_crypto_sha1su0v4si): Ditto.
5897 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
5898 (aarch64_crypto_sha256su0v4si): Ditto.
5899 (aarch64_crypto_sha256su1v4si): Ditto.
5900 (doc/invoke.texi): Document new aes and sha2 options.
5902 2018-01-10 Martin Sebor <msebor@redhat.com>
5904 PR tree-optimization/83781
5905 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
5908 2018-01-11 Martin Sebor <msebor@gmail.com>
5909 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5911 PR tree-optimization/83501
5912 PR tree-optimization/81703
5914 * tree-ssa-strlen.c (get_string_cst): Rename...
5915 (get_string_len): ...to this. Handle global constants.
5916 (handle_char_store): Adjust.
5918 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
5919 Jim Wilson <jimw@sifive.com>
5921 * config/riscv/riscv-protos.h (riscv_output_return): New.
5922 * config/riscv/riscv.c (struct machine_function): New naked_p field.
5923 (riscv_attribute_table, riscv_output_return),
5924 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
5925 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
5926 (riscv_compute_frame_info): Only compute frame->mask if not a naked
5928 (riscv_expand_prologue): Add early return for naked function.
5929 (riscv_expand_epilogue): Likewise.
5930 (riscv_function_ok_for_sibcall): Return false for naked function.
5931 (riscv_set_current_function): New.
5932 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
5933 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
5934 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
5935 * doc/extend.texi (RISC-V Function Attributes): New.
5937 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
5939 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
5940 check for 128-bit long double before checking TCmode.
5941 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
5942 128-bit long doubles before checking TFmode or TCmode.
5943 (FLOAT128_IBM_P): Likewise.
5945 2018-01-10 Martin Sebor <msebor@redhat.com>
5947 PR tree-optimization/83671
5948 * builtins.c (c_strlen): Unconditionally return zero for the empty
5950 Use -Warray-bounds for warnings.
5951 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
5952 for non-constant array indices with COMPONENT_REF, arrays of
5953 arrays, and pointers to arrays.
5954 (gimple_fold_builtin_strlen): Determine and set length range for
5955 non-constant character arrays.
5957 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
5960 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
5963 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
5965 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
5967 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5970 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
5971 VECTOR_MEM_ALTIVEC_OR_VSX_P.
5972 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
5973 indexed_or_indirect_operand predicate.
5974 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
5975 (*vsx_le_perm_load_v8hi): Likewise.
5976 (*vsx_le_perm_load_v16qi): Likewise.
5977 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
5978 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
5979 (*vsx_le_perm_store_v8hi): Likewise.
5980 (*vsx_le_perm_store_v16qi): Likewise.
5981 (eight unnamed splitters): Likewise.
5983 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5985 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
5986 * config/rs6000/emmintrin.h: Likewise.
5987 * config/rs6000/mmintrin.h: Likewise.
5988 * config/rs6000/xmmintrin.h: Likewise.
5990 2018-01-10 David Malcolm <dmalcolm@redhat.com>
5993 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
5995 * tree.c (tree_nop_conversion): Return true for location wrapper
5997 (maybe_wrap_with_location): New function.
5998 (selftest::check_strip_nops): New function.
5999 (selftest::test_location_wrappers): New function.
6000 (selftest::tree_c_tests): Call it.
6001 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
6002 (maybe_wrap_with_location): New decl.
6003 (EXPR_LOCATION_WRAPPER_P): New macro.
6004 (location_wrapper_p): New inline function.
6005 (tree_strip_any_location_wrapper): New inline function.
6007 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
6010 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
6011 stack_realign_offset for the largest alignment of stack slot
6013 (ix86_find_max_used_stack_alignment): New function.
6014 (ix86_finalize_stack_frame_flags): Use it. Set
6015 max_used_stack_alignment if we don't realign stack.
6016 * config/i386/i386.h (machine_function): Add
6017 max_used_stack_alignment.
6019 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
6021 * config/arm/arm.opt (-mbranch-cost): New option.
6022 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
6025 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
6028 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
6029 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
6031 2018-01-10 Richard Biener <rguenther@suse.de>
6034 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
6035 early out so it also covers the case where we have a non-NULL
6038 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6040 PR tree-optimization/83753
6041 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
6042 for non-strided grouped accesses if the number of elements is 1.
6044 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6047 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
6048 * i386.h (TARGET_USE_GATHER): Define.
6049 * x86-tune.def (X86_TUNE_USE_GATHER): New.
6051 2018-01-10 Martin Liska <mliska@suse.cz>
6054 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
6055 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
6057 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
6058 CLEANUP_NO_PARTITIONING is not set.
6060 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6062 * doc/rtl.texi: Remove documentation of (const ...) wrappers
6063 for vectors, as a partial revert of r254296.
6064 * rtl.h (const_vec_p): Delete.
6065 (const_vec_duplicate_p): Don't test for vector CONSTs.
6066 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
6067 * expmed.c (make_tree): Likewise.
6070 * common.md (E, F): Use CONSTANT_P instead of checking for
6072 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
6073 checking for CONST_VECTOR.
6075 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6078 * predict.c (force_edge_cold): Handle in more sane way edges
6081 2018-01-09 Carl Love <cel@us.ibm.com>
6083 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
6085 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
6086 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
6087 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
6088 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
6089 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
6090 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
6091 * config/rs6000/rs6000-protos.h: Add extern defition for
6092 rs6000_generate_float2_double_code.
6093 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
6095 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
6096 (float2_v2df): Add define_expand.
6098 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
6101 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
6102 op_mode in the force_to_mode call.
6104 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6106 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
6107 instead of checking each element individually.
6108 (aarch64_evpc_uzp): Likewise.
6109 (aarch64_evpc_zip): Likewise.
6110 (aarch64_evpc_ext): Likewise.
6111 (aarch64_evpc_rev): Likewise.
6112 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
6113 instead of checking each element individually. Return true without
6115 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
6116 whether all selected elements come from the same input, instead of
6117 checking each element individually. Remove calls to gen_rtx_REG,
6118 start_sequence and end_sequence and instead assert that no rtl is
6121 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6123 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
6124 order of HIGH and CONST checks.
6126 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6128 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
6129 if the destination isn't an SSA_NAME.
6131 2018-01-09 Richard Biener <rguenther@suse.de>
6133 PR tree-optimization/83668
6134 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
6136 (canonicalize_loop_form): ... here, renamed from ...
6137 (canonicalize_loop_closed_ssa_form): ... this and amended to
6138 swap successor edges for loop exit blocks to make us use
6139 the RPO order we need for initial schedule generation.
6141 2018-01-09 Joseph Myers <joseph@codesourcery.com>
6143 PR tree-optimization/64811
6144 * match.pd: When optimizing comparisons with Inf, avoid
6145 introducing or losing exceptions from comparisons with NaN.
6147 2018-01-09 Martin Liska <mliska@suse.cz>
6150 * asan.c (shadow_mem_size): Add gcc_assert.
6152 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
6154 Don't save registers in main().
6157 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
6158 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
6159 * config/avr/avr.c (avr_set_current_function): Don't error if
6160 naked, OS_task or OS_main are specified at the same time.
6161 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
6163 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
6165 * common/config/avr/avr-common.c (avr_option_optimization_table):
6166 Switch on -mmain-is-OS_task for optimizing compilations.
6168 2018-01-09 Richard Biener <rguenther@suse.de>
6170 PR tree-optimization/83572
6171 * graphite.c: Include cfganal.h.
6172 (graphite_transform_loops): Connect infinite loops to exit
6173 and remove fake edges at the end.
6175 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6177 * ipa-inline.c (edge_badness): Revert accidental checkin.
6179 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6182 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
6183 symbols; not inline clones.
6185 2018-01-09 Jakub Jelinek <jakub@redhat.com>
6188 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
6189 hard registers. Formatting fixes.
6191 PR preprocessor/83722
6192 * gcc.c (try_generate_repro): Pass
6193 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
6194 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
6197 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
6198 Kito Cheng <kito.cheng@gmail.com>
6200 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
6201 (riscv_leaf_function_p): Delete.
6202 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
6204 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6206 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
6208 (do_ifelse): New function.
6209 (do_isel): New function.
6210 (do_sub3): New function.
6211 (do_add3): New function.
6212 (do_load_mask_compare): New function.
6213 (do_overlap_load_compare): New function.
6214 (expand_compare_loop): New function.
6215 (expand_block_compare): Call expand_compare_loop() when appropriate.
6216 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
6218 (-mblock-compare-inline-loop-limit): New option.
6220 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6223 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
6224 Reverse order of second and third operands in first alternative.
6225 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
6226 of first and second elements in UNSPEC_VPERMR vector.
6227 (altivec_expand_vec_perm_le): Likewise.
6229 2018-01-08 Jeff Law <law@redhat.com>
6231 PR rtl-optimizatin/81308
6232 * tree-switch-conversion.c (cfg_altered): New file scoped static.
6233 (process_switch): If group_case_labels makes a change, then set
6235 (pass_convert_switch::execute): If a switch is converted, then
6236 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
6238 PR rtl-optimization/81308
6239 * recog.c (split_all_insns): Conditionally cleanup the CFG after
6242 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
6244 PR target/83663 - Revert r255946
6245 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
6246 generation for cases where splatting a value is not useful.
6247 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
6248 across a vec_duplicate and a paradoxical subreg forming a vector
6249 mode to a vec_concat.
6251 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6253 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
6254 -march=armv8.3-a variants.
6255 * config/arm/t-multilib: Likewise.
6256 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
6258 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6260 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
6262 (cceq_ior_compare_complement): Give it a name so I can use it, and
6263 change boolean_or_operator predicate to boolean_operator so it can
6264 be used to generate a crand.
6265 (eqne): New code iterator.
6266 (bd/bd_neg): New code_attrs.
6267 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
6268 a single define_insn.
6269 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
6270 decrement (bdnzt/bdnzf/bdzt/bdzf).
6271 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
6272 with the new names of the branch decrement patterns, and added the
6273 names of the branch decrement conditional patterns.
6275 2018-01-08 Richard Biener <rguenther@suse.de>
6277 PR tree-optimization/83563
6278 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
6281 2018-01-08 Richard Biener <rguenther@suse.de>
6284 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
6286 2018-01-08 Richard Biener <rguenther@suse.de>
6288 PR tree-optimization/83685
6289 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
6290 references to abnormals.
6292 2018-01-08 Richard Biener <rguenther@suse.de>
6295 * dwarf2out.c (output_indirect_strings): Handle empty
6296 skeleton_debug_str_hash.
6297 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
6299 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6301 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
6302 (emit_store_direct): Likewise.
6303 (arc_trampoline_adjust_address): Likewise.
6304 (arc_asm_trampoline_template): New function.
6305 (arc_initialize_trampoline): Use asm_trampoline_template.
6306 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
6307 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
6308 * config/arc/arc.md (flush_icache): Delete pattern.
6310 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6312 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
6313 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
6316 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6319 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
6320 by not USED_FOR_TARGET.
6321 (make_pass_resolve_sw_modes): Likewise.
6323 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6325 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
6328 2018-01-08 Richard Biener <rguenther@suse.de>
6331 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
6333 2018-01-08 Richard Biener <rguenther@suse.de>
6336 * match.pd ((t * 2) / 2) -> t): Add missing :c.
6338 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
6341 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
6342 basic blocks with a small number of successors.
6343 (convert_control_dep_chain_into_preds): Improve handling of
6345 (dump_predicates): Split apart into...
6346 (dump_pred_chain): ...here...
6347 (dump_pred_info): ...and here.
6348 (can_one_predicate_be_invalidated_p): Add debugging printfs.
6349 (can_chain_union_be_invalidated_p): Improve check for invalidation
6351 (uninit_uses_cannot_happen): Avoid unnecessary if
6352 convert_control_dep_chain_into_preds yielded nothing.
6354 2018-01-06 Martin Sebor <msebor@redhat.com>
6356 PR tree-optimization/83640
6357 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
6358 subtracting negative offset from size.
6359 (builtin_access::overlap): Adjust offset bounds of the access to fall
6360 within the size of the object if possible.
6362 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
6364 PR rtl-optimization/83699
6365 * expmed.c (extract_bit_field_1): Restrict the vector usage of
6366 extract_bit_field_as_subreg to cases in which the extracted
6367 value is also a vector.
6369 * lra-constraints.c (process_alt_operands): Test for the equivalence
6370 substitutions when detecting a possible reload cycle.
6372 2018-01-06 Jakub Jelinek <jakub@redhat.com>
6375 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
6376 by default if flag_selective_schedling{,2}. Formatting fixes.
6378 PR rtl-optimization/83682
6379 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
6380 if it has non-VECTOR_MODE element mode.
6381 (vec_duplicate_p): Likewise.
6384 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
6385 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
6387 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6390 * config/i386/i386-builtin.def
6391 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
6392 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
6393 Require also OPTION_MASK_ISA_AVX512F in addition to
6394 OPTION_MASK_ISA_GFNI.
6395 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
6396 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
6397 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
6398 to OPTION_MASK_ISA_GFNI.
6399 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
6400 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
6401 OPTION_MASK_ISA_AVX512BW.
6402 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
6403 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
6404 addition to OPTION_MASK_ISA_GFNI.
6405 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
6406 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
6407 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
6408 to OPTION_MASK_ISA_GFNI.
6409 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
6410 a requirement for all ISAs rather than any of them with a few
6412 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
6414 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
6415 bitmasks to be enabled with 3 exceptions, instead of requiring any
6416 enabled ISA with lots of exceptions.
6417 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
6418 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
6419 Change avx512bw in isa attribute to avx512f.
6420 * config/i386/sgxintrin.h: Add license boilerplate.
6421 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
6422 to __AVX512F__ and __AVX512VL to __AVX512VL__.
6423 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
6424 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
6426 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
6427 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
6428 temporarily sse2 rather than sse if not enabled already.
6431 * config/i386/sse.md (VI248_VLBW): Rename to ...
6432 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
6433 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
6434 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
6435 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
6436 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
6437 mode iterator instead of VI248_VLBW.
6439 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
6441 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
6442 (record_modified): Skip clobbers; add debug output.
6443 (param_change_prob): Use sreal frequencies.
6445 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6447 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
6448 punt for user-aligned variables.
6450 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6452 * tree-chrec.c (chrec_contains_symbols): Return true for
6455 2018-01-05 Sudakshina Das <sudi.das@arm.com>
6458 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
6459 of (x|y) == x for BICS pattern.
6461 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6463 PR tree-optimization/83605
6464 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
6465 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
6468 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
6470 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
6471 * config/epiphany/rtems.h: New file.
6473 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6474 Uros Bizjak <ubizjak@gmail.com>
6477 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
6478 QIreg_operand instead of register_operand predicate.
6479 * config/i386/i386.c (ix86_rop_should_change_byte_p,
6480 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
6481 comments instead of -fmitigate[-_]rop.
6483 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6486 * cgraphunit.c (symbol_table::compile): Switch to text_section
6487 before calling assembly_start debug hook.
6488 * run-rtl-passes.c (run_rtl_passes): Likewise.
6491 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6493 * tree-vrp.c (extract_range_from_binary_expr_1): Check
6494 range_int_cst_p rather than !symbolic_range_p before calling
6495 extract_range_from_multiplicative_op_1.
6497 2018-01-04 Jeff Law <law@redhat.com>
6499 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
6500 redundant test in assertion.
6502 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6504 * doc/rtl.texi: Document machine_mode wrapper classes.
6506 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6508 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
6511 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6513 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
6514 the VEC_PERM_EXPR fold to fail.
6516 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6519 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
6520 to switched_sections.
6522 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6525 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
6528 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
6531 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
6532 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
6534 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6537 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
6538 is BLKmode and bitpos not zero or mode change is needed.
6540 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6543 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
6546 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
6549 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
6550 instead of MULT rtx. Update all corresponding splitters.
6552 (*ssub<modesuffix>): Ditto.
6554 (*cmp_sadd_di): Update split patterns.
6555 (*cmp_sadd_si): Ditto.
6556 (*cmp_sadd_sidi): Ditto.
6557 (*cmp_ssub_di): Ditto.
6558 (*cmp_ssub_si): Ditto.
6559 (*cmp_ssub_sidi): Ditto.
6560 * config/alpha/predicates.md (const23_operand): New predicate.
6561 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
6562 Look for ASHIFT, not MULT inner operand.
6563 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
6565 2018-01-04 Martin Liska <mliska@suse.cz>
6567 PR gcov-profile/83669
6568 * gcov.c (output_intermediate_file): Add version to intermediate
6570 * doc/gcov.texi: Document new field 'version' in intermediate
6571 file format. Fix location of '-k' option of gcov command.
6573 2018-01-04 Martin Liska <mliska@suse.cz>
6576 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
6578 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6580 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
6582 2018-01-03 Martin Sebor <msebor@redhat.com>
6584 PR tree-optimization/83655
6585 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
6586 checking calls with invalid arguments.
6588 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6590 * tree-vect-stmts.c (vect_get_store_rhs): New function.
6591 (vectorizable_mask_load_store): Delete.
6592 (vectorizable_call): Return false for masked loads and stores.
6593 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
6594 instead of gimple_assign_rhs1.
6595 (vectorizable_load): Handle IFN_MASK_LOAD.
6596 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
6598 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6600 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
6602 (vectorizable_mask_load_store): ...here.
6603 (vectorizable_load): ...and here.
6605 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6607 * tree-vect-stmts.c (vect_build_all_ones_mask)
6608 (vect_build_zero_merge_argument): New functions, split out from...
6609 (vectorizable_load): ...here.
6611 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6613 * tree-vect-stmts.c (vect_check_store_rhs): New function,
6615 (vectorizable_mask_load_store): ...here.
6616 (vectorizable_store): ...and here.
6618 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6620 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
6622 (vectorizable_mask_load_store): ...here.
6624 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6626 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
6627 (vect_model_store_cost): Take a vec_load_store_type instead of a
6629 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
6630 (vect_model_store_cost): Take a vec_load_store_type instead of a
6632 (vectorizable_mask_load_store): Update accordingly.
6633 (vectorizable_store): Likewise.
6634 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
6636 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6638 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
6639 IFN_MASK_LOAD calls here rather than...
6640 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
6642 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6643 Alan Hayward <alan.hayward@arm.com>
6644 David Sherwood <david.sherwood@arm.com>
6646 * expmed.c (extract_bit_field_1): For vector extracts,
6647 fall back to extract_bit_field_as_subreg if vec_extract
6650 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6651 Alan Hayward <alan.hayward@arm.com>
6652 David Sherwood <david.sherwood@arm.com>
6654 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
6655 they are variable or constant sized.
6656 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
6657 slots for constant-sized data.
6659 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6660 Alan Hayward <alan.hayward@arm.com>
6661 David Sherwood <david.sherwood@arm.com>
6663 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
6664 handling COND_EXPRs with boolean comparisons, try to find a better
6665 basis for the mask type than the boolean itself.
6667 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6669 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
6670 is calculated and how it can be overridden.
6671 * genmodes.c (max_bitsize_mode_any_mode): New variable.
6672 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
6674 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
6677 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6678 Alan Hayward <alan.hayward@arm.com>
6679 David Sherwood <david.sherwood@arm.com>
6681 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
6682 Remove the mode argument.
6683 (aarch64_simd_valid_immediate): Remove the mode and inverse
6685 * config/aarch64/iterators.md (bitsize): New iterator.
6686 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
6687 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
6688 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
6689 aarch64_simd_valid_immediate.
6690 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
6691 (aarch64_reg_or_bic_imm): Likewise.
6692 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
6693 with an insn_type enum and msl with a modifier_type enum.
6694 Replace element_width with a scalar_mode. Change the shift
6695 to unsigned int. Add constructors for scalar_float_mode and
6696 scalar_int_mode elements.
6697 (aarch64_vect_float_const_representable_p): Delete.
6698 (aarch64_can_const_movi_rtx_p)
6699 (aarch64_simd_scalar_immediate_valid_for_move)
6700 (aarch64_simd_make_constant): Update call to
6701 aarch64_simd_valid_immediate.
6702 (aarch64_advsimd_valid_immediate_hs): New function.
6703 (aarch64_advsimd_valid_immediate): Likewise.
6704 (aarch64_simd_valid_immediate): Remove mode and inverse
6705 arguments. Rewrite to use the above. Use const_vec_duplicate_p
6706 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
6707 and aarch64_float_const_representable_p on the result.
6708 (aarch64_output_simd_mov_immediate): Remove mode argument.
6709 Update call to aarch64_simd_valid_immediate and use of
6710 simd_immediate_info.
6711 (aarch64_output_scalar_simd_mov_immediate): Update call
6714 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6715 Alan Hayward <alan.hayward@arm.com>
6716 David Sherwood <david.sherwood@arm.com>
6718 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
6719 (mode_nunits): Likewise CONST_MODE_NUNITS.
6720 * machmode.def (ADJUST_NUNITS): Document.
6721 * genmodes.c (mode_data::need_nunits_adj): New field.
6722 (blank_mode): Update accordingly.
6723 (adj_nunits): New variable.
6724 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
6726 (emit_mode_size_inline): Set need_bytesize_adj for all modes
6727 listed in adj_nunits.
6728 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
6729 listed in adj_nunits. Don't emit case statements for such modes.
6730 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
6731 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
6732 nothing if adj_nunits is nonnull.
6733 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
6734 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
6735 (emit_mode_fbit): Update use of print_maybe_const_decl.
6736 (emit_move_size): Likewise. Treat the array as non-const
6738 (emit_mode_adjustments): Handle adj_nunits.
6740 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6742 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
6743 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
6744 (VECTOR_MODES): Use it.
6745 (make_vector_modes): Take the prefix as an argument.
6747 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6748 Alan Hayward <alan.hayward@arm.com>
6749 David Sherwood <david.sherwood@arm.com>
6751 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
6752 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
6753 for MODE_VECTOR_BOOL.
6754 * machmode.def (VECTOR_BOOL_MODE): Document.
6755 * genmodes.c (VECTOR_BOOL_MODE): New macro.
6756 (make_vector_bool_mode): New function.
6757 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
6759 * lto-streamer-in.c (lto_input_mode_table): Likewise.
6760 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
6762 * stor-layout.c (int_mode_for_mode): Likewise.
6763 * tree.c (build_vector_type_for_mode): Likewise.
6764 * varasm.c (output_constant_pool_2): Likewise.
6765 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
6766 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
6767 for MODE_VECTOR_BOOL.
6768 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
6769 of mode class checks.
6770 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
6771 instead of a list of mode class checks.
6772 (expand_vector_scalar_condition): Likewise.
6773 (type_for_widest_vector_mode): Handle BImode as an inner mode.
6775 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6776 Alan Hayward <alan.hayward@arm.com>
6777 David Sherwood <david.sherwood@arm.com>
6779 * machmode.h (mode_size): Change from unsigned short to
6781 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
6782 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6783 or if measurement_type is not polynomial.
6784 (fixed_size_mode::includes_p): Check for constant-sized modes.
6785 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
6786 return a poly_uint16 rather than an unsigned short.
6787 (emit_mode_size): Change the type of mode_size from unsigned short
6788 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
6789 (emit_mode_adjustments): Cope with polynomial vector sizes.
6790 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6792 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6794 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
6795 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
6796 * caller-save.c (setup_save_areas): Likewise.
6797 (replace_reg_with_saved_mem): Likewise.
6798 * calls.c (emit_library_call_value_1): Likewise.
6799 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
6800 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
6801 (gen_lowpart_for_combine): Likewise.
6802 * convert.c (convert_to_integer_1): Likewise.
6803 * cse.c (equiv_constant, cse_insn): Likewise.
6804 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
6805 (cselib_subst_to_values): Likewise.
6806 * dce.c (word_dce_process_block): Likewise.
6807 * df-problems.c (df_word_lr_mark_ref): Likewise.
6808 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
6809 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
6810 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
6811 (rtl_for_decl_location): Likewise.
6812 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
6813 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
6814 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
6815 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
6816 (expand_expr_real_1): Likewise.
6817 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
6818 (pad_below): Likewise.
6819 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6820 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
6821 * ira.c (get_subreg_tracking_sizes): Likewise.
6822 * ira-build.c (ira_create_allocno_objects): Likewise.
6823 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
6824 (ira_sort_regnos_for_alter_reg): Likewise.
6825 * ira-costs.c (record_operand_costs): Likewise.
6826 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
6827 (resolve_simple_move): Likewise.
6828 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
6829 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
6830 (lra_constraints): Likewise.
6831 (CONST_POOL_OK_P): Reject variable-sized modes.
6832 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
6833 (add_pseudo_to_slot, lra_spill): Likewise.
6834 * omp-low.c (omp_clause_aligned_alignment): Likewise.
6835 * optabs-query.c (get_best_extraction_insn): Likewise.
6836 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6837 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
6838 (expand_mult_highpart, valid_multiword_target_p): Likewise.
6839 * recog.c (offsettable_address_addr_space_p): Likewise.
6840 * regcprop.c (maybe_mode_change): Likewise.
6841 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
6842 * regrename.c (build_def_use): Likewise.
6843 * regstat.c (dump_reg_info): Likewise.
6844 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
6845 (find_reloads, find_reloads_subreg_address): Likewise.
6846 * reload1.c (eliminate_regs_1): Likewise.
6847 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
6848 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
6849 (simplify_binary_operation_1, simplify_subreg): Likewise.
6850 * targhooks.c (default_function_arg_padding): Likewise.
6851 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
6852 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
6853 (verify_gimple_assign_ternary): Likewise.
6854 * tree-inline.c (estimate_move_cost): Likewise.
6855 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6856 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
6857 (get_address_cost_ainc): Likewise.
6858 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
6859 (vect_supportable_dr_alignment): Likewise.
6860 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6861 (vectorizable_reduction): Likewise.
6862 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
6863 (vectorizable_operation, vectorizable_load): Likewise.
6864 * tree.c (build_same_sized_truth_vector_type): Likewise.
6865 * valtrack.c (cleanup_auto_inc_dec): Likewise.
6866 * var-tracking.c (emit_note_insn_var_location): Likewise.
6867 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
6868 (ADDR_VEC_ALIGN): Likewise.
6870 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6871 Alan Hayward <alan.hayward@arm.com>
6872 David Sherwood <david.sherwood@arm.com>
6874 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
6876 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6877 or if measurement_type is polynomial.
6878 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
6879 * combine.c (make_extraction): Likewise.
6880 * dse.c (find_shift_sequence): Likewise.
6881 * dwarf2out.c (mem_loc_descriptor): Likewise.
6882 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
6883 (extract_bit_field, extract_low_bits): Likewise.
6884 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
6885 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
6886 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
6887 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
6888 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6889 * reload.c (find_reloads): Likewise.
6890 * reload1.c (alter_reg): Likewise.
6891 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
6892 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
6893 * tree-if-conv.c (predicate_mem_writes): Likewise.
6894 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
6895 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
6896 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
6897 * valtrack.c (dead_debug_insert_temp): Likewise.
6898 * varasm.c (mergeable_constant_section): Likewise.
6899 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
6901 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6902 Alan Hayward <alan.hayward@arm.com>
6903 David Sherwood <david.sherwood@arm.com>
6905 * expr.c (expand_assignment): Cope with polynomial mode sizes
6906 when assigning to a CONCAT.
6908 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6909 Alan Hayward <alan.hayward@arm.com>
6910 David Sherwood <david.sherwood@arm.com>
6912 * machmode.h (mode_precision): Change from unsigned short to
6914 (mode_to_precision): Return a poly_uint16 rather than an unsigned
6916 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
6917 or if measurement_type is not polynomial.
6918 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
6919 in which the mode is already known to be a scalar_int_mode.
6920 * genmodes.c (emit_mode_precision): Change the type of mode_precision
6921 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
6923 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6924 for GET_MODE_PRECISION.
6925 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6926 for GET_MODE_PRECISION.
6927 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
6929 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
6930 (expand_field_assignment, make_extraction): Likewise.
6931 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
6932 (get_last_value): Likewise.
6933 * convert.c (convert_to_integer_1): Likewise.
6934 * cse.c (cse_insn): Likewise.
6935 * expr.c (expand_expr_real_1): Likewise.
6936 * lra-constraints.c (simplify_operand_subreg): Likewise.
6937 * optabs-query.c (can_atomic_load_p): Likewise.
6938 * optabs.c (expand_atomic_load): Likewise.
6939 (expand_atomic_store): Likewise.
6940 * ree.c (combine_reaching_defs): Likewise.
6941 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
6942 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
6943 * tree.h (type_has_mode_precision_p): Likewise.
6944 * ubsan.c (instrument_si_overflow): Likewise.
6946 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6947 Alan Hayward <alan.hayward@arm.com>
6948 David Sherwood <david.sherwood@arm.com>
6950 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
6951 polynomial numbers of units.
6952 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
6953 (valid_vector_subparts_p): New function.
6954 (build_vector_type): Remove temporary shim and take the number
6955 of units as a poly_uint64 rather than an int.
6956 (build_opaque_vector_type): Take the number of units as a
6957 poly_uint64 rather than an int.
6958 * tree.c (build_vector_from_ctor): Handle polynomial
6959 TYPE_VECTOR_SUBPARTS.
6960 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
6961 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
6962 (build_vector_from_val): If the number of units is variable,
6963 use build_vec_duplicate_cst for constant operands and
6964 VEC_DUPLICATE_EXPR otherwise.
6965 (make_vector_type): Remove temporary is_constant ().
6966 (build_vector_type, build_opaque_vector_type): Take the number of
6967 units as a poly_uint64 rather than an int.
6968 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
6970 * cfgexpand.c (expand_debug_expr): Likewise.
6971 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
6972 (store_constructor, expand_expr_real_1): Likewise.
6973 (const_scalar_mask_from_tree): Likewise.
6974 * fold-const-call.c (fold_const_reduction): Likewise.
6975 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
6976 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
6977 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
6978 (fold_relational_const): Likewise.
6979 (native_interpret_vector): Likewise. Change the size from an
6980 int to an unsigned int.
6981 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
6982 TYPE_VECTOR_SUBPARTS.
6983 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
6984 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
6985 duplicating a non-constant operand into a variable-length vector.
6986 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
6987 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
6988 * ipa-icf.c (sem_variable::equals): Likewise.
6989 * match.pd: Likewise.
6990 * omp-simd-clone.c (simd_clone_subparts): Likewise.
6991 * print-tree.c (print_node): Likewise.
6992 * stor-layout.c (layout_type): Likewise.
6993 * targhooks.c (default_builtin_vectorization_cost): Likewise.
6994 * tree-cfg.c (verify_gimple_comparison): Likewise.
6995 (verify_gimple_assign_binary): Likewise.
6996 (verify_gimple_assign_ternary): Likewise.
6997 (verify_gimple_assign_single): Likewise.
6998 * tree-pretty-print.c (dump_generic_node): Likewise.
6999 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7000 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
7001 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
7002 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
7003 (vect_shift_permute_load_chain): Likewise.
7004 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
7005 (expand_vector_condition, optimize_vector_constructor): Likewise.
7006 (lower_vec_perm, get_compute_type): Likewise.
7007 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7008 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
7009 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
7010 (vect_recog_mask_conversion_pattern): Likewise.
7011 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
7012 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
7013 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7014 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
7015 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
7016 (vectorizable_shift, vectorizable_operation, vectorizable_store)
7017 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
7018 (supportable_widening_operation): Likewise.
7019 (supportable_narrowing_operation): Likewise.
7020 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
7022 * varasm.c (output_constant): Likewise.
7024 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7025 Alan Hayward <alan.hayward@arm.com>
7026 David Sherwood <david.sherwood@arm.com>
7028 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
7029 so that both the length == 3 and length != 3 cases set up their
7030 own permute vectors. Add comments explaining why we know the
7031 number of elements is constant.
7032 (vect_permute_load_chain): Likewise.
7034 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7035 Alan Hayward <alan.hayward@arm.com>
7036 David Sherwood <david.sherwood@arm.com>
7038 * machmode.h (mode_nunits): Change from unsigned char to
7040 (ONLY_FIXED_SIZE_MODES): New macro.
7041 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
7042 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
7043 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
7045 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
7046 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
7047 or if measurement_type is not polynomial.
7048 * genmodes.c (ZERO_COEFFS): New macro.
7049 (emit_mode_nunits_inline): Make mode_nunits_inline return a
7051 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
7052 Use ZERO_COEFFS when emitting initializers.
7053 * data-streamer.h (bp_pack_poly_value): New function.
7054 (bp_unpack_poly_value): Likewise.
7055 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7056 for GET_MODE_NUNITS.
7057 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7058 for GET_MODE_NUNITS.
7059 * tree.c (make_vector_type): Remove temporary shim and make
7060 the real function take the number of units as a poly_uint64
7062 (build_vector_type_for_mode): Handle polynomial nunits.
7063 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
7064 * emit-rtl.c (const_vec_series_p_1): Likewise.
7065 (gen_rtx_CONST_VECTOR): Likewise.
7066 * fold-const.c (test_vec_duplicate_folding): Likewise.
7067 * genrecog.c (validate_pattern): Likewise.
7068 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
7069 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7070 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
7071 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
7072 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
7073 * rtlanal.c (subreg_get_info): Likewise.
7074 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7075 (vect_grouped_load_supported): Likewise.
7076 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
7077 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
7078 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
7079 (simplify_const_unary_operation, simplify_binary_operation_1)
7080 (simplify_const_binary_operation, simplify_ternary_operation)
7081 (test_vector_ops_duplicate, test_vector_ops): Likewise.
7082 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
7083 instead of CONST_VECTOR_NUNITS.
7084 * varasm.c (output_constant_pool_2): Likewise.
7085 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
7086 explicit-encoded elements in the XVEC for variable-length vectors.
7088 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7090 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
7092 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7093 Alan Hayward <alan.hayward@arm.com>
7094 David Sherwood <david.sherwood@arm.com>
7096 * coretypes.h (fixed_size_mode): Declare.
7097 (fixed_size_mode_pod): New typedef.
7098 * builtins.h (target_builtins::x_apply_args_mode)
7099 (target_builtins::x_apply_result_mode): Change type to
7100 fixed_size_mode_pod.
7101 * builtins.c (apply_args_size, apply_result_size, result_vector)
7102 (expand_builtin_apply_args_1, expand_builtin_apply)
7103 (expand_builtin_return): Update accordingly.
7105 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7107 * cse.c (hash_rtx_cb): Hash only the encoded elements.
7108 * cselib.c (cselib_hash_rtx): Likewise.
7109 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
7110 CONST_VECTOR encoding.
7112 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7113 Jeff Law <law@redhat.com>
7116 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
7117 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
7118 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
7119 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
7122 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
7123 explicitly probe *sp in a noreturn function if there were any callee
7124 register saves or frame pointer is needed.
7126 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7129 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
7130 BLKmode for ternary, binary or unary expressions.
7133 * var-tracking.c (delete_vta_debug_insn): New inline function.
7134 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
7135 insns from get_insns () to NULL instead of each bb separately.
7136 Use delete_vta_debug_insn. No longer static.
7137 (vt_debug_insns_local, variable_tracking_main_1): Adjust
7138 delete_vta_debug_insns callers.
7139 * rtl.h (delete_vta_debug_insns): Declare.
7140 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
7141 instead of variable_tracking_main.
7143 2018-01-03 Martin Sebor <msebor@redhat.com>
7145 PR tree-optimization/83603
7146 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
7147 arguments past the endof the argument list in functions declared
7148 without a prototype.
7149 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7150 Avoid checking when arguments are null.
7152 2018-01-03 Martin Sebor <msebor@redhat.com>
7155 * doc/extend.texi (attribute const): Fix a typo.
7156 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
7157 issuing -Wsuggest-attribute for void functions.
7159 2018-01-03 Martin Sebor <msebor@redhat.com>
7161 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
7162 offset_int::from instead of wide_int::to_shwi.
7163 (maybe_diag_overlap): Remove assertion.
7164 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
7165 * gimple-ssa-sprintf.c (format_directive): Same.
7166 (parse_directive): Same.
7167 (sprintf_dom_walker::compute_format_length): Same.
7168 (try_substitute_return_value): Same.
7170 2018-01-03 Jeff Law <law@redhat.com>
7173 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
7174 non-constant residual for zero at runtime and avoid probing in
7175 that case. Reorganize code for trailing problem to mirror handling
7178 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7180 PR tree-optimization/83501
7181 * tree-ssa-strlen.c (get_string_cst): New.
7182 (handle_char_store): Call get_string_cst.
7184 2018-01-03 Martin Liska <mliska@suse.cz>
7186 PR tree-optimization/83593
7187 * tree-ssa-strlen.c: Include tree-cfg.h.
7188 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
7189 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
7190 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
7192 (strlen_dom_walker::before_dom_children): Call
7193 gimple_purge_dead_eh_edges. Dump tranformation with details
7195 (strlen_dom_walker::before_dom_children): Update call by adding
7196 new argument cleanup_eh.
7197 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
7199 2018-01-03 Martin Liska <mliska@suse.cz>
7202 * cif-code.def (VARIADIC_THUNK): New enum value.
7203 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
7206 2018-01-03 Jan Beulich <jbeulich@suse.com>
7208 * sse.md (mov<mode>_internal): Tighten condition for when to use
7209 vmovdqu<ssescalarsize> for TI and OI modes.
7211 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7213 Update copyright years.
7215 2018-01-03 Martin Liska <mliska@suse.cz>
7218 * ipa-visibility.c (function_and_variable_visibility): Skip
7219 functions with noipa attribure.
7221 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7223 * gcc.c (process_command): Update copyright notice dates.
7224 * gcov-dump.c (print_version): Ditto.
7225 * gcov.c (print_version): Ditto.
7226 * gcov-tool.c (print_version): Ditto.
7227 * gengtype.c (create_file): Ditto.
7228 * doc/cpp.texi: Bump @copying's copyright year.
7229 * doc/cppinternals.texi: Ditto.
7230 * doc/gcc.texi: Ditto.
7231 * doc/gccint.texi: Ditto.
7232 * doc/gcov.texi: Ditto.
7233 * doc/install.texi: Ditto.
7234 * doc/invoke.texi: Ditto.
7236 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7238 * vector-builder.h (vector_builder::m_full_nelts): Change from
7239 unsigned int to poly_uint64.
7240 (vector_builder::full_nelts): Update prototype accordingly.
7241 (vector_builder::new_vector): Likewise.
7242 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
7243 (vector_builder::operator ==): Likewise.
7244 (vector_builder::finalize): Likewise.
7245 * int-vector-builder.h (int_vector_builder::int_vector_builder):
7246 Take the number of elements as a poly_uint64 rather than an
7248 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
7249 from unsigned int to poly_uint64.
7250 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
7251 (vec_perm_indices::new_vector): Likewise.
7252 (vec_perm_indices::length): Likewise.
7253 (vec_perm_indices::nelts_per_input): Likewise.
7254 (vec_perm_indices::input_nelts): Likewise.
7255 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
7256 number of elements per input as a poly_uint64 rather than an
7257 unsigned int. Use the original encoding for variable-length
7258 vectors, rather than clamping each individual element.
7259 For the second and subsequent elements in each pattern,
7260 clamp the step and base before clamping their sum.
7261 (vec_perm_indices::series_p): Handle polynomial element counts.
7262 (vec_perm_indices::all_in_range_p): Likewise.
7263 (vec_perm_indices_to_tree): Likewise.
7264 (vec_perm_indices_to_rtx): Likewise.
7265 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
7266 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
7267 (tree_vector_builder::new_binary_operation): Handle polynomial
7268 element counts. Return false if we need to know the number
7269 of elements at compile time.
7270 * fold-const.c (fold_vec_perm): Punt if the number of elements
7271 isn't known at compile time.
7273 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7275 * vec-perm-indices.h (vec_perm_builder): Change element type
7276 from HOST_WIDE_INT to poly_int64.
7277 (vec_perm_indices::element_type): Update accordingly.
7278 (vec_perm_indices::clamp): Handle polynomial element_types.
7279 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7280 (vec_perm_indices::all_in_range_p): Likewise.
7281 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
7283 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
7284 polynomial vec_perm_indices element types.
7285 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
7286 * fold-const.c (fold_vec_perm): Likewise.
7287 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
7288 * tree-vect-generic.c (lower_vec_perm): Likewise.
7289 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7290 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
7291 element type to HOST_WIDE_INT.
7293 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7294 Alan Hayward <alan.hayward@arm.com>
7295 David Sherwood <david.sherwood@arm.com>
7297 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
7298 rather than an int. Use plus_constant.
7299 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
7300 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
7302 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7303 Alan Hayward <alan.hayward@arm.com>
7304 David Sherwood <david.sherwood@arm.com>
7306 * calls.c (emit_call_1, expand_call): Change struct_value_size from
7307 a HOST_WIDE_INT to a poly_int64.
7309 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7310 Alan Hayward <alan.hayward@arm.com>
7311 David Sherwood <david.sherwood@arm.com>
7313 * calls.c (load_register_parameters): Cope with polynomial
7314 mode sizes. Require a constant size for BLKmode parameters
7315 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
7316 forces a parameter to be padded at the lsb end in order to
7317 fill a complete number of words, require the parameter size
7318 to be ordered wrt UNITS_PER_WORD.
7320 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7321 Alan Hayward <alan.hayward@arm.com>
7322 David Sherwood <david.sherwood@arm.com>
7324 * reload1.c (spill_stack_slot_width): Change element type
7325 from unsigned int to poly_uint64_pod.
7326 (alter_reg): Treat mode sizes as polynomial.
7328 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7329 Alan Hayward <alan.hayward@arm.com>
7330 David Sherwood <david.sherwood@arm.com>
7332 * reload.c (complex_word_subreg_p): New function.
7333 (reload_inner_reg_of_subreg, push_reload): Use it.
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 * lra-constraints.c (process_alt_operands): Reject matched
7340 operands whose sizes aren't ordered.
7341 (match_reload): Refer to this check here.
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 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
7348 that the mode size is in the set {1, 2, 4, 8, 16}.
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 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
7355 Use plus_constant instead of gen_rtx_PLUS.
7357 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7358 Alan Hayward <alan.hayward@arm.com>
7359 David Sherwood <david.sherwood@arm.com>
7361 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
7362 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
7363 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
7364 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
7365 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
7366 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
7367 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
7368 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
7369 * config/i386/i386.c (ix86_push_rounding): ...this new function.
7370 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
7372 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
7373 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
7374 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
7375 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
7376 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
7377 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
7378 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
7379 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
7380 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
7381 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
7383 * expr.c (emit_move_resolve_push): Treat the input and result
7384 of PUSH_ROUNDING as a poly_int64.
7385 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
7386 (emit_push_insn): Likewise.
7387 * lra-eliminations.c (mark_not_eliminable): Likewise.
7388 * recog.c (push_operand): Likewise.
7389 * reload1.c (elimination_effects): Likewise.
7390 * rtlanal.c (nonzero_bits1): Likewise.
7391 * calls.c (store_one_arg): Likewise. Require the padding to be
7392 known at compile time.
7394 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7395 Alan Hayward <alan.hayward@arm.com>
7396 David Sherwood <david.sherwood@arm.com>
7398 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
7399 Use plus_constant instead of gen_rtx_PLUS.
7401 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7402 Alan Hayward <alan.hayward@arm.com>
7403 David Sherwood <david.sherwood@arm.com>
7405 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
7408 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7409 Alan Hayward <alan.hayward@arm.com>
7410 David Sherwood <david.sherwood@arm.com>
7412 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
7413 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
7414 via stack temporaries. Treat the mode size as polynomial too.
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 * expr.c (expand_expr_real_2): When handling conversions involving
7421 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
7422 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
7423 as a poly_uint64 too.
7425 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7426 Alan Hayward <alan.hayward@arm.com>
7427 David Sherwood <david.sherwood@arm.com>
7429 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
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 * combine.c (can_change_dest_mode): Handle polynomial
7436 REGMODE_NATURAL_SIZE.
7437 * expmed.c (store_bit_field_1): Likewise.
7438 * expr.c (store_constructor): Likewise.
7439 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
7440 and polynomial REGMODE_NATURAL_SIZE.
7441 (gen_lowpart_common): Likewise.
7442 * reginfo.c (record_subregs_of_mode): Likewise.
7443 * rtlanal.c (read_modify_subreg_p): Likewise.
7445 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7446 Alan Hayward <alan.hayward@arm.com>
7447 David Sherwood <david.sherwood@arm.com>
7449 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
7450 numbers of elements.
7452 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7453 Alan Hayward <alan.hayward@arm.com>
7454 David Sherwood <david.sherwood@arm.com>
7456 * match.pd: Cope with polynomial numbers of vector elements.
7458 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7459 Alan Hayward <alan.hayward@arm.com>
7460 David Sherwood <david.sherwood@arm.com>
7462 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
7463 in a POINTER_PLUS_EXPR.
7465 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7466 Alan Hayward <alan.hayward@arm.com>
7467 David Sherwood <david.sherwood@arm.com>
7469 * omp-simd-clone.c (simd_clone_subparts): New function.
7470 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
7471 (ipa_simd_modify_function_body): Likewise.
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 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
7478 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
7479 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
7480 (expand_vector_condition, vector_element): Likewise.
7481 (subparts_gt): New function.
7482 (get_compute_type): Use subparts_gt.
7483 (count_type_subparts): Delete.
7484 (expand_vector_operations_1): Use subparts_gt instead of
7485 count_type_subparts.
7487 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7488 Alan Hayward <alan.hayward@arm.com>
7489 David Sherwood <david.sherwood@arm.com>
7491 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
7492 (vect_compile_time_alias): ...this new function. Do the calculation
7493 on poly_ints rather than trees.
7494 (vect_prune_runtime_alias_test_list): Update call accordingly.
7496 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7497 Alan Hayward <alan.hayward@arm.com>
7498 David Sherwood <david.sherwood@arm.com>
7500 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
7502 (vect_schedule_slp_instance): Likewise.
7504 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7505 Alan Hayward <alan.hayward@arm.com>
7506 David Sherwood <david.sherwood@arm.com>
7508 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
7509 constant and extern definitions for variable-length vectors.
7510 (vect_get_constant_vectors): Note that the number of units
7511 is known to be constant.
7513 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7514 Alan Hayward <alan.hayward@arm.com>
7515 David Sherwood <david.sherwood@arm.com>
7517 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
7518 of units as polynomial. Choose between WIDE and NARROW based
7521 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7522 Alan Hayward <alan.hayward@arm.com>
7523 David Sherwood <david.sherwood@arm.com>
7525 * tree-vect-stmts.c (simd_clone_subparts): New function.
7526 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
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_call): Treat the number of
7533 vectors as polynomial. Use build_index_vector for
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 (get_load_store_type): Treat the number of
7541 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
7542 for variable-length vectors.
7543 (vectorizable_mask_load_store): Treat the number of units as
7544 polynomial, asserting that it is constant if the condition has
7545 already been enforced.
7546 (vectorizable_store, vectorizable_load): Likewise.
7548 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7549 Alan Hayward <alan.hayward@arm.com>
7550 David Sherwood <david.sherwood@arm.com>
7552 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
7553 of units as polynomial. Punt if we can't tell at compile time
7554 which vector contains the final result.
7556 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7557 Alan Hayward <alan.hayward@arm.com>
7558 David Sherwood <david.sherwood@arm.com>
7560 * tree-vect-loop.c (vectorizable_induction): Treat the number
7561 of units as polynomial. Punt on SLP inductions. Use an integer
7562 VEC_SERIES_EXPR for variable-length integer reductions. Use a
7563 cast of such a series for variable-length floating-point
7566 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7567 Alan Hayward <alan.hayward@arm.com>
7568 David Sherwood <david.sherwood@arm.com>
7570 * tree.h (build_index_vector): Declare.
7571 * tree.c (build_index_vector): New function.
7572 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
7573 of units as polynomial, forcibly converting it to a constant if
7574 vectorizable_reduction has already enforced the condition.
7575 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
7576 to create a {1,2,3,...} vector.
7577 (vectorizable_reduction): Treat the number of units as polynomial.
7578 Choose vectype_in based on the largest scalar element size rather
7579 than the smallest number of units. Enforce the restrictions
7582 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7583 Alan Hayward <alan.hayward@arm.com>
7584 David Sherwood <david.sherwood@arm.com>
7586 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
7587 number of units as polynomial.
7589 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7590 Alan Hayward <alan.hayward@arm.com>
7591 David Sherwood <david.sherwood@arm.com>
7593 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
7594 * target.def (autovectorize_vector_sizes): Return the vector sizes
7595 by pointer, using vector_sizes rather than a bitmask.
7596 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
7597 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
7598 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
7600 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
7601 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
7602 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
7603 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
7604 * omp-general.c (omp_max_vf): Likewise.
7605 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7606 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
7607 * tree-vect-loop.c (vect_analyze_loop): Likewise.
7608 * tree-vect-slp.c (vect_slp_bb): Likewise.
7609 * doc/tm.texi: Regenerate.
7610 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
7612 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
7613 the vector size as a poly_uint64 rather than an unsigned int.
7614 (current_vector_size): Change from an unsigned int to a poly_uint64.
7615 (get_vectype_for_scalar_type): Update accordingly.
7616 * tree.h (build_truth_vector_type): Take the size and number of
7617 units as a poly_uint64 rather than an unsigned int.
7618 (build_vector_type): Add a temporary overload that takes
7619 the number of units as a poly_uint64 rather than an unsigned int.
7620 * tree.c (make_vector_type): Likewise.
7621 (build_truth_vector_type): Take the number of units as a poly_uint64
7622 rather than an unsigned int.
7624 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7625 Alan Hayward <alan.hayward@arm.com>
7626 David Sherwood <david.sherwood@arm.com>
7628 * target.def (get_mask_mode): Take the number of units and length
7629 as poly_uint64s rather than unsigned ints.
7630 * targhooks.h (default_get_mask_mode): Update accordingly.
7631 * targhooks.c (default_get_mask_mode): Likewise.
7632 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
7633 * doc/tm.texi: Regenerate.
7635 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7636 Alan Hayward <alan.hayward@arm.com>
7637 David Sherwood <david.sherwood@arm.com>
7639 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
7640 * omp-general.c (omp_max_vf): Likewise.
7641 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
7642 (expand_omp_simd): Handle polynomial safelen.
7643 * omp-low.c (omplow_simd_context): Add a default constructor.
7644 (omplow_simd_context::max_vf): Change from int to poly_uint64.
7645 (lower_rec_simd_input_clauses): Update accordingly.
7646 (lower_rec_input_clauses): Likewise.
7648 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7649 Alan Hayward <alan.hayward@arm.com>
7650 David Sherwood <david.sherwood@arm.com>
7652 * tree-vectorizer.h (vect_nunits_for_cost): New function.
7653 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
7654 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
7655 (vect_analyze_slp_cost): Likewise.
7656 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
7657 (vect_model_load_cost): Likewise.
7659 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7660 Alan Hayward <alan.hayward@arm.com>
7661 David Sherwood <david.sherwood@arm.com>
7663 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
7664 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
7665 from an unsigned int * to a poly_uint64_pod *.
7666 (calculate_unrolling_factor): New function.
7667 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
7669 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7670 Alan Hayward <alan.hayward@arm.com>
7671 David Sherwood <david.sherwood@arm.com>
7673 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
7674 from an unsigned int to a poly_uint64.
7675 (_loop_vec_info::slp_unrolling_factor): Likewise.
7676 (_loop_vec_info::vectorization_factor): Change from an int
7678 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
7679 (vect_get_num_vectors): New function.
7680 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
7681 (vect_get_num_copies): Use vect_get_num_vectors.
7682 (vect_analyze_data_ref_dependences): Change max_vf from an int *
7683 to an unsigned int *.
7684 (vect_analyze_data_refs): Change min_vf from an int * to a
7686 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7687 than an unsigned HOST_WIDE_INT.
7688 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7689 (vect_analyze_data_ref_dependence): Change max_vf from an int *
7690 to an unsigned int *.
7691 (vect_analyze_data_ref_dependences): Likewise.
7692 (vect_compute_data_ref_alignment): Handle polynomial vf.
7693 (vect_enhance_data_refs_alignment): Likewise.
7694 (vect_prune_runtime_alias_test_list): Likewise.
7695 (vect_shift_permute_load_chain): Likewise.
7696 (vect_supportable_dr_alignment): Likewise.
7697 (dependence_distance_ge_vf): Take the vectorization factor as a
7698 poly_uint64 rather than an unsigned HOST_WIDE_INT.
7699 (vect_analyze_data_refs): Change min_vf from an int * to a
7701 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
7702 vfm1 as a poly_uint64 rather than an int. Make the same change
7703 for the returned bound_scalar.
7704 (vect_gen_vector_loop_niters): Handle polynomial vf.
7705 (vect_do_peeling): Likewise. Update call to
7706 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
7707 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
7709 * tree-vect-loop.c (vect_determine_vectorization_factor)
7710 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
7711 (vect_get_known_peeling_cost): Likewise.
7712 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
7713 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
7714 (vect_transform_loop): Likewise. Use the lowest possible VF when
7715 updating the upper bounds of the loop.
7716 (vect_min_worthwhile_factor): Make static. Return an unsigned int
7718 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
7719 polynomial unroll factors.
7720 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
7721 (vect_make_slp_decision): Likewise.
7722 (vect_supported_load_permutation_p): Likewise, and polynomial
7724 (vect_analyze_slp_cost): Handle polynomial vf.
7725 (vect_slp_analyze_node_operations): Likewise.
7726 (vect_slp_analyze_bb_1): Likewise.
7727 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7728 than an unsigned HOST_WIDE_INT.
7729 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
7730 (vectorizable_load): Handle polynomial vf.
7731 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
7733 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
7735 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7736 Alan Hayward <alan.hayward@arm.com>
7737 David Sherwood <david.sherwood@arm.com>
7739 * match.pd: Handle bit operations involving three constants
7740 and try to fold one pair.
7742 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7744 * tree-vect-loop-manip.c: Include gimple-fold.h.
7745 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
7746 niters_maybe_zero parameters. Handle other cases besides a step of 1.
7747 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
7748 Add a path that uses a step of VF instead of 1, but disable it
7750 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
7751 and niters_no_overflow parameters. Update calls to
7752 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
7753 Create a new SSA name if the latter choses to use a ste other
7754 than zero, and return it via niters_vector_mult_vf_var.
7755 * tree-vect-loop.c (vect_transform_loop): Update calls to
7756 vect_do_peeling, vect_gen_vector_loop_niters and
7757 slpeel_make_loop_iterate_ntimes.
7758 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
7759 (vect_gen_vector_loop_niters): Update declarations after above changes.
7761 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
7763 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
7764 128-bit round to integer instructions.
7765 (ceil<mode>2): Likewise.
7766 (btrunc<mode>2): Likewise.
7767 (round<mode>2): Likewise.
7769 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7771 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
7772 unaligned VSX load/store on P8/P9.
7773 (expand_block_clear): Allow the use of unaligned VSX
7774 load/store on P8/P9.
7776 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7778 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
7780 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
7781 swap associated with both a load and a store.
7783 2018-01-02 Andrew Waterman <andrew@sifive.com>
7785 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
7786 * config/riscv/riscv.md (clear_cache): Use it.
7788 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
7790 * web.c: Remove out-of-date comment.
7792 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7794 * expr.c (fixup_args_size_notes): Check that any existing
7795 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
7796 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
7797 (emit_single_push_insn): ...here.
7799 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7801 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
7802 (const_vector_encoded_nelts): New function.
7803 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
7804 (const_vector_int_elt, const_vector_elt): Declare.
7805 * emit-rtl.c (const_vector_int_elt_1): New function.
7806 (const_vector_elt): Likewise.
7807 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
7808 of CONST_VECTOR_ELT.
7810 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7812 * expr.c: Include rtx-vector-builder.h.
7813 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
7814 directly on the tree encoding.
7815 (const_vector_from_tree): Likewise.
7816 * optabs.c: Include rtx-vector-builder.h.
7817 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
7818 sequence of "u" values.
7819 * vec-perm-indices.c: Include rtx-vector-builder.h.
7820 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
7821 directly on the vec_perm_indices encoding.
7823 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7825 * doc/rtl.texi (const_vector): Describe new encoding scheme.
7826 * Makefile.in (OBJS): Add rtx-vector-builder.o.
7827 * rtx-vector-builder.h: New file.
7828 * rtx-vector-builder.c: Likewise.
7829 * rtl.h (rtx_def::u2): Add a const_vector field.
7830 (CONST_VECTOR_NPATTERNS): New macro.
7831 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
7832 (CONST_VECTOR_DUPLICATE_P): Likewise.
7833 (CONST_VECTOR_STEPPED_P): Likewise.
7834 (CONST_VECTOR_ENCODED_ELT): Likewise.
7835 (const_vec_duplicate_p): Check for a duplicated vector encoding.
7836 (unwrap_const_vec_duplicate): Likewise.
7837 (const_vec_series_p): Check for a non-duplicated vector encoding.
7838 Say that the function only returns true for integer vectors.
7839 * emit-rtl.c: Include rtx-vector-builder.h.
7840 (gen_const_vec_duplicate_1): Delete.
7841 (gen_const_vector): Call gen_const_vec_duplicate instead of
7842 gen_const_vec_duplicate_1.
7843 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
7844 (gen_const_vec_duplicate): Use rtx_vector_builder.
7845 (gen_const_vec_series): Likewise.
7846 (gen_rtx_CONST_VECTOR): Likewise.
7847 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
7848 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7849 Build a new vector rather than modifying a CONST_VECTOR in-place.
7850 (handle_special_swappables): Update call accordingly.
7851 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
7852 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7853 Build a new vector rather than modifying a CONST_VECTOR in-place.
7854 (handle_special_swappables): Update call accordingly.
7856 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7858 * simplify-rtx.c (simplify_const_binary_operation): Use
7859 CONST_VECTOR_ELT instead of XVECEXP.
7861 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7863 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
7864 the selector elements to be different from the data elements
7865 if the selector is a VECTOR_CST.
7866 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
7867 ssizetype for the selector.
7869 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7871 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
7872 before testing each element individually.
7873 * tree-vect-generic.c (lower_vec_perm): Likewise.
7875 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7877 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
7878 * selftest-run-tests.c (selftest::run_tests): Call it.
7879 * vector-builder.h (vector_builder::operator ==): New function.
7880 (vector_builder::operator !=): Likewise.
7881 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
7882 (vec_perm_indices::all_from_input_p): New function.
7883 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7884 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
7885 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
7886 instead of reading the VECTOR_CST directly. Detect whether both
7887 vector inputs are the same before constructing the vec_perm_indices,
7888 and update the number of inputs argument accordingly. Use the
7889 utility functions added above. Only construct sel2 if we need to.
7891 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7893 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
7894 the broadcast of the low byte.
7895 (expand_mult_highpart): Use an explicit encoding for the permutes.
7896 * optabs-query.c (can_mult_highpart_p): Likewise.
7897 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
7898 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7899 (vectorizable_bswap): Likewise.
7900 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
7901 explicit encoding for the power-of-2 permutes.
7902 (vect_permute_store_chain): Likewise.
7903 (vect_grouped_load_supported): Likewise.
7904 (vect_permute_load_chain): Likewise.
7906 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7908 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
7909 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
7910 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
7911 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7912 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
7913 (vect_gen_perm_mask_any): Likewise.
7915 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7917 * int-vector-builder.h: New file.
7918 * vec-perm-indices.h: Include int-vector-builder.h.
7919 (vec_perm_indices): Redefine as an int_vector_builder.
7920 (auto_vec_perm_indices): Delete.
7921 (vec_perm_builder): Redefine as a stand-alone class.
7922 (vec_perm_indices::vec_perm_indices): New function.
7923 (vec_perm_indices::clamp): Likewise.
7924 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
7925 (vec_perm_indices::new_vector): New function.
7926 (vec_perm_indices::new_expanded_vector): Update for new
7927 vec_perm_indices class.
7928 (vec_perm_indices::rotate_inputs): New function.
7929 (vec_perm_indices::all_in_range_p): Operate directly on the
7930 encoded form, without computing elided elements.
7931 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
7932 encoding. Update for new vec_perm_indices class.
7933 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
7934 the given vec_perm_builder.
7935 (expand_vec_perm_var): Update vec_perm_builder constructor.
7936 (expand_mult_highpart): Use vec_perm_builder instead of
7937 auto_vec_perm_indices.
7938 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
7939 vec_perm_indices instead of auto_vec_perm_indices. Use a single
7940 or double series encoding as appropriate.
7941 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
7942 vec_perm_indices instead of auto_vec_perm_indices.
7943 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7944 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7945 (vect_permute_store_chain): Likewise.
7946 (vect_grouped_load_supported): Likewise.
7947 (vect_permute_load_chain): Likewise.
7948 (vect_shift_permute_load_chain): Likewise.
7949 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
7950 (vect_transform_slp_perm_load): Likewise.
7951 (vect_schedule_slp_instance): Likewise.
7952 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7953 (vectorizable_mask_load_store): Likewise.
7954 (vectorizable_bswap): Likewise.
7955 (vectorizable_store): Likewise.
7956 (vectorizable_load): Likewise.
7957 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
7958 vec_perm_indices instead of auto_vec_perm_indices. Use
7959 tree_to_vec_perm_builder to read the vector from a tree.
7960 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
7961 vec_perm_builder instead of a vec_perm_indices.
7962 (have_whole_vector_shift): Use vec_perm_builder and
7963 vec_perm_indices instead of auto_vec_perm_indices. Leave the
7964 truncation to calc_vec_perm_mask_for_shift.
7965 (vect_create_epilog_for_reduction): Likewise.
7966 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
7967 from auto_vec_perm_indices to vec_perm_indices.
7968 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7969 instead of changing individual elements.
7970 (aarch64_vectorize_vec_perm_const): Use new_vector to install
7971 the vector in d.perm.
7972 * config/arm/arm.c (expand_vec_perm_d::perm): Change
7973 from auto_vec_perm_indices to vec_perm_indices.
7974 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7975 instead of changing individual elements.
7976 (arm_vectorize_vec_perm_const): Use new_vector to install
7977 the vector in d.perm.
7978 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
7979 Update vec_perm_builder constructor.
7980 (rs6000_expand_interleave): Likewise.
7981 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
7982 (rs6000_expand_interleave): Likewise.
7984 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7986 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
7987 to qimode could truncate the indices.
7988 * optabs.c (expand_vec_perm_var): Likewise.
7990 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7992 * Makefile.in (OBJS): Add vec-perm-indices.o.
7993 * vec-perm-indices.h: New file.
7994 * vec-perm-indices.c: Likewise.
7995 * target.h (vec_perm_indices): Replace with a forward class
7997 (auto_vec_perm_indices): Move to vec-perm-indices.h.
7998 * optabs.h: Include vec-perm-indices.h.
7999 (expand_vec_perm): Delete.
8000 (selector_fits_mode_p, expand_vec_perm_var): Declare.
8001 (expand_vec_perm_const): Declare.
8002 * target.def (vec_perm_const_ok): Replace with...
8003 (vec_perm_const): ...this new hook.
8004 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
8005 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
8006 * doc/tm.texi: Regenerate.
8007 * optabs.def (vec_perm_const): Delete.
8008 * doc/md.texi (vec_perm_const): Likewise.
8009 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
8010 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
8011 expand_vec_perm for constant permutation vectors. Assert that
8012 the mode of variable permutation vectors is the integer equivalent
8013 of the mode that is being permuted.
8014 * optabs-query.h (selector_fits_mode_p): Declare.
8015 * optabs-query.c: Include vec-perm-indices.h.
8016 (selector_fits_mode_p): New function.
8017 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
8018 is defined, instead of checking whether the vec_perm_const_optab
8019 exists. Use targetm.vectorize.vec_perm_const instead of
8020 targetm.vectorize.vec_perm_const_ok. Check whether the indices
8021 fit in the vector mode before using a variable permute.
8022 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
8023 vec_perm_indices instead of an rtx.
8024 (expand_vec_perm): Replace with...
8025 (expand_vec_perm_const): ...this new function. Take the selector
8026 as a vec_perm_indices rather than an rtx. Also take the mode of
8027 the selector. Update call to shift_amt_for_vec_perm_mask.
8028 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
8029 Use vec_perm_indices::new_expanded_vector to expand the original
8030 selector into bytes. Check whether the indices fit in the vector
8031 mode before using a variable permute.
8032 (expand_vec_perm_var): Make global.
8033 (expand_mult_highpart): Use expand_vec_perm_const.
8034 * fold-const.c: Includes vec-perm-indices.h.
8035 * tree-ssa-forwprop.c: Likewise.
8036 * tree-vect-data-refs.c: Likewise.
8037 * tree-vect-generic.c: Likewise.
8038 * tree-vect-loop.c: Likewise.
8039 * tree-vect-slp.c: Likewise.
8040 * tree-vect-stmts.c: Likewise.
8041 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
8043 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
8044 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
8045 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
8046 (aarch64_vectorize_vec_perm_const): ...this new function.
8047 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8048 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8049 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
8050 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
8051 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8052 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8053 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
8055 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
8056 check for NEON modes.
8057 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
8058 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
8059 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
8060 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
8062 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
8063 the old VEC_PERM_CONST conditions.
8064 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
8065 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
8066 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
8067 (ia64_vectorize_vec_perm_const_ok): Merge into...
8068 (ia64_vectorize_vec_perm_const): ...this new function.
8069 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
8070 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
8071 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
8072 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
8073 * config/mips/mips.c (mips_expand_vec_perm_const)
8074 (mips_vectorize_vec_perm_const_ok): Merge into...
8075 (mips_vectorize_vec_perm_const): ...this new function.
8076 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
8077 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
8078 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
8079 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
8080 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
8081 (rs6000_expand_vec_perm_const): Delete.
8082 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
8084 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8085 (altivec_expand_vec_perm_const_le): Take each operand individually.
8086 Operate on constant selectors rather than rtxes.
8087 (altivec_expand_vec_perm_const): Likewise. Update call to
8088 altivec_expand_vec_perm_const_le.
8089 (rs6000_expand_vec_perm_const): Delete.
8090 (rs6000_vectorize_vec_perm_const_ok): Delete.
8091 (rs6000_vectorize_vec_perm_const): New function.
8092 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8093 an element count and rtx array.
8094 (rs6000_expand_extract_even): Update call accordingly.
8095 (rs6000_expand_interleave): Likewise.
8096 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
8097 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
8098 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
8099 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
8100 (rs6000_expand_vec_perm_const): Delete.
8101 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8102 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8103 (altivec_expand_vec_perm_const_le): Take each operand individually.
8104 Operate on constant selectors rather than rtxes.
8105 (altivec_expand_vec_perm_const): Likewise. Update call to
8106 altivec_expand_vec_perm_const_le.
8107 (rs6000_expand_vec_perm_const): Delete.
8108 (rs6000_vectorize_vec_perm_const_ok): Delete.
8109 (rs6000_vectorize_vec_perm_const): New function. Remove stray
8110 reference to the SPE evmerge intructions.
8111 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8112 an element count and rtx array.
8113 (rs6000_expand_extract_even): Update call accordingly.
8114 (rs6000_expand_interleave): Likewise.
8115 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
8116 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
8118 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8120 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8122 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
8123 vector mode and that that mode matches the mode of the data
8125 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
8126 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
8127 directly using expand_vec_perm_1 when forcing selectors into
8129 (expand_vec_perm_var): New function, split out from expand_vec_perm.
8131 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8133 * optabs-query.h (can_vec_perm_p): Delete.
8134 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
8135 * optabs-query.c (can_vec_perm_p): Split into...
8136 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
8137 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
8138 particular selector is valid.
8139 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8140 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8141 (vect_grouped_load_supported): Likewise.
8142 (vect_shift_permute_load_chain): Likewise.
8143 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8144 (vect_transform_slp_perm_load): Likewise.
8145 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8146 (vectorizable_bswap): Likewise.
8147 (vect_gen_perm_mask_checked): Likewise.
8148 * fold-const.c (fold_ternary_loc): Likewise. Don't take
8149 implementations of variable permutation vectors into account
8150 when deciding which selector to use.
8151 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
8152 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
8153 with a false third argument.
8154 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
8155 to test whether the constant selector is valid and can_vec_perm_var_p
8156 to test whether a variable selector is valid.
8158 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8160 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
8161 * optabs-query.c (can_vec_perm_p): Likewise.
8162 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
8163 instead of vec_perm_indices.
8164 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
8165 (vect_gen_perm_mask_checked): Likewise,
8166 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
8167 (vect_gen_perm_mask_checked): Likewise,
8169 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8171 * optabs-query.h (qimode_for_vec_perm): Declare.
8172 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
8173 (qimode_for_vec_perm): ...this new function.
8174 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
8176 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8178 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
8179 does not have a conditional at the top.
8181 2018-01-02 Richard Biener <rguenther@suse.de>
8183 * ipa-inline.c (big_speedup_p): Fix expression.
8185 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8188 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
8191 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8195 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
8196 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
8197 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
8198 cond_taken_branch_cost 3->4.
8200 2018-01-01 Jakub Jelinek <jakub@redhat.com>
8202 PR tree-optimization/83581
8203 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
8204 TODO_cleanup_cfg if any changes have been made.
8207 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
8208 convert_modes if target mode has the right side, but different mode
8212 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
8213 last argument when extracting from CONCAT. If either from_real or
8214 from_imag is NULL, use expansion through memory. If result is not
8215 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
8216 the parts directly to inner mode, if even that fails, use expansion
8220 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
8221 check for bswap in mode rather than HImode and use that in expand_unop
8224 Copyright (C) 2018 Free Software Foundation, Inc.
8226 Copying and distribution of this file, with or without modification,
8227 are permitted in any medium without royalty provided the copyright
8228 notice and this notice are preserved.