1 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3 PR rtl-optimization/85342
4 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
5 a bool scalar var inside of the loop instead. Don't try to update
6 recog_data.operand after failed apply_change_group.
8 2018-04-12 Tom de Vries <tom@codesourcery.com>
11 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
12 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
13 array with flexible array member as array without given dimension.
14 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
15 argument for undefined param to true.
17 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
20 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
21 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
23 * config/rs6000/sysv4.opt (mcall-): Improve help text.
24 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
25 help text that is too long.
26 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
27 help text that is too long.
28 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
29 help text that is too long.
31 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
33 * config/alpha/alpha.md (stack_probe_internal): Rename
34 from "probe_stack". Update all callers.
36 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
38 PR rtl-optimization/84566
39 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
40 sched_macro_fuse_insns.
42 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
45 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
46 (compute_block_dependences): ... from here.
48 2018-04-11 Jakub Jelinek <jakub@redhat.com>
50 PR tree-optimization/85331
51 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
52 from int to HOST_WIDE_INT.
54 2018-04-11 Martin Jambor <mjambor@suse.cz>
57 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
58 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
59 not the same as the source val.
60 (cgraph_edge_brings_value_p): New parameter.
61 (gather_edges_for_value): Pass destination value to
62 cgraph_edge_brings_value_p.
63 (perhaps_add_new_callers): Likewise.
64 (get_info_about_necessary_edges): Likewise and exclude values brought
65 only by self-recursive edges.
66 (create_specialized_node): Redirect only clones of self-calling edges.
67 (+self_recursive_pass_through_p): New function.
68 (find_more_scalar_values_for_callers_subset): Use it.
69 (find_aggregate_values_for_callers_subset): Likewise.
70 (known_aggs_to_agg_replacement_list): Removed.
71 (decide_whether_version_node): Re-calculate known constants for all
72 remaining context clones.
74 2018-04-11 Richard Biener <rguenther@suse.de>
77 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
78 from early DWARF output.
79 (dwarf2out_early_finish): Output line info unconditionally into
80 early DWARF and add reference to it.
82 2018-04-11 Jakub Jelinek <jakub@redhat.com>
85 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
86 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
87 other than V2DFmode using iptr mode attribute.
88 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
90 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
92 PR rtl-optimization/84659
93 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
95 2018-04-11 Jakub Jelinek <jakub@redhat.com>
98 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
100 (output_loc_list): Pass address of a dummy size variable even in the
101 locview handling loop.
102 (index_location_lists): Add comment on why skip_loc_list_entry can't
105 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
108 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
111 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
114 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
115 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
116 and -mstring-compare-inline-limit.
118 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
121 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
122 for stack clash protection in a register whenever we need it to be in
125 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
127 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
128 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
130 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
133 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
135 (mlong-double-): Ditto.
136 * config/rs6000/sysv4.opt (msdata=): Ditto.
139 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
141 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
142 erroneous entries for
143 "vector int vec_ldl (int, long int *)", and
144 "vector unsigned int vec_ldl (int, unsigned long int *)".
145 Add comments and entries for
146 "vector bool char vec_ldl (int, bool char *)",
147 "vector bool short vec_ldl (int, bool short *)",
148 "vector bool int vec_ldl (int, bool int *)",
149 "vector bool long long vec_ldl (int, bool long long *)",
150 "vector pixel vec_ldl (int, pixel *)",
151 "vector long long vec_ldl (int, long long *)",
152 "vector unsigned long long vec_ldl (int, unsigned long long *)".
153 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
154 type tree bool_long_long_type_node and correct definition of
155 bool_V2DI_type_node to make reference to this new type tree.
156 (rs6000_mangle_type): Replace erroneous reference to
157 bool_long_type_node with bool_long_long_type_node.
158 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
159 comments to emphasize sign distinctions for char and int types and
160 replace RS6000_BTI_bool_long constant with
161 RS6000_BTI_bool_long_long constant. Also add comment to restrict
162 use of RS6000_BTI_pixel.
163 (bool_long_type_node): Remove this macro definition.
164 (bool_long_long_type_node): New macro definition
166 2018-04-10 Jakub Jelinek <jakub@redhat.com>
168 PR rtl-optimization/85300
169 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
170 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
171 simplify_unary_operation fails.
173 2018-04-10 Martin Liska <mliska@suse.cz>
175 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
176 cgraph_edge and ipa_ref.
178 2018-04-10 Jakub Jelinek <jakub@redhat.com>
183 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
184 computation of the VEC_MERGE selector from mask.
185 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
186 Fix decoding of the VEC_MERGE selector into mask.
188 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
190 PR tree-optimization/85286
191 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
193 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
195 * final.c (final_1): Set insn_last_address as well as
196 insn_current_address.
198 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
201 * explow.c (emit_stack_probe): Call validize_mem on memory location
202 before passing it to gen_probe_stack. Create address operand and
203 legitimize it for the probe_stack_address case.
205 2018-04-09 Jan Hubicka <jh@suse.cz>
208 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
209 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
210 * tree.c (free_lang_data_in_type): Fix handling of binfos;
212 (free_lang_data): Rebuild type inheritance graph.
214 2018-04-09 Martin Sebor <msebor@redhat.com>
216 * invoke.texi (-finline-small-functions): Mention other optimization
218 (-findirect-inlining, -fpartial-inlining): Same.
219 (-finline-functions-called-once): Same.
220 (-freorder-blocks-and-partition): Same.
222 2018-04-09 Jan Hubicka <jh@suse.cz>
225 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
226 jumps; choose last target that matches the criteria (i.e.
227 no partition changes for non-crossing jumps).
228 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
229 support for redirecting crossing jumps to non-crossing.
231 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
233 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
234 also for naked functions.
236 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
238 * config/arc/arc.md (add_shift): New pattern.
239 (add_shift2): Likewise.
240 (sub_shift): Likewise.
241 (sub_shift_cmp0_noout): Likewise.
242 (compare_si_ashiftsi): Likewise.
243 (xbfu_cmp0_noout): New combine pattern.
244 (xbfu_cmp0"): Likewise.
245 (movsi_set_cc_insn): Place the predicable variant first.
246 (commutative_binary_cmp0_noout): Remove clobber.
247 (commutative_binary_cmp0): New pattern.
248 (noncommutative_binary_cmp0): Likewise.
249 (noncommutative_binary_cmp0_noout): Likewise.
250 (noncommutative_binary_comparison_result_used): Removed.
251 (rsub_cmp0): New pattern.
252 (rsub_cmp0_noout): Likewise.
253 (extzvsi): Changed, keep only meaningful variants.
254 (SQH, SEZ): New iterators.
255 (SQH_postfix): New mode attribute.
256 (SEZ_prefix): New code attribute.
257 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
258 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
259 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
261 (noncommutative_operator): Check the availability of barrel
264 2018-04-09 Richard Biener <rguenther@suse.de>
266 PR tree-optimization/85284
267 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
268 Only use the niter constraining form of simple_iv when the exit
271 2018-04-09 Tom de Vries <tom@codesourcery.com>
274 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
275 (define_expand "*memory_barrier"): New define_expand.
276 (define_insn "memory_barrier"): New insn.
278 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
280 PR rtl-optimization/80463
281 PR rtl-optimization/83972
282 PR rtl-optimization/83480
284 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
285 correct producer for the insn.
286 (tidy_control_flow): Fixup seqnos in case of debug insns.
288 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
290 PR rtl-optimization/83913
292 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
293 different sched-times when merging exprs.
295 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
297 PR rtl-optimization/83962
299 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
300 tidy_fallthru_edge and tidy_control_flow.
302 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
304 PR rtl-optimization/83530
306 * sel-sched.c (force_next_insn): New global variable.
307 (remove_insn_for_debug): When force_next_insn is true, also leave only
308 next insn in the ready list.
309 (sel_sched_region): When the region wasn't scheduled, make another pass
310 over it with force_next_insn set to 1.
312 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
314 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
316 * config/nds32/constants.md (unspec_volatile_element): Add enum values
317 for interrupt control.
318 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
319 functions for interrupt control.
320 * config/nds32/nds32-intrinsic.md: Likewise.
321 * config/nds32/nds32_intrinsic.h: Likewise.
322 * config/nds32/nds32.h (nds32_builtins): Likewise.
324 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
326 * config/nds32/nds32.c (nds32_init_machine_status,
327 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
328 strict_aligned_p field.
329 (nds32_expand_to_rtl_hook): New function.
330 (TARGET_EXPAND_TO_RTL_HOOK): Define.
331 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
333 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
334 Chung-Ju Wu <jasonwucj@gmail.com>
336 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
337 * config/nds32/nds32-n7.md: New file.
338 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
339 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
341 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
342 * config/nds32/nds32.md (pipeline_model): Add n7.
343 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
344 * config/nds32/pipelines.md: Include n7 settings.
346 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
347 Chung-Ju Wu <jasonwucj@gmail.com>
349 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
350 * config/nds32/nds32-e8.md: New file.
351 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
352 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
354 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
355 * config/nds32/nds32.md (pipeline_model): Add e8.
356 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
357 * config/nds32/pipelines.md: Include e8 settings.
359 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
360 Chung-Ju Wu <jasonwucj@gmail.com>
362 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
363 * config/nds32/nds32-n8.md: New file.
364 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
365 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
367 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
368 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
369 * config/nds32/nds32.md (pipeline_model): Add n8.
370 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
371 * config/nds32/pipelines.md: Include n8 settings.
373 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
374 Chung-Ju Wu <jasonwucj@gmail.com>
376 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
377 * config/nds32/nds32-n9-2r1w.md: New file.
378 * config/nds32/nds32-n9-3r2w.md: New file.
379 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
380 nds32_register_ports): New or modify for cpu n9.
381 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
383 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
384 * config/nds32/nds32-utils.c: New file.
385 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
386 TARGET_MUL_SLOW): Define.
387 * config/nds32/nds32.md (pipeline_model): New attribute.
388 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
389 New options that support cpu n9.
390 * config/nds32/pipelines.md: Include n9 settings.
391 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
393 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
395 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
396 information if necessary.
397 (output_cond_branch_compare_zero): Likewise.
398 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
399 (nds32_target_alignment): Refine for alignment.
400 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
401 (FUNCTION_BOUNDARY): Modify.
402 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
404 * config/nds32/nds32.opt (malways-align, malign-functions): New.
406 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
408 * config/nds32/constants.md (unspec_volatile_element): Add values for
409 TLB operation and data prefetch.
410 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
411 functions for TLB operation and data prefetch.
412 * config/nds32/nds32-intrinsic.md: Likewise.
413 * config/nds32/nds32_intrinsic.h: Likewise.
414 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
415 (nds32_print_operand): Likewise.
416 * config/nds32/nds32.h (nds32_builtins): Likewise.
418 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
419 Andrew Pinski <pinsika@gcc.gnu.org>
422 * match.pd: Use constant_boolean_node of correct type instead of
423 boolean_true_node or boolean_false_node for simplifying
424 pointer comparisons to zero.
426 2018-04-07 Jakub Jelinek <jakub@redhat.com>
428 PR tree-optimization/80021
429 * tree.c (verify_type_variant): Make error call in verify_variant_match
430 translatable and remove final full stop.
432 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
434 * config/nds32/constants.md (unspec_volatile_element): Add
435 UNSPEC_VOLATILE_EH_RETURN.
436 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
437 nds32_output_stack_pop): Support dwarf exception handling process.
438 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
439 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
440 exception handling process.
441 (nds32_compute_stack_frame): Likewise.
442 (nds32_return_addr_rtx): Likewise.
443 (nds32_initial_elimination_offset): Likewise.
444 (nds32_expand_prologue): Likewise.
445 (nds32_expand_epilogue): Likewise.
446 (nds32_dynamic_chain_address): New function.
447 * config/nds32/nds32.h (machine_function): Add fields for dwarf
449 (DYNAMIC_CHAIN_ADDRESS): Define.
450 (EH_RETURN_DATA_REGNO): Define.
451 (EH_RETURN_STACKADJ_RTX): Define.
452 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
453 patterns for dwarf exception handling.
455 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
457 * config/nds32/nds32.h: Clean up obsolete macros.
459 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
461 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
462 Add enum values for particular instructions.
463 * config/nds32/nds32-intrinsic.c: Implementation of expanding
464 particular intrinsic functions.
465 * config/nds32/nds32-intrinsic.md: Likewise.
466 * config/nds32/nds32_intrinsic.h: Likewise.
467 * config/nds32/nds32.h (nds32_builtins): Likewise.
468 * config/nds32/nds32.md (type): Add pbsad and pbsada.
469 (btst, ave): New patterns for particular instructions.
471 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
473 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
474 Add enum values for atomic load/store and memory sync.
475 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
477 * config/nds32/nds32-intrinsic.md: Likewise.
478 * config/nds32/nds32_intrinsic.h: Likewise.
479 * config/nds32/nds32.h (nds32_builtins): Likewise.
481 2018-04-07 Jakub Jelinek <jakub@redhat.com>
483 PR tree-optimization/85257
484 * fold-const.c (native_encode_vector): If not all elts could fit
485 and off is -1, return 0 rather than offset.
486 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
487 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
488 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
489 adjust buffer in native_interpret_expr call.
491 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
493 * config/nds32/constants.md (unspec_volatile_element): Add cache
495 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
496 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
497 * config/nds32/nds32.c (nds32_cctl_names): New.
498 (nds32_print_operand): Handle cache control register names.
499 * config/nds32/nds32.h (nds32_builtins): New enum values.
500 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
502 * config/nds32/nds32.md (type): Add mmu.
503 * config/nds32/pipelines.md (simple_insn): Add mmu.
505 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
507 * config/nds32/nds32.md (type): Remove call.
508 * config/nds32/pipelines.md (simple_insn): Likewise.
510 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
512 * config/nds32/constants.md (unspec_volatile_element): Add
513 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
514 UNSPEC_VOLATILE_FMFCFG.
515 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
516 description for fmfcfg and fmfcsr.
517 (bdesc_1arg): Add fmtcsr.
518 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
519 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
520 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
521 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
522 unspec_fmfcfg): New patterns.
523 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
524 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
525 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
526 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
527 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
528 __nds32__fmfcfg): Define.
530 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
532 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
533 intrinsic register names.
534 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
535 intrinsic register enum values and macros.
537 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
539 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
540 for load/store addressing form.
541 (nds32_print_operand_address): Likewise.
543 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
546 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
547 based on LABEL_REF. Remove useless assertion.
548 (pic_address_needs_scratch): Fix formatting.
549 (sparc_legitimize_pic_address): Minor tweaks.
550 (sparc_delegitimize_address): Adjust assertion accordingly.
551 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
552 into symbolic_operand.
553 (movsi_high_pic_label_ref): Likewise.
554 (movsi_lo_sum_pic_label_ref): Likewise.
555 (movdi_pic_label_ref): Likewise.
556 (movdi_high_pic_label_ref): Likewise.
557 (movdi_lo_sum_pic_label_ref): Likewise.
559 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
561 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
562 custom LIB_SPEC setup.
564 2018-04-06 Ruslan Bukin <br@bsdpad.com>
565 Kito Cheng <kito.cheng@gmail.com>
567 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
568 * config/riscv/freebsd.h: New.
570 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
572 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
573 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
576 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
577 Kito Cheng <kito.cheng@gmail.com>
579 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
580 nds32_output_call, nds32_symbol_binds_local_p): New functions.
581 * config/nds32/nds32-protos.h (nds32_output_call,
582 nds32_output_return): Declare.
583 * config/nds32/nds32.md: Refine all the call and return patterns.
585 2018-04-06 Jakub Jelinek <jakub@redhat.com>
588 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
589 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
591 PR rtl-optimization/84872
592 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
593 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
596 2018-04-06 Tamar Christina <tamar.christina@arm.com>
598 * expr.c (copy_blkmode_to_reg): Revert 254862.
599 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
601 2018-04-06 Richard Biener <rguenther@suse.de>
604 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
605 after seeing a component reference with an adjacent field. Treat
606 refs to arrays at struct end of external decls similar to
607 refs to unconstrained commons.
609 2018-04-06 Jakub Jelinek <jakub@redhat.com>
612 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
613 look through SAVE_EXPRs with non-side-effects argument. Adjust
615 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
618 2018-04-06 Richard Biener <rguenther@suse.de>
621 * alias.c (find_base_term): New wrapper around find_base_term
622 unwinding CSELIB_VAL_PTR changes.
623 (find_base_term): Do not restore CSELIB_VAL_PTR during the
626 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
628 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
630 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
631 constant definitions.
632 ("nop"): lr 0,0 -> nopr r0
633 ("nop_lr0", "nop_lr1"): New insn definitions.
635 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
637 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
638 NDS32_V3PUSH_AVAILABLE_P macro.
640 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
641 Chung-Ju Wu <jasonwucj@gmail.com>
643 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
644 (nds32*-*-*): Add float and fpu_config into supported_defaults.
645 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
646 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
647 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
648 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
649 * config/nds32/constraints.md: New constraints and checking for hard
651 * config/nds32/iterators.md: New mode iterator and attribute for hard
653 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
655 * config/nds32/nds32-fpu.md: New file.
656 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
657 deal with hard float code generation.
658 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
660 (abi_type, float_reg_number): New enum type.
661 * config/nds32/nds32-predicates.c: New predicates for hard float.
662 * config/nds32/nds32-protos.h: Declare functions for hard float.
663 * config/nds32/nds32.c: Implementation for hard float configuration.
664 * config/nds32/nds32.h: Definitions for hard float configuration.
665 * config/nds32/nds32.md: Include hard float machine description and
666 modify patterns for hard float configuration.
667 * config/nds32/nds32.opt: New options for hard float configuration.
668 * config/nds32/predicates.md: New predicates for hard float
671 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
673 * common/config/nds32/nds32-common.c
674 (nds32_option_optimization_table): Enable -mreleax-hint by default.
676 2018-04-05 Jakub Jelinek <jakub@redhat.com>
679 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
680 CONSTRUCTOR_ELT (ctor, ...)->value.
682 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
685 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
687 2018-04-05 Tom de Vries <tom@codesourcery.com>
690 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
693 2018-04-05 Shiva Chen <shiva0217@gmail.com>
694 Kito Cheng <kito.cheng@gmail.com>
696 * config/nds32/constraints.md (U33): Fine-tune checking condition.
697 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
698 * config/nds32/nds32.h (nds32_16bit_address_type): Add
699 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
701 2018-04-05 Shiva Chen <shiva0217@gmail.com>
702 Kito Cheng <kito.cheng@gmail.com>
704 * config/nds32/constraints.md (Ufe): New memory constraint.
705 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
706 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
707 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
709 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
710 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
712 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
714 * config/nds32/nds32.md: Use optimize_size in the condition for
715 alu-shift instructions.
717 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
719 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
721 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
723 * config/nds32/nds32.md (negsi2): Refine pattern.
725 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
726 Chung-Ju Wu <jasonwucj@gmail.com>
728 * config/nds32/iterators.md (shift_rotate): New code iterator.
729 (shift): New code attribute.
730 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
731 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
732 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
733 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
735 (andsi3, *andsi3): Ditto.
736 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
737 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
738 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
739 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
740 nds32_ior_operand, nds32_xor_operand): New predicates.
742 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
744 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
745 (addsi3, subsi3): ... this.
747 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
749 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
751 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
753 * config/nds32/nds32.md: Adjust indention.
755 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
757 * config/nds32/nds32.md (feature): New attribute.
759 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
761 * config/nds32/nds32.md (subtype): New attribute.
763 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
766 * config/arm/arm-builtins.c (arm_expand_builtin): Change
767 expansion to perform a bitwise AND of the argument followed by a
768 boolean negation of the result.
770 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
772 PR rtl-optimization/84878
773 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
774 the basic block. Assert the use reference is not artificial and that
775 it has an associated insn.
777 2018-04-04 Michael Matz <matz@suse.de>
779 * builtins.c (compute_objsize): Pass correct operand
780 to array_at_struct_end_p.
782 2018-04-04 Richard Biener <rguenther@suse.de>
785 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
786 from contexts for DINFO_LEVEL_TERSE and below.
788 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
790 * config/nds32/nds32-doubleword.md (move_<mode>): Require
791 resiter_operand condition.
792 * config/nds32/nds32.md (*move<mode>): Ditto.
794 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
795 Monk Chiang <sh.chiang04@gmail.com>
797 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
799 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
801 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
803 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
804 Kito Cheng <kito.cheng@gmail.com>
806 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
807 nds32_cond_code_str, output_cond_branch,
808 output_cond_branch_compare_zero, nds32_expand_cbranch,
809 nds32_expand_cstore, nds32_expand_movcc,
810 nds32_output_cbranchsi4_equality_zero,
811 nds32_output_cbranchsi4_equality_reg,
812 nds32_output_cbranchsi4_equality_reg_or_const_int,
813 nds32_output_cbranchsi4_greater_less_zero: New functions.
814 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
815 nds32_expand_cstore, nds32_expand_movcc,
816 nds32_output_cbranchsi4_equality_zero,
817 nds32_output_cbranchsi4_equality_reg,
818 nds32_output_cbranchsi4_equality_reg_or_const_int,
819 nds32_output_cbranchsi4_greater_less_zero): Declare.
820 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
821 nds32_rimm11s_operand): New predicates.
822 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
823 * config/nds32/nds32.md: Rewrite all the branch and conditional move
826 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
828 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
829 * config/nds32/nds32.md: Ditto.
830 * config/nds32/pipelines.md: Ditto.
832 2018-04-04 Richard Biener <rguenther@suse.de>
834 PR tree-optimization/85168
835 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
836 propagating abnormals.
838 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
840 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
842 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
843 Kito Cheng <kito.cheng@gmail.com>
845 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
846 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
847 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
848 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
849 * config/nds32/nds32.md (sibcall_internal): New.
850 (sibcall_register): Remove.
851 (sibcall_immediate): Remove.
852 (sibcall_value_internal): New.
853 (sibcall_value_register): Remove.
854 (sibcall_value_immediate): Remove.
855 * config/nds32/predicates.md (nds32_general_register_operand): New.
856 (nds32_call_address_operand): New.
858 2018-04-03 Jakub Jelinek <jakub@redhat.com>
860 PR rtl-optimization/85167
861 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
862 bb_defs if *split_p, instead preinitialize it to NULL.
864 PR tree-optimization/85156
865 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
866 evaluating the argument multiple times.
868 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
870 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
872 (_mm_cvtpd_ps): Likewise.
873 (_mm_cvttpd_epi32): Likewise.
874 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
875 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
876 vector, pixel, and bool following altivec.h include.
878 2018-04-03 Martin Sebor <msebor@redhat.com>
880 * doc/extend.texi (Common Function Attributes): Clarify.
881 (const attribute): Likewise.
882 (pure attribute): Likewise.
884 2018-04-03 Jakub Jelinek <jakub@redhat.com>
887 * config/i386/i386.c (ix86_expand_vector_set): Use
888 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
890 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
892 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
893 instructions when changing rounding bits to preserve precision bits
894 in the x87 control word.
896 2018-04-03 Martin Liska <mliska@suse.cz>
898 PR tree-optimization/82491
899 * rtl.h (strip_offset_and_add): Replace += suboffset with
900 poly_uint64 () + suboffset.
902 2018-03-29 Martin Liska <mliska@suse.cz>
903 Martin Jambor <mjambor@suse.cz>
906 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
907 param_type is not an integral or pointer type.
909 2018-04-03 Richard Biener <rguenther@suse.de>
911 * sese.h (recompute_all_dominators): Remove.
913 2018-04-02 Martin Sebor <msebor@redhat.com>
915 * doc/invoke.texi (-Wrestrict): Fix typos.
917 2018-04-02 Jim Wilson <jimw@sifive.com>
919 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
920 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
921 (<optab>di3, <optab>si3_extend): Likewise.
922 (<optab>si3_mask, <optab>si3_mask_1): New.
923 (<optab>di3_mask, <optab>di3_mask_1): New.
924 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
925 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
926 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
928 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
930 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
933 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
935 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
936 (nds32_canonicalize_comparison): New function.
938 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
939 Kito Cheng <kito.cheng@gmail.com>
940 Kuan-Lin Chen <kuanlinchentw@gmail.com>
942 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
943 * config/nds32/constants.md (unspec_volatile_element): Add
944 UNSPEC_VOLATILE_RELAX_GROUP.
945 * config/nds32/nds32-relax-opt.c: New file.
946 * config/nds32/nds32-predicates.c
947 (nds32_symbol_load_store_p): New function.
948 * config/nds32/nds32-protos.h
949 (nds32_symbol_load_store_p): Declare function.
950 (make_pass_nds32_relax_opt): Declare new rtl pass function.
951 * config/nds32/nds32.c
952 (nds32_register_pass): New function to register pass.
953 (nds32_register_passes): New function to register passes.
954 * config/nds32/nds32.md (relax_group): New pattern.
955 * config/nds32/nds32.opt (mrelax-hint): New option.
956 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
958 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
960 * config/nds32/t-nds32: Modify files dependency.
962 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
964 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
965 (PROFILE_HOOK): Define its implementation.
967 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
969 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
970 type and 32-bit size.
972 2018-04-01 Jakub Jelinek <jakub@redhat.com>
975 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
976 (V_128_256): New mode iterator.
977 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
978 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
979 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
981 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
982 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
984 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
987 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
988 NaN inputs correctly.
990 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
993 * config/rs6000/vsx.md (??r): New mode attribute.
994 (*vsx_mov<mode>_64bit): Use it.
995 (*vsx_mov<mode>_32bit): Likewise.
997 2018-03-30 Martin Sebor <msebor@redhat.com>
999 PR tree-optimization/84818
1000 * builtins.c (check_access): Use warning_n.
1002 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1005 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
1007 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
1010 2018-03-30 Julia Koval <julia.koval@intel.com>
1013 * x86-tune.def (movx, partial_reg_dependency): Enable for
1016 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
1019 * lra-constraints.c (process_alt_operands): Move setting
1020 this_alternative_matches below.
1022 2018-03-29 Martin Liska <mliska@suse.cz>
1025 * doc/invoke.texi: Document how LTO works with debug info.
1026 Describe auto-load support of binutils. Mention 'x86-64'
1027 as valid option value of -march option.
1029 2018-03-29 Jakub Jelinek <jakub@redhat.com>
1031 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
1034 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
1035 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
1036 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
1039 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1042 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
1043 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
1044 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
1045 for RS6000_BTM_POWERPC64.
1046 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
1047 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
1048 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
1053 2018-03-28 Carl Love <cel@us.ibm.com>
1056 2017-09-27 Carl Love <cel@us.ibm.com>
1058 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
1059 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
1060 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
1063 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1065 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
1066 instead of __vector bool.
1067 (_mm_max_pu8): Likewise.
1068 (_mm_min_pi16): Likewise.
1070 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1073 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
1074 (DIVWEUO): Likewise.
1076 (DIVDEUO): Likewise.
1077 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
1078 DIVWEUO and DIVDEUO.
1079 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
1080 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
1081 (div_extend): Likewise.
1082 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
1084 (__builtin_divweuo): Likewise.
1085 (__builtin_divdeo): Likewise.
1086 (__builtin_divdeuo): Likewise.
1088 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1091 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
1092 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
1094 PR tree-optimization/82004
1095 * gimple-match-head.c (optimize_pow_to_exp): New function.
1096 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
1097 Don't fold to exp if optimize_pow_to_exp is false.
1099 2018-03-28 Martin Liska <mliska@suse.cz>
1102 * calls.c (initialize_argument_information): Fix trailing space.
1103 * common.opt: Fix typo and provide better explanation for
1104 -fsanitize-coverage option.
1105 * config/i386/i386.opt: Fix typo.
1107 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1108 Martin Liska <mliska@suse.cz>
1111 * gimplify.c (asan_poison_variable): Don't do the check for
1112 gimplify_omp_ctxp here.
1113 (gimplify_decl_expr): Do it here.
1114 (gimplify_target_expr): Likewise.
1116 2018-03-28 Martin Liska <mliska@suse.cz>
1119 * config/i386/i386.c (ix86_function_arg_advance): Do not call
1120 chkp_type_bounds_count if MPX is not enabled.
1122 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
1124 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
1126 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
1129 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
1130 function to create the function decl for complex long double
1131 multiply and divide for -mabi=ieeelongdouble.
1132 (init_float128_ieee): Call it.
1134 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
1137 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
1138 -fcf-protection=branch -mibt.
1139 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
1141 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1144 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
1146 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
1149 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
1150 extern array declarations.
1152 2018-03-27 Richard Biener <rguenther@suse.de>
1155 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
1156 explicit single_use checks.
1158 2018-03-27 Richard Biener <rguenther@suse.de>
1160 PR tree-optimization/85082
1161 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
1164 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1166 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
1167 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
1168 Turn on fasynchronous-unwind-tables and funwind-tables.
1170 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
1173 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
1174 (*bmi_blsr_<mode>_ccz): Ditto.
1176 2018-03-26 Tom de Vries <tom@codesourcery.com>
1178 PR tree-optimization/85063
1179 * omp-general.c (offloading_function_p): New function. Factor out
1181 * omp-offload.c (pass_omp_target_link::gate): ... here.
1182 * omp-general.h (offloading_function_p): Declare.
1183 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
1184 with attribute omp declare target for offloading functions.
1186 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
1188 PR tree-optimization/84005
1189 * tree-data-ref.h (get_base_for_alignment): Declare.
1190 * tree-data-ref.c (get_base_for_alignment_1): New function.
1191 (get_base_for_alignment): Likewise.
1192 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
1193 get_base_for_alignment to find a suitable base object, instead
1194 of always using drb->base_address.
1196 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1199 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
1200 known size by default.
1202 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
1205 * lra-constraints.c (process_alt_operands): Don't match BLKmode
1206 and non BLKmode operands.
1208 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1211 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
1212 Clean up attributes.
1214 2018-03-23 Richard Biener <rguenther@suse.de>
1217 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
1218 we are going to emit early debug for LTO.
1220 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1223 * function.c (match_asm_constraints_1): Don't optimize if input
1224 doesn't satisfy general_operand predicate for output's mode.
1227 * alias.c (write_dependence_p): Don't require for x_canonicalized
1228 non-VOIDmode if x has VOIDmode.
1231 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
1232 just don't try to optimize it rather than assert it never happens.
1234 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1236 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
1237 macro expansions for definition of ST_INTERNAL_<mode> and
1238 LD_INTERNAL_<mode> builtins.
1239 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
1241 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
1243 (altivec_expand_st_builtin): Likewise.
1244 (altivec_expand_builtin): Remove calls to deleted functions.
1245 (rs6000_address_for_altivec): Delete this function.
1246 * config/rs6000/vector.md: Remove expands for
1247 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
1249 2018-03-22 Sudakshina Das <sudi.das@arm.com>
1252 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
1253 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
1254 re-computing once computed.
1255 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
1256 (arm_init_machine_status): Initialize
1257 machine->static_chain_stack_bytes.
1259 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1262 * doc/extend.texi: Add four new prototypes for vec_ld.
1263 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
1264 definitions for more logical presentation.
1265 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
1266 entries for V1TI variants of __builtin_altivec_ld builtin.
1267 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
1268 handling of V1TI variant of LVX icode pattern.
1269 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
1270 (rs6000_gimple_fold_builtin): Likewise.
1271 (altivec_init_builtins): Add code to define
1272 __builtin_altivec_lvx_v1ti function.
1274 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1277 * function.c (match_asm_constraints_1): Don't do the optimization
1278 if input isn't a REG, SUBREG, MEM or constant.
1280 2018-03-22 Tom de Vries <tom@codesourcery.com>
1282 PR tree-optimization/84956
1283 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
1284 bb_has_abnormal_pred.
1286 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1289 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
1290 DECL_INITIAL (decl) to decl at the end.
1291 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
1294 2018-03-21 Joseph Myers <joseph@codesourcery.com>
1296 * doc/extend.texi (__builtin_tgmath): Document when complex
1297 integer types are treated as _Complex _Float64.
1299 2018-03-21 Tom de Vries <tom@codesourcery.com>
1301 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
1303 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1305 PR tree-optimization/84960
1306 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
1307 if it is ENTRY block, move them into single succ of ENTRY in that case.
1309 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
1311 PR tree-optimization/84811
1312 * poly-int.h (poly_span_traits): Remove the T3 parameter and
1313 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
1314 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
1315 (known_subrange_p): Update accordingly. Cast each value involved
1316 in the size comparison, rather than casting the result of the
1319 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1321 PR tree-optimization/84982
1322 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
1323 by flipping the least significant bit rather than all bits from
1324 bitpos to bitpos + bitsize - 1.
1326 2018-03-21 Nathan Sidwell <nathan@acm.org>
1328 * doc/extend.texi (Deprecated Features): Remove mention of
1329 long-deleted deprecations.
1331 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1334 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
1335 * configure: Regenerate.
1337 2018-03-21 Tom de Vries <tom@codesourcery.com>
1339 PR tree-optimization/83126
1340 * tree-parloops.c (num_phis): New function.
1341 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
1343 2018-03-21 Nathan Sidwell <nathan@acm.org>
1345 * doc/extend.texi (Deprecated Features): Update deprecated flags,
1346 mention anon-struct/union members and trailing attributes.
1348 2018-03-21 Bin Cheng <bin.cheng@arm.com>
1350 PR tree-optimization/84969
1351 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
1352 builtin memset partitions if they set different rhs values.
1354 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1356 PR rtl-optimization/84989
1357 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
1358 VEC_DUPLICATE with scalar result mode.
1360 2018-03-21 Martin Liska <mliska@suse.cz>
1363 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
1364 not intended return statement.
1366 2018-03-21 Martin Liska <mliska@suse.cz>
1369 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
1370 (chkp_find_bound_slots_1): Limit number of iterations.
1372 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
1375 * Minor grammar fixes for x86 options.
1377 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1380 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
1381 holding REG_CFA_RESTORE notes, instead turn them into a USE.
1383 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
1386 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
1387 (altivec_lvx_<mode>_1op): Likewise.
1388 (altivec_stvx_<mode>_2op): Likewise.
1389 (altivec_stvx_<mode>_1op): Likewise.
1390 (altivec_lvx_<VM2:mode>): New define_expand.
1391 (altivec_stvx_<VM2:mode>): Likewise.
1392 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
1393 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1394 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
1395 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1396 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
1397 (rs6000_gen_lvx): Likewise.
1398 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
1399 (altivec_expand_stv_builtin): Likewise.
1400 (altivec_expand_builtin): Likewise.
1401 * config/rs6000/vector.md: Likewise.
1403 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1406 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
1409 2018-03-20 Richard Biener <rguenther@suse.de>
1412 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
1413 sign-conversions as zero, fall back to standard scalar_stmt
1416 2018-03-20 Martin Liska <mliska@suse.cz>
1419 * predict.c (rebuild_frequencies): Handle case when we have
1420 PROFILE_ABSENT, but flag_guess_branch_prob is false.
1422 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1425 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
1426 flag_section_anchors.
1427 * varasm.c (use_blocks_for_decl_p): Remove hack for
1428 dw2_force_const_mem.
1431 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
1433 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
1434 be created, use lowpart_subreg of operands[0] rather than operands[0]
1436 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
1437 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
1438 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
1439 and n constraint instead of aarch64_shift_imm_di and Usd.
1440 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
1441 (*aarch64_<optab>_reg_minus<mode>3): ... this.
1443 2018-03-20 Sudakshina Das <sudi.das@arm.com>
1446 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
1447 to favor GPR over NEON registers.
1448 (<shift>di3_neon): Likewise.
1450 2018-03-20 Tom de Vries <tom@codesourcery.com>
1453 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
1454 (nvptx_process_pars): Emit bar.sync asap and alap.
1456 2018-03-20 Tom de Vries <tom@codesourcery.com>
1459 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
1460 seen_label if seen_label is already set.
1462 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1465 * config/i386/i386.c (fold_builtin_cpu): For features above 31
1466 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
1467 Use 1U instead of 1. Formatting fixes.
1470 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
1471 instead of TREE_TYPE (s1) for the return value.
1473 2018-03-19 Jakub Jelinek <jakub@redhat.com>
1475 PR tree-optimization/84946
1476 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
1477 bitsize + bitsize in poly_uint64 rather than poly_int64.
1480 * dwarf2asm.c: Include fold-const.c.
1481 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
1482 of decl rather than decl itself.
1484 PR rtl-optimization/84643
1485 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
1487 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
1490 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
1491 calling assemble_variable.
1493 2018-03-19 Sudakshina Das <sudi.das@arm.com>
1496 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
1497 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
1499 2018-03-19 Jim Wilson <jimw@sifive.com>
1502 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
1503 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
1504 (riscv_first_stack_step): Likewise.
1505 (riscv_option_override): Use STACK_BOUNDARY instead of
1507 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
1509 (BIGGEST_ALIGNMENT): Set to 128.
1510 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
1511 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
1514 2018-03-19 Richard Biener <rguenther@suse.de>
1516 PR tree-optimization/84933
1517 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
1518 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
1520 2018-03-19 Richard Biener <rguenther@suse.de>
1522 PR tree-optimization/84859
1523 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
1524 (cond_if_else_store_replacement): Perform sinking operation on
1525 single-store BBs regardless of MAX_STORES_TO_SINK setting.
1526 Generalize what a BB with a single eligible store is.
1528 2018-03-19 Richard Biener <rguenther@suse.de>
1530 PR tree-optimization/84929
1531 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
1532 chrec_is_positive against non-chrec arg.
1534 2018-03-19 Tamar Christina <tamar.christina@arm.com>
1537 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
1539 2018-03-18 Martin Liska <mliska@suse.cz>
1541 PR rtl-optimization/84635
1542 * regrename.c (build_def_use): Use matches_mode only when
1545 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
1547 PR tree-optimization/84913
1548 * tree-vect-loop.c (vectorizable_reduction): Don't try to
1549 vectorize chains of COND_EXPRs.
1551 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1553 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
1555 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1557 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
1559 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1561 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
1563 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1564 Kito Cheng <kito.cheng@gmail.com>
1566 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
1567 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
1568 (nds32_adjust_reg_alloc_order): New function.
1569 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
1571 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
1573 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
1574 nds32_print_operand, nds32_print_operand_address): Use
1575 HOST_WIDE_INT_PRINT_DEC instead.
1577 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1579 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
1581 2018-03-17 Jakub Jelinek <jakub@redhat.com>
1584 * config/i386/i386.c (initial_ix86_tune_features,
1585 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
1587 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
1588 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
1589 rather than 1u << ix86_tune. Formatting fix.
1590 (ix86_option_override_internal): Change ix86_arch_mask from
1591 unsigned int to unsigned HOST_WIDE_INT, initialize to
1592 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
1593 (ix86_function_specific_restore): Likewise.
1595 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1598 * postreload.c (reload_combine_recognize_pattern): Perform
1599 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
1600 truncate_int_for_mode the result for the destination's mode.
1603 * hsa-gen.c (mem_type_for_type): Fix comment typo.
1604 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
1606 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1609 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
1612 * lra-assigns.c (lra_split_hard_reg_for): Don't use
1613 regno_allocno_class_array and sorted_pseudos.
1614 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
1615 insns where regno is used.
1617 2018-03-16 Martin Liska <mliska@suse.cz>
1620 * multiple_target.c (create_dispatcher_calls): Redirect
1621 reference in the symbol table.
1623 2018-03-16 Martin Liska <mliska@suse.cz>
1626 * multiple_target.c (create_dispatcher_calls): Redirect also
1629 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1633 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
1634 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
1635 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
1637 2018-03-16 Julia Koval <julia.koval@intel.com>
1639 * doc/invoke.texi (Skylake Server): Add CLWB.
1640 Cannonlake): Remove CLWB.
1642 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1644 PR tree-optimization/84841
1645 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
1647 (FLOAT_ONE_CONST_TYPE): Define.
1648 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
1649 (sort_by_operand_rank): Put entries with higher constant_type last
1650 rather than first to match comments.
1652 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
1654 * config/nios2/nios2.md (movsi_internal): Fix thinko in
1657 2018-03-15 Jakub Jelinek <jakub@redhat.com>
1660 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
1661 check and use address of target always.
1663 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
1666 * config/i386/i386.c (indirect_thunk_needed): Update comments.
1667 (indirect_thunk_bnd_needed): Likewise.
1668 (indirect_thunks_used): Likewise.
1669 (indirect_thunks_bnd_used): Likewise.
1670 (indirect_return_needed): New.
1671 (indirect_return_bnd_needed): Likewise.
1672 (output_indirect_thunk_function): Add a bool argument for
1674 (output_indirect_thunk_function): Don't generate alias for
1675 function return thunk.
1676 (ix86_code_end): Call output_indirect_thunk_function to generate
1677 function return thunks.
1678 (ix86_output_function_return): Set indirect_return_bnd_needed
1679 and indirect_return_needed instead of indirect_thunk_bnd_needed
1680 and indirect_thunk_needed.
1682 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
1684 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
1685 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
1686 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
1688 2018-03-15 David Malcolm <dmalcolm@redhat.com>
1689 Paul Hua <paul.hua.gm@gmail.com>
1692 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
1694 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
1696 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
1697 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
1700 2018-03-15 Tamar Christina <tamar.christina@arm.com>
1703 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
1704 instead of GET_MODE_SIZE when comparing Units.
1706 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1709 * varasm.c (hash_section): Return an unchangeble hash value
1710 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
1711 Return !aarch64_can_use_per_function_literal_pools_p ().
1713 2018-03-15 Jakub Jelinek <jakub@redhat.com>
1716 * optabs.c (emit_conditional_move): Pass address of cmode's copy
1717 rather than address of cmode as last argument to prepare_cmp_insn.
1719 2018-03-15 Julia Koval <julia.koval@intel.com>
1721 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
1722 F_AVX512VNNI, F_AVX512BITALG): New.
1724 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
1727 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
1728 insn for floating-point loads and stores.
1730 2018-03-14 Carl Love <cel@us.ibm.com>
1732 * config/rs6000/rs6000-c.c: Add macro definitions for
1733 ALTIVEC_BUILTIN_VEC_PERMXOR.
1734 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
1735 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
1736 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
1737 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
1739 * config/doc/extend.texi: Add prototypes for vec_permxor.
1741 2018-03-14 David Malcolm <dmalcolm@redhat.com>
1744 * diagnostic-show-locus.c (class layout_point): Convert m_line
1745 from int to linenum_type.
1746 (line_span::comparator): Use linenum "compare" function when
1747 comparing line numbers.
1748 (test_line_span): New function.
1749 (layout_range::contains_point): Convert param "row" from int to
1751 (layout_range::intersects_line_p): Likewise.
1752 (layout::will_show_line_p): Likewise.
1753 (layout::print_source_line): Likewise.
1754 (layout::should_print_annotation_line_p): Likewise.
1755 (layout::print_annotation_line): Likewise.
1756 (layout::print_leading_fixits): Likewise.
1757 (layout::annotation_line_showed_range_p): Likewise.
1758 (struct line_corrections): Likewise for field m_row.
1759 (line_corrections::line_corrections): Likewise for param "row".
1760 (layout::print_trailing_fixits): Likewise.
1761 (layout::get_state_at_point): Likewise.
1762 (layout::get_x_bound_for_row): Likewise.
1763 (layout::print_line): Likewise.
1764 (diagnostic_show_locus): Likewise for locals "last_line" and
1766 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
1767 * input.c (selftest::test_linenum_comparisons): New function.
1768 (selftest::input_c_tests): Call it.
1769 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
1770 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
1771 * selftest.h (ASSERT_GT): New macro.
1772 (ASSERT_GT_AT): New macro.
1773 (ASSERT_LT): New macro.
1774 (ASSERT_LT_AT): New macro.
1776 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
1778 PR rtl-optimization/84780
1779 * combine.c (distribute_links): Don't make a link based on pc_rtx.
1781 2018-03-14 Martin Liska <mliska@suse.cz>
1783 * tree.c (record_node_allocation_statistics): Use
1784 get_stats_node_kind.
1785 (get_stats_node_kind): New function extracted from
1786 record_node_allocation_statistics.
1787 (free_node): Use get_stats_node_kind.
1789 2018-03-14 Richard Biener <rguenther@suse.de>
1791 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
1792 that the value-set of ANTIC_IN doesn't grow.
1795 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1797 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1798 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1800 2018-03-14 Julia Koval <julia.koval@intel.com>
1802 * config.gcc (icelake-client, icelake-server): New.
1804 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
1805 (initial_ix86_arch_features): Ditto.
1806 (PTA_SKYLAKE): Add SGX.
1807 (PTA_ICELAKE): Remove.
1808 (PTA_ICELAKE_CLIENT): New.
1809 (PTA_ICELAKE_SERVER): New.
1810 (ix86_option_override_internal): Split up icelake on icelake client and
1812 (get_builtin_code_for_version): Ditto.
1813 (fold_builtin_cpu): Ditto.
1814 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
1815 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
1816 * config/i386/i386.h (processor_type): Ditto.
1817 * doc/invoke.texi: Ditto.
1819 2018-03-14 Jakub Jelinek <jakub@redhat.com>
1822 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
1823 INTEGER_CST offset, add it together with bitpos / 8 and
1824 sign extend based on POINTER_SIZE.
1828 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
1831 * config/i386/constraints.md (Yc): New register constraint.
1832 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
1833 Use Yc constraint for alternative 2 of operand 0. Remove
1834 preferred_for_speed attribute.
1836 2018-03-14 Richard Biener <rguenther@suse.de>
1838 PR tree-optimization/84830
1839 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
1840 with the old one to avoid oscillations.
1842 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
1845 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
1847 (assign_by_spills): Return a flag of reload assignment failure.
1848 Do not process the reload assignment failures. Do not spill other
1849 reload pseudos if they has the same reg class. Update n if
1851 (lra_assign): Add a return arg. Set up from the result of
1852 assign_by_spills call.
1853 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1854 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1855 usage_insns if it is not NULL.
1856 (spill_hard_reg_in_range): New function.
1857 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1858 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1859 function prototypes.
1860 (lra_assign): Change prototype.
1861 * lra.c (lra): Add code to deal with fails by splitting hard reg
1864 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
1866 * config/riscv/riscv.opt (mrelax): New option.
1867 * config/riscv/riscv.c (riscv_file_start): Emit ".option
1868 "norelax" when riscv_mrelax is disabled.
1869 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
1871 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1874 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
1875 reassociation for int modes.
1877 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1879 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
1880 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
1882 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
1883 * config/aarch64/aarch64-sve.md
1884 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
1885 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
1886 (*extend<mode><Vwide>2): Rename to...
1887 (aarch64_sve_extend<mode><Vwide>2): ...this.
1888 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
1889 renaming the old pattern to...
1890 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
1892 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
1893 define_expand, renaming the old pattern to...
1894 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
1895 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
1896 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
1897 account when deciding which SVE instruction the optab should use.
1898 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
1900 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1902 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
1903 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
1904 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
1905 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
1906 (tlsdesc_small_<mode>): Turn a define_expand and use
1907 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
1908 (tlsdesc_small_advsimd_<mode>): ...this.
1909 (tlsdesc_small_sve_<mode>): New pattern.
1911 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1913 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
1914 (UNSPEC_UMUL_HIGHPART): New constants.
1915 (MUL_HIGHPART): New int iteraor.
1916 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
1917 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
1919 (*<su>mul<mode>3_highpart): New define_insn.
1921 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1924 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1927 2018-03-13 Martin Liska <mliska@suse.cz>
1930 * (sem_item_optimizer::sem_item_optimizer): Initialize new
1932 (sem_item_optimizer::~sem_item_optimizer): Release it.
1933 (sem_item_optimizer::merge_classes): Register variable aliases.
1934 (sem_item_optimizer::fixup_pt_set): New function.
1935 (sem_item_optimizer::fixup_points_to_sets): Likewise.
1936 * ipa-icf.h: Declare new variables and functions.
1938 2018-03-13 Jakub Jelinek <jakub@redhat.com>
1941 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
1942 integer_pow2p@2 and test integer_pow2p in condition.
1943 (A < 0 ? C : 0): Similarly for @1.
1946 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
1947 characters starting at p contain '\0' character, don't look beyond
1951 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
1952 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
1955 * reg-stack.c (change_stack): Change update_end var from int to
1956 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
1957 also call set_block_for_insn on the newly added insns and rescan.
1960 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
1961 on the last operand.
1964 * tree.c (stabilize_reference_1): Return save_expr (e) for
1965 STATEMENT_LIST even if it doesn't have side-effects.
1967 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
1969 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
1971 2018-03-12 Renlin Li <renlin.li@arm.com>
1973 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
1974 aarch64_output_scalar_simd_mov_immediate.
1976 2018-03-12 Martin Sebor <msebor@redhat.com>
1978 PR tree-optimization/83456
1979 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
1980 for perfectly overlapping calls to memcpy.
1981 (gimple_fold_builtin_memory_chk): Same.
1982 (gimple_fold_builtin_strcpy): Handle no-warning.
1983 (gimple_fold_builtin_stxcpy_chk): Same.
1984 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
1986 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
1988 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
1989 parameter. Use it for SFmode.
1990 (rs6000_function_arg_advance_1): Adjust.
1991 (rs6000_function_arg): Adjust.
1992 (rs6000_gimplify_va_arg): Pass false for that new parameter.
1994 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
1996 PR rtl-optimization/84169
1997 PR rtl-optimization/84780
1998 * combine.c (can_combine_p): Check for a 2-insn combination whether
1999 the destination register is used between the two insns, too.
2001 2018-03-12 Richard Biener <rguenther@suse.de>
2003 PR tree-optimization/84803
2004 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
2005 for refs DR analysis didn't process.
2007 2018-03-12 Richard Biener <rguenther@suse.de>
2009 PR tree-optimization/84777
2010 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
2011 force-vectorize loops ignore whether we are optimizing for size.
2013 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
2015 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
2016 (TARGET_MD_ASM_ADJUST): Define.
2018 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
2019 Kito Cheng <kito.cheng@gmail.com>
2020 Chung-Ju Wu <jasonwucj@gmail.com>
2022 * config/nds32/nds32.c (nds32_compute_stack_frame,
2023 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
2024 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
2025 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
2026 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
2027 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
2028 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
2029 * config/nds32/nds32.md (prologue, epilogue): Use macro
2030 NDS32_V3PUSH_AVAILABLE_P to do checking.
2032 2018-03-11 Jakub Jelinek <jakub@redhat.com>
2035 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
2036 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
2037 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
2038 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
2039 addition of most attributes on !orig_type_die or the attribute not
2040 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
2042 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2043 Chung-Ju Wu <jasonwucj@gmail.com>
2045 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
2047 * config/nds32/nds32.opt (mvh): New option.
2049 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2050 Chung-Ju Wu <jasonwucj@gmail.com>
2052 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
2054 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
2055 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
2058 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2059 Chung-Ju Wu <jasonwucj@gmail.com>
2061 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
2063 * config/nds32/nds32-multiple.md (strlensi): New pattern.
2064 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
2066 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
2067 Kito Cheng <kito.cheng@gmail.com>
2068 Chung-Ju Wu <jasonwucj@gmail.com>
2070 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
2071 UNSPEC_FFMISM and UNSPEC_FLMISM.
2072 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
2073 for ffb, ffmism and flmism.
2074 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
2075 (unspec_ffmism): Ditto.
2076 (unspec_flmism): Ditto.
2077 (nds32_expand_builtin_impl): Check if string extension is available.
2078 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
2079 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
2081 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
2084 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2087 * lra-assigns.c (assign_by_spills): Return a flag of reload
2088 assignment failure. Do not process the reload assignment
2089 failures. Do not spill other reload pseudos if they has the same
2091 (lra_assign): Add a return arg. Set up from the result of
2092 assign_by_spills call.
2093 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2094 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2095 usage_insns if it is not NULL.
2096 (spill_hard_reg_in_range): New function.
2097 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2098 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2099 function prototypes.
2100 (lra_assign): Change prototype.
2101 * lra.c (lra): Add code to deal with fails by splitting hard reg
2104 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
2107 * config/i386/i386.opt: Replace Enforcment with Enforcement.
2109 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
2112 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
2113 (dw_val_node): Add val_symbolic_view.
2114 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
2115 (symview_upper_bound): New.
2116 (new_line_info_table): Initialize symviews_since_reset.
2117 (dwarf2out_source_line): Count symviews_since_reset and set
2118 symview_upper_bound.
2119 (dw_val_equal_p): Handle symview.
2120 (add_AT_symview): New.
2121 (print_dw_val): Handle symview.
2122 (attr_checksum, attr_checksum_ordered): Likewise.
2123 (same_dw_val_p, size_of_die): Likewise.
2124 (value_format, output_die): Likewise.
2125 (add_high_low_attributes): Use add_AT_symview for entry_view.
2126 (dwarf2out_finish): Reset symview_upper_bound, clear
2129 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
2132 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
2133 Add strict argument and use it.
2134 (rs6000_split_multireg_move): Update for new strict argument.
2135 (mem_operand_gpr): Disallow all non-offsettable addresses.
2136 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
2138 2018-03-09 Jakub Jelinek <jakub@redhat.com>
2141 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
2142 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
2143 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
2146 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
2147 decl, use remap_type if we want to use the type.
2149 2018-03-09 Martin Sebor <msebor@redhat.com>
2151 PR tree-optimization/84526
2152 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2154 (builtin_access::generic_overlap): Be prepared to handle non-array
2157 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
2159 PR rtl-optimization/84682
2160 * lra-constraints.c (process_address_1): Check is_address flag
2161 for address constraints.
2162 (process_alt_operands): Likewise.
2163 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
2164 preprocess_constraints.
2165 * recog.h (preprocess_constraints): Add oploc parameter.
2167 * recog.c (preprocess_constraints): Test address_operand for
2168 CT_ADDRESS constraints.
2170 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2173 * lra-assigns.c (assign_by_spills): Return a flag of reload
2174 assignment failure. Do not process the reload assignment
2175 failures. Do not spill other reload pseudos if they has the same
2177 (lra_assign): Add a return arg. Set up from the result of
2178 assign_by_spills call.
2179 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2180 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2181 usage_insns if it is not NULL.
2182 (spill_hard_reg_in_range): New function.
2183 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2184 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2185 function prototypes.
2186 (lra_assign): Change prototype.
2187 * lra.c (lra): Add code to deal with fails by splitting hard reg
2190 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2193 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
2194 Accept complain bool parameter. Only emit errors if it is true.
2195 (arm_parse_cpu_option_name): Likewise.
2196 (arm_target_thumb_only): Adjust callers of the above.
2197 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
2198 prototype to take a default true bool parameter.
2199 (arm_parse_arch_option_name): Likewise.
2201 2018-03-09 David Malcolm <dmalcolm@redhat.com>
2202 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
2206 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
2207 * configure: Regenerate.
2208 * configure.ac ("linker --version-script option"): New.
2209 ("linker soname option"): New.
2211 2018-03-09 Richard Biener <rguenther@suse.de>
2213 PR tree-optimization/84775
2214 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
2215 immediate uses of predicate stmts and mark them modified.
2218 PR tree-optimization/84178
2219 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2221 (version_loop_for_if_conversion): Delay update_ssa call.
2222 (tree_if_conversion): Delay update_ssa until after predicate
2225 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
2228 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
2229 when the function accesses prior frames.
2231 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2234 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
2235 gen_llsym, otherwise call maybe_gen_llsym.
2238 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
2239 has ',' character inside of it.
2241 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2244 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
2245 as clobbering CC_REGNUM.
2247 2018-03-08 Richard Biener <rguenther@suse.de>
2250 * tree-scalar-evolution.c: Include tree-into-ssa.h.
2251 (follow_copies_to_constant): Do not follow SSA names registered
2254 2018-03-08 Richard Biener <rguenther@suse.de>
2256 PR tree-optimization/84178
2257 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2259 (version_loop_for_if_conversion): Delay update_ssa call.
2260 (tree_if_conversion): Delay update_ssa until after predicate
2263 2018-03-08 David Malcolm <dmalcolm@redhat.com>
2265 PR tree-optimization/84178
2266 * tree-if-conv.c (release_bb_predicate): Remove the
2267 the assertion that the stmts have NULL use_ops.
2268 Discard the statements, asserting that they haven't
2269 yet been added to a BB.
2271 2018-03-08 Richard Biener <rguenther@suse.de>
2273 PR tree-optimization/84746
2274 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
2275 (phi_translate): Pass in destination ANTIC_OUT set.
2276 (phi_translate_1): Likewise. For a simplified result lookup
2277 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
2278 (phi_translate_set): Adjust.
2279 (do_pre_regular_insertion): Likewise.
2280 (do_pre_partial_partial_insertion): Likewise.
2282 2018-03-08 Martin Liska <mliska@suse.cz>
2284 PR gcov-profile/84735
2285 * doc/gcov.texi: Document usage of profile files.
2286 * gcov-io.h: Document changes in the format.
2288 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
2292 * dwarf2out.c (struct dw_line_info_table): Update comments for
2294 (FORCE_RESET_NEXT_VIEW): New.
2295 (FORCE_RESETTING_VIEW_P): New.
2296 (RESETTING_VIEW_P): Check for -1 too.
2297 (ZERO_VIEW_P): Likewise.
2298 (new_line_info_table): Force-reset next view.
2299 (dwarf2out_begin_function): Likewise.
2300 (dwarf2out_source_line): Simplify zero_view_p initialization.
2301 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
2302 view directly. Omit view when omitting .loc at line 0.
2304 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2306 PR tree-optimization/84740
2307 * tree-switch-conversion.c (process_switch): Call build_constructors
2308 only if info.phi_count is non-zero.
2310 PR tree-optimization/84739
2311 * tree-tailcall.c (find_tail_calls): Check call arguments against
2312 DECL_ARGUMENTS (current_function_decl) rather than
2313 DECL_ARGUMENTS (func) when checking for tail recursion.
2315 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2317 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
2318 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
2319 Volker Reichelt's entry and add entries for people that perform
2320 GCC fuzzy testing and report numerous bugs.
2322 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
2325 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
2326 readonly data in sdata, if that is disabled.
2327 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
2328 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
2329 -mreadonly-in-sdata option.
2331 2018-03-07 Martin Sebor <msebor@redhat.com>
2333 PR tree-optimization/84468
2334 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
2335 basic block when looking for nul assignment.
2337 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
2340 * except.h (output_function_exception_table): Adjust prototype.
2341 * except.c (output_function_exception_table): Remove FNNAME parameter
2342 and add SECTION parameter. Ouput one part of the table at a time.
2343 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
2344 the first part of the exception table and emit unwind directives.
2345 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
2346 (i386_pe_seh_cold_init): Likewise.
2347 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
2348 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
2349 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
2350 (ix86_output_call_insn): Emit a nop in one more case for SEH.
2351 * config/i386/winnt.c: Include except.h.
2352 (struct seh_frame_state): Add reg_offset, after_prologue and
2353 in_cold_section fields.
2354 (i386_pe_seh_end_prologue): Set seh->after_prologue.
2355 (i386_pe_seh_cold_init): New function.
2356 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
2357 to seh->in_cold_section.
2358 (seh_emit_push): Record the offset of the push.
2359 (seh_emit_save): Record the offet of the save.
2360 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
2361 Test seh->after_prologue to disregard the epilogue.
2362 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
2363 (i386_pe_end_cold_function): New function.
2365 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2368 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
2369 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
2372 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
2374 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
2375 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
2378 * multiple_target.c: Include tree-inline.h and intl.h.
2379 (expand_target_clones): Diagnose and fail if node->definition and
2380 !tree_versionable_function_p (node->decl).
2382 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
2384 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
2386 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
2387 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
2388 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
2390 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2393 * combine.c (try_combine): Use reg_or_subregno instead of handling
2394 just paradoxical SUBREGs and REGs.
2396 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
2398 * config/arc/arc.c (arc_finalize_pic): Remove function.
2399 (arc_must_save_register): We use single base PIC register, remove
2400 checks to save/restore the PIC register.
2401 (arc_expand_prologue): Likewise.
2402 * config/arc/arc-protos.h (arc_set_default_type_attributes):
2404 (arc_verify_short): Likewise.
2405 (arc_attr_type): Likewise.
2406 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
2407 (walk_stores): Likewise.
2408 (arc_address_cost): Make it static.
2409 (arc_verify_short): Likewise.
2410 (branch_dest): Likewise.
2411 (arc_attr_type): Likewise.
2412 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
2413 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
2414 (arc_final_prescan_insn): Remove inserting the nops due to
2415 hardware hazards. It is done in reorg step.
2416 (insn_length_variant_t): Remove.
2417 (insn_length_parameters_t): Likewise.
2418 (arc_insn_length_parameters): Likewise.
2419 (arc_get_insn_variants): Likewise.
2420 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
2422 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2425 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
2428 PR tree-optimization/84687
2429 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
2431 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
2433 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2435 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
2436 Rename to ppc_speculation_barrier.
2437 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
2438 __builtin_ppc_speculation_barrier.
2440 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2443 * combine.c (combine_simplify_rtx): Don't try to simplify if
2444 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
2447 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
2449 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
2450 to 32 bytes when compiling for POWER9.
2452 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2455 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
2456 regparm >= 3 with no arg reg available also for calls with
2457 flag_force_indirect_call. Pass decl to ix86_function_regparm.
2460 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
2462 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
2464 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
2467 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
2469 2018-03-05 Richard Biener <rguenther@suse.de>
2471 PR tree-optimization/84486
2472 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
2473 When inserting a __builtin_assume_aligned call set the LHS
2474 SSA name alignment info accordingly.
2476 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
2478 PR tree-optimization/84114
2479 * config/aarch64/aarch64.c (aarch64_reassociation_width)
2480 Avoid reassociation of FLOAT_MODE addition.
2482 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
2484 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
2485 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
2486 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
2487 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
2488 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
2489 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
2490 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
2492 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
2493 __builtin_ia32_wbinvd): New builtins.
2494 (SPECIAL_ARGS2): New.
2495 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
2496 (SPECIAL_ARGS2): New.
2497 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
2498 (ix86_valid_target_attribute_inner_p): Ditto.
2499 (ix86_init_mmx_sse_builtins): Add special_args2.
2500 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
2501 TARGET_WBNOINVD_P): New.
2502 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
2503 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
2504 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
2505 * config/i386/immintrin.h (_wbinvd): New intrinsic.
2506 * config/i386/pconfigintrin.h: New file.
2507 * config/i386/wbnoinvdintrin.h: Ditto.
2508 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
2509 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
2511 2018-03-05 Richard Biener <rguenther@suse.de>
2513 PR tree-optimization/84670
2514 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2516 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2517 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2518 (compute_antic_aux): Only assert the number of values in ANTIC_IN
2519 doesn't grow if all successors (recursively) were visited at least
2522 2018-03-05 Richard Biener <rguenther@suse.de>
2524 PR tree-optimization/84650
2525 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
2526 if executed in the loop pipeline.
2528 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
2530 * doc/configfiles.texi (Configuration Files): Move info about
2531 conditionalizing $target-protos.h to...
2532 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
2533 differs from $target-protos.h.
2535 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
2536 Chung-Ju Wu <jasonwucj@gmail.com>
2538 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
2539 * config/nds32/nds32-multiple.md (setmemsi): Define.
2540 * config/nds32/nds32-memory-manipulation.c
2541 (nds32_gen_dup_4_byte_to_word_value): New.
2542 (emit_setmem_word_loop): New.
2543 (emit_setmem_byte_loop): New.
2544 (nds32_expand_setmem_loop): New.
2545 (nds32_expand_setmem_loop_v3m): New.
2546 (nds32_expand_setmem_unroll): New.
2547 (nds32_expand_setmem): New.
2549 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2550 Chung-Ju Wu <jasonwucj@gmail.com>
2552 * config/nds32/nds32-memory-manipulation.c
2553 (nds32_emit_load_store): New.
2554 (nds32_emit_post_inc_load_store): New.
2555 (nds32_emit_mem_move): New.
2556 (nds32_emit_mem_move_block): New.
2557 (nds32_expand_movmemsi_loop_unknown_size): New.
2558 (nds32_expand_movmemsi_loop_known_size): New.
2559 (nds32_expand_movmemsi_loop): New.
2560 (nds32_expand_movmemsi_unroll): New.
2561 (nds32_expand_movmemqi): Rename ...
2562 (nds32_expand_movmemsi): ... to this.
2563 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
2564 (movmemsi): ... to this.
2565 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
2566 (nds32_expand_movmemsi): ... to this.
2568 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2569 Monk Chiang <sh.chiang04@gmail.com>
2570 Chung-Ju Wu <jasonwucj@gmail.com>
2572 * config/nds32/nds32-protos.h
2573 (nds32_expand_load_multiple): New arguments.
2574 (nds32_expand_store_multiple): Ditto.
2575 (nds32_valid_multiple_load_store): Rename ...
2576 (nds32_valid_multiple_load_store_p): ... to this.
2577 * config/nds32/nds32-memory-manipulation.c
2578 (nds32_expand_load_multiple): Refine implementation.
2579 (nds32_expand_store_multiple): Ditto.
2580 * config/nds32/nds32-multiple.md
2581 (load_multiple): Update nds32_expand_load_multiple interface.
2582 (store_multiple): Update nds32_expand_store_multiple interface.
2583 * config/nds32/nds32-predicates.c
2584 (nds32_valid_multiple_load_store): Rename ...
2585 (nds32_valid_multiple_load_store_p): ... to this and refine
2587 * config/nds32/predicates.md
2588 (nds32_load_multiple_and_update_address_operation): New predicate.
2589 (nds32_store_multiple_and_update_address_operation): New predicate.
2591 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2592 Chung-Ju Wu <jasonwucj@gmail.com>
2594 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
2595 (combo): New attribute.
2596 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
2598 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
2600 * config/nds32/nds32.opt: Change -mcmodel= default value.
2602 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
2603 Monk Chiang <sh.chiang04@gmail.com>
2604 Chung-Ju Wu <jasonwucj@gmail.com>
2606 * config/nds32/constants.md (unspec_element): New enum.
2607 * config/nds32/constraints.md (Umw): New constraint.
2608 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
2609 * config/nds32/nds32-intrinsic.md: Likewise.
2610 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
2611 (nds32_valid_smw_lwm_base_p): New.
2612 (nds32_output_smw_single_word): New.
2613 (nds32_output_lmw_single_word): New.
2614 (nds32_expand_unaligned_load): New.
2615 (nds32_expand_unaligned_store): New.
2616 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
2617 (nds32_output_smw_single_word): Declare.
2618 (nds32_output_lmw_single_word): Declare.
2619 (nds32_expand_unaligned_load): Declare.
2620 (nds32_expand_unaligned_store): Declare.
2621 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
2622 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
2623 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
2624 NDS32_BUILTIN_UASTORE_DW.
2625 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
2628 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
2629 Kito Cheng <kito.cheng@gmail.com>
2630 Chung-Ju Wu <jasonwucj@gmail.com>
2632 * config/nds32/nds32-intrinsic.c
2633 (nds32_expand_builtin_null_ftype_reg): Delete.
2634 (nds32_expand_builtin_reg_ftype_imm): Ditto.
2635 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
2636 (nds32_read_argument): New.
2637 (nds32_legitimize_target): Ditto.
2638 (nds32_legitimize_argument): Ditto.
2639 (nds32_check_constant_argument): Ditto.
2640 (nds32_expand_unop_builtin): Ditto.
2641 (nds32_expand_unopimm_builtin): Ditto.
2642 (nds32_expand_binop_builtin): Ditto.
2643 (nds32_builtin_decl_impl): Ditto.
2644 (builtin_description): Ditto.
2645 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
2646 (nds32_init_builtins_impl): Ditto.
2647 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
2648 (nds32_builtin_decl): New.
2649 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
2650 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
2652 2018-03-02 Jeff Law <law@redhat.com>
2654 * reorg.c (stop_search_p): Handle DEBUG_INSN.
2655 (redundant_insn, fill_simple_delay_slots): Likewise.
2656 (fill_slots_from_thread): Likewise.
2657 * resource.c (mark_referenced_resources): Likewise.
2658 (mark_set_resources, find_dead_or_set_registers): Likewise.
2660 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2662 * substring-locations.h (format_warning_va): Formatting fix for
2664 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
2666 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
2667 * substring-locations.c: Include intl.h.
2668 (format_warning_va): Turned into small wrapper around
2669 format_warning_n_va, renamed to ...
2670 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
2671 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
2673 (format_warning_at_substring_n): New function.
2674 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
2675 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
2676 format_warning_at_substring with just a shorter name instead of
2677 const function pointer.
2678 (fmtwarn_n): New function.
2679 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
2680 appropriate, get rid of all the fmtstr temporaries, move conditionals
2681 with G_() wrapped string literals directly into fmtwarn arguments,
2682 cast dir.len to (int), formatting fixes.
2684 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
2686 * doc/invoke.texi: Remove "Cilk Plus" references.
2688 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2689 Richard Biener <rguenther@suse.de>
2692 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
2693 for error or warning attributes if CALL_FROM_THUNK_P is set.
2696 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2699 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
2700 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
2703 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
2704 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
2706 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
2707 instead of -1U in last predictors element's probability member.
2709 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2712 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2713 arguments if they are comparable.
2715 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
2717 PR tree-optimization/84634
2718 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
2719 masks and masked_loop_p with a single loop_masks, making sure it's
2720 null for bb vectorization.
2722 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
2724 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
2725 (vect_analyze_data_ref_access): Use loop->safe_len rather than
2726 loop->force_vectorize to check whether there is no alias.
2728 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2731 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
2733 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
2735 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
2736 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
2737 instead of a loop around prev_real_insn.
2738 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
2742 * config/i386/i386.c (ix86_print_operand): Use conditional
2743 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
2746 2018-03-02 Richard Biener <rguenther@suse.de>
2748 PR tree-optimization/84427
2749 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
2750 (bitmap_set_subtract_values): Rewrite to handle multiple
2753 (prune_clobbered_mems): Likewise.
2754 (phi_translate): Take edge instead of pred/phiblock.
2755 (phi_translate_1): Likewise.
2756 (phi_translate_set): Likewise. Insert all translated
2757 exprs for a value into the set, keeping possibly multiple
2758 expressions per value.
2759 (compute_antic_aux): Adjust for phi_translate changes.
2760 When intersecting union the expressions and prune those
2761 not in the final value set, keeping possibly multiple
2762 expressions per value. Do not use value-insertion
2763 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
2764 all expressions. Add verification that the value-sets
2765 only shrink during iteration.
2766 (compute_partial_antic_aux): Adjust for the phi_translate changes.
2767 (do_pre_regular_insertion): Likewise.
2768 (do_pre_partial_partial_insertion): Likewise.
2770 2018-03-02 Richard Biener <rguenther@suse.de>
2773 * config/darwin.c (saved_debug_info_level): New static global.
2774 (darwin_asm_lto_start): Disable debug info generation for LTO out.
2775 (darwin_asm_lto_end): Restore debug info generation settings.
2777 2018-03-01 Martin Liska <mliska@suse.cz>
2780 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
2783 2018-03-01 Richard Biener <rguenther@suse.de>
2786 * dwarf2out.c (gen_variable_die): Properly handle late VLA
2787 type annotation with LTO when debug was disabled at compile-time.
2789 2018-03-01 Matthew Fortune <mfortune@gmail.com>
2791 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
2793 (mips_final_postscan_insn): Likewise.
2795 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
2797 PR rtl-optimization/84528
2798 * alias.c (init_alias_target): Add commentary.
2799 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
2800 a unique base value if the frame pointer is not eliminated
2801 to the stack pointer.
2803 2018-03-01 Tom de Vries <tom@codesourcery.com>
2805 PR rtl-optimization/83327
2806 * lra-int.h (hard_regs_spilled_into): Declare.
2807 * lra.c (hard_regs_spilled_into): Define.
2808 (init_reg_info): Init hard_regs_spilled_into.
2809 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
2810 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
2811 (process_bb_lives): Handle hard_regs_spilled_into.
2812 (lra_create_live_ranges_1): Before doing liveness propagation, clear
2813 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
2815 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
2817 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
2818 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
2819 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
2820 * config/rs6000/aix72.h: New file.
2822 2018-02-28 Jakub Jelinek <jakub@redhat.com>
2824 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
2825 instead of warning_at with conditional singular and plural messages
2829 * stor-layout.c (update_alignment_for_field): For
2830 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
2831 && !DECL_PACKED (field), do the alignment update, just use
2832 only desired_align instead of MAX (type_align, desired_align)
2834 (place_field): Don't do known_align < desired_align handling
2835 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
2836 is non-NULL, instead do it after rli->prev_field handling and
2837 only if not within a bitfield word. For DECL_PACKED (field)
2838 use type_align of BITS_PER_UNIT.
2840 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
2842 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
2843 superfluous parentheses and trailing spaces.
2845 2018-02-28 Richard Biener <rguenther@suse.de>
2847 PR tree-optimization/84584
2848 * graphite-scop-detection.c (scop_detection::add_scop): Discard
2849 SCoPs with fake exit edge.
2851 2018-02-28 Martin Liska <mliska@suse.cz>
2854 * timevar.c (timer::print): Fix format to properly print 100%
2857 2018-02-28 Richard Biener <rguenther@suse.de>
2860 * genmatch.c (capture_info::walk_match): Do not mark
2861 captured expressions without operands as expr_p given
2862 they act more like predicates and should be subject to
2863 "lost tail" side-effect preserving.
2865 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
2867 PR rtl-optimization/81611
2868 * auto-inc-dec.c (attempt_change): Move dead note from
2869 mem_insn if it's the next use of regno
2870 (find_address): Take address use of reg holding
2871 non-incremented value. Add parm to limit search to the named
2873 (merge_in_block): Attempt to use a mem insn that is the next
2874 use of the original regno.
2876 2018-02-27 Martin Sebor <msebor@redhat.com>
2879 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
2880 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
2882 2018-02-27 Martin Sebor <msebor@redhat.com>
2884 PR translation/84207
2885 * diagnostic-core.h (warning_n, error_n, inform_n): Change
2886 n argument to unsigned HOST_WIDE_INT.
2887 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
2888 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
2889 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
2890 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
2892 2018-02-27 Richard Biener <rguenther@suse.de>
2894 PR tree-optimization/84512
2895 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
2896 Do not use the estimate returned from record_stmt_cost for
2897 the scalar iteration cost but sum properly using add_stmt_cost.
2899 2018-02-27 Richard Biener <rguenther@suse.de>
2901 PR tree-optimization/84466
2902 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2903 Adjust last change to less strictly validate use operands.
2905 2018-02-27 Martin Liska <mliska@suse.cz>
2907 PR gcov-profile/84548
2908 * gcov.c (process_file): Allow partial overlap and consider it
2909 also as group functions.
2910 (output_lines): Properly calculate range of lines for a group.
2912 2018-02-27 Martin Liska <mliska@suse.cz>
2914 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
2915 'ggc' suffixes. Change first column width.
2916 (timer::print): Fix formatting of the column.
2918 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
2920 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
2921 preserve inline entry blocks for the sake of debug inline
2922 entry point markers alone.
2923 (remove_unused_locals): Suggest in comments a better place to
2924 force the preservation of inline entry blocks that are
2925 otherwise unused, but do not preserve them.
2927 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2929 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
2931 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2934 * config/i386/constraints.md (Bs): Replace
2935 ix86_indirect_branch_register with
2936 TARGET_INDIRECT_BRANCH_REGISTER.
2938 * config/i386/i386.md (indirect_jump): Likewise.
2939 (tablejump): Likewise.
2940 (*sibcall_memory): Likewise.
2941 (*sibcall_value_memory): Likewise.
2942 Peepholes of indirect call and jump via memory: Likewise.
2943 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
2944 (*sibcall_value_GOT_32): Likewise.
2945 * config/i386/predicates.md (indirect_branch_operand): Likewise.
2946 (GOT_memory_operand): Likewise.
2947 (call_insn_operand): Likewise.
2948 (sibcall_insn_operand): Likewise.
2949 (GOT32_symbol_operand): Likewise.
2950 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
2952 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2954 PR rtl-optimization/83496
2955 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
2956 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
2957 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
2958 redundant insn, if any.
2959 (relax_delay_slots): Likewise.
2960 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
2962 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
2964 PR tree-optimization/83965
2965 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
2966 that grouped statements are part of a reduction chain. Return
2967 true if the statement is not marked as a reduction itself but
2969 (vect_recog_dot_prod_pattern): Don't check whether the statement
2970 is part of a group here.
2971 (vect_recog_sad_pattern): Likewise.
2972 (vect_recog_widen_sum_pattern): Likewise.
2974 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2977 * final.c (rest_of_clean_state): Also look for calls inside sequences.
2979 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2982 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
2984 (ix86_output_indirect_function_return): New prototype.
2985 (ix86_split_simple_return_pop_internal): Likewise.
2986 * config/i386/i386.c (indirect_return_via_cx): New.
2987 (indirect_return_via_cx_bnd): Likewise.
2988 (indirect_thunk_name): Handle return va CX_REG.
2989 (output_indirect_thunk_function): Create alias for
2990 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
2991 (ix86_output_indirect_jmp): Remove the bool argument.
2992 (ix86_output_indirect_function_return): New function.
2993 (ix86_split_simple_return_pop_internal): Likewise.
2994 * config/i386/i386.md (*indirect_jump): Don't pass false
2995 to ix86_output_indirect_jmp.
2996 (*tablejump_1): Likewise.
2997 (simple_return_pop_internal): Change it to define_insn_and_split.
2998 Call ix86_split_simple_return_pop_internal to split it for
3000 (simple_return_indirect_internal): Call
3001 ix86_output_indirect_function_return instead of
3002 ix86_output_indirect_jmp.
3004 2018-02-26 Jakub Jelinek <jakub@redhat.com>
3007 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
3008 memset and value initialization afterwards.
3010 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
3012 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
3014 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3017 * common/config/aarch64/aarch64-common.c
3018 (aarch_option_optimization_table[]): Switch
3019 off fomit-frame-pointer
3021 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3022 Chung-Ju Wu <jasonwucj@gmail.com>
3024 * config/nds32/nds32-multiple.md (load_multiple): Disallow
3026 (store_multiple): Ditto.
3028 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3030 * config.gcc: Add --with-cpu support for nds32 target.
3031 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
3032 * config/nds32/nds32.opt: Add -mcpu= option.
3034 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
3036 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
3037 isel=yes): Warn for these deprecated options.
3039 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
3041 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
3042 ISA_2_5_MASKS_EMBEDDED.
3044 2018-02-23 Jakub Jelinek <jakub@redhat.com>
3046 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
3047 p->max as pointers rather than using iterative_hash_expr.
3049 2018-02-23 Carl Love <cel@us.ibm.com>
3051 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
3052 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
3054 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
3055 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
3056 P8V_BUILTIN_VEC_VUNSIGNED2.
3058 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
3061 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
3062 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
3064 * lra-constraints.c (curr_insn_transform): Set up
3065 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
3067 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
3068 * lra-eliminations.c (spill_pseudos): Ditto.
3069 (process_insn_for_elimination): Ditto.
3070 * lra-lives.c (reg_early_clobber_p): Use the new macros.
3071 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
3072 LRA_NON_CLOBBERED_ALT.
3074 2018-02-22 Martin Sebor <msebor@redhat.com>
3076 PR tree-optimization/84480
3077 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
3078 to maybe_diag_stxncpy_trunc. Call it.
3079 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
3080 from gimple_fold_builtin_strcpy. Print inlining stack.
3081 (handle_builtin_stxncpy): Print inlining stack.
3082 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
3084 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
3087 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
3088 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
3089 and -fcheck-pointer-bounds are used together.
3090 (indirect_thunk_prefix): New enum.
3091 (indirect_thunk_need_prefix): New function.
3092 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
3093 "_nt" instead of "_bnd" for NOTRACK prefix.
3094 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
3095 (output_indirect_thunk_function): Likewise.
3097 (ix86_code_end): Update output_indirect_thunk_function calls.
3098 (ix86_output_indirect_branch_via_reg): Replace
3099 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
3100 (ix86_output_indirect_branch_via_push): Likewise.
3101 (ix86_output_function_return): Likewise.
3102 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
3103 incompatible with -fcf-protection=branch and
3104 -fcheck-pointer-bounds.
3106 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3109 * config/aarch64/aarch64.c (aarch64_print_address_internal):
3110 Change gcc_assert call to output_operand_lossage.
3112 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3114 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
3116 2018-02-22 DJ Delorie <dj@redhat.com>
3117 Sebastian Perta <sebastian.perta@renesas.com>
3118 Oleg Endo <olegendo@gcc.gnu.org>
3120 * config/rx/rx.c (rx_rtx_costs): New function.
3121 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
3123 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
3125 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
3127 2018-02-22 Martin Liska <mliska@suse.cz>
3130 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
3131 Add "native" as a possible value.
3133 2018-02-22 Martin Liska <mliska@suse.cz>
3136 * config/i386/i386.c (ix86_option_override_internal):
3137 Add "native" as a possible value for -march and -mtune.
3139 2018-02-22 Jakub Jelinek <jakub@redhat.com>
3142 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
3143 to all type variants.
3145 PR tree-optimization/84503
3146 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
3147 width as info->bitpos + info->bitsize - start.
3148 (merged_store_group::merge_overlapping): Simplify width computation.
3149 (check_no_overlap): New function.
3150 (imm_store_chain_info::try_coalesce_bswap): Compute expected
3151 start + width and last_order of the group, fail if check_no_overlap
3153 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
3154 to group if check_no_overlap fails.
3156 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3158 * config/rs6000/altivec.md: Delete contraint arguments to
3159 define_expand, define_split, and define_peephole2, and in
3160 define_insn_and_split if always unused.
3161 * config/rs6000/darwin.md: Ditto.
3162 * config/rs6000/dfp.md: Ditto.
3163 * config/rs6000/rs6000.md: Ditto.
3164 * config/rs6000/sync.md: Ditto.
3165 * config/rs6000/vector.md: Ditto.
3166 * config/rs6000/vsx.md: Ditto.
3168 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3170 * config/rs6000/altivec.md: Write output control strings as braced
3171 blocks instead of double-quoted strings.
3172 * config/rs6000/darwin.md: Ditto.
3173 * config/rs6000/rs6000.md: Ditto.
3174 * config/rs6000/vector.md: Ditto.
3175 * config/rs6000/vsx.md: Ditto.
3177 2018-02-21 Jason Merrill <jason@redhat.com>
3179 PR c++/84314 - ICE with templates and fastcall attribute.
3180 * attribs.c (build_type_attribute_qual_variant): Remove assert.
3182 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3184 * ipa-cp.c (determine_versionability): Fix comment typos.
3186 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3189 * ipa-cp.c (determine_versionability): Do not version functions caling
3192 2018-02-21 Martin Liska <mliska@suse.cz>
3195 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
3196 Add "native" as a possible value.
3197 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
3198 the macro when native cpu detection is available.
3200 2018-02-21 Martin Liska <mliska@suse.cz>
3203 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
3204 Add "native" as a possible value.
3205 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
3206 when native cpu detection is available.
3208 2018-02-21 Jakub Jelinek <jakub@redhat.com>
3209 Martin Sebor <msebor@redhat.com>
3211 PR tree-optimization/84478
3212 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
3214 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
3215 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
3216 support which is conservatively correct, for 2 only stay conservative
3217 for maxlen. Formatting and comment capitalization fixes. Add STRICT
3218 argument to the 2 argument get_range_strlen, adjust 6 arg
3219 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
3221 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
3222 (gimple_fold_builtin_strlen): Pass true as last argument to
3225 2018-02-20 Martin Sebor <msebor@redhat.com>
3228 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
3229 (builtin_memref::set_base_and_offset): Same. Handle inner references.
3230 (builtin_memref::builtin_memref): Factor out parts into
3231 set_base_and_offset and call it.
3233 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3236 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
3237 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
3238 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
3239 search at the associated MODE_INT.
3241 2018-02-20 Jeff Law <law@redhat.com>
3244 PR tree-optimization/81592
3246 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
3247 for range data rather than using global data.
3248 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
3249 range data rather than using global data.
3250 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
3251 pass it to children as needed.
3252 (struct directive::fmtresult): Similarly.
3253 (struct directive::set_width): Similarly.
3254 (struct directive::set_precision): Similarly.
3255 (format_integer, format_directive, parse_directive): Similarly.
3256 (format_none): Accept unnamed vr_values parameter.
3257 (format_percent, format_floating, format_character): Similarly.
3258 (format_string, format_plain): Similarly.
3259 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
3260 the EVRP range analyzer for range data rather than using global data.
3261 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
3262 gimple-ssa-evrp-analyze.h
3263 (class sprintf_dom_walker): Add after_dom_children member function.
3264 Add evrp_range_analyzer member.
3265 (sprintf_dom_walker::before_dom_children): Call into the EVRP
3266 range analyzer as needed.
3267 (sprintf_dom_walker::after_dom_children): New member function.
3268 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
3270 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
3271 (evrp_range_analyzer::pop_to_marker): Likewise.
3273 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3275 PR tree-optimization/84419
3276 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
3277 with the required type if its current type is compatible but
3280 2018-02-20 Jakub Jelinek <jakub@redhat.com>
3283 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
3284 after vectorization.
3286 2018-02-20 Martin Liska <mliska@suse.cz>
3289 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
3290 possible values if we don't have a hint.
3292 2018-02-20 Martin Liska <mliska@suse.cz>
3296 * final.c (shorten_branches): Build align_tab array with one
3298 * opts.c (finish_options): Add alignment option limit check.
3299 (MAX_CODE_ALIGN): Likewise.
3300 (MAX_CODE_ALIGN_VALUE): Likewise.
3301 * doc/invoke.texi: Document maximum allowed option value for
3302 all -falign-* options.
3304 2018-02-19 Jakub Jelinek <jakub@redhat.com>
3307 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
3308 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
3309 * var-tracking.c (emit_note_insn_var_location): Remove all references
3310 to NOTE_INSN_CALL_ARG_LOCATION.
3311 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
3312 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
3313 Use copy_rtx_if_shared.
3314 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
3315 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
3316 (dwarf2out_var_location): Remove handling of
3317 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
3319 * final.c (final_scan_insn): Remove all references to
3320 NOTE_INSN_CALL_ARG_LOCATION.
3321 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
3322 before dumping final insns.
3323 * except.c (emit_note_eh_region_end): Remove all references to
3324 NOTE_INSN_CALL_ARG_LOCATION.
3325 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
3326 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
3327 * config/arc/arc.c (hwloop_optimize): Likewise.
3328 * config/arm/arm.c (create_fix_barrier): Likewise.
3329 * config/s390/s390.c (s390_chunkify_start): Likewise.
3330 * config/sh/sh.c (find_barrier): Likewise.
3331 * config/i386/i386.c (rest_of_insert_endbranch,
3332 ix86_seh_fixup_eh_fallthru): Likewise.
3333 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
3334 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
3335 * config/frv/frv.c (frv_function_prologue): Likewise.
3336 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
3338 (note_outside_basic_block_p): Remove all references to
3339 NOTE_INSN_CALL_ARG_LOCATION.
3340 * gengtype.c (adjust_field_rtx_def): Likewise.
3341 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
3343 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
3344 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
3347 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
3350 PR tree-optimization/84452
3351 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
3352 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
3355 2018-02-19 Martin Liska <mliska@suse.cz>
3358 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
3360 2018-02-19 Martin Liska <mliska@suse.cz>
3361 Richard Sandiford <richard.sandiford@linaro.org>
3363 PR tree-optimization/82491
3364 * gimple-fold.c (get_base_constructor): Make earlier bail out
3367 2018-02-19 Carl Love <cel@us.ibm.com>
3369 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
3370 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
3372 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
3373 P8V_BUILTIN_VEC_NEG.
3375 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
3377 * config/rl78/rl78.md (movdf): New define expand.
3379 2018-02-19 Martin Liska <mliska@suse.cz>
3382 * doc/invoke.texi: Fix typo.
3383 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
3385 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
3387 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
3388 handle rs6000_single_float and rs6000_double_float specially for
3391 2018-02-16 Jeff Law <law@redhat.com>
3393 * config/rx/rx.c (add_pop_cfi_notes): New function.;
3396 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3399 * ipa-inline.c (inline_small_functions): Fix a typo.
3401 2018-02-16 Nathan Sidwell <nathan@acm.org>
3403 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
3405 2018-02-16 Carl Love <cel@us.ibm.com>
3407 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
3408 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
3409 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
3410 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
3411 expansion to P8V_BUILTIN_VEC_FLOAT2.
3413 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
3415 PR rtl-optimization/70023
3416 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
3417 src_regno into account.
3419 2018-02-16 Carl Love <cel@us.ibm.com>
3421 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
3422 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
3423 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
3424 * config/rs6000/rs6000.c: Remove case statements for
3425 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
3426 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
3427 and P9V_BUILTIN_VEC_VINSERT4B.
3428 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
3429 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
3430 * config/rs6000/vsx.md:
3431 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
3434 2018-02-16 Carl Love <cel@us.ibm.com>
3436 * config/rs6000/altivec.h: Add builtin names vec_extract4b
3438 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
3440 * config/rs6000/rs6000-c.c: Add the definitions for
3441 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
3442 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
3443 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
3444 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
3445 definition for insert4b and define insn *insert3b_internal.
3446 * doc/extend.texi: Add documentation for vec_extract4b.
3448 2018-02-16 Nathan Sidwell <nathan@acm.org>
3450 * doc/extend.texi (Backwards Compatibility): Mention friend
3451 injection. Note for-scope is deprecated.
3452 * doc/invoke.texi (-ffriend-injection): Deprecate.
3454 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
3456 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
3457 that moved to I2, also allow destinations that are a paradoxical
3458 subreg (instead of a normal reg).
3460 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
3463 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
3466 2018-02-16 Richard Biener <rguenther@suse.de>
3468 PR tree-optimization/84037
3469 PR tree-optimization/84016
3471 * config/i386/i386.c (ix86_builtin_vectorization_cost):
3472 Adjust vec_construct for the fact we need additional higher latency
3473 128bit inserts for AVX256 and AVX512 vector builds.
3474 (ix86_add_stmt_cost): Scale vector construction cost for
3477 2018-02-16 Richard Biener <rguenther@suse.de>
3479 PR tree-optimization/84417
3480 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
3481 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
3482 (non_rewritable_lvalue_p): Likewise, use poly-ints.
3484 2018-02-16 Martin Liska <mliska@suse.cz>
3487 * internal-fn.def (ASAN_CHECK): Set proper flags.
3488 (ASAN_MARK): Likewise.
3490 2018-02-16 Julia Koval <julia.koval@intel.com>
3492 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
3493 from PTA_CANNONLAKE.
3495 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3498 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
3499 Use ++iter rather than iter++ for std::list iterators.
3500 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
3501 defer deleting them until all nodes in the forest are processed. Do
3502 free even leaf nodes. Change to_process into auto_vec.
3505 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
3506 * vec.h (vec_default_construct): Use memset instead of placement new
3507 if BROKEN_VALUE_INITIALIZATION is defined.
3508 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
3509 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
3512 PR rtl-optimization/83723
3513 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
3514 * lra.c (lra_substitute_pseudo): Likewise. If true, use
3515 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
3517 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
3519 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
3521 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
3523 PR rtl-optimization/81443
3524 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
3525 from inner REGs to paradoxical SUBREGs.
3527 2018-02-16 Richard Biener <rguenther@suse.de>
3529 PR tree-optimization/84399
3530 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3531 For operands we can analyze at their definition make sure we can
3532 analyze them at each use as well.
3534 2018-02-16 Richard Biener <rguenther@suse.de>
3536 PR tree-optimization/84190
3537 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
3538 volatile accesses if the decl isn't volatile.
3540 2018-02-15 Jason Merrill <jason@redhat.com>
3542 PR c++/84314 - ICE with templates and fastcall attribute.
3543 * attribs.c (build_type_attribute_qual_variant): Don't clobber
3544 TYPE_CANONICAL on an existing type.
3546 2018-02-15 Jakub Jelinek <jakub@redhat.com>
3548 PR tree-optimization/84383
3549 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
3550 dstoff nor call operand_equal_p if dstbase is NULL.
3552 PR tree-optimization/84334
3553 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
3554 also a CONSTANT_CLASS_P, punt.
3556 2018-02-14 Jim Wilson <jimw@sifive.com>
3558 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
3559 first SMALL_OPERAND check. New local min_second_step. Move assert
3560 to where locals are set. Add TARGET_RVC support.
3561 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
3563 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
3565 * doc/invoke.texi: Correct -Wformat-overflow code sample.
3567 2018-02-14 Martin Sebor <msebor@redhat.com>
3569 PR tree-optimization/83698
3570 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
3571 arrays constrain the offset range to their bounds.
3572 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
3573 (builtin_access::overlap): Avoid setting the size of overlap if it's
3575 (maybe_diag_overlap): Also consider arrays when deciding what values
3576 of offsets to include in diagnostics.
3578 2018-02-14 Martin Sebor <msebor@redhat.com>
3581 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
3582 that correspond to the kind of a declaration.
3584 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
3587 * config/pa/pa.md: Load address of PIC label using the linkage table
3588 if the label is nonlocal.
3590 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
3592 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
3593 warning message if user requests -maltivec=be.
3594 * doc/invoke.texi: Document deprecation of -maltivec=be.
3596 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
3599 * config/rs6000/rs6000-c.c: Update definitions for
3600 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
3601 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
3603 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3606 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
3607 add _get_ssp intrinsics. Remove argument from
3608 __builtin_ia32_rdssp[d|q].
3609 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
3610 * config/i386/i386-builtin.def: Remove argument from
3611 __builtin_ia32_rdssp[d|q].
3612 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
3613 ix86_expand_special_args_builtin for _rdssp[d|q].
3614 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
3615 Clear register before usage.
3616 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
3617 Add documentation for new _get_ssp and _inc_ssp intrinsics.
3619 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
3621 PR tree-optimization/84357
3622 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
3623 operand 1 of an ARRAY_REF too.
3625 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
3628 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
3629 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3631 (set_of_reg): New struct.
3632 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
3633 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
3634 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3636 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
3637 Split into bitclr, bitset, bitinvert patterns if appropriate.
3638 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
3639 use rx_fuse_in_memory_bitop.
3640 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
3641 to named insn, correct maximum insn length.
3643 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
3646 * machmode.def: Define a complex mode for PARTIAL_INT.
3647 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
3649 * doc/rtl.texi: Document CSPImode.
3650 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
3652 (msp430_hard_regno_nregs_with_padding): Likewise.
3654 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
3657 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
3659 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
3661 PR rtl-optimization/84169
3662 * combine.c (try_combine): New variable split_i2i3. Set it to true if
3663 we generated a parallel as new i3 and we split that to new i2 and i3
3664 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
3665 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
3666 those to i2, not i1. Partially rewrite this scan code.
3668 2018-02-13 Jakub Jelinek <jakub@redhat.com>
3671 * stor-layout.c (place_field): For variable length fields, adjust
3672 offset_align afterwards not just based on the field's alignment,
3673 but also on the size.
3676 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
3677 of exps and logs in the use_exp2 case.
3679 2018-02-13 Jeff Law <law@redhat.com>
3681 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
3684 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
3687 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
3691 * common.opt (gas-loc-support, gas-locview-support): New.
3692 (ginline-points, ginternal-reset-location-views): New.
3693 * doc/invoke.texi: Document them. Use @itemx where intended.
3694 (gvariable-location-views): Adjust.
3695 * target.def (reset_location_view): New.
3696 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
3697 (TARGET_RESET_LOCATION_VIEW): New.
3698 * doc/tm.texi: Rebuilt.
3699 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
3700 (dwarf2out_default_as_locview_support): New.
3701 (output_asm_line_debug_info): Use option variables.
3702 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
3703 (output_loc_list): Likewise.
3704 (add_high_low_attributes): Check option variables.
3705 Don't output entry view attribute in strict mode.
3706 (gen_inlined_subroutine_die): Check option variables.
3707 (dwarf2out_inline_entry): Likewise.
3708 (init_sections_and_labels): Likewise.
3709 (dwarf2out_early_finish): Likewise.
3710 (maybe_reset_location_view): New, from...
3711 (dwarf2out_var_location): ... here. Call it.
3712 * debug.h (dwarf2out_default_as_loc_support): Declare.
3713 (dwarf2out_default_as_locview_support): Declare.
3714 * hooks.c (hook_int_rtx_insn_0): New.
3715 * hooks.h (hook_int_rtx_insn_0): Declare.
3716 * toplev.c (process_options): Take -gas-loc-support and
3717 -gas-locview-support from dwarf2out. Enable
3718 -gvariable-location-views by default only with locview
3719 assembler support. Enable -ginternal-reset-location-views by
3720 default only if the target defines the corresponding hook.
3721 Enable -ginline-points by default if location views are
3722 enabled; force it disabled if statement frontiers are
3724 * tree-inline.c (expand_call_inline): Check option variables.
3725 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
3727 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
3729 PR tree-optimization/84321
3730 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
3731 handling. Also check whether the anti-range contains any values
3732 that satisfy the mask; switch to a VR_RANGE if not.
3734 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
3737 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
3739 2018-02-13 Martin Jambor <mjambor@suse.cz>
3742 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
3743 of call statements, also set location of a load to a temporary.
3745 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
3747 * config/rl78/rl78.c (add_vector_labels): New function.
3748 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
3749 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
3750 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
3751 which checks that no arguments are passed.
3752 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
3753 * doc/extend.texi: Documentation for the new attribute.
3755 2018-02-13 Andreas Schwab <schwab@suse.de>
3757 * config/riscv/linux.h (CPP_SPEC): Define.
3759 2018-02-13 Jakub Jelinek <jakub@redhat.com>
3762 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
3763 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
3764 OPTION_MASK_ISA_AES as first argument to def_builtin_const
3765 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
3766 instead of OPTION_MASK_ISA_PCLMUL as first argument to
3767 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
3768 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
3769 temporarily for AES and PCLMUL builtins.
3771 PR tree-optimization/84339
3772 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
3773 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
3777 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
3778 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
3779 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
3781 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
3783 * omp-simd-clone.h: New file.
3784 * omp-simd-clone.c: Include omp-simd-clone.h.
3785 (expand_simd_clones): No longer static.
3786 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
3787 cgraph.h and omp-simd-clone.h.
3788 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
3789 (vect_recog_widen_shift_pattern): Formatting fix.
3790 (vect_pattern_recog_1): Don't check optab for calls.
3793 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
3794 operands[2] into a REG before using gen_lowpart on it.
3796 2018-02-12 Jeff Law <law@redhat.com>
3799 * config/sh/sh.c (find_barrier): Consider a sibling call
3802 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
3803 successfully back substituting a reg.
3805 2018-02-12 Richard Biener <rguenther@suse.de>
3807 PR tree-optimization/84037
3808 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
3809 parameter, move visited init to caller.
3810 (vect_slp_analyze_operations): Separate cost from validity
3811 check, initialize visited once for all instances.
3812 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
3814 * tree-vect-stmts.c (vect_model_simple_cost): Make early
3816 (vect_model_promotion_demotion_cost): Likewise.
3817 (vectorizable_bswap): Guard cost modeling with !slp_node
3818 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
3820 (vectorizable_call): Likewise.
3821 (vectorizable_conversion): Likewise.
3822 (vectorizable_assignment): Likewise.
3823 (vectorizable_shift): Likewise.
3824 (vectorizable_operation): Likewise.
3825 (vectorizable_store): Likewise.
3826 (vectorizable_load): Likewise.
3827 (vectorizable_condition): Likewise.
3828 (vectorizable_comparison): Likewise.
3830 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
3833 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
3834 (ASAN_MARK): Fix fnspec to account for return value, change pointer
3835 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
3837 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3840 * params.def (inline-min-speedup): Increase from 8 to 15.
3841 (max-inline-insns-auto): Decrease from 40 to 30.
3842 * ipa-split.c (consider_split): Add some buffer for function to
3843 be considered inlining candidate.
3844 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
3847 2018-02-12 Richard Biener <rguenther@suse.de>
3849 PR tree-optimization/84037
3850 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
3851 matched stmts if we cannot swap the non-matched ones.
3853 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3855 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
3856 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
3857 _mm_maskz_scalef_round_ss): New intrinsics.
3858 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
3859 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
3860 __builtin_ia32_scalefss_round): Remove.
3861 (__builtin_ia32_scalefsd_mask_round,
3862 __builtin_ia32_scalefss_mask_round): New intrinsics.
3863 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
3864 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
3865 ((match_operand:VF_128 2 "<round_nimm_predicate>"
3866 "<round_constraint>")): Changed to ...
3867 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
3868 "<round_scalar_constraint>")): ... this.
3869 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
3870 %0, %1, %2<round_op3>}"): Changed to ...
3871 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
3872 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
3873 %2<round_scalar_mask_op3>}"): ... this.
3874 * config/i386/subst.md (round_scalar_nimm_predicate): New.
3876 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3878 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
3879 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
3880 (_mm_maskz_sqrt_round_ss): New intrinsics.
3881 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
3882 (__builtin_ia32_sqrtsd_mask_round)
3883 (__builtin_ia32_sqrtss_mask_round): New builtins.
3884 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
3885 (__builtin_ia32_sqrtss_round): Remove.
3886 (__builtin_ia32_sqrtsd_mask_round)
3887 (__builtin_ia32_sqrtss_mask_round): New builtins.
3888 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
3889 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
3890 ((match_operand:VF_128 1 "vector_operand"
3891 "xBm,<round_constraint>")): Changed to ...
3892 ((match_operand:VF_128 1 "vector_operand"
3893 "xBm,<round_scalar_constraint>")): ... this.
3894 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
3895 %0, %2, %<iptr>1<round_op3>}): Changed to ...
3896 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
3897 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
3898 %<iptr>1<round_scalar_mask_op3>}): ... this.
3899 ((set_attr "prefix" "<round_prefix>")): Changed to ...
3900 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
3902 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
3905 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
3906 Cast vec_cmpeq result to correct type.
3907 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
3908 Cast vec_cmpgt result to correct type.
3910 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
3912 * final.c (final_scan_insn_1): Renamed from...
3913 (final_scan_insn): ... this. New wrapper, to recover
3914 seen from the outermost call in recursive ones.
3915 * config/sparc/sparc.c (output_return): Drop seen from call.
3916 (output_sibcall): Likewise.
3917 * config/visium/visium.c (output_branch): Likewise.
3919 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
3921 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
3924 2018-02-10 Alan Modra <amodra@gmail.com>
3927 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
3928 Specify LR as an input.
3930 2018-02-10 Jakub Jelinek <jakub@redhat.com>
3933 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
3934 remove_member_access_dummy_vars): New functions.
3935 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
3936 lower_omp_1, execute_lower_omp): Use them.
3938 PR rtl-optimization/84308
3939 * shrink-wrap.c (spread_components): Release todo vector.
3941 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
3943 PR rtl-optimization/57193
3944 * ira-color.c (struct allocno_color_data): Add member
3945 conflict_allocno_hard_prefs.
3946 (update_conflict_allocno_hard_prefs): New.
3947 (bucket_allocno_compare_func): Add a preference based on
3948 conflict_allocno_hard_prefs.
3949 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
3950 (color_allocnos): Remove a dead code. Initiate
3951 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
3953 2018-02-09 Jakub Jelinek <jakub@redhat.com>
3956 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
3957 constraint from =wa to wa. Avoid a subreg on the output operand,
3958 instead use a pseudo and subreg it in a move.
3959 (p9_xxbrd_<mode>): Changed to ...
3960 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
3961 (p9_xxbrd_v2df): New expander.
3962 (p9_xxbrw_<mode>): Changed to ...
3963 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
3964 (p9_xxbrw_v4sf): New expander.
3966 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
3968 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
3970 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
3973 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
3974 multiply in 32-bit mode.
3975 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
3976 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
3979 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
3981 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
3982 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
3983 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
3984 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
3986 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
3989 * dwarf2out.c (is_trivial_indirect_ref): New function.
3990 (dwarf2out_late_global_decl): Do not generate a location
3991 attribute for variables that have a non-trivial DECL_VALUE_EXPR
3992 and that are not defined in the current unit.
3994 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
3996 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
3997 instead of a libcall for UNORDERED.
3999 2018-02-09 Tamar Christina <tamar.christina@arm.com>
4002 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
4003 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
4005 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4008 * config/s390/s390.c (s390_set_current_function): Invoke
4009 s390_indirect_branch_settings also if fndecl didn't change.
4011 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4013 * config/rs6000/rs6000.md (blockage): Set length to zero.
4015 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4017 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
4019 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4022 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
4023 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
4027 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
4028 PARALLEL incoming that failed vt_get_decl_and_offset check.
4031 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
4032 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
4034 (get_variable_section): For decls in named .bss* sections pass true as
4035 second argument to bss_initializer_p.
4037 2018-02-09 Marek Polacek <polacek@redhat.com>
4038 Jakub Jelinek <jakub@redhat.com>
4041 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
4042 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
4043 Sync some changes from cxx_fold_indirect_ref.
4045 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4047 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
4049 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
4050 (BLOCK_INLINE_ENTRY_LABEL): New.
4051 (dwarf2out_var_location): Disregard inline entry markers.
4052 (inline_entry_data): New struct.
4053 (inline_entry_data_hasher): New hashtable type.
4054 (inline_entry_data_hasher::hash): New.
4055 (inline_entry_data_hasher::equal): New.
4056 (inline_entry_data_table): New variable.
4057 (add_high_low_attributes): Add DW_AT_entry_pc and
4058 DW_AT_GNU_entry_view attributes if a pending entry is found
4059 in inline_entry_data_table. Add old entry_pc attribute only
4060 if debug nonbinding markers are disabled.
4061 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
4062 markers are enabled.
4063 (block_within_block_p, dwarf2out_inline_entry): New.
4064 (dwarf2out_finish): Check that no entries remained in
4065 inline_entry_data_table.
4066 * final.c (reemit_insn_block_notes): Handle inline entry notes.
4067 (final_scan_insn, notice_source_line): Likewise.
4068 (rest_of_clean_state): Skip inline entry markers.
4069 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
4071 * gimple.c (gimple_build_debug_inline_entry): New.
4072 * gimple.h (enum gimple_debug_subcode): Add
4073 GIMPLE_DEBUG_INLINE_ENTRY.
4074 (gimple_build_debug_inline_entry): Declare.
4075 (gimple_debug_inline_entry_p): New.
4076 (gimple_debug_nonbind_marker_p): Adjust.
4077 * insn-notes.def (INLINE_ENTRY): New.
4078 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
4079 inline entry marker notes.
4080 (print_insn): Likewise.
4081 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
4082 (INSN_DEBUG_MARKER_KIND): Likewise.
4083 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
4084 * tree-inline.c (expand_call_inline): Build and insert
4085 debug_inline_entry stmt.
4086 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
4087 inline entry blocks early, if nonbind markers are enabled.
4088 (dump_scope_block): Dump fragment info.
4089 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
4090 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
4091 (gimple_build_debug_inline_entry): New.
4092 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
4093 Enable/disable inline entry points too.
4094 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
4095 (DEBUG_INSN): Describe inline entry markers.
4097 * common.opt (gvariable-location-views): New.
4098 (gvariable-location-views=incompat5): New.
4099 * config.in: Rebuilt.
4100 * configure: Rebuilt.
4101 * configure.ac: Test assembler for view support.
4102 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
4103 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
4104 * dwarf2out.c (var_loc_view): New typedef.
4105 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
4106 (dwarf2out_locviews_in_attribute): New.
4107 (dwarf2out_locviews_in_loclist): New.
4108 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
4109 (enum dw_line_info_opcode): Add LI_adv_address.
4110 (struct dw_line_info_table): Add view.
4111 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
4112 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
4113 (zero_view_p): New variable.
4114 (ZERO_VIEW_P): New macro.
4115 (output_asm_line_debug_info): New.
4116 (struct var_loc_node): Add view.
4117 (add_AT_view_list, AT_loc_list): New.
4118 (add_var_loc_to_decl): Add view param. Test it against last.
4119 (new_loc_list): Add view params. Record them.
4120 (AT_loc_list_ptr): Handle loc and view lists.
4121 (view_list_to_loc_list_val_node): New.
4122 (print_dw_val): Handle dw_val_class_view_list.
4123 (size_of_die): Likewise.
4124 (value_format): Likewise.
4125 (loc_list_has_views): New.
4126 (gen_llsym): Set vl_symbol too.
4127 (maybe_gen_llsym, skip_loc_list_entry): New.
4128 (dwarf2out_maybe_output_loclist_view_pair): New.
4129 (output_loc_list): Output view list or entries too.
4130 (output_view_list_offset): New.
4131 (output_die): Handle dw_val_class_view_list.
4132 (output_dwarf_version): New.
4133 (output_compilation_unit_header): Use it.
4134 (output_skeleton_debug_sections): Likewise.
4135 (output_rnglists, output_line_info): Likewise.
4136 (output_pubnames, output_aranges): Update version comments.
4137 (output_one_line_info_table): Output view numbers in asm comments.
4138 (dw_loc_list): Determine current endview, pass it to new_loc_list.
4139 Call maybe_gen_llsym.
4140 (loc_list_from_tree_1): Adjust.
4141 (add_AT_location_description): Create view list attribute if
4142 needed, check it's absent otherwise.
4143 (convert_cfa_to_fb_loc_list): Adjust.
4144 (maybe_emit_file): Call output_asm_line_debug_info for test.
4145 (dwarf2out_var_location): Reset views as needed. Precompute
4146 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
4147 attribute. Set view.
4148 (new_line_info_table): Reset next view.
4149 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
4150 (dwarf2out_source_line): Likewise. Output view resets and labels to
4151 the assembler, or select appropriate line info opcodes.
4152 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
4153 (optimize_string_length): Catch it. Adjust.
4154 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
4155 dw_val_class_view_list, and remove it if no longer needed.
4156 (hash_loc_list): Hash view numbers.
4157 (loc_list_hasher::equal): Compare them.
4158 (optimize_location_lists): Check whether a view list symbol is
4159 needed, and whether the locview attribute is present, and
4160 whether they match. Remove the locview attribute if no longer
4162 (index_location_lists): Call skip_loc_list_entry for test.
4163 (dwarf2out_finish): Call output_asm_line_debug_info for test.
4164 Use output_dwarf_version.
4165 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
4166 (struct dw_val_node): Add val_view_list.
4167 * final.c (SEEN_NEXT_VIEW): New.
4168 (set_next_view_needed): New.
4169 (clear_next_view_needed): New.
4170 (maybe_output_next_view): New.
4171 (final_start_function): Rename to...
4172 (final_start_function_1): ... this. Take pointer to FIRST,
4173 add SEEN parameter. Emit param bindings in the initial view.
4174 (final_start_function): Reintroduce SEEN-less interface.
4175 (final): Rename to...
4176 (final_1): ... this. Take SEEN parameter. Output final pending
4177 next view at the end.
4178 (final): Reintroduce seen-less interface.
4179 (final_scan_insn): Output pending next view before switching
4180 sections or ending a block. Mark the next view as needed when
4181 outputting variable locations. Notify debug backend of section
4182 changes, and of location view changes.
4183 (rest_of_handle_final): Adjust.
4184 * toplev.c (process_options): Autodetect value for debug variable
4185 location views option. Warn on incompat5 without -gdwarf-5.
4186 * doc/invoke.texi (gvariable-location-views): New.
4187 (gvariable-location-views=incompat5): New.
4188 (gno-variable-location-views): New.
4190 2018-02-08 David Malcolm <dmalcolm@redhat.com>
4192 PR tree-optimization/84136
4193 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
4194 that the result of find_edge is non-NULL.
4196 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4199 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
4200 storing integer register in SImode. Fix cost of 256 and 512
4201 byte aligned SSE register store.
4203 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4205 * config/i386/i386.c (ix86_multiplication_cost): Fix
4206 multiplication cost for TARGET_AVX512DQ.
4208 2018-02-08 Marek Polacek <polacek@redhat.com>
4210 PR tree-optimization/84238
4211 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
4214 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4216 PR tree-optimization/84265
4217 * tree-vect-stmts.c (vectorizable_store): Don't treat
4218 VMAT_CONTIGUOUS accesses as grouped.
4219 (vectorizable_load): Likewise.
4221 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4223 PR tree-optimization/81635
4224 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
4225 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
4226 (test_round_for_mask): New functions.
4227 (wide_int_cc_tests): Call test_round_for_mask.
4228 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
4229 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
4230 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
4231 range returned by get_range_info.
4233 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4236 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
4237 * symtab.c: Include builtins.h
4238 (symtab_node::output_to_lto_symbol_table_p): Move here
4239 from lto-streamer-out.c:output_symbol_p.
4240 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
4241 (output_symbol_p): Move all logic to symtab.c
4242 (produce_symtab): Update.
4244 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4246 * config/s390/s390-opts.h (enum indirect_branch): Define.
4247 * config/s390/s390-protos.h (s390_return_addr_from_memory)
4248 (s390_indirect_branch_via_thunk)
4249 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
4250 (enum s390_indirect_branch_type): Define.
4251 * config/s390/s390.c (struct s390_frame_layout, struct
4252 machine_function): Remove.
4253 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
4254 (indirect_branch_table_label_no, indirect_branch_table_name):
4256 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
4257 (enum s390_indirect_branch_option): Define.
4258 (s390_return_addr_from_memory): New function.
4259 (s390_handle_string_attribute): New function.
4260 (s390_attribute_table): Add new attribute handler.
4261 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
4262 (s390_indirect_branch_via_thunk): New function.
4263 (s390_indirect_branch_via_inline_thunk): New function.
4264 (s390_function_ok_for_sibcall): When jumping via thunk disallow
4265 sibling call optimization for non z10 compiles.
4266 (s390_emit_call): Force indirect branch target to be a single
4267 register. Add r1 clobber for non-z10 compiles.
4268 (s390_emit_epilogue): Emit return jump via return_use expander.
4269 (s390_reorg): Handle JUMP_INSNs as execute targets.
4270 (s390_option_override_internal): Perform validity checks for the
4271 new command line options.
4272 (s390_indirect_branch_attrvalue): New function.
4273 (s390_indirect_branch_settings): New function.
4274 (s390_set_current_function): Invoke s390_indirect_branch_settings.
4275 (s390_output_indirect_thunk_function): New function.
4276 (s390_code_end): Implement target hook.
4277 (s390_case_values_threshold): Implement target hook.
4278 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
4280 * config/s390/s390.h (struct s390_frame_layout)
4281 (struct machine_function): Move here from s390.c.
4282 (TARGET_INDIRECT_BRANCH_NOBP_RET)
4283 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
4284 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
4285 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
4286 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
4287 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
4288 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
4289 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
4290 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
4291 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
4292 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
4293 (mnemonic attribute): Add values which aren't recognized
4295 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
4296 pattern for branch conversion. Fix mnemonic attribute.
4297 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
4298 indirect branch via thunk if requested.
4299 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
4300 ("*indirect_jump"): Disable for branch conversion using out of
4302 ("indirect_jump_via_thunk<mode>_z10")
4303 ("indirect_jump_via_thunk<mode>")
4304 ("indirect_jump_via_inlinethunk<mode>_z10")
4305 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
4306 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
4307 ("casesi_jump_via_inlinethunk<mode>_z10")
4308 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
4309 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
4310 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
4311 ("*indirect2_jump"): Disable for branch conversion.
4312 ("casesi_jump"): Turn into expander and expand patterns for branch
4314 ("return_use"): New expander.
4315 ("*return"): Emit return via thunk and rename it to ...
4316 ("*return<mode>"): ... this one.
4317 * config/s390/s390.opt: Add new options and and enum for the
4320 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4322 * lra-constraints.c (match_reload): Unconditionally use
4323 gen_lowpart_SUBREG, rather than selecting between that
4324 and equivalent gen_rtx_SUBREG code.
4326 2018-02-08 Richard Biener <rguenther@suse.de>
4328 PR tree-optimization/84233
4329 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
4330 changed flag instead of boguously re-using phi_inserted.
4332 2018-02-08 Martin Jambor <mjambor@suse.cz>
4334 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
4335 static local variables.
4337 2018-02-08 Richard Biener <rguenther@suse.de>
4339 PR tree-optimization/84278
4340 * tree-vect-stmts.c (vectorizable_store): When looking for
4341 smaller vector types to perform grouped strided loads/stores
4342 make sure the mode is supported by the target.
4343 (vectorizable_load): Likewise.
4345 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4347 * config/aarch64/aarch64.c (aarch64_components_for_bb):
4348 Increase LDP/STP opportunities by adding adjacent callee-saves.
4350 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4352 PR rtl-optimization/84068
4353 PR rtl-optimization/83459
4354 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
4356 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
4358 PR tree-optimization/84224
4359 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
4360 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
4363 2018-02-07 Iain Sandoe <iain@codesourcery.com>
4366 * config/rs6000/altivec.md (*restore_world): Remove LR use.
4367 * config/rs6000/predicates.md (restore_world_operation): Adjust op
4368 count, remove one USE.
4370 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
4372 * doc/install.texi (Configuration): Document the
4373 --with-long-double-format={ibm,ieee} PowerPC configuration
4377 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
4378 Convert from define_expand to be define_insn_and_split. Rework
4379 float/double/_Float128 conversions to QI/HI/SImode to work with
4380 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
4381 conversions to QI/HImode types did a store and then a load to
4382 truncate the value. For conversions to VSX registers, don't split
4383 the insn, instead emit the code directly. Use the code iterator
4384 any_fix to combine signed and unsigned conversions.
4385 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
4386 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
4387 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4388 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
4389 (fix_<mode>di2_hw): Likewise.
4390 (fixuns_<mode>di2_hw): Likewise.
4391 (fix_<mode>si2_hw): Likewise.
4392 (fixuns_<mode>si2_hw): Likewise.
4393 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
4394 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4395 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
4396 fix<uns>_trunc<SFDF:mode>si2_p8.
4397 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
4399 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
4400 (fix<uns>_<mode>_mem): Likewise.
4401 (fctiw<u>z_<mode>_mem): Likewise.
4402 (fix<uns>_<mode>_mem): Likewise.
4403 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
4404 the register allocator from doing a direct move to the GPRs to do
4405 a store, and instead use the ISA 3.0 store byte/half-word from
4406 vector register instruction. For IEEE 128-bit floating point,
4407 also optimize stores of 32-bit ints.
4408 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
4410 2018-02-07 Alan Hayward <alan.hayward@arm.com>
4412 * genextract.c (push_pathstr_operand): New function to support
4414 (walk_rtx): Call push_pathstr_operand.
4415 (print_path): Support [a-zA-Z].
4417 2018-02-07 Richard Biener <rguenther@suse.de>
4419 PR tree-optimization/84037
4420 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
4421 (cse_and_gimplify_to_preheader): Declare.
4422 (vect_get_place_in_interleaving_chain): Likewise.
4423 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4425 (_loop_vec_info::~_loop_vec_info): Delete it.
4426 (cse_and_gimplify_to_preheader): New function.
4427 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
4428 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
4429 (vectorizable_load): Likewise. For grouped stores always base
4430 the IV on the first element.
4431 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
4432 condition before gimplifying.
4434 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4436 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
4437 *DIV_EXPR and *MOD_EXPR.
4439 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
4442 * config/i386/i386.c (ix86_option_override_internal): Mask out
4443 the CF_SET bit when checking -fcf-protection.
4445 2018-02-07 Tom de Vries <tom@codesourcery.com>
4448 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
4451 2018-02-07 Richard Biener <rguenther@suse.de>
4453 PR tree-optimization/84204
4454 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
4457 PR tree-optimization/84205
4458 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
4459 special-case isl_ast_op_zdiv_r.
4461 PR tree-optimization/84223
4462 * graphite-scop-detection.c (gather_bbs::before_dom_children):
4463 Only add conditions from within the region.
4464 (gather_bbs::after_dom_children): Adjust.
4466 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
4469 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
4470 * config/avr/avr.md: Only post-reload split REG-REG moves if
4471 either register is GENERAL_REG_P.
4473 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4475 PR tree-optimization/84235
4476 * tree-ssa-scopedtables.c
4477 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
4478 if the subtraction is performed in floating point type where NaNs are
4479 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
4480 build 1. Formatting fix.
4482 2018-02-06 Jakub Jelinek <jakub@redhat.com>
4485 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
4486 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
4487 and skip it regardless of bb boundaries. Use CALL_P macro,
4488 don't test INSN_P (insn) together with CALL_P or JUMP_P check
4489 unnecessarily, formatting fix.
4491 2018-02-06 Michael Collison <michael.collison@arm.com>
4493 * config/arm/thumb2.md:
4494 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
4495 (*thumb_mov_notscc): Ditto.
4497 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
4500 * config/rs6000/rs6000.md (su code attribute): Use "u" for
4501 unsigned_fix, not "s".
4503 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4505 * configure.ac (gcc_fn_eh_frame_ro): New function.
4506 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
4507 correct .eh_frame permissions.
4508 * configure: Regenerate.
4510 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
4512 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
4515 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4517 * config/rs6000/rs6000.c (rs6000_option_override_internal):
4518 Display warning message for -mno-speculate-indirect-jumps.
4520 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
4522 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
4524 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
4526 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
4528 PR tree-optimization/84225
4529 * tree-eh.c (find_trapping_overflow): Only call
4530 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
4532 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4535 * config/i386/i386.c: Reimplement the check of possible options
4536 -mibt/-mshstk conbination. Change error messages.
4537 * doc/invoke.texi: Fix a typo: remove extra '='.
4539 2018-02-06 Marek Polacek <polacek@redhat.com>
4541 PR tree-optimization/84228
4542 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
4544 2018-02-06 Tamar Christina <tamar.christina@arm.com>
4547 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
4548 emitted arch directives.
4549 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
4550 __ARM_FEATURE_COPROC before changing architectures.
4552 2018-02-06 Richard Biener <rguenther@suse.de>
4554 * config/i386/i386.c (print_reg): Fix typo.
4555 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
4557 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
4559 * configure: Regenerate.
4561 2018-02-05 Martin Sebor <msebor@redhat.com>
4563 PR tree-optimization/83369
4564 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
4567 2018-02-05 Martin Liska <mliska@suse.cz>
4569 * doc/invoke.texi: Cherry-pick upstream r323995.
4571 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
4573 * ira.c (ira_init_register_move_cost): Adjust comment.
4575 2018-02-05 Martin Liska <mliska@suse.cz>
4577 PR gcov-profile/84137
4578 * doc/gcov.texi: Fix typo in documentation.
4580 2018-02-05 Martin Liska <mliska@suse.cz>
4582 PR gcov-profile/83879
4583 * doc/gcov.texi: Document necessity of --dynamic-list-data when
4584 using dlopen functionality.
4586 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
4588 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
4589 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
4590 _mm_maskz_range_ss, _mm_mask_range_round_ss,
4591 _mm_maskz_range_round_ss): New intrinsics.
4592 (__builtin_ia32_rangesd128_round)
4593 (__builtin_ia32_rangess128_round): Remove.
4594 (__builtin_ia32_rangesd128_mask_round,
4595 __builtin_ia32_rangess128_mask_round): New builtins.
4596 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
4597 __builtin_ia32_rangess128_round): Remove.
4598 (__builtin_ia32_rangesd128_mask_round,
4599 __builtin_ia32_rangess128_mask_round): New builtins.
4600 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
4601 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
4602 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
4603 "<round_saeonly_constraint>")): Changed to ...
4604 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
4605 "<round_saeonly_scalar_constraint>")): ... this.
4606 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
4607 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
4608 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
4609 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
4610 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
4612 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
4614 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
4616 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
4617 Remove all values except native, 8540 and 8548.
4619 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
4621 * config/i386/i386.c (ix86_output_function_return): Pass
4622 INVALID_REGNUM, instead of -1, as invalid register number to
4623 indirect_thunk_name and output_indirect_thunk.
4625 2018-02-02 Julia Koval <julia.koval@intel.com>
4627 * config.gcc: Add -march=icelake.
4628 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
4629 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
4630 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
4631 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
4632 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
4633 (processor_target_table): Add icelake.
4634 (ix86_option_override_internal): Handle new PTAs.
4635 (get_builtin_code_for_version): Handle icelake.
4636 (M_INTEL_COREI7_ICELAKE): New.
4637 (fold_builtin_cpu): Handle icelake.
4638 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
4639 * doc/invoke.texi: Add -march=icelake.
4641 2018-02-02 Julia Koval <julia.koval@intel.com>
4643 * config/i386/i386.c (ix86_option_override_internal): Change flags type
4644 to wide_int_bitmask.
4645 * wide-int-bitmask.h: New.
4647 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4650 * config/i386/i386.md: Replace Pmode with word_mode in
4651 builtin_setjmp_setup and builtin_longjmp to support x32.
4653 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
4657 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
4659 (rs6000_supported_cpu_names): New static variable.
4660 (linux_cpu_translation_table): Likewise.
4661 (elf_platform) <cpu>: Define new static variable and use it.
4662 Translate kernel AT_PLATFORM name to canonical name if needed.
4663 Error if platform name is unknown.
4665 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
4668 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
4670 2018-02-01 Jeff Law <law@redhat.com>
4673 * config/i386/i386.c (release_scratch_register_on_entry): Add new
4674 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
4675 the scratch if RELEASE_VIA_POP is false.
4676 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
4677 If we have to save a temporary register, decrement SIZE appropriately.
4678 Pass new arguments to release_scratch_register_on_entry.
4679 (ix86_adjust_stack_and_probe): Likewise.
4680 (ix86_emit_probe_stack_range): Pass new arguments to
4681 release_scratch_register_on_entry.
4683 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
4685 PR rtl-optimization/84157
4686 * combine.c (change_zero_ext): Use REG_P predicate in
4687 front of HARD_REGISTER_P predicate.
4689 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
4691 * config/avr/avr.c (avr_option_override): Move disabling of
4692 -fdelete-null-pointer-checks to...
4693 * common/config/avr/avr-common.c (avr_option_optimization_table):
4696 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4698 PR tree-optimization/81635
4699 * tree-data-ref.c (split_constant_offset_1): For types that
4700 wrap on overflow, try to use range info to prove that wrapping
4703 2018-02-01 Renlin Li <renlin.li@arm.com>
4706 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
4708 (aarch64_register_move_cost): Likewise.
4709 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
4711 (REG_CLASS_NAMES): Likewise.
4712 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
4713 TAILCALL_ADDR_REGS. Remove IP registers.
4714 * config/aarch64/aarch64.md (Ucs): Update register constraint.
4716 2018-02-01 Richard Biener <rguenther@suse.de>
4718 * domwalk.h (dom_walker::dom_walker): Add additional constructor
4719 for specifying RPO order and allow NULL for that.
4720 * domwalk.c (dom_walker::dom_walker): Likewise.
4721 (dom_walker::walk): Handle NULL RPO order.
4722 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
4724 (rewrite_update_dom_walker): Likewise.
4725 (mark_def_dom_walker): Likewise.
4727 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4729 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
4730 (aarch64_maybe_expand_sve_subreg_move): Declare.
4731 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
4732 * config/aarch64/predicates.md (aarch64_any_register_operand): New
4734 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
4735 that are semantically a reverse operation.
4736 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
4737 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
4738 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
4740 (aarch64_can_change_mode_class): For big-endian, forbid changes
4741 between two SVE modes if they have different element sizes.
4743 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4745 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
4746 the TImode handling for big-endian targets.
4748 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4750 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
4751 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
4753 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
4754 Remove BSWAP handing for big-endian targets and use the form of
4755 LD1RQ appropariate for the mode.
4757 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4759 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
4760 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
4763 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4766 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
4767 check for operands that need to go through aarch64_sve_reload_be.
4769 2018-02-01 Jakub Jelinek <jakub@redhat.com>
4771 PR tree-optimization/81661
4772 PR tree-optimization/84117
4773 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
4774 * tree-eh.c: Include gimplify.h.
4775 (find_trapping_overflow, replace_trapping_overflow,
4776 rewrite_to_non_trapping_overflow): New functions.
4777 * tree-vect-loop.c: Include tree-eh.h.
4778 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
4779 * tree-data-ref.c: Include tree-eh.h.
4780 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
4782 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
4784 PR rtl-optimization/84123
4785 * combine.c (change_zero_ext): Check if hard register satisfies
4786 can_change_dest_mode before calling gen_lowpart_SUBREG.
4788 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
4791 * ira.c (ira_init_register_move_cost): Remove assert.
4793 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4795 PR rtl-optimization/84071
4796 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
4797 * doc/tm.texi: Regenerate.
4799 2018-01-31 Richard Biener <rguenther@suse.de>
4801 PR tree-optimization/84132
4802 * tree-data-ref.c (analyze_miv_subscript): Properly
4803 check whether evolution_function_is_affine_multivariate_p
4804 before calling gcd_of_steps_may_divide_p.
4806 2018-01-31 Julia Koval <julia.koval@intel.com>
4809 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
4810 * config/i386/i386.md (rdpid_rex64) New.
4811 (rdpid): Make 32bit only.
4813 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
4816 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
4817 an IDENTIFIER_NODE for FUNCTION_TYPE's.
4819 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4822 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4824 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
4826 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4828 PR rtl-optimization/84071
4829 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
4830 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
4832 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4834 * config/arc/arc.c (arc_handle_aux_attribute): New function.
4835 (arc_attribute_table): Add 'aux' attribute.
4836 (arc_in_small_data_p): Consider aux like variables.
4837 (arc_is_aux_reg_p): New function.
4838 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
4839 (arc_get_aux_arg): New function.
4840 (prepare_move_operands): Handle aux-register access.
4841 (arc_handle_aux_attribute): New function.
4842 * doc/extend.texi (ARC Variable attributes): Add subsection.
4844 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4846 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
4847 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
4848 (arc_attribute_table): Add 'uncached' attribute.
4849 (arc_print_operand): Print '.di' flag for uncached memory
4851 (arc_in_small_data_p): Do not consider for small data the uncached
4853 (arc_is_uncached_mem_p): New function.
4854 * config/arc/predicates.md (compact_store_memory_operand): Check
4855 for uncached memory accesses.
4856 (nonvol_nonimm_operand): Likewise.
4857 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
4859 2018-01-31 Jakub Jelinek <jakub@redhat.com>
4862 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
4863 falign-loops=): Add Optimization flag.
4865 2018-01-30 Jeff Law <law@redhat.com>
4868 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
4869 INT_REGISTERS_SAVED. Check it prior to calling
4870 get_scratch_register_on_entry.
4871 (ix86_adjust_stack_and_probe): Similarly.
4872 (ix86_emit_probe_stack_range): Similarly.
4873 (ix86_expand_prologue): Corresponding changes.
4875 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4878 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
4879 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
4881 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
4884 * lra-constraints.c (curr_insn_transform): Process AND in the
4887 2018-01-30 Jakub Jelinek <jakub@redhat.com>
4889 PR rtl-optimization/83986
4890 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
4891 dependence against last_pending_memory_flush in addition to
4894 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
4896 PR tree-optimization/81611
4897 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
4900 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4903 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
4906 2018-01-30 Richard Biener <rguenther@suse.de>
4907 Jakub Jelinek <jakub@redhat.com>
4909 PR tree-optimization/84111
4910 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
4911 inner loops added during recursion, as they don't have up-to-date
4914 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4917 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
4918 (can_inline_edge_by_limits_p): ... here.
4919 (can_early_inline_edge_p, check_callers,
4920 update_caller_keys, update_callee_keys, recursive_inlining,
4921 add_new_edges_to_heap, speculation_useful_p,
4922 inline_small_functions,
4923 inline_small_functions, flatten_function,
4924 inline_to_all_callers_1): Update.
4926 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4928 * profile-count.c (profile_count::combine_with_ipa_count): Handle
4931 2018-01-30 Richard Biener <rguenther@suse.de>
4933 PR tree-optimization/83008
4934 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
4935 invariant and constant vector uses in stmts when they need
4938 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4941 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
4942 * configure: Regenerate.
4944 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4946 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
4948 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
4949 Use gen_rtx_REG rather than gen_lowpart.
4951 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4953 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
4954 rather than 0 when creating partial subregs.
4956 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4958 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
4961 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
4964 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
4965 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
4966 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
4967 flags. This restores the settings used before the 2017-07-24.
4968 Turning off pre increment/decrement/modify allows IVOPTS to
4969 optimize DF/SF loops where the index is an int.
4971 2018-01-29 Richard Biener <rguenther@suse.de>
4972 Kelvin Nilsen <kelvin@gcc.gnu.org>
4975 * tree-vect-stmts.c (vectorizable_call): Don't call
4976 targetm.vectorize_builtin_md_vectorized_function if callee is
4979 2018-01-22 Carl Love <cel@us.ibm.com>
4981 * doc/extend.tex: Fix typo in second arg in
4982 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
4984 2018-01-29 Richard Biener <rguenther@suse.de>
4986 PR tree-optimization/84086
4987 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
4988 (flush_ssaname_freelist): When SSA names were released reset
4989 the SCEV hash table.
4991 2018-01-29 Richard Biener <rguenther@suse.de>
4993 PR tree-optimization/84057
4994 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
4995 removed paths when removing edges.
4997 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
4999 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
5000 -mfunction-return=@var{choice}.
5002 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
5005 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
5007 (layout::print_source_line): Likewise.
5008 (test_get_line_width_without_trailing_whitespace): Add test cases.
5010 2018-01-27 Jakub Jelinek <jakub@redhat.com>
5013 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
5016 2018-01-26 Jim Wilson <jimw@sifive.com>
5018 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
5020 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
5023 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5025 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
5026 and CMP + SUB-immediate -> SUBS.
5028 2018-01-26 Martin Sebor <msebor@redhat.com>
5030 PR tree-optimization/83896
5031 * tree-ssa-strlen.c (get_string_len): Rename...
5032 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
5033 Avoid assuming length is constant.
5034 (handle_char_store): Use HOST_WIDE_INT for string length.
5036 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
5039 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
5040 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
5042 2018-01-26 Richard Biener <rguenther@suse.de>
5044 PR rtl-optimization/84003
5045 * dse.c (record_store): Only record redundant stores when
5046 the earlier store aliases at least all accesses the later one does.
5048 2018-01-26 Jakub Jelinek <jakub@redhat.com>
5050 PR rtl-optimization/83985
5051 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
5052 REG_CFA_RESTORE insns.
5053 (delete_unmarked_insns): Don't ignore separate shrink wrapping
5054 REG_CFA_RESTORE insns here.
5057 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
5058 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
5060 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5062 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
5063 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
5064 (arc_init): Likewise.
5065 (arc_override_options): Likewise.
5066 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
5068 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
5070 * config/arc/arc.h (TARGET_DBNZ): Define.
5071 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
5072 properly set the tune attribute.
5073 (dbnz): Use TARGET_DBNZ guard.
5074 * config/arc/arc.opt (mtune): Add core3 option.
5076 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5078 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
5079 recognize new pic like addresses.
5080 (arc_delegitimize_address): Clean up.
5082 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5084 * config/arc/arc-arches.def: Option mrf16 valid for all
5086 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
5087 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
5088 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
5089 * config/arc/arc-tables.opt: Regenerate.
5090 * config/arc/arc.c (arc_conditional_register_usage): Handle
5091 reduced register file case.
5092 (arc_file_start): Set must have build attributes.
5093 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
5095 * config/arc/arc.opt (mrf16): Add new option.
5096 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
5097 * config/arc/genmultilib.awk: Handle new mrf16 option.
5098 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
5099 * config/arc/t-multilib: Regenerate.
5100 * doc/invoke.texi (ARC Options): Document mrf16 option.
5102 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5104 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
5105 * config/arc/arc.c (arc_handle_secure_attribute): New function.
5106 (arc_attribute_table): Add 'secure_call' attribute.
5107 (arc_print_operand): Print secure call operand.
5108 (arc_function_ok_for_sibcall): Don't optimize tail calls when
5110 (arc_is_secure_call_p): New function. * config/arc/arc.md
5111 (call_i): Add support for sjli instruction.
5112 (call_value_i): Likewise.
5113 * config/arc/constraints.md (Csc): New constraint.
5115 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5116 John Eric Martin <John.Martin@emmicro-us.com>
5118 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
5119 * config/arc/arc.c (_arc_jli_section): New struct.
5120 (arc_jli_section): New type.
5121 (rc_jli_sections): New static variable.
5122 (arc_handle_jli_attribute): New function.
5123 (arc_attribute_table): Add jli_always and jli_fixed attribute.
5124 (arc_file_end): New function.
5125 (TARGET_ASM_FILE_END): Define.
5126 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
5127 (arc_add_jli_section): New function.
5128 (jli_call_scan): Likewise.
5129 (arc_reorg): Call jli_call_scan.
5130 (arc_output_addsi): Remove 'S' from printing asm operand.
5131 (arc_is_jli_call_p): New function.
5132 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
5134 (movhi_insn): Likewise.
5135 (movsi_insn): Likewise.
5136 (movsi_set_cc_insn): Likewise.
5137 (loadqi_update): Likewise.
5138 (load_zeroextendqisi_update): Likewise.
5139 (load_signextendqisi_update): Likewise.
5140 (loadhi_update): Likewise.
5141 (load_zeroextendhisi_update): Likewise.
5142 (load_signextendhisi_update): Likewise.
5143 (loadsi_update): Likewise.
5144 (loadsf_update): Likewise.
5145 (movsicc_insn): Likewise.
5146 (bset_insn): Likewise.
5147 (bxor_insn): Likewise.
5148 (bclr_insn): Likewise.
5149 (bmsk_insn): Likewise.
5150 (bicsi3_insn): Likewise.
5151 (cmpsi_cc_c_insn): Likewise.
5152 (movsi_ne): Likewise.
5153 (movsi_cond_exec): Likewise.
5154 (clrsbsi2): Likewise.
5165 (call_i): Remove 'S' asm letter, add jli instruction.
5166 (call_value_i): Likewise.
5167 * config/arc/arc.op (mjli-always): New option.
5168 * config/arc/constraints.md (Cji): New constraint.
5169 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
5171 (subsf3_fpx): Likewise.
5172 (mulsf3_fpx): Likewise.
5173 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
5175 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
5177 * doc/invoke.texi (ARC): Document mjli-always option.
5179 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
5181 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
5182 avoid addition with 0 and use incw and decw where possible.
5184 2018-01-26 Richard Biener <rguenther@suse.de>
5186 PR tree-optimization/81082
5187 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
5188 association if it requires casting to unsigned.
5189 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
5190 from fold_plusminus_mult_expr to catch important cases late when
5191 range info is available.
5193 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5195 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
5196 * configure.ac (hidden_linkonce): New test.
5197 * configure: Regenerate.
5198 * config.in: Regenerate.
5200 2018-01-26 Julia Koval <julia.koval@intel.com>
5202 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
5203 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
5204 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
5205 _mm_mask_bitshuffle_epi64_mask): Fix type.
5206 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
5207 USI_FTYPE_V4DI_V4DI_USI): Remove.
5208 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
5209 __builtin_ia32_vpshufbitqmb256_mask,
5210 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
5211 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
5212 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
5214 2018-01-26 Alan Modra <amodra@gmail.com>
5217 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
5218 UNSPEC_VBPERMQ. Sort other unspecs.
5220 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
5222 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
5224 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
5227 * predict.c (drop_profile): Do not push/pop cfun; update also
5229 (handle_missing_profiles): Fix logic looking for zero profiles.
5231 2018-01-25 Jakub Jelinek <jakub@redhat.com>
5234 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
5235 on functions with #pragma omp declare simd or functions with simd
5237 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
5238 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
5239 Remove trailing \n from warning_at calls.
5241 2018-01-25 Tom de Vries <tom@codesourcery.com>
5244 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5245 for neutered workers.
5247 2018-01-24 Joseph Myers <joseph@codesourcery.com>
5250 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
5251 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
5253 2018-01-24 Jeff Law <law@redhat.com>
5256 * i386.c (get_probe_interval): Move to earlier point.
5257 (ix86_compute_frame_layout): If -fstack-clash-protection and
5258 the frame is larger than the probe interval, then use pushes
5259 to save registers rather than reg->mem moves.
5260 (ix86_expand_prologue): Remove conditional for int_registers_saved
5263 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
5266 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
5267 min/max for never referenced object.
5269 2018-01-24 Jakub Jelinek <jakub@redhat.com>
5272 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
5274 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
5275 attributes from DECL_ATTRIBUTES (decl) without affecting
5276 DECL_ATTRIBUTES (current_function_decl).
5277 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
5278 functions with non-NULL DECL_ABSTRACT_ORIGIN.
5280 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
5282 PR tree-optimization/83979
5283 * fold-const.c (fold_comparison): Use constant_boolean_node
5284 instead of boolean_{true,false}_node.
5286 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
5288 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
5291 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5293 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5294 Simplify the clause that sets the length attribute.
5295 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5296 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
5297 clause that sets the length attribute.
5298 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5300 2018-01-24 Tom de Vries <tom@codesourcery.com>
5303 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
5304 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
5305 Add strict parameter.
5306 (prevent_branch_around_nothing): Insert dummy insn between branch to
5307 label and label with no ptx insn inbetween.
5308 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
5310 2018-01-24 Tom de Vries <tom@codesourcery.com>
5313 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5314 for neutered threads in warp.
5315 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
5317 2018-01-24 Richard Biener <rguenther@suse.de>
5319 PR tree-optimization/83176
5320 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
5323 2018-01-24 Richard Biener <rguenther@suse.de>
5325 PR tree-optimization/82819
5326 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
5327 code generating pluses that are no-ops in the target precision.
5329 2018-01-24 Richard Biener <rguenther@suse.de>
5332 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
5334 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5336 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
5337 to merge probabilities.
5338 * predict.c (probably_never_executed): Also mark as cold functions
5339 with global 0 profile and guessed local profile.
5340 * profile-count.c (profile_probability::combine_with_count): New
5342 * profile-count.h (profile_probability::operator*,
5343 profile_probability::operator*=, profile_probability::operator/,
5344 profile_probability::operator/=): Reduce precision to adjusted
5345 and set value to guessed on contradictory divisions.
5346 (profile_probability::combine_with_freq): Remove.
5347 (profile_probability::combine_wiht_count): Declare.
5348 (profile_count::force_nonzero):: Set to adjusted.
5349 (profile_count::probability_in):: Set quality to adjusted.
5350 * tree-ssa-tail-merge.c (replace_block_by): Use
5353 2018-01-23 Andrew Waterman <andrew@sifive.com>
5354 Jim Wilson <jimw@sifive.com>
5356 * config/riscv/riscv.c (riscv_stack_boundary): New.
5357 (riscv_option_override): Set riscv_stack_boundary. Handle
5358 riscv_preferred_stack_boundary_arg.
5359 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
5360 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
5361 (STACK_BOUNDARY): Set to riscv_stack_boundary.
5362 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
5363 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
5364 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
5366 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
5369 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
5370 of struct ix86_frame.
5371 (ix86_expand_epilogue): Likewise. Add a local variable for
5372 the reg_save_offset field in struct ix86_frame.
5374 2018-01-23 Bin Cheng <bin.cheng@arm.com>
5376 PR tree-optimization/82604
5377 * tree-loop-distribution.c (enum partition_kind): New enum item
5378 PKIND_PARTIAL_MEMSET.
5379 (partition_builtin_p): Support above new enum item.
5380 (generate_code_for_partition): Ditto.
5381 (compute_access_range): Differentiate cases that equality can be
5382 proven at all loops, the innermost loops or no loops.
5383 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
5384 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
5385 (finalize_partitions, distribute_loop): Don't fuse partition of
5386 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
5387 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
5390 2018-01-23 Martin Liska <mliska@suse.cz>
5392 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
5393 order to ignore the predictor.
5394 (PRED_POLYMORPHIC_CALL): Likewise.
5395 (PRED_RECURSIVE_CALL): Likewise.
5397 2018-01-23 Martin Liska <mliska@suse.cz>
5399 * tree-profile.c (tree_profiling): Print function header to
5400 aware reader which function we are working on.
5401 * value-prof.c (gimple_find_values_to_profile): Do not print
5402 not interesting value histograms.
5404 2018-01-23 Martin Liska <mliska@suse.cz>
5406 * profile-count.h (enum profile_quality): Add
5407 profile_uninitialized as the first value. Do not number values
5408 as they are zero based.
5409 (profile_count::verify): Update sanity check.
5410 (profile_probability::verify): Likewise.
5412 2018-01-23 Nathan Sidwell <nathan@acm.org>
5414 * doc/invoke.texi (ffor-scope): Deprecate.
5416 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5418 PR tree-optimization/83510
5419 * domwalk.c (set_all_edges_as_executable): New function.
5420 (dom_walker::dom_walker): Convert bool param
5421 "skip_unreachable_blocks" to enum reachability. Move setup of
5422 edge flags to set_all_edges_as_executable and only do it when
5423 reachability is REACHABLE_BLOCKS.
5424 * domwalk.h (enum dom_walker::reachability): New enum.
5425 (dom_walker::dom_walker): Convert bool param
5426 "skip_unreachable_blocks" to enum reachability.
5427 (set_all_edges_as_executable): New decl.
5428 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
5429 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
5431 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
5432 but converting true to REACHABLE_BLOCKS.
5433 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
5435 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
5436 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
5437 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
5439 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
5440 if check_all_array_refs will be called.
5442 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5444 * tree.c (selftest::test_location_wrappers): Add more test
5447 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5449 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
5450 (selftest::test_bit_in_range): Likewise.
5452 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5455 * doc/sourcebuild.texi (vect_float): Say that the selector
5456 only describes the situation when -funsafe-math-optimizations is on.
5457 (vect_float_strict): Document.
5459 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5461 PR tree-optimization/83965
5462 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
5463 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
5464 instead of checking only for a reduction.
5465 (vect_recog_widen_sum_pattern): Likewise.
5467 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5469 * predict.c (probably_never_executed): Only use precise profile info.
5470 (compute_function_frequency): Skip after inlining hack since we now
5471 have quality checking.
5473 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5475 * profile-count.h (profile_probability::very_unlikely,
5476 profile_probability::unlikely, profile_probability::even): Set
5477 precision to guessed.
5479 2018-01-23 Richard Biener <rguenther@suse.de>
5481 PR tree-optimization/83963
5482 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
5483 Properly terminate dominator walk when crossing the exit edge not
5484 when visiting its source block.
5486 2018-01-23 Jakub Jelinek <jakub@redhat.com>
5489 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
5490 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
5492 2018-01-22 Jakub Jelinek <jakub@redhat.com>
5494 PR tree-optimization/83957
5495 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
5496 semicolon after for body surrounded by braces.
5498 PR tree-optimization/83081
5499 * profile-count.h (profile_probability::split): New method.
5500 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
5501 Use profile_probability::split.
5502 (do_compare_rtx_and_jump): Fix adjustment of probabilities
5503 when splitting a single conditional jump into 2.
5505 2018-01-22 David Malcolm <dmalcolm@redhat.com>
5507 PR tree-optimization/69452
5508 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
5511 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5513 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
5514 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
5515 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
5517 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5519 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
5520 * config/rl78/rl78.md (movdi): New define_expand.
5521 * config/rl78/rl78.c (rl78_split_movdi): New function.
5523 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
5526 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
5528 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
5529 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
5530 128-bit to produce an UNSPEC move to get the double word with the
5531 signbit and then a shift directly to do signbit.
5532 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
5533 implementation with a new version that just does either a direct
5534 move or a regular move. Move memory interface to separate insns.
5535 Move insns so they are next to the expander.
5536 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
5537 with signbit move. Split big and little endian case.
5538 (signbit<mode>2_dm_mem_le): Likewise.
5539 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
5540 (signbit<mode>2_dm2): Likewise.
5542 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5544 * config/rl78/rl78.md (anddi3): New define_expand.
5546 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5548 * config/rl78/rl78.md (umindi3): New define_expand.
5550 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5552 * config/rl78/rl78.md (smindi3): New define_expand.
5554 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5556 * config/rl78/rl78.md (smaxdi3): New define_expand.
5558 2018-01-22 Carl Love <cel@us.ibm.com>
5560 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
5561 LVX_V1TI): Add macro expansion.
5562 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
5563 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
5564 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
5565 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
5566 Change check to determine if the instruction is a byte reversing
5567 entry. Fix typo in comment.
5568 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
5569 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
5570 Add def_builtin calls for new builtins.
5571 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
5572 Add define_insn expansion.
5574 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5576 * config/rl78/rl78.md (umaxdi3): New define_expand.
5578 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5580 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
5581 for non-QImode registers.
5583 2018-01-22 Richard Biener <rguenther@suse.de>
5585 PR tree-optimization/83963
5586 * graphite-scop-detection.c (scop_detection::get_sese): Delay
5587 including the loop exit block.
5588 (scop_detection::merge_sese): Likewise.
5589 (scop_detection::add_scop): Do it here instead.
5591 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5593 * doc/sourcebuild.texi (arm_softfloat): Document.
5595 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
5598 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
5599 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
5600 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
5602 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5603 David Edelsohn <dje.gcc@gmail.com>
5606 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5607 Change "crset eq" to "crset 2".
5608 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5609 (*call_indirect_aix<mode>_nospec): Likewise.
5610 (*call_value_indirect_aix<mode>_nospec): Likewise.
5611 (*call_indirect_elfv2<mode>_nospec): Likewise.
5612 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
5613 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
5614 change assembly output from . to $.
5615 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5616 (indirect_jump<mode>_nospec): Change assembly output from . to $.
5617 (*tablejump<mode>_internal1_nospec): Likewise.
5619 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
5622 * config/sh/sh_optimize_sett_clrt.cc:
5623 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
5625 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
5627 PR tree-optimization/83940
5628 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
5629 offset_dt to vect_constant_def rather than vect_unknown_def_type.
5630 (vect_check_load_store_mask): Add a mask_dt_out parameter and
5631 use it to pass back the definition type.
5632 (vect_check_store_rhs): Likewise rhs_dt_out.
5633 (vect_build_gather_load_calls): Add a mask_dt argument and use
5634 it instead of a call to vect_is_simple_use.
5635 (vectorizable_store): Update calls to vect_check_load_store_mask
5636 and vect_check_store_rhs. Use the dt returned by the latter instead
5637 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
5638 instead of calls to vect_is_simple_use. Pass the scalar rather
5639 than the vector operand to vect_is_simple_use when handling
5640 second and subsequent copies of an rhs value.
5641 (vectorizable_load): Update calls to vect_check_load_store_mask
5642 and vect_build_gather_load_calls. Use the cached mask_dt and
5643 gs_info.offset_dt instead of calls to vect_is_simple_use.
5645 2018-01-20 Jakub Jelinek <jakub@redhat.com>
5648 * tree-emutls.c: Include gimplify.h.
5649 (lower_emutls_2): New function.
5650 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
5651 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
5652 it before further processing.
5655 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
5656 UINTVAL (trueop1) instead of INTVAL (op1).
5658 2018-01-19 Jakub Jelinek <jakub@redhat.com>
5662 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
5663 INCOMING_FRAME_SP_OFFSET if not defined.
5664 (scan_trace): Add ENTRY argument. If true and
5665 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
5666 emit a note to adjust the CFA offset.
5667 (create_cfi_notes): Adjust scan_trace callers.
5668 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
5669 INCOMING_FRAME_SP_OFFSET in the CIE.
5670 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
5671 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
5673 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
5674 * doc/tm.texi: Regenerated.
5676 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5678 PR rtl-optimization/83147
5679 * lra-constraints.c (remove_inheritance_pseudos): Use
5680 lra_substitute_pseudo_within_insn.
5682 2018-01-19 Tom de Vries <tom@codesourcery.com>
5683 Cesar Philippidis <cesar@codesourcery.com>
5686 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
5688 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
5691 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
5692 spaces for function labels.
5694 2018-01-19 Martin Liska <mliska@suse.cz>
5696 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
5697 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
5698 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
5699 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
5700 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
5701 (PRED_CONST_RETURN): Change from 69 to 65.
5702 (PRED_NULL_RETURN): Change from 91 to 71.
5703 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
5704 (PRED_LOOP_GUARD): Change from 66 to 73.
5706 2018-01-19 Martin Liska <mliska@suse.cz>
5708 * predict.c (predict_insn_def): Add new assert.
5709 (struct branch_predictor): Change type to signed integer.
5710 (test_prediction_value_range): Amend test to cover
5712 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
5713 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
5714 (PRED_LOOP_ITERATIONS_MAX): Likewise.
5715 (PRED_LOOP_IV_COMPARE): Likewise.
5716 * predict.h (PROB_UNINITIALIZED): Define new constant.
5718 2018-01-19 Martin Liska <mliska@suse.cz>
5720 * predict.c (dump_prediction): Add new format for
5721 analyze_brprob.py script which is enabled with -details
5723 * profile-count.h (precise_p): New function.
5725 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
5727 PR tree-optimization/83922
5728 * tree-vect-loop.c (vect_verify_full_masking): Return false if
5729 there are no statements that need masking.
5730 (vect_active_double_reduction_p): New function.
5731 (vect_analyze_loop_operations): Use it when handling phis that
5732 are not in the loop header.
5734 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
5736 PR tree-optimization/83914
5737 * tree-vect-loop.c (vectorizable_induction): Don't convert
5738 init_expr or apply the peeling adjustment for inductions
5739 that are nested within the vectorized loop.
5741 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5743 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
5746 2018-01-18 Jakub Jelinek <jakub@redhat.com>
5750 * function.h (gimplify_parameters): Add gimple_seq * argument.
5751 * function.c: Include gimple.h and options.h.
5752 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
5753 for the added local temporaries if needed.
5754 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
5755 if there are any parameter cleanups, wrap whole body into a
5756 try/finally with the cleanups.
5758 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
5761 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
5762 Use GET_MODE_CLASS for scalar floating point.
5764 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
5768 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
5769 Fix call of call_cgraph_insertion_hooks.
5771 2018-01-18 Martin Sebor <msebor@redhat.com>
5773 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
5775 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
5778 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
5781 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
5784 * common.opt: (-ffile-prefix-map): New option.
5785 * opts.c (common_handle_option): Defer it.
5786 * opts-global.c (handle_common_deferred_options): Handle it.
5787 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
5788 * file-prefix-map.h: New file.
5789 (remap_debug_filename, add_debug_prefix_map): ...here.
5790 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
5791 * final.c (debug_prefix_map, add_debug_prefix_map
5792 remap_debug_filename): Move to...
5793 * file-prefix-map.c: New file.
5794 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
5795 generalize, get rid of alloca(), use strrchr() instead of strchr().
5796 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
5797 Implement in terms of add_prefix_map().
5798 (remap_macro_filename, remap_debug_filename): Implement in term of
5800 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
5801 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
5802 * dbxout.c: Include file-prefix-map.h.
5803 * varasm.c: Likewise.
5804 * vmsdbgout.c: Likewise.
5805 * xcoffout.c: Likewise.
5806 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
5807 * doc/cppopts.texi (-fmacro-prefix-map): Document.
5808 * doc/invoke.texi (-ffile-prefix-map): Document.
5809 (-fdebug-prefix-map): Update description.
5811 2018-01-18 Martin Liska <mliska@suse.cz>
5813 * config/i386/i386.c (indirect_thunk_name): Document that also
5815 (output_indirect_thunk): Document why both instructions
5816 (pause and lfence) are generated.
5818 2018-01-18 Richard Biener <rguenther@suse.de>
5820 PR tree-optimization/83887
5821 * graphite-scop-detection.c
5822 (scop_detection::get_nearest_dom_with_single_entry): Remove.
5823 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
5824 (scop_detection::merge_sese): Re-implement with a flood-fill
5825 algorithm that properly finds a SESE region if it exists.
5827 2018-01-18 Jakub Jelinek <jakub@redhat.com>
5830 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
5831 pointer_diff optimizations use view_convert instead of convert.
5833 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5835 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5836 Generate different code for -mno-speculate-indirect-jumps.
5837 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5838 (*call_indirect_aix<mode>): Disable for
5839 -mno-speculate-indirect-jumps.
5840 (*call_indirect_aix<mode>_nospec): New define_insn.
5841 (*call_value_indirect_aix<mode>): Disable for
5842 -mno-speculate-indirect-jumps.
5843 (*call_value_indirect_aix<mode>_nospec): New define_insn.
5844 (*sibcall_nonlocal_sysv<mode>): Generate different code for
5845 -mno-speculate-indirect-jumps.
5846 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5848 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
5850 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
5851 long double type, set the flags for noting the default long double
5852 type, even if we don't pass or return a long double type.
5854 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
5857 * ipa-inline.c (flatten_function): Do not overwrite final inlining
5860 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
5862 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
5863 support for merge[hl].
5864 (fold_mergehl_helper): New helper function.
5865 (tree-vector-builder.h): New #include for tree_vector_builder usage.
5866 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
5867 (altivec_vmrglw_direct): Add xxmrglw insn.
5869 2018-01-17 Andrew Waterman <andrew@sifive.com>
5871 * config/riscv/riscv.c (riscv_conditional_register_usage): If
5872 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
5874 2018-01-17 David Malcolm <dmalcolm@redhat.com>
5877 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
5878 call the lto_location_cache before reading the
5879 DECL_SOURCE_LOCATION of the types.
5881 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
5882 Richard Sandiford <richard.sandiford@linaro.org>
5884 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
5885 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
5886 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
5887 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
5888 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
5890 * config/aarch64/constraints.md (aarch64_movti_operand):
5892 * config/aarch64/predicates.md (Uti): Add new constraint.
5894 2018-01-17 Carl Love <cel@us.ibm.com>
5896 * config/rs6000/vsx.md (define_expand xl_len_r,
5897 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
5898 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
5900 (define_expand, define_insn): Move the shift left from the
5901 define_insn to the define_expand for lxvl and stxvl instructions.
5902 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
5903 and XL_LEN_R definitions to PURE.
5905 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
5907 * config/i386/i386.c (indirect_thunk_name): Declare regno
5908 as unsigned int. Compare regno with INVALID_REGNUM.
5909 (output_indirect_thunk): Ditto.
5910 (output_indirect_thunk_function): Ditto.
5911 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
5912 in the call to output_indirect_thunk_function.
5914 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
5917 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
5918 rather than the size of inner_type to determine the stack slot size
5919 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
5921 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
5924 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
5927 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
5929 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
5930 endian Linux systems to optionally enable multilibs for selecting
5931 the long double type if the user configured an explicit type.
5932 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
5933 have no long double multilibs if not defined.
5934 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
5935 warn if the user used -mabi={ieee,ibm}longdouble and we built
5936 multilibs for long double.
5937 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
5938 appropriate multilib option.
5939 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
5941 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
5942 for building long double multilibs.
5943 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
5945 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
5947 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
5950 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
5952 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
5955 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
5958 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
5961 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5963 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
5964 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
5966 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
5968 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
5969 different rtl trees depending on TARGET_64BIT.
5970 (rs6000_gen_lvx): Likewise.
5972 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5974 * config/visium/visium.md (nop): Tweak comment.
5975 (hazard_nop): Likewise.
5977 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5979 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
5980 -mspeculate-indirect-jumps.
5981 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
5982 for -mno-speculate-indirect-jumps.
5983 (*call_indirect_elfv2<mode>_nospec): New define_insn.
5984 (*call_value_indirect_elfv2<mode>): Disable for
5985 -mno-speculate-indirect-jumps.
5986 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
5987 (indirect_jump): Emit different RTL for
5988 -mno-speculate-indirect-jumps.
5989 (*indirect_jump<mode>): Disable for
5990 -mno-speculate-indirect-jumps.
5991 (*indirect_jump<mode>_nospec): New define_insn.
5992 (tablejump): Emit different RTL for
5993 -mno-speculate-indirect-jumps.
5994 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
5995 (tablejumpsi_nospec): New define_expand.
5996 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
5997 (tablejumpdi_nospec): New define_expand.
5998 (*tablejump<mode>_internal1): Disable for
5999 -mno-speculate-indirect-jumps.
6000 (*tablejump<mode>_internal1_nospec): New define_insn.
6001 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
6004 2018-01-16 Artyom Skrobov tyomitch@gmail.com
6006 * caller-save.c (insert_save): Drop unnecessary parameter. All
6009 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6010 Richard Biener <rguenth@suse.de>
6013 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
6014 return early, inline manually is_gimple_sizepos. Make sure if we
6015 call gimplify_expr we don't end up with a gimple constant.
6016 * tree.c (variably_modified_type_p): Don't return true for
6017 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
6018 * gimplify.h (is_gimple_sizepos): Remove.
6020 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6022 PR tree-optimization/83857
6023 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
6024 vectorizable_live_operation for pure SLP statements.
6025 (vectorizable_live_operation): Handle PHIs.
6027 2018-01-16 Richard Biener <rguenther@suse.de>
6029 PR tree-optimization/83867
6030 * tree-vect-stmts.c (vect_transform_stmt): Precompute
6031 nested_in_vect_loop_p since the scalar stmt may get invalidated.
6033 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6036 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
6037 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
6038 If off is not INTEGER_CST, issue a may not be aligned warning
6039 rather than isn't aligned. Use isn%'t rather than isn't.
6040 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
6042 <case MULT_EXPR>: Improve the case when bottom and one of the
6043 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
6044 operand, in that case check if the other operand is multiple of
6045 bottom divided by the INTEGER_CST operand.
6047 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6050 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
6051 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
6052 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
6053 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
6054 * config/pa/pa.c (pa_function_arg_advance): Likewise.
6055 (pa_function_arg, pa_arg_partial_bytes): Likewise.
6056 (pa_function_arg_size): New function.
6058 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6060 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
6061 in a separate statement.
6063 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6065 PR tree-optimization/83847
6066 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
6067 group gathers and scatters.
6069 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6071 PR rtl-optimization/86620
6072 * params.def (max-sched-ready-insns): Bump minimum value to 1.
6074 PR rtl-optimization/83213
6075 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
6076 to last if both are JUMP_INSNs.
6078 PR tree-optimization/83843
6079 * gimple-ssa-store-merging.c
6080 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
6081 store_immediate_info for bswap/nop orig_stores.
6083 2018-01-15 Andrew Waterman <andrew@sifive.com>
6085 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
6087 <UDIV>: Increase cost if !TARGET_DIV.
6089 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
6091 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
6092 (define_attr "cr_logical_3op"): New.
6093 (cceq_ior_compare): Adjust.
6094 (cceq_ior_compare_complement): Adjust.
6095 (*cceq_rev_compare): Adjust.
6096 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
6097 (is_cracked_insn): Adjust.
6098 (insn_must_be_first_in_group): Adjust.
6099 * config/rs6000/40x.md: Adjust.
6100 * config/rs6000/440.md: Adjust.
6101 * config/rs6000/476.md: Adjust.
6102 * config/rs6000/601.md: Adjust.
6103 * config/rs6000/603.md: Adjust.
6104 * config/rs6000/6xx.md: Adjust.
6105 * config/rs6000/7450.md: Adjust.
6106 * config/rs6000/7xx.md: Adjust.
6107 * config/rs6000/8540.md: Adjust.
6108 * config/rs6000/cell.md: Adjust.
6109 * config/rs6000/e300c2c3.md: Adjust.
6110 * config/rs6000/e500mc.md: Adjust.
6111 * config/rs6000/e500mc64.md: Adjust.
6112 * config/rs6000/e5500.md: Adjust.
6113 * config/rs6000/e6500.md: Adjust.
6114 * config/rs6000/mpc.md: Adjust.
6115 * config/rs6000/power4.md: Adjust.
6116 * config/rs6000/power5.md: Adjust.
6117 * config/rs6000/power6.md: Adjust.
6118 * config/rs6000/power7.md: Adjust.
6119 * config/rs6000/power8.md: Adjust.
6120 * config/rs6000/power9.md: Adjust.
6121 * config/rs6000/rs64.md: Adjust.
6122 * config/rs6000/titan.md: Adjust.
6124 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6126 * config/i386/predicates.md (indirect_branch_operand): Rewrite
6127 ix86_indirect_branch_register logic.
6129 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6131 * config/i386/constraints.md (Bs): Update
6132 ix86_indirect_branch_register check. Don't check
6133 ix86_indirect_branch_register with GOT_memory_operand.
6135 * config/i386/predicates.md (GOT_memory_operand): Don't check
6136 ix86_indirect_branch_register here.
6137 (GOT32_symbol_operand): Likewise.
6139 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6141 * config/i386/predicates.md (constant_call_address_operand):
6142 Rewrite ix86_indirect_branch_register logic.
6143 (sibcall_insn_operand): Likewise.
6145 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6147 * config/i386/constraints.md (Bs): Replace
6148 ix86_indirect_branch_thunk_register with
6149 ix86_indirect_branch_register.
6151 * config/i386/i386.md (indirect_jump): Likewise.
6152 (tablejump): Likewise.
6153 (*sibcall_memory): Likewise.
6154 (*sibcall_value_memory): Likewise.
6155 Peepholes of indirect call and jump via memory: Likewise.
6156 * config/i386/i386.opt: Likewise.
6157 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6158 (GOT_memory_operand): Likewise.
6159 (call_insn_operand): Likewise.
6160 (sibcall_insn_operand): Likewise.
6161 (GOT32_symbol_operand): Likewise.
6163 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6166 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
6167 type rather than type addr's type points to.
6168 (expand_omp_atomic_mutex): Likewise.
6169 (expand_omp_atomic): Likewise.
6171 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6174 * config/i386/i386.c (output_indirect_thunk_function): Use
6175 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
6176 for __x86_return_thunk.
6178 2018-01-15 Richard Biener <rguenther@suse.de>
6181 * expmed.c (extract_bit_field_1): Fix typo.
6183 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6186 * config/arm/iterators.md (VF): New mode iterator.
6187 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
6188 Remove integer-related logic from pattern.
6189 (neon_vabd<mode>_3): Likewise.
6191 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6194 * common.opt (fstrict-overflow): No longer an alias.
6195 (fwrapv-pointer): New option.
6196 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
6197 also for pointer types based on flag_wrapv_pointer.
6198 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
6199 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
6200 opts->x_flag_wrapv got set.
6201 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
6202 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
6203 POINTER_TYPE_OVERFLOW_UNDEFINED.
6204 * match.pd: Likewise in address comparison pattern.
6205 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
6207 2018-01-15 Richard Biener <rguenther@suse.de>
6210 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
6211 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
6212 Reset type names to their identifier if their TYPE_DECL doesn't
6213 have linkage (and thus is used for ODR and devirt).
6214 (save_debug_info_for_decl): Remove.
6215 (save_debug_info_for_type): Likewise.
6216 (add_tree_to_fld_list): Adjust.
6217 * tree-pretty-print.c (dump_generic_node): Make dumping of
6218 type names more robust.
6220 2018-01-15 Richard Biener <rguenther@suse.de>
6222 * BASE-VER: Bump to 8.0.1.
6224 2018-01-14 Martin Sebor <msebor@redhat.com>
6227 * builtins.c (check_access): Avoid warning when the no-warning bit
6230 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
6232 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
6233 * ira-color (allocno_hard_regs_compare): Likewise.
6235 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
6238 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
6239 Use .pushsection/.popsection.
6241 2018-01-14 Martin Sebor <msebor@redhat.com>
6244 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
6246 2018-01-14 Jakub Jelinek <jakub@redhat.com>
6248 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
6249 entry from extra_headers.
6250 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
6251 extra_headers, make the list bitwise identical to the i?86-*-* one.
6253 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6255 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
6256 -mcmodel=large with -mindirect-branch=thunk,
6257 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
6258 -mfunction-return=thunk-extern.
6259 * doc/invoke.texi: Document -mcmodel=large is incompatible with
6260 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
6261 -mfunction-return=thunk and -mfunction-return=thunk-extern.
6263 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6265 * config/i386/i386.c (print_reg): Print the name of the full
6266 integer register without '%'.
6267 (ix86_print_operand): Handle 'V'.
6268 * doc/extend.texi: Document 'V' modifier.
6270 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6272 * config/i386/constraints.md (Bs): Disallow memory operand for
6273 -mindirect-branch-register.
6275 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6276 (GOT_memory_operand): Likewise.
6277 (call_insn_operand): Likewise.
6278 (sibcall_insn_operand): Likewise.
6279 (GOT32_symbol_operand): Likewise.
6280 * config/i386/i386.md (indirect_jump): Call convert_memory_address
6281 for -mindirect-branch-register.
6282 (tablejump): Likewise.
6283 (*sibcall_memory): Likewise.
6284 (*sibcall_value_memory): Likewise.
6285 Disallow peepholes of indirect call and jump via memory for
6286 -mindirect-branch-register.
6287 (*call_pop): Replace m with Bw.
6288 (*call_value_pop): Likewise.
6289 (*sibcall_pop_memory): Replace m with Bs.
6290 * config/i386/i386.opt (mindirect-branch-register): New option.
6291 * doc/invoke.texi: Document -mindirect-branch-register option.
6293 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6295 * config/i386/i386-protos.h (ix86_output_function_return): New.
6296 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
6297 set function_return_type.
6298 (indirect_thunk_name): Add ret_p to indicate thunk for function
6300 (output_indirect_thunk_function): Pass false to
6301 indirect_thunk_name.
6302 (ix86_output_indirect_branch_via_reg): Likewise.
6303 (ix86_output_indirect_branch_via_push): Likewise.
6304 (output_indirect_thunk_function): Create alias for function
6305 return thunk if regno < 0.
6306 (ix86_output_function_return): New function.
6307 (ix86_handle_fndecl_attribute): Handle function_return.
6308 (ix86_attribute_table): Add function_return.
6309 * config/i386/i386.h (machine_function): Add
6310 function_return_type.
6311 * config/i386/i386.md (simple_return_internal): Use
6312 ix86_output_function_return.
6313 (simple_return_internal_long): Likewise.
6314 * config/i386/i386.opt (mfunction-return=): New option.
6315 (indirect_branch): Mention -mfunction-return=.
6316 * doc/extend.texi: Document function_return function attribute.
6317 * doc/invoke.texi: Document -mfunction-return= option.
6319 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6321 * config/i386/i386-opts.h (indirect_branch): New.
6322 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
6323 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
6324 with local indirect jump when converting indirect call and jump.
6325 (ix86_set_indirect_branch_type): New.
6326 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
6327 (indirectlabelno): New.
6328 (indirect_thunk_needed): Likewise.
6329 (indirect_thunk_bnd_needed): Likewise.
6330 (indirect_thunks_used): Likewise.
6331 (indirect_thunks_bnd_used): Likewise.
6332 (INDIRECT_LABEL): Likewise.
6333 (indirect_thunk_name): Likewise.
6334 (output_indirect_thunk): Likewise.
6335 (output_indirect_thunk_function): Likewise.
6336 (ix86_output_indirect_branch_via_reg): Likewise.
6337 (ix86_output_indirect_branch_via_push): Likewise.
6338 (ix86_output_indirect_branch): Likewise.
6339 (ix86_output_indirect_jmp): Likewise.
6340 (ix86_code_end): Call output_indirect_thunk_function if needed.
6341 (ix86_output_call_insn): Call ix86_output_indirect_branch if
6343 (ix86_handle_fndecl_attribute): Handle indirect_branch.
6344 (ix86_attribute_table): Add indirect_branch.
6345 * config/i386/i386.h (machine_function): Add indirect_branch_type
6346 and has_local_indirect_jump.
6347 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
6349 (tablejump): Likewise.
6350 (*indirect_jump): Use ix86_output_indirect_jmp.
6351 (*tablejump_1): Likewise.
6352 (simple_return_indirect_internal): Likewise.
6353 * config/i386/i386.opt (mindirect-branch=): New option.
6354 (indirect_branch): New.
6357 (thunk-inline): Likewise.
6358 (thunk-extern): Likewise.
6359 * doc/extend.texi: Document indirect_branch function attribute.
6360 * doc/invoke.texi: Document -mindirect-branch= option.
6362 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
6365 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
6367 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
6369 * ipa-inline.c (want_inline_small_function_p): Return false if
6370 inlining has already failed with CIF_FINAL_ERROR.
6371 (update_caller_keys): Call want_inline_small_function_p before
6373 (update_callee_keys): Likewise.
6375 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6377 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
6379 (rs6000_quadword_masked_address_p): Likewise.
6380 (quad_aligned_load_p): Likewise.
6381 (quad_aligned_store_p): Likewise.
6382 (const_load_sequence_p): Add comment to describe the outer-most loop.
6383 (mimic_memory_attributes_and_flags): New function.
6384 (rs6000_gen_stvx): Likewise.
6385 (replace_swapped_aligned_store): Likewise.
6386 (rs6000_gen_lvx): Likewise.
6387 (replace_swapped_aligned_load): Likewise.
6388 (replace_swapped_load_constant): Capitalize argument name in
6389 comment describing this function.
6390 (rs6000_analyze_swaps): Add a third pass to search for vector loads
6391 and stores that access quad-word aligned addresses and replace
6392 with stvx or lvx instructions when appropriate.
6393 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
6394 New function prototype.
6395 (rs6000_quadword_masked_address_p): Likewise.
6396 (rs6000_gen_lvx): Likewise.
6397 (rs6000_gen_stvx): Likewise.
6398 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
6399 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
6400 when memory address is aligned.
6401 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
6402 this split to select lvx instruction when memory address is aligned.
6403 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
6404 instruction when memory address is aligned.
6405 (*vsx_le_perm_load_v16qi): Likewise.
6406 (four unnamed splitters): Modify to select the stvx instruction
6407 when memory is aligned.
6409 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
6411 * predict.c (determine_unlikely_bbs): Handle correctly BBs
6412 which appears in the queue multiple times.
6414 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6415 Alan Hayward <alan.hayward@arm.com>
6416 David Sherwood <david.sherwood@arm.com>
6418 * tree-vectorizer.h (vec_lower_bound): New structure.
6419 (_loop_vec_info): Add check_nonzero and lower_bounds.
6420 (LOOP_VINFO_CHECK_NONZERO): New macro.
6421 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
6422 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
6423 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
6424 fields. Make seg_len the distance travelled, not including the
6426 (dr_direction_indicator): Declare.
6427 (dr_zero_step_indicator): Likewise.
6428 (dr_known_forward_stride_p): Likewise.
6429 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
6431 (runtime_alias_check_p): Allow runtime alias checks with
6433 (operator ==): Compare access_size and align.
6434 (prune_runtime_alias_test_list): Rework for new distinction between
6435 the access_size and seg_len.
6436 (create_intersect_range_checks_index): Likewise. Cope with polynomial
6438 (get_segment_min_max): New function.
6439 (create_intersect_range_checks): Use it.
6440 (dr_step_indicator): New function.
6441 (dr_direction_indicator): Likewise.
6442 (dr_zero_step_indicator): Likewise.
6443 (dr_known_forward_stride_p): Likewise.
6444 * tree-loop-distribution.c (data_ref_segment_size): Return
6445 DR_STEP * (niters - 1).
6446 (compute_alias_check_pairs): Update call to the dr_with_seg_len
6448 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
6449 (vect_preserves_scalar_order_p): New function, split out from...
6450 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
6451 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
6452 (vect_vfa_access_size): New function.
6453 (vect_vfa_align): Likewise.
6454 (vect_compile_time_alias): Take access_size_a and access_b arguments.
6455 (dump_lower_bound): New function.
6456 (vect_check_lower_bound): Likewise.
6457 (vect_small_gap_p): Likewise.
6458 (vectorizable_with_step_bound_p): Likewise.
6459 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
6460 depencies if the vectorization factor is 1. Convert the checks
6461 for nonzero steps into checks on the bounds of DR_STEP. Try using
6462 a bunds check for variable steps if the minimum required step is
6463 relatively small. Update calls to the dr_with_seg_len
6464 constructor and to vect_compile_time_alias.
6465 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
6467 (vect_loop_versioning): Call it.
6468 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
6470 (vect_estimate_min_profitable_iters): Account for any bounds checks.
6472 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6473 Alan Hayward <alan.hayward@arm.com>
6474 David Sherwood <david.sherwood@arm.com>
6476 * doc/sourcebuild.texi (vect_scatter_store): Document.
6477 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
6479 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
6481 * genopinit.c (main): Add supports_vec_scatter_store and
6482 supports_vec_scatter_store_cached to target_optabs.
6483 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
6484 IFN_MASK_SCATTER_STORE.
6485 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
6487 * internal-fn.h (internal_store_fn_p): Declare.
6488 (internal_fn_stored_value_index): Likewise.
6489 * internal-fn.c (scatter_store_direct): New macro.
6490 (expand_scatter_store_optab_fn): New function.
6491 (direct_scatter_store_optab_supported_p): New macro.
6492 (internal_store_fn_p): New function.
6493 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
6494 IFN_MASK_SCATTER_STORE.
6495 (internal_fn_mask_index): Likewise.
6496 (internal_fn_stored_value_index): New function.
6497 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
6499 * optabs-query.h (supports_vec_scatter_store_p): Declare.
6500 * optabs-query.c (supports_vec_scatter_store_p): New function.
6501 * tree-vectorizer.h (vect_get_store_rhs): Declare.
6502 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
6503 true for scatter stores.
6504 (vect_gather_scatter_fn_p): Handle scatter stores too.
6505 (vect_check_gather_scatter): Consider using scatter stores if
6506 supports_vec_scatter_store_p.
6507 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
6509 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6510 internal_fn_stored_value_index.
6511 (check_load_store_masking): Handle scatter stores too.
6512 (vect_get_store_rhs): Make public.
6513 (vectorizable_call): Use internal_store_fn_p.
6514 (vectorizable_store): Handle scatter store internal functions.
6515 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
6516 when deciding whether the end of the group has been reached.
6517 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
6518 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
6519 (mask_scatter_store<mode>): New insns.
6521 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6522 Alan Hayward <alan.hayward@arm.com>
6523 David Sherwood <david.sherwood@arm.com>
6525 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
6526 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
6527 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
6529 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
6530 Use vect_truncate_gather_scatter_offset if we can't treat the
6531 operation as a normal gather load or scatter store.
6532 (get_group_load_store_type): Take the gather_scatter_info
6533 as argument. Try using a gather load or scatter store for
6534 single-element groups.
6535 (get_load_store_type): Update calls to get_group_load_store_type
6536 and vect_use_strided_gather_scatters_p.
6538 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6539 Alan Hayward <alan.hayward@arm.com>
6540 David Sherwood <david.sherwood@arm.com>
6542 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
6543 optional tree argument.
6544 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
6546 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
6547 but continue to use the current value as a fallback.
6548 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
6549 to compare the updates.
6550 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
6551 (get_load_store_type): Use it when handling a strided access.
6552 (vect_get_strided_load_store_ops): New function.
6553 (vect_get_data_ptr_increment): Likewise.
6554 (vectorizable_load): Handle strided gather loads. Always pass
6555 a step to vect_create_data_ref_ptr and bump_vector_ptr.
6557 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6558 Alan Hayward <alan.hayward@arm.com>
6559 David Sherwood <david.sherwood@arm.com>
6561 * doc/md.texi (gather_load@var{m}): Document.
6562 (mask_gather_load@var{m}): Likewise.
6563 * genopinit.c (main): Add supports_vec_gather_load and
6564 supports_vec_gather_load_cached to target_optabs.
6565 * optabs-tree.c (init_tree_optimization_optabs): Use
6566 ggc_cleared_alloc to allocate target_optabs.
6567 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
6568 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
6570 * internal-fn.h (internal_load_fn_p): Declare.
6571 (internal_gather_scatter_fn_p): Likewise.
6572 (internal_fn_mask_index): Likewise.
6573 (internal_gather_scatter_fn_supported_p): Likewise.
6574 * internal-fn.c (gather_load_direct): New macro.
6575 (expand_gather_load_optab_fn): New function.
6576 (direct_gather_load_optab_supported_p): New macro.
6577 (direct_internal_fn_optab): New function.
6578 (internal_load_fn_p): Likewise.
6579 (internal_gather_scatter_fn_p): Likewise.
6580 (internal_fn_mask_index): Likewise.
6581 (internal_gather_scatter_fn_supported_p): Likewise.
6582 * optabs-query.c (supports_at_least_one_mode_p): New function.
6583 (supports_vec_gather_load_p): Likewise.
6584 * optabs-query.h (supports_vec_gather_load_p): Declare.
6585 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
6586 and memory_type field.
6587 (NUM_PATTERNS): Bump to 15.
6588 * tree-vect-data-refs.c: Include internal-fn.h.
6589 (vect_gather_scatter_fn_p): New function.
6590 (vect_describe_gather_scatter_call): Likewise.
6591 (vect_check_gather_scatter): Try using internal functions for
6592 gather loads. Recognize existing calls to a gather load function.
6593 (vect_analyze_data_refs): Consider using gather loads if
6594 supports_vec_gather_load_p.
6595 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
6596 (vect_get_gather_scatter_offset_type): Likewise.
6597 (vect_convert_mask_for_vectype): Likewise.
6598 (vect_add_conversion_to_patterm): Likewise.
6599 (vect_try_gather_scatter_pattern): Likewise.
6600 (vect_recog_gather_scatter_pattern): New pattern recognizer.
6601 (vect_vect_recog_func_ptrs): Add it.
6602 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6603 internal_fn_mask_index and internal_gather_scatter_fn_p.
6604 (check_load_store_masking): Take the gather_scatter_info as an
6605 argument and handle gather loads.
6606 (vect_get_gather_scatter_ops): New function.
6607 (vectorizable_call): Check internal_load_fn_p.
6608 (vectorizable_load): Likewise. Handle gather load internal
6610 (vectorizable_store): Update call to check_load_store_masking.
6611 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
6612 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
6613 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
6614 (aarch64_gather_scale_operand_d): New predicates.
6615 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
6616 (mask_gather_load<mode>): New insns.
6618 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6619 Alan Hayward <alan.hayward@arm.com>
6620 David Sherwood <david.sherwood@arm.com>
6622 * optabs.def (fold_left_plus_optab): New optab.
6623 * doc/md.texi (fold_left_plus_@var{m}): Document.
6624 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
6625 * internal-fn.c (fold_left_direct): Define.
6626 (expand_fold_left_optab_fn): Likewise.
6627 (direct_fold_left_optab_supported_p): Likewise.
6628 * fold-const-call.c (fold_const_fold_left): New function.
6629 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
6630 * tree-parloops.c (valid_reduction_p): New function.
6631 (gather_scalar_reductions): Use it.
6632 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
6633 (vect_finish_replace_stmt): Declare.
6634 * tree-vect-loop.c (fold_left_reduction_fn): New function.
6635 (needs_fold_left_reduction_p): New function, split out from...
6636 (vect_is_simple_reduction): ...here. Accept reductions that
6637 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
6638 (vect_force_simple_reduction): Also store the reduction type in
6639 the assignment's STMT_VINFO_REDUC_TYPE.
6640 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
6641 (merge_with_identity): New function.
6642 (vect_expand_fold_left): Likewise.
6643 (vectorize_fold_left_reduction): Likewise.
6644 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
6645 scalar phi in place for it. Check for target support and reject
6646 cases that would reassociate the operation. Defer the transform
6647 phase to vectorize_fold_left_reduction.
6648 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
6649 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
6650 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
6652 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6654 * tree-if-conv.c (predicate_mem_writes): Remove redundant
6655 call to ifc_temp_var.
6657 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6658 Alan Hayward <alan.hayward@arm.com>
6659 David Sherwood <david.sherwood@arm.com>
6661 * target.def (legitimize_address_displacement): Take the original
6662 offset as a poly_int.
6663 * targhooks.h (default_legitimize_address_displacement): Update
6665 * targhooks.c (default_legitimize_address_displacement): Likewise.
6666 * doc/tm.texi: Regenerate.
6667 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
6668 as an argument, moving assert of ad->disp == ad->disp_term to...
6669 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
6670 Try calling targetm.legitimize_address_displacement before expanding
6671 the address rather than afterwards, and adjust for the new interface.
6672 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
6673 Match the new hook interface. Handle SVE addresses.
6674 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
6677 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6679 * Makefile.in (OBJS): Add early-remat.o.
6680 * target.def (select_early_remat_modes): New hook.
6681 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
6682 * doc/tm.texi: Regenerate.
6683 * targhooks.h (default_select_early_remat_modes): Declare.
6684 * targhooks.c (default_select_early_remat_modes): New function.
6685 * timevar.def (TV_EARLY_REMAT): New timevar.
6686 * passes.def (pass_early_remat): New pass.
6687 * tree-pass.h (make_pass_early_remat): Declare.
6688 * early-remat.c: New file.
6689 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
6691 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
6693 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6694 Alan Hayward <alan.hayward@arm.com>
6695 David Sherwood <david.sherwood@arm.com>
6697 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
6698 vfm1 with a bound_epilog parameter.
6699 (vect_do_peeling): Update calls accordingly, and move the prologue
6700 call earlier in the function. Treat the base bound_epilog as 0 for
6701 fully-masked loops and retain vf - 1 for other loops. Add 1 to
6702 this base when peeling for gaps.
6703 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
6704 with fully-masked loops.
6705 (vect_estimate_min_profitable_iters): Handle the single peeled
6706 iteration in that case.
6708 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6709 Alan Hayward <alan.hayward@arm.com>
6710 David Sherwood <david.sherwood@arm.com>
6712 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
6713 single-element interleaving even if the size is not a power of 2.
6714 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
6715 accesses for single-element interleaving if the group size is
6718 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6719 Alan Hayward <alan.hayward@arm.com>
6720 David Sherwood <david.sherwood@arm.com>
6722 * doc/md.texi (fold_extract_last_@var{m}): Document.
6723 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
6724 * optabs.def (fold_extract_last_optab): New optab.
6725 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
6726 * internal-fn.c (fold_extract_direct): New macro.
6727 (expand_fold_extract_optab_fn): Likewise.
6728 (direct_fold_extract_optab_supported_p): Likewise.
6729 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
6730 * tree-vect-loop.c (vect_model_reduction_cost): Handle
6731 EXTRACT_LAST_REDUCTION.
6732 (get_initial_def_for_reduction): Do not create an initial vector
6733 for EXTRACT_LAST_REDUCTION reductions.
6734 (vectorizable_reduction): Leave the scalar phi in place for
6735 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
6736 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
6737 epilogue code for EXTRACT_LAST_REDUCTION and defer the
6738 transform phase to vectorizable_condition.
6739 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
6741 (vect_finish_stmt_generation): ...here.
6742 (vect_finish_replace_stmt): New function.
6743 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
6744 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
6746 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
6748 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6749 Alan Hayward <alan.hayward@arm.com>
6750 David Sherwood <david.sherwood@arm.com>
6752 * doc/md.texi (extract_last_@var{m}): Document.
6753 * optabs.def (extract_last_optab): New optab.
6754 * internal-fn.def (EXTRACT_LAST): New internal function.
6755 * internal-fn.c (cond_unary_direct): New macro.
6756 (expand_cond_unary_optab_fn): Likewise.
6757 (direct_cond_unary_optab_supported_p): Likewise.
6758 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
6759 loops using EXTRACT_LAST.
6760 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
6761 (extract_last_<mode>): ...this optab.
6762 (vec_extract<mode><Vel>): Update accordingly.
6764 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6765 Alan Hayward <alan.hayward@arm.com>
6766 David Sherwood <david.sherwood@arm.com>
6768 * target.def (empty_mask_is_expensive): New hook.
6769 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
6770 * doc/tm.texi: Regenerate.
6771 * targhooks.h (default_empty_mask_is_expensive): Declare.
6772 * targhooks.c (default_empty_mask_is_expensive): New function.
6773 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
6774 if the target says that empty masks are expensive.
6775 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
6777 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
6779 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6780 Alan Hayward <alan.hayward@arm.com>
6781 David Sherwood <david.sherwood@arm.com>
6783 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
6784 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
6785 (vect_use_loop_mask_for_alignment_p): New function.
6786 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
6787 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
6788 niters_skip argument. Make sure that the first niters_skip elements
6789 of the first iteration are inactive.
6790 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
6791 Update call to vect_set_loop_masks_directly.
6792 (get_misalign_in_elems): New function, split out from...
6793 (vect_gen_prolog_loop_niters): ...here.
6794 (vect_update_init_of_dr): Take a code argument that specifies whether
6795 the adjustment should be added or subtracted.
6796 (vect_update_init_of_drs): Likewise.
6797 (vect_prepare_for_masked_peels): New function.
6798 (vect_do_peeling): Skip prologue peeling if we're using a mask
6799 instead. Update call to vect_update_inits_of_drs.
6800 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6802 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
6803 alignment. Do not include the number of peeled iterations in
6804 the minimum threshold in that case.
6805 (vectorizable_induction): Adjust the start value down by
6806 LOOP_VINFO_MASK_SKIP_NITERS iterations.
6807 (vect_transform_loop): Call vect_prepare_for_masked_peels.
6808 Take the number of skipped iterations into account when calculating
6810 * tree-vect-stmts.c (vect_gen_while_not): New function.
6812 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6813 Alan Hayward <alan.hayward@arm.com>
6814 David Sherwood <david.sherwood@arm.com>
6816 * doc/sourcebuild.texi (vect_fully_masked): Document.
6817 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
6819 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
6821 (vect_analyze_loop_2): ...here. Don't check the vectorization
6822 factor against the number of loop iterations if the loop is
6825 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6826 Alan Hayward <alan.hayward@arm.com>
6827 David Sherwood <david.sherwood@arm.com>
6829 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
6830 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
6831 (dump_groups): Update accordingly.
6832 (iv_use::mem_type): New member variable.
6833 (address_p): New function.
6834 (record_use): Add a mem_type argument and initialize the new
6836 (record_group_use): Add a mem_type argument. Use address_p.
6837 Remove obsolete null checks of base_object. Update call to record_use.
6838 (find_interesting_uses_op): Update call to record_group_use.
6839 (find_interesting_uses_cond): Likewise.
6840 (find_interesting_uses_address): Likewise.
6841 (get_mem_type_for_internal_fn): New function.
6842 (find_address_like_use): Likewise.
6843 (find_interesting_uses_stmt): Try find_address_like_use before
6844 calling find_interesting_uses_op.
6845 (addr_offset_valid_p): Use the iv mem_type field as the type
6846 of the addressed memory.
6847 (add_autoinc_candidates): Likewise.
6848 (get_address_cost): Likewise.
6849 (split_small_address_groups_p): Use address_p.
6850 (split_address_groups): Likewise.
6851 (add_iv_candidate_for_use): Likewise.
6852 (autoinc_possible_for_pair): Likewise.
6853 (rewrite_groups): Likewise.
6854 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
6855 (determine_group_iv_cost): Update after split of USE_ADDRESS.
6856 (get_alias_ptr_type_for_ptr_address): New function.
6857 (rewrite_use_address): Rewrite address uses in calls that were
6858 identified by find_address_like_use.
6860 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6861 Alan Hayward <alan.hayward@arm.com>
6862 David Sherwood <david.sherwood@arm.com>
6864 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
6866 * gimple-expr.h (is_gimple_addressable: Likewise.
6867 * gimple-expr.c (is_gimple_address): Likewise.
6868 * internal-fn.c (expand_call_mem_ref): New function.
6869 (expand_mask_load_optab_fn): Use it.
6870 (expand_mask_store_optab_fn): Likewise.
6872 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6873 Alan Hayward <alan.hayward@arm.com>
6874 David Sherwood <david.sherwood@arm.com>
6876 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
6877 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
6878 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
6879 (cond_umax@var{mode}): Document.
6880 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
6881 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
6882 (cond_umin_optab, cond_umax_optab): New optabs.
6883 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
6884 (COND_IOR, COND_XOR): New internal functions.
6885 * internal-fn.h (get_conditional_internal_fn): Declare.
6886 * internal-fn.c (cond_binary_direct): New macro.
6887 (expand_cond_binary_optab_fn): Likewise.
6888 (direct_cond_binary_optab_supported_p): Likewise.
6889 (get_conditional_internal_fn): New function.
6890 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
6891 Cope with reduction statements that are vectorized as calls rather
6893 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
6894 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
6895 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
6896 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
6897 (UNSPEC_COND_EOR): New unspecs.
6898 (optab): Add mappings for them.
6899 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
6900 (sve_int_op, sve_fp_op): New int attributes.
6902 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6903 Alan Hayward <alan.hayward@arm.com>
6904 David Sherwood <david.sherwood@arm.com>
6906 * optabs.def (while_ult_optab): New optab.
6907 * doc/md.texi (while_ult@var{m}@var{n}): Document.
6908 * internal-fn.def (WHILE_ULT): New internal function.
6909 * internal-fn.h (direct_internal_fn_supported_p): New override
6910 that takes two types as argument.
6911 * internal-fn.c (while_direct): New macro.
6912 (expand_while_optab_fn): New function.
6913 (convert_optab_supported_p): Likewise.
6914 (direct_while_optab_supported_p): New macro.
6915 * wide-int.h (wi::udiv_ceil): New function.
6916 * tree-vectorizer.h (rgroup_masks): New structure.
6917 (vec_loop_masks): New typedef.
6918 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
6920 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
6921 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
6922 (vect_max_vf): New function.
6923 (slpeel_make_loop_iterate_ntimes): Delete.
6924 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
6925 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
6926 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
6927 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
6928 internal-fn.h, stor-layout.h and optabs-query.h.
6929 (vect_set_loop_mask): New function.
6930 (add_preheader_seq): Likewise.
6931 (add_header_seq): Likewise.
6932 (interleave_supported_p): Likewise.
6933 (vect_maybe_permute_loop_masks): Likewise.
6934 (vect_set_loop_masks_directly): Likewise.
6935 (vect_set_loop_condition_masked): Likewise.
6936 (vect_set_loop_condition_unmasked): New function, split out from
6937 slpeel_make_loop_iterate_ntimes.
6938 (slpeel_make_loop_iterate_ntimes): Rename to..
6939 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
6940 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
6941 (vect_do_peeling): Update call accordingly.
6942 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
6944 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6945 mask_compare_type, can_fully_mask_p and fully_masked_p.
6946 (release_vec_loop_masks): New function.
6947 (_loop_vec_info): Use it to free the loop masks.
6948 (can_produce_all_loop_masks_p): New function.
6949 (vect_get_max_nscalars_per_iter): Likewise.
6950 (vect_verify_full_masking): Likewise.
6951 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
6952 retries, and free the mask rgroups before retrying. Check loop-wide
6953 reasons for disallowing fully-masked loops. Make the final decision
6954 about whether use a fully-masked loop or not.
6955 (vect_estimate_min_profitable_iters): Do not assume that peeling
6956 for the number of iterations will be needed for fully-masked loops.
6957 (vectorizable_reduction): Disable fully-masked loops.
6958 (vectorizable_live_operation): Likewise.
6959 (vect_halve_mask_nunits): New function.
6960 (vect_double_mask_nunits): Likewise.
6961 (vect_record_loop_mask): Likewise.
6962 (vect_get_loop_mask): Likewise.
6963 (vect_transform_loop): Handle the case in which the final loop
6964 iteration might handle a partial vector. Call vect_set_loop_condition
6965 instead of slpeel_make_loop_iterate_ntimes.
6966 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
6967 (check_load_store_masking): New function.
6968 (prepare_load_store_mask): Likewise.
6969 (vectorizable_store): Handle fully-masked loops.
6970 (vectorizable_load): Likewise.
6971 (supportable_widening_operation): Use vect_halve_mask_nunits for
6973 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
6974 (vect_gen_while): New function.
6975 * config/aarch64/aarch64.md (umax<mode>3): New expander.
6976 (aarch64_uqdec<mode>): New insn.
6978 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6979 Alan Hayward <alan.hayward@arm.com>
6980 David Sherwood <david.sherwood@arm.com>
6982 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
6983 (reduc_xor_scal_optab): New optabs.
6984 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
6985 (reduc_xor_scal_@var{m}): Document.
6986 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
6987 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
6989 * fold-const-call.c (fold_const_call): Handle them.
6990 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
6991 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
6992 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
6993 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
6994 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
6995 (UNSPEC_XORV): New unspecs.
6996 (optab): Add entries for them.
6997 (BITWISEV): New int iterator.
6998 (bit_reduc_op): New int attributes.
7000 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7001 Alan Hayward <alan.hayward@arm.com>
7002 David Sherwood <david.sherwood@arm.com>
7004 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
7005 * internal-fn.def (VEC_SHL_INSERT): New internal function.
7006 * optabs.def (vec_shl_insert_optab): New optab.
7007 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
7008 (duplicate_and_interleave): Likewise.
7009 * tree-vect-loop.c: Include internal-fn.h.
7010 (neutral_op_for_slp_reduction): New function, split out from
7011 get_initial_defs_for_reduction.
7012 (get_initial_def_for_reduction): Handle option 2 for variable-length
7013 vectors by loading the neutral value into a vector and then shifting
7014 the initial value into element 0.
7015 (get_initial_defs_for_reduction): Replace the code argument with
7016 the neutral value calculated by neutral_op_for_slp_reduction.
7017 Use gimple_build_vector for constant-length vectors.
7018 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
7019 but the first group_size elements have a neutral value.
7020 Use duplicate_and_interleave otherwise.
7021 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
7022 Update call to get_initial_defs_for_reduction. Handle SLP
7023 reductions for variable-length vectors by creating one vector
7024 result for each scalar result, with the elements associated
7025 with other scalar results stubbed out with the neutral value.
7026 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
7027 Require IFN_VEC_SHL_INSERT for double reductions on
7028 variable-length vectors, or SLP reductions that have
7029 a neutral value. Require can_duplicate_and_interleave_p
7030 support for variable-length unchained SLP reductions if there
7031 is no neutral value, such as for MIN/MAX reductions. Also require
7032 the number of vector elements to be a multiple of the number of
7033 SLP statements when doing variable-length unchained SLP reductions.
7034 Update call to vect_create_epilog_for_reduction.
7035 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
7036 and remove initial values.
7037 (duplicate_and_interleave): Make public.
7038 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
7039 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
7041 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7042 Alan Hayward <alan.hayward@arm.com>
7043 David Sherwood <david.sherwood@arm.com>
7045 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
7046 (can_duplicate_and_interleave_p): New function.
7047 (vect_get_and_check_slp_defs): Take the vector of statements
7048 rather than just the current one. Remove excess parentheses.
7049 Restriction rejectinon of vect_constant_def and vect_external_def
7050 for variable-length vectors to boolean types, or types for which
7051 can_duplicate_and_interleave_p is false.
7052 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
7053 (duplicate_and_interleave): New function.
7054 (vect_get_constant_vectors): Use gimple_build_vector for
7055 constant-length vectors and suitable variable-length constant
7056 vectors. Use duplicate_and_interleave for other variable-length
7057 vectors. Don't defer the update when inserting new statements.
7059 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7060 Alan Hayward <alan.hayward@arm.com>
7061 David Sherwood <david.sherwood@arm.com>
7063 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
7064 min_profitable_iters doesn't go negative.
7066 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7067 Alan Hayward <alan.hayward@arm.com>
7068 David Sherwood <david.sherwood@arm.com>
7070 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
7071 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
7072 * optabs.def (vec_mask_load_lanes_optab): New optab.
7073 (vec_mask_store_lanes_optab): Likewise.
7074 * internal-fn.def (MASK_LOAD_LANES): New internal function.
7075 (MASK_STORE_LANES): Likewise.
7076 * internal-fn.c (mask_load_lanes_direct): New macro.
7077 (mask_store_lanes_direct): Likewise.
7078 (expand_mask_load_optab_fn): Handle masked operations.
7079 (expand_mask_load_lanes_optab_fn): New macro.
7080 (expand_mask_store_optab_fn): Handle masked operations.
7081 (expand_mask_store_lanes_optab_fn): New macro.
7082 (direct_mask_load_lanes_optab_supported_p): Likewise.
7083 (direct_mask_store_lanes_optab_supported_p): Likewise.
7084 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
7086 (vect_load_lanes_supported): Likewise.
7087 * tree-vect-data-refs.c (strip_conversion): New function.
7088 (can_group_stmts_p): Likewise.
7089 (vect_analyze_data_ref_accesses): Use it instead of checking
7090 for a pair of assignments.
7091 (vect_store_lanes_supported): Take a masked_p parameter.
7092 (vect_load_lanes_supported): Likewise.
7093 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
7094 vect_store_lanes_supported and vect_load_lanes_supported.
7095 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7096 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
7097 parameter. Don't allow gaps for masked accesses.
7098 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
7099 and vect_load_lanes_supported.
7100 (get_load_store_type): Take a masked_p parameter and update
7101 call to get_group_load_store_type.
7102 (vectorizable_store): Update call to get_load_store_type.
7103 Handle IFN_MASK_STORE_LANES.
7104 (vectorizable_load): Update call to get_load_store_type.
7105 Handle IFN_MASK_LOAD_LANES.
7107 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7108 Alan Hayward <alan.hayward@arm.com>
7109 David Sherwood <david.sherwood@arm.com>
7111 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
7113 * config/aarch64/aarch64-protos.h
7114 (aarch64_sve_struct_memory_operand_p): Declare.
7115 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
7116 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
7117 (VPRED, vpred): Handle SVE structure modes.
7118 * config/aarch64/constraints.md (Utx): New constraint.
7119 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
7120 (aarch64_sve_struct_nonimmediate_operand): New predicates.
7121 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
7122 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
7123 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
7124 structure modes. Split into pieces after RA.
7125 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
7126 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
7128 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
7129 SVE structure modes.
7130 (aarch64_classify_address): Likewise.
7131 (sizetochar): Move earlier in file.
7132 (aarch64_print_operand): Handle SVE register lists.
7133 (aarch64_array_mode): New function.
7134 (aarch64_sve_struct_memory_operand_p): Likewise.
7135 (TARGET_ARRAY_MODE): Redefine.
7137 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7138 Alan Hayward <alan.hayward@arm.com>
7139 David Sherwood <david.sherwood@arm.com>
7141 * target.def (array_mode): New target hook.
7142 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
7143 * doc/tm.texi: Regenerate.
7144 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
7145 * hooks.c (hook_optmode_mode_uhwi_none): New function.
7146 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
7148 * stor-layout.c (mode_for_array): Likewise. Support polynomial
7151 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7152 Alan Hayward <alan.hayward@arm.com>
7153 David Sherwood <david.sherwood@arm.com>
7155 * fold-const.c (fold_binary_loc): Check the argument types
7156 rather than the result type when testing for a vector operation.
7158 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7160 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
7161 * doc/tm.texi: Regenerate.
7163 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7164 Alan Hayward <alan.hayward@arm.com>
7165 David Sherwood <david.sherwood@arm.com>
7167 * doc/invoke.texi (-msve-vector-bits=): Document new option.
7168 (sve): Document new AArch64 extension.
7169 * doc/md.texi (w): Extend the description of the AArch64
7170 constraint to include SVE vectors.
7171 (Upl, Upa): Document new AArch64 predicate constraints.
7172 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
7174 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
7175 (msve-vector-bits=): New option.
7176 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
7177 SVE when these are disabled.
7178 (sve): New extension.
7179 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
7180 modes. Adjust their number of units based on aarch64_sve_vg.
7181 (MAX_BITSIZE_MODE_ANY_MODE): Define.
7182 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
7183 aarch64_addr_query_type.
7184 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
7185 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
7186 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
7187 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
7188 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
7189 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
7190 (aarch64_simd_imm_zero_p): Delete.
7191 (aarch64_check_zero_based_sve_index_immediate): Declare.
7192 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7193 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7194 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7195 (aarch64_sve_float_mul_immediate_p): Likewise.
7196 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7198 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
7199 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
7200 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
7201 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
7202 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
7203 (aarch64_regmode_natural_size): Likewise.
7204 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
7205 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
7207 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
7208 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
7209 for VG and the SVE predicate registers.
7210 (V_ALIASES): Add a "z"-prefixed alias.
7211 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
7212 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
7213 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
7214 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
7215 (REG_CLASS_NAMES): Add entries for them.
7216 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
7217 and the predicate registers.
7218 (aarch64_sve_vg): Declare.
7219 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
7220 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
7221 (REGMODE_NATURAL_SIZE): Define.
7222 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
7224 * config/aarch64/aarch64.c: Include cfgrtl.h.
7225 (simd_immediate_info): Add a constructor for series vectors,
7226 and an associated step field.
7227 (aarch64_sve_vg): New variable.
7228 (aarch64_dbx_register_number): Handle VG and the predicate registers.
7229 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
7230 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
7231 (VEC_ANY_DATA, VEC_STRUCT): New constants.
7232 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
7233 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
7234 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
7235 (aarch64_get_mask_mode): New functions.
7236 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
7237 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
7238 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
7239 predicate modes and predicate registers. Explicitly restrict
7240 GPRs to modes of 16 bytes or smaller. Only allow FP registers
7241 to store a vector mode if it is recognized by
7242 aarch64_classify_vector_mode.
7243 (aarch64_regmode_natural_size): New function.
7244 (aarch64_hard_regno_caller_save_mode): Return the original mode
7246 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
7247 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
7248 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
7249 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
7251 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
7252 does not overlap dest if the function is frame-related. Handle
7254 (aarch64_split_add_offset): New function.
7255 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
7256 them aarch64_add_offset.
7257 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
7258 and update call to aarch64_sub_sp.
7259 (aarch64_add_cfa_expression): New function.
7260 (aarch64_expand_prologue): Pass extra temporary registers to the
7261 functions above. Handle the case in which we need to emit new
7262 DW_CFA_expressions for registers that were originally saved
7263 relative to the stack pointer, but now have to be expressed
7264 relative to the frame pointer.
7265 (aarch64_output_mi_thunk): Pass extra temporary registers to the
7267 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
7268 IP0 and IP1 values for SVE frames.
7269 (aarch64_expand_vec_series): New function.
7270 (aarch64_expand_sve_widened_duplicate): Likewise.
7271 (aarch64_expand_sve_const_vector): Likewise.
7272 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
7273 Handle SVE constants. Use emit_move_insn to move a force_const_mem
7274 into the register, rather than emitting a SET directly.
7275 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
7276 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
7277 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
7278 (offset_9bit_signed_scaled_p): New functions.
7279 (aarch64_replicate_bitmask_imm): New function.
7280 (aarch64_bitmask_imm): Use it.
7281 (aarch64_cannot_force_const_mem): Reject expressions involving
7282 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
7283 (aarch64_classify_index): Handle SVE indices, by requiring
7284 a plain register index with a scale that matches the element size.
7285 (aarch64_classify_address): Handle SVE addresses. Assert that
7286 the mode of the address is VOIDmode or an integer mode.
7287 Update call to aarch64_classify_symbol.
7288 (aarch64_classify_symbolic_expression): Update call to
7289 aarch64_classify_symbol.
7290 (aarch64_const_vec_all_in_range_p): New function.
7291 (aarch64_print_vector_float_operand): Likewise.
7292 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
7293 "vN" for FP registers with SVE modes. Handle (const ...) vectors
7294 and the FP immediates 1.0 and 0.5.
7295 (aarch64_print_address_internal): Handle SVE addresses.
7296 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
7297 (aarch64_regno_regclass): Handle predicate registers.
7298 (aarch64_secondary_reload): Handle big-endian reloads of SVE
7300 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
7301 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
7302 (aarch64_convert_sve_vector_bits): New function.
7303 (aarch64_override_options): Use it to handle -msve-vector-bits=.
7304 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7306 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
7307 Handle SVE vector and predicate modes. Accept VL-based constants
7308 that need only one temporary register, and VL offsets that require
7309 no temporary registers.
7310 (aarch64_conditional_register_usage): Mark the predicate registers
7311 as fixed if SVE isn't available.
7312 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
7313 Return true for SVE vector and predicate modes.
7314 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
7315 rather than an unsigned int. Handle SVE modes.
7316 (aarch64_preferred_simd_mode): Update call accordingly. Handle
7318 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
7320 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7321 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7322 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7323 (aarch64_sve_float_mul_immediate_p): New functions.
7324 (aarch64_sve_valid_immediate): New function.
7325 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
7326 Explicitly reject structure modes. Check for INDEX constants.
7327 Handle PTRUE and PFALSE constants.
7328 (aarch64_check_zero_based_sve_index_immediate): New function.
7329 (aarch64_simd_imm_zero_p): Delete.
7330 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
7331 vector modes. Accept constants in the range of CNT[BHWD].
7332 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
7333 ask for an Advanced SIMD mode.
7334 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
7335 (aarch64_simd_vector_alignment): Handle SVE predicates.
7336 (aarch64_vectorize_preferred_vector_alignment): New function.
7337 (aarch64_simd_vector_alignment_reachable): Use it instead of
7339 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
7340 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
7342 (MAX_VECT_LEN): Delete.
7343 (expand_vec_perm_d): Add a vec_flags field.
7344 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
7345 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
7346 (aarch64_evpc_ext): Don't apply a big-endian lane correction
7348 (aarch64_evpc_rev): Rename to...
7349 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
7350 (aarch64_evpc_rev_global): New function.
7351 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
7352 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
7354 (aarch64_evpc_sve_tbl): New function.
7355 (aarch64_expand_vec_perm_const_1): Update after rename of
7356 aarch64_evpc_rev. Handle SVE permutes too, trying
7357 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
7358 than aarch64_evpc_tbl.
7359 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
7360 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
7361 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
7362 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
7363 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
7364 (aarch64_expand_sve_vcond): New functions.
7365 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
7366 of aarch64_vector_mode_p.
7367 (aarch64_dwarf_poly_indeterminate_value): New function.
7368 (aarch64_compute_pressure_classes): Likewise.
7369 (aarch64_can_change_mode_class): Likewise.
7370 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
7371 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
7372 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
7373 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
7374 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
7375 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
7376 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
7377 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
7379 (Dn, Dl, Dr): Accept const as well as const_vector.
7380 (Dz): Likewise. Compare against CONST0_RTX.
7381 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
7382 of "vector" where appropriate.
7383 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
7384 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
7385 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
7386 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
7387 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
7388 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
7389 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
7390 (v_int_equiv): Extend to SVE modes.
7391 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
7393 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
7394 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
7395 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
7396 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
7397 (SVE_COND_FP_CMP): New int iterators.
7398 (perm_hilo): Handle the new unpack unspecs.
7399 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
7401 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
7402 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
7403 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
7404 (aarch64_equality_operator, aarch64_constant_vector_operand)
7405 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
7406 (aarch64_sve_nonimmediate_operand): Likewise.
7407 (aarch64_sve_general_operand): Likewise.
7408 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
7409 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
7410 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
7411 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
7412 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
7413 (aarch64_sve_float_arith_immediate): Likewise.
7414 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
7415 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
7416 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
7417 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
7418 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
7419 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
7420 (aarch64_sve_float_arith_operand): Likewise.
7421 (aarch64_sve_float_arith_with_sub_operand): Likewise.
7422 (aarch64_sve_float_mul_operand): Likewise.
7423 (aarch64_sve_vec_perm_operand): Likewise.
7424 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
7425 (aarch64_mov_operand): Accept const_poly_int and const_vector.
7426 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
7427 as well as const_vector.
7428 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
7429 in file. Use CONST0_RTX and CONSTM1_RTX.
7430 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
7431 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
7432 Use aarch64_simd_imm_zero.
7433 * config/aarch64/aarch64-sve.md: New file.
7434 * config/aarch64/aarch64.md: Include it.
7435 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
7436 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
7437 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
7438 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
7439 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
7440 (sve): New attribute.
7441 (enabled): Disable instructions with the sve attribute unless
7443 (movqi, movhi): Pass CONST_POLY_INT operaneds through
7444 aarch64_expand_mov_immediate.
7445 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
7446 CNT[BHSD] immediates.
7447 (movti): Split CONST_POLY_INT moves into two halves.
7448 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
7449 Split additions that need a temporary here if the destination
7450 is the stack pointer.
7451 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
7452 (*add<mode>3_poly_1): New instruction.
7453 (set_clobber_cc): New expander.
7455 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7457 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
7458 parameter and use it instead of GET_MODE_SIZE (innermode). Use
7459 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
7460 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
7461 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
7462 Change innermode from fixed_mode_size to machine_mode.
7463 (simplify_subreg): Update call accordingly. Handle a constant-sized
7464 subreg of a variable-length CONST_VECTOR.
7466 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7467 Alan Hayward <alan.hayward@arm.com>
7468 David Sherwood <david.sherwood@arm.com>
7470 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
7471 (add_offset_to_base): New function, split out from...
7472 (create_mem_ref): ...here. When handling a scale other than 1,
7473 check first whether the address is valid without the offset.
7474 Add it into the base if so, leaving the index and scale as-is.
7476 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7479 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
7480 fold_for_warn before checking if arg2 is INTEGER_CST.
7482 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
7484 * config/rs6000/predicates.md (load_multiple_operation): Delete.
7485 (store_multiple_operation): Delete.
7486 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
7487 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
7488 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
7489 guarded by TARGET_STRING.
7490 (rs6000_output_load_multiple): Delete.
7491 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
7492 OPTION_MASK_STRING / TARGET_STRING handling.
7493 (print_operand) <'N', 'O'>: Add comment that these are unused now.
7494 (const rs6000_opt_masks) <"string">: Change mask to 0.
7495 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
7496 (MASK_STRING): Delete.
7497 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
7499 (load_multiple): Delete.
7506 (store_multiple): Delete.
7513 (movmemsi_8reg): Delete.
7514 (corresponding unnamed define_insn): Delete.
7515 (movmemsi_6reg): Delete.
7516 (corresponding unnamed define_insn): Delete.
7517 (movmemsi_4reg): Delete.
7518 (corresponding unnamed define_insn): Delete.
7519 (movmemsi_2reg): Delete.
7520 (corresponding unnamed define_insn): Delete.
7521 (movmemsi_1reg): Delete.
7522 (corresponding unnamed define_insn): Delete.
7523 * config/rs6000/rs6000.opt (mno-string): New.
7524 (mstring): Replace by deprecation warning stub.
7525 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
7527 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7529 * regrename.c (regrename_do_replace): If replacing the same
7530 reg multiple times, try to reuse last created gen_raw_REG.
7533 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
7534 main to workaround a bug in GDB.
7536 2018-01-12 Tom de Vries <tom@codesourcery.com>
7539 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
7541 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
7543 PR rtl-optimization/80481
7544 * ira-color.c (get_cap_member): New function.
7545 (allocnos_conflict_by_live_ranges_p): Use it.
7546 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
7547 (setup_slot_coalesced_allocno_live_ranges): Ditto.
7549 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
7552 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
7553 (*saddl_se_1): Ditto.
7555 (*ssubl_se_1): Ditto.
7557 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7559 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
7560 rather than wi::to_widest for DR_INITs.
7561 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
7562 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
7563 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
7565 (vect_analyze_group_access_1): Note that here.
7567 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7569 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
7570 polynomial type sizes.
7572 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7574 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
7575 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
7576 (gimple_add_tmp_var): Likewise.
7578 2018-01-12 Martin Liska <mliska@suse.cz>
7580 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
7581 (gimple_alloc_sizes): Likewise.
7582 (dump_gimple_statistics): Use PRIu64 in printf format.
7583 * gimple.h: Change uint64_t to int.
7585 2018-01-12 Martin Liska <mliska@suse.cz>
7587 * tree-core.h: Use uint64_t instead of int.
7588 * tree.c (tree_node_counts): Likewise.
7589 (tree_node_sizes): Likewise.
7590 (dump_tree_statistics): Use PRIu64 in printf format.
7592 2018-01-12 Martin Liska <mliska@suse.cz>
7594 * Makefile.in: As qsort_chk is implemented in vec.c, add
7595 vec.o to linkage of gencfn-macros.
7596 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
7597 passing the info to record_node_allocation_statistics.
7598 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
7600 * ggc-common.c (struct ggc_usage): Add operator== and use
7601 it in operator< and compare function.
7602 * mem-stats.h (struct mem_usage): Likewise.
7603 * vec.c (struct vec_usage): Remove operator< and compare
7604 function. Can be simply inherited.
7606 2018-01-12 Martin Jambor <mjambor@suse.cz>
7609 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
7610 * tree-ssa-math-opts.c: Include domwalk.h.
7611 (convert_mult_to_fma_1): New function.
7612 (fma_transformation_info): New type.
7613 (fma_deferring_state): Likewise.
7614 (cancel_fma_deferring): New function.
7615 (result_of_phi): Likewise.
7616 (last_fma_candidate_feeds_initial_phi): Likewise.
7617 (convert_mult_to_fma): Added deferring logic, split actual
7618 transformation to convert_mult_to_fma_1.
7619 (math_opts_dom_walker): New type.
7620 (math_opts_dom_walker::after_dom_children): New method, body moved
7621 here from pass_optimize_widening_mul::execute, added deferring logic
7623 (pass_optimize_widening_mul::execute): Moved most of code to
7624 math_opts_dom_walker::after_dom_children.
7625 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
7626 * config/i386/i386.c (ix86_option_override_internal): Added
7627 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
7629 2018-01-12 Richard Biener <rguenther@suse.de>
7632 * dwarf2out.c (gen_variable_die): Do not reset old_die for
7633 inline instance vars.
7635 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
7638 * config/rx/rx.c (rx_is_restricted_memory_address):
7641 2018-01-12 Richard Biener <rguenther@suse.de>
7643 PR tree-optimization/80846
7644 * target.def (split_reduction): New target hook.
7645 * targhooks.c (default_split_reduction): New function.
7646 * targhooks.h (default_split_reduction): Declare.
7647 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
7648 target requests first reduce vectors by combining low and high
7650 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
7651 (get_vectype_for_scalar_type_and_size): Export.
7652 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
7653 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
7654 * doc/tm.texi: Regenerate.
7655 * config/i386/i386.c (ix86_split_reduction): Implement
7656 TARGET_VECTORIZE_SPLIT_REDUCTION.
7658 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7661 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
7662 in PIC mode except for TARGET_VXWORKS_RTP.
7663 * config/sparc/sparc.c: Include cfgrtl.h.
7664 (TARGET_INIT_PIC_REG): Define.
7665 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
7666 (sparc_pic_register_p): New predicate.
7667 (sparc_legitimate_address_p): Use it.
7668 (sparc_legitimize_pic_address): Likewise.
7669 (sparc_delegitimize_address): Likewise.
7670 (sparc_mode_dependent_address_p): Likewise.
7671 (gen_load_pcrel_sym): Remove 4th parameter.
7672 (load_got_register): Adjust call to above. Remove obsolete stuff.
7673 (sparc_expand_prologue): Do not call load_got_register here.
7674 (sparc_flat_expand_prologue): Likewise.
7675 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
7676 (sparc_use_pseudo_pic_reg): New function.
7677 (sparc_init_pic_reg): Likewise.
7678 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7679 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
7681 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
7683 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
7684 Add item for branch_cost.
7686 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7688 PR rtl-optimization/83565
7689 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
7690 not extend the result to a larger mode for rotate operations.
7691 (num_sign_bit_copies1): Likewise.
7693 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7696 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
7698 Use values-Xc.o for -pedantic.
7699 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
7701 2018-01-12 Martin Liska <mliska@suse.cz>
7704 * ipa-devirt.c (final_warning_record::grow_type_warnings):
7706 (possible_polymorphic_call_targets): Use it.
7707 (ipa_devirt): Likewise.
7709 2018-01-12 Martin Liska <mliska@suse.cz>
7711 * profile-count.h (enum profile_quality): Use 0 as invalid
7712 enum value of profile_quality.
7714 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
7716 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
7717 -mext-string options.
7719 2018-01-12 Richard Biener <rguenther@suse.de>
7721 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
7722 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
7723 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
7725 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
7727 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
7729 * configure.ac (--with-long-double-format): Add support for the
7730 configuration option to change the default long double format on
7732 * config.gcc (powerpc*-linux*-*): Likewise.
7733 * configure: Regenerate.
7734 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
7735 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
7736 used without modification.
7738 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7740 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
7741 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
7742 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
7743 MISC_BUILTIN_SPEC_BARRIER.
7744 (rs6000_init_builtins): Likewise.
7745 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
7747 (speculation_barrier): New define_insn.
7748 * doc/extend.texi: Document __builtin_speculation_barrier.
7750 2018-01-11 Jakub Jelinek <jakub@redhat.com>
7753 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
7754 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
7755 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
7757 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
7758 integral modes instead of "ss" and "sd".
7759 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
7760 vectors with 32-bit and 64-bit elements.
7761 (vecdupssescalarmodesuffix): New mode attribute.
7762 (vec_dup<mode>): Use it.
7764 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
7767 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
7768 frame if argument is passed on stack.
7770 2018-01-11 Jakub Jelinek <jakub@redhat.com>
7773 * ree.c (combine_reaching_defs): Optimize also
7774 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
7775 reg2=any_extend(exp); reg1=reg2;, formatting fix.
7777 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
7780 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
7782 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
7785 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
7786 after they are computed.
7788 2018-01-11 Bin Cheng <bin.cheng@arm.com>
7790 PR tree-optimization/83695
7791 * gimple-loop-linterchange.cc
7792 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
7793 reset cached scev information after interchange.
7794 (pass_linterchange::execute): Remove call to scev_reset_htab.
7796 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7798 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
7799 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
7800 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
7801 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
7802 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
7803 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
7804 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
7805 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
7806 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
7807 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
7808 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
7809 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
7810 (V_lane_reg): Likewise.
7811 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
7813 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
7814 (vfmal_lane_low<mode>_intrinsic,
7815 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
7816 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
7817 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
7818 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
7819 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
7820 vfmsl_lane_high<mode>_intrinsic): New define_insns.
7822 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7824 * config/arm/arm-cpus.in (fp16fml): New feature.
7825 (ALL_SIMD): Add fp16fml.
7826 (armv8.2-a): Add fp16fml as an option.
7827 (armv8.3-a): Likewise.
7828 (armv8.4-a): Add fp16fml as part of fp16.
7829 * config/arm/arm.h (TARGET_FP16FML): Define.
7830 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
7832 * config/arm/arm-modes.def (V2HF): Define.
7833 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
7834 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
7835 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
7836 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
7837 vfmsl_low, vfmsl_high): New set of builtins.
7838 * config/arm/iterators.md (PLUSMINUS): New code iterator.
7839 (vfml_op): New code attribute.
7840 (VFMLHALVES): New int iterator.
7841 (VFML, VFMLSEL): New mode attributes.
7842 (V_reg): Define mapping for V2HF.
7843 (V_hi, V_lo): New mode attributes.
7844 (VF_constraint): Likewise.
7845 (vfml_half, vfml_half_selector): New int attributes.
7846 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
7848 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
7849 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
7851 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
7852 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
7853 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
7854 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
7856 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
7857 Document new effective target and option set.
7859 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7861 * config/arm/arm-cpus.in (armv8_4): New feature.
7862 (ARMv8_4a): New fgroup.
7863 (armv8.4-a): New arch.
7864 * config/arm/arm-tables.opt: Regenerate.
7865 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
7866 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
7867 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
7868 Add matching rules for -march=armv8.4-a and extensions.
7869 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
7871 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
7874 * config/rx/rx.md (BW): New mode attribute.
7875 (sync_lock_test_and_setsi): Add mode suffix to insn output.
7877 2018-01-11 Richard Biener <rguenther@suse.de>
7879 PR tree-optimization/83435
7880 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
7881 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
7882 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
7884 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7885 Alan Hayward <alan.hayward@arm.com>
7886 David Sherwood <david.sherwood@arm.com>
7888 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
7890 (aarch64_classify_address): Initialize it. Track polynomial offsets.
7891 (aarch64_print_address_internal): Use it to check for a zero offset.
7893 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7894 Alan Hayward <alan.hayward@arm.com>
7895 David Sherwood <david.sherwood@arm.com>
7897 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
7898 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
7899 Return a poly_int64 rather than a HOST_WIDE_INT.
7900 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
7901 rather than a HOST_WIDE_INT.
7902 * config/aarch64/aarch64.h (aarch64_frame): Protect with
7903 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
7904 hard_fp_offset, frame_size, initial_adjust, callee_offset and
7905 final_offset from HOST_WIDE_INT to poly_int64.
7906 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
7907 to_constant when getting the number of units in an Advanced SIMD
7909 (aarch64_builtin_vectorized_function): Check for a constant number
7911 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
7913 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
7914 attribute instead of GET_MODE_NUNITS.
7915 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
7916 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
7917 GET_MODE_SIZE for fixed-size registers.
7918 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
7919 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
7920 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
7921 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
7922 (aarch64_print_operand, aarch64_print_address_internal)
7923 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
7924 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
7925 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
7926 Handle polynomial GET_MODE_SIZE.
7927 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
7928 wider than SImode without modification.
7929 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
7930 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
7931 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
7932 passing and returning SVE modes.
7933 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
7934 rather than GEN_INT.
7935 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
7936 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
7937 (aarch64_allocate_and_probe_stack_space): Likewise.
7938 (aarch64_layout_frame): Cope with polynomial offsets.
7939 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
7940 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
7942 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
7943 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
7944 poly_int64 rather than a HOST_WIDE_INT.
7945 (aarch64_get_separate_components, aarch64_process_components)
7946 (aarch64_expand_prologue, aarch64_expand_epilogue)
7947 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
7948 (aarch64_anchor_offset): New function, split out from...
7949 (aarch64_legitimize_address): ...here.
7950 (aarch64_builtin_vectorization_cost): Handle polynomial
7951 TYPE_VECTOR_SUBPARTS.
7952 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
7954 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
7955 number of elements from the PARALLEL rather than the mode.
7956 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
7957 rather than GET_MODE_BITSIZE.
7958 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
7959 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
7960 (aarch64_expand_vec_perm_const_1): Handle polynomial
7961 d->perm.length () and d->perm elements.
7962 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
7963 Apply to_constant to d->perm elements.
7964 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
7965 polynomial CONST_VECTOR_NUNITS.
7966 (aarch64_move_pointer): Take amount as a poly_int64 rather
7968 (aarch64_progress_pointer): Avoid temporary variable.
7969 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
7970 the mode attribute instead of GET_MODE.
7972 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7973 Alan Hayward <alan.hayward@arm.com>
7974 David Sherwood <david.sherwood@arm.com>
7976 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
7977 x exists before using it.
7978 (aarch64_add_constant_internal): Rename to...
7979 (aarch64_add_offset_1): ...this. Replace regnum with separate
7980 src and dest rtxes. Handle the case in which they're different,
7981 including when the offset is zero. Replace scratchreg with an rtx.
7982 Use 2 additions if there is no spare register into which we can
7983 move a 16-bit constant.
7984 (aarch64_add_constant): Delete.
7985 (aarch64_add_offset): Replace reg with separate src and dest
7986 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
7987 Use aarch64_add_offset_1.
7988 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
7989 an rtx rather than an int. Take the delta as a poly_int64
7990 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
7991 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
7992 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
7993 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
7994 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
7996 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
7997 aarch64_add_constant.
7999 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8001 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
8002 Use scalar_float_mode.
8004 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8006 * config/aarch64/aarch64-simd.md
8007 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
8008 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
8009 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
8010 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
8011 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
8012 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
8013 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
8014 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
8015 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
8016 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
8018 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8021 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
8022 targ_options->x_arm_arch_string is non NULL.
8024 2018-01-11 Tamar Christina <tamar.christina@arm.com>
8026 * config/aarch64/aarch64.h
8027 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
8029 2018-01-11 Sudakshina Das <sudi.das@arm.com>
8032 * expmed.c (emit_store_flag_force): Swap if const op0
8033 and change VOIDmode to mode of op0.
8035 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8037 PR rtl-optimization/83761
8038 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
8039 than bytes to mode_for_size.
8041 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8044 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
8045 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
8048 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8051 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
8052 when in layout mode.
8053 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
8054 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
8057 2018-01-10 Michael Collison <michael.collison@arm.com>
8059 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
8060 * config/aarch64/aarch64-option-extension.def: Add
8061 AARCH64_OPT_EXTENSION of 'fp16fml'.
8062 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8063 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
8064 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
8065 * config/aarch64/constraints.md (Ui7): New constraint.
8066 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
8067 (VFMLA_SEL_W): Ditto.
8070 (VFMLA16_LOW): New int iterator.
8071 (VFMLA16_HIGH): Ditto.
8072 (UNSPEC_FMLAL): New unspec.
8073 (UNSPEC_FMLSL): Ditto.
8074 (UNSPEC_FMLAL2): Ditto.
8075 (UNSPEC_FMLSL2): Ditto.
8076 (f16mac): New code attribute.
8077 * config/aarch64/aarch64-simd-builtins.def
8078 (aarch64_fmlal_lowv2sf): Ditto.
8079 (aarch64_fmlsl_lowv2sf): Ditto.
8080 (aarch64_fmlalq_lowv4sf): Ditto.
8081 (aarch64_fmlslq_lowv4sf): Ditto.
8082 (aarch64_fmlal_highv2sf): Ditto.
8083 (aarch64_fmlsl_highv2sf): Ditto.
8084 (aarch64_fmlalq_highv4sf): Ditto.
8085 (aarch64_fmlslq_highv4sf): Ditto.
8086 (aarch64_fmlal_lane_lowv2sf): Ditto.
8087 (aarch64_fmlsl_lane_lowv2sf): Ditto.
8088 (aarch64_fmlal_laneq_lowv2sf): Ditto.
8089 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
8090 (aarch64_fmlalq_lane_lowv4sf): Ditto.
8091 (aarch64_fmlsl_lane_lowv4sf): Ditto.
8092 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
8093 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
8094 (aarch64_fmlal_lane_highv2sf): Ditto.
8095 (aarch64_fmlsl_lane_highv2sf): Ditto.
8096 (aarch64_fmlal_laneq_highv2sf): Ditto.
8097 (aarch64_fmlsl_laneq_highv2sf): Ditto.
8098 (aarch64_fmlalq_lane_highv4sf): Ditto.
8099 (aarch64_fmlsl_lane_highv4sf): Ditto.
8100 (aarch64_fmlalq_laneq_highv4sf): Ditto.
8101 (aarch64_fmlsl_laneq_highv4sf): Ditto.
8102 * config/aarch64/aarch64-simd.md:
8103 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
8104 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8105 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
8106 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8107 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
8108 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
8109 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
8110 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
8111 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
8112 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
8113 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
8114 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
8115 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
8116 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
8117 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
8118 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
8119 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
8120 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
8121 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
8122 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
8123 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
8124 (vfmlsl_low_u32): Ditto.
8125 (vfmlalq_low_u32): Ditto.
8126 (vfmlslq_low_u32): Ditto.
8127 (vfmlal_high_u32): Ditto.
8128 (vfmlsl_high_u32): Ditto.
8129 (vfmlalq_high_u32): Ditto.
8130 (vfmlslq_high_u32): Ditto.
8131 (vfmlal_lane_low_u32): Ditto.
8132 (vfmlsl_lane_low_u32): Ditto.
8133 (vfmlal_laneq_low_u32): Ditto.
8134 (vfmlsl_laneq_low_u32): Ditto.
8135 (vfmlalq_lane_low_u32): Ditto.
8136 (vfmlslq_lane_low_u32): Ditto.
8137 (vfmlalq_laneq_low_u32): Ditto.
8138 (vfmlslq_laneq_low_u32): Ditto.
8139 (vfmlal_lane_high_u32): Ditto.
8140 (vfmlsl_lane_high_u32): Ditto.
8141 (vfmlal_laneq_high_u32): Ditto.
8142 (vfmlsl_laneq_high_u32): Ditto.
8143 (vfmlalq_lane_high_u32): Ditto.
8144 (vfmlslq_lane_high_u32): Ditto.
8145 (vfmlalq_laneq_high_u32): Ditto.
8146 (vfmlslq_laneq_high_u32): Ditto.
8147 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
8148 (AARCH64_FL_FOR_ARCH8_4): New.
8149 (AARCH64_ISA_F16FML): New ISA flag.
8150 (TARGET_F16FML): New feature flag for fp16fml.
8151 (doc/invoke.texi): Document new fp16fml option.
8153 2018-01-10 Michael Collison <michael.collison@arm.com>
8155 * config/aarch64/aarch64-builtins.c:
8156 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
8157 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8158 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
8159 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
8160 (AARCH64_ISA_SHA3): New ISA flag.
8161 (TARGET_SHA3): New feature flag for sha3.
8162 * config/aarch64/iterators.md (sha512_op): New int attribute.
8163 (CRYPTO_SHA512): New int iterator.
8164 (UNSPEC_SHA512H): New unspec.
8165 (UNSPEC_SHA512H2): Ditto.
8166 (UNSPEC_SHA512SU0): Ditto.
8167 (UNSPEC_SHA512SU1): Ditto.
8168 * config/aarch64/aarch64-simd-builtins.def
8169 (aarch64_crypto_sha512hqv2di): New builtin.
8170 (aarch64_crypto_sha512h2qv2di): Ditto.
8171 (aarch64_crypto_sha512su0qv2di): Ditto.
8172 (aarch64_crypto_sha512su1qv2di): Ditto.
8173 (aarch64_eor3qv8hi): Ditto.
8174 (aarch64_rax1qv2di): Ditto.
8175 (aarch64_xarqv2di): Ditto.
8176 (aarch64_bcaxqv8hi): Ditto.
8177 * config/aarch64/aarch64-simd.md:
8178 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
8179 (aarch64_crypto_sha512su0qv2di): Ditto.
8180 (aarch64_crypto_sha512su1qv2di): Ditto.
8181 (aarch64_eor3qv8hi): Ditto.
8182 (aarch64_rax1qv2di): Ditto.
8183 (aarch64_xarqv2di): Ditto.
8184 (aarch64_bcaxqv8hi): Ditto.
8185 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
8186 (vsha512h2q_u64): Ditto.
8187 (vsha512su0q_u64): Ditto.
8188 (vsha512su1q_u64): Ditto.
8189 (veor3q_u16): Ditto.
8190 (vrax1q_u64): Ditto.
8192 (vbcaxq_u16): Ditto.
8193 * config/arm/types.md (crypto_sha512): New type attribute.
8194 (crypto_sha3): Ditto.
8195 (doc/invoke.texi): Document new sha3 option.
8197 2018-01-10 Michael Collison <michael.collison@arm.com>
8199 * config/aarch64/aarch64-builtins.c:
8200 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
8201 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8202 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
8203 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
8204 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
8205 (AARCH64_ISA_SM4): New ISA flag.
8206 (TARGET_SM4): New feature flag for sm4.
8207 * config/aarch64/aarch64-simd-builtins.def
8208 (aarch64_sm3ss1qv4si): Ditto.
8209 (aarch64_sm3tt1aq4si): Ditto.
8210 (aarch64_sm3tt1bq4si): Ditto.
8211 (aarch64_sm3tt2aq4si): Ditto.
8212 (aarch64_sm3tt2bq4si): Ditto.
8213 (aarch64_sm3partw1qv4si): Ditto.
8214 (aarch64_sm3partw2qv4si): Ditto.
8215 (aarch64_sm4eqv4si): Ditto.
8216 (aarch64_sm4ekeyqv4si): Ditto.
8217 * config/aarch64/aarch64-simd.md:
8218 (aarch64_sm3ss1qv4si): Ditto.
8219 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
8220 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
8221 (aarch64_sm4eqv4si): Ditto.
8222 (aarch64_sm4ekeyqv4si): Ditto.
8223 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
8224 (sm3part_op): Ditto.
8225 (CRYPTO_SM3TT): Ditto.
8226 (CRYPTO_SM3PART): Ditto.
8227 (UNSPEC_SM3SS1): New unspec.
8228 (UNSPEC_SM3TT1A): Ditto.
8229 (UNSPEC_SM3TT1B): Ditto.
8230 (UNSPEC_SM3TT2A): Ditto.
8231 (UNSPEC_SM3TT2B): Ditto.
8232 (UNSPEC_SM3PARTW1): Ditto.
8233 (UNSPEC_SM3PARTW2): Ditto.
8234 (UNSPEC_SM4E): Ditto.
8235 (UNSPEC_SM4EKEY): Ditto.
8236 * config/aarch64/constraints.md (Ui2): New constraint.
8237 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
8238 * config/arm/types.md (crypto_sm3): New type attribute.
8239 (crypto_sm4): Ditto.
8240 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
8241 (vsm3tt1aq_u32): Ditto.
8242 (vsm3tt1bq_u32): Ditto.
8243 (vsm3tt2aq_u32): Ditto.
8244 (vsm3tt2bq_u32): Ditto.
8245 (vsm3partw1q_u32): Ditto.
8246 (vsm3partw2q_u32): Ditto.
8247 (vsm4eq_u32): Ditto.
8248 (vsm4ekeyq_u32): Ditto.
8249 (doc/invoke.texi): Document new sm4 option.
8251 2018-01-10 Michael Collison <michael.collison@arm.com>
8253 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
8254 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
8255 (AARCH64_FL_FOR_ARCH8_4): New.
8256 (AARCH64_FL_V8_4): New flag.
8257 (doc/invoke.texi): Document new armv8.4-a option.
8259 2018-01-10 Michael Collison <michael.collison@arm.com>
8261 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8262 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
8263 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
8264 * config/aarch64/aarch64-option-extension.def: Add
8265 AARCH64_OPT_EXTENSION of 'sha2'.
8266 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
8267 (crypto): Disable sha2 and aes if crypto disabled.
8268 (crypto): Enable aes and sha2 if enabled.
8269 (simd): Disable sha2 and aes if simd disabled.
8270 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
8272 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
8273 (TARGET_SHA2): New feature flag for sha2.
8274 (TARGET_AES): New feature flag for aes.
8275 * config/aarch64/aarch64-simd.md:
8276 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
8277 conditional on TARGET_AES.
8278 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
8279 (aarch64_crypto_sha1hsi): Make pattern conditional
8281 (aarch64_crypto_sha1hv4si): Ditto.
8282 (aarch64_be_crypto_sha1hv4si): Ditto.
8283 (aarch64_crypto_sha1su1v4si): Ditto.
8284 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
8285 (aarch64_crypto_sha1su0v4si): Ditto.
8286 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
8287 (aarch64_crypto_sha256su0v4si): Ditto.
8288 (aarch64_crypto_sha256su1v4si): Ditto.
8289 (doc/invoke.texi): Document new aes and sha2 options.
8291 2018-01-10 Martin Sebor <msebor@redhat.com>
8293 PR tree-optimization/83781
8294 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
8297 2018-01-11 Martin Sebor <msebor@gmail.com>
8298 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8300 PR tree-optimization/83501
8301 PR tree-optimization/81703
8303 * tree-ssa-strlen.c (get_string_cst): Rename...
8304 (get_string_len): ...to this. Handle global constants.
8305 (handle_char_store): Adjust.
8307 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
8308 Jim Wilson <jimw@sifive.com>
8310 * config/riscv/riscv-protos.h (riscv_output_return): New.
8311 * config/riscv/riscv.c (struct machine_function): New naked_p field.
8312 (riscv_attribute_table, riscv_output_return),
8313 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
8314 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
8315 (riscv_compute_frame_info): Only compute frame->mask if not a naked
8317 (riscv_expand_prologue): Add early return for naked function.
8318 (riscv_expand_epilogue): Likewise.
8319 (riscv_function_ok_for_sibcall): Return false for naked function.
8320 (riscv_set_current_function): New.
8321 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
8322 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
8323 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
8324 * doc/extend.texi (RISC-V Function Attributes): New.
8326 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
8328 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
8329 check for 128-bit long double before checking TCmode.
8330 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
8331 128-bit long doubles before checking TFmode or TCmode.
8332 (FLOAT128_IBM_P): Likewise.
8334 2018-01-10 Martin Sebor <msebor@redhat.com>
8336 PR tree-optimization/83671
8337 * builtins.c (c_strlen): Unconditionally return zero for the empty
8339 Use -Warray-bounds for warnings.
8340 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
8341 for non-constant array indices with COMPONENT_REF, arrays of
8342 arrays, and pointers to arrays.
8343 (gimple_fold_builtin_strlen): Determine and set length range for
8344 non-constant character arrays.
8346 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
8349 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
8352 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
8354 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
8356 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8359 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
8360 VECTOR_MEM_ALTIVEC_OR_VSX_P.
8361 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
8362 indexed_or_indirect_operand predicate.
8363 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
8364 (*vsx_le_perm_load_v8hi): Likewise.
8365 (*vsx_le_perm_load_v16qi): Likewise.
8366 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
8367 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
8368 (*vsx_le_perm_store_v8hi): Likewise.
8369 (*vsx_le_perm_store_v16qi): Likewise.
8370 (eight unnamed splitters): Likewise.
8372 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8374 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
8375 * config/rs6000/emmintrin.h: Likewise.
8376 * config/rs6000/mmintrin.h: Likewise.
8377 * config/rs6000/xmmintrin.h: Likewise.
8379 2018-01-10 David Malcolm <dmalcolm@redhat.com>
8382 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
8384 * tree.c (tree_nop_conversion): Return true for location wrapper
8386 (maybe_wrap_with_location): New function.
8387 (selftest::check_strip_nops): New function.
8388 (selftest::test_location_wrappers): New function.
8389 (selftest::tree_c_tests): Call it.
8390 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
8391 (maybe_wrap_with_location): New decl.
8392 (EXPR_LOCATION_WRAPPER_P): New macro.
8393 (location_wrapper_p): New inline function.
8394 (tree_strip_any_location_wrapper): New inline function.
8396 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
8399 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
8400 stack_realign_offset for the largest alignment of stack slot
8402 (ix86_find_max_used_stack_alignment): New function.
8403 (ix86_finalize_stack_frame_flags): Use it. Set
8404 max_used_stack_alignment if we don't realign stack.
8405 * config/i386/i386.h (machine_function): Add
8406 max_used_stack_alignment.
8408 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
8410 * config/arm/arm.opt (-mbranch-cost): New option.
8411 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
8414 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
8417 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
8418 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
8420 2018-01-10 Richard Biener <rguenther@suse.de>
8423 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
8424 early out so it also covers the case where we have a non-NULL
8427 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8429 PR tree-optimization/83753
8430 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
8431 for non-strided grouped accesses if the number of elements is 1.
8433 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8436 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
8437 * i386.h (TARGET_USE_GATHER): Define.
8438 * x86-tune.def (X86_TUNE_USE_GATHER): New.
8440 2018-01-10 Martin Liska <mliska@suse.cz>
8443 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
8444 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
8446 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
8447 CLEANUP_NO_PARTITIONING is not set.
8449 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8451 * doc/rtl.texi: Remove documentation of (const ...) wrappers
8452 for vectors, as a partial revert of r254296.
8453 * rtl.h (const_vec_p): Delete.
8454 (const_vec_duplicate_p): Don't test for vector CONSTs.
8455 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
8456 * expmed.c (make_tree): Likewise.
8459 * common.md (E, F): Use CONSTANT_P instead of checking for
8461 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
8462 checking for CONST_VECTOR.
8464 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8467 * predict.c (force_edge_cold): Handle in more sane way edges
8470 2018-01-09 Carl Love <cel@us.ibm.com>
8472 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
8474 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
8475 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
8476 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
8477 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
8478 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
8479 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
8480 * config/rs6000/rs6000-protos.h: Add extern defition for
8481 rs6000_generate_float2_double_code.
8482 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
8484 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
8485 (float2_v2df): Add define_expand.
8487 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
8490 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
8491 op_mode in the force_to_mode call.
8493 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8495 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
8496 instead of checking each element individually.
8497 (aarch64_evpc_uzp): Likewise.
8498 (aarch64_evpc_zip): Likewise.
8499 (aarch64_evpc_ext): Likewise.
8500 (aarch64_evpc_rev): Likewise.
8501 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
8502 instead of checking each element individually. Return true without
8504 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
8505 whether all selected elements come from the same input, instead of
8506 checking each element individually. Remove calls to gen_rtx_REG,
8507 start_sequence and end_sequence and instead assert that no rtl is
8510 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8512 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
8513 order of HIGH and CONST checks.
8515 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8517 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
8518 if the destination isn't an SSA_NAME.
8520 2018-01-09 Richard Biener <rguenther@suse.de>
8522 PR tree-optimization/83668
8523 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
8525 (canonicalize_loop_form): ... here, renamed from ...
8526 (canonicalize_loop_closed_ssa_form): ... this and amended to
8527 swap successor edges for loop exit blocks to make us use
8528 the RPO order we need for initial schedule generation.
8530 2018-01-09 Joseph Myers <joseph@codesourcery.com>
8532 PR tree-optimization/64811
8533 * match.pd: When optimizing comparisons with Inf, avoid
8534 introducing or losing exceptions from comparisons with NaN.
8536 2018-01-09 Martin Liska <mliska@suse.cz>
8539 * asan.c (shadow_mem_size): Add gcc_assert.
8541 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
8543 Don't save registers in main().
8546 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
8547 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
8548 * config/avr/avr.c (avr_set_current_function): Don't error if
8549 naked, OS_task or OS_main are specified at the same time.
8550 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
8552 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
8554 * common/config/avr/avr-common.c (avr_option_optimization_table):
8555 Switch on -mmain-is-OS_task for optimizing compilations.
8557 2018-01-09 Richard Biener <rguenther@suse.de>
8559 PR tree-optimization/83572
8560 * graphite.c: Include cfganal.h.
8561 (graphite_transform_loops): Connect infinite loops to exit
8562 and remove fake edges at the end.
8564 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8566 * ipa-inline.c (edge_badness): Revert accidental checkin.
8568 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8571 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
8572 symbols; not inline clones.
8574 2018-01-09 Jakub Jelinek <jakub@redhat.com>
8577 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
8578 hard registers. Formatting fixes.
8580 PR preprocessor/83722
8581 * gcc.c (try_generate_repro): Pass
8582 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
8583 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
8586 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
8587 Kito Cheng <kito.cheng@gmail.com>
8589 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
8590 (riscv_leaf_function_p): Delete.
8591 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
8593 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8595 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
8597 (do_ifelse): New function.
8598 (do_isel): New function.
8599 (do_sub3): New function.
8600 (do_add3): New function.
8601 (do_load_mask_compare): New function.
8602 (do_overlap_load_compare): New function.
8603 (expand_compare_loop): New function.
8604 (expand_block_compare): Call expand_compare_loop() when appropriate.
8605 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
8607 (-mblock-compare-inline-loop-limit): New option.
8609 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8612 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
8613 Reverse order of second and third operands in first alternative.
8614 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
8615 of first and second elements in UNSPEC_VPERMR vector.
8616 (altivec_expand_vec_perm_le): Likewise.
8618 2018-01-08 Jeff Law <law@redhat.com>
8620 PR rtl-optimizatin/81308
8621 * tree-switch-conversion.c (cfg_altered): New file scoped static.
8622 (process_switch): If group_case_labels makes a change, then set
8624 (pass_convert_switch::execute): If a switch is converted, then
8625 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
8627 PR rtl-optimization/81308
8628 * recog.c (split_all_insns): Conditionally cleanup the CFG after
8631 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
8633 PR target/83663 - Revert r255946
8634 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
8635 generation for cases where splatting a value is not useful.
8636 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
8637 across a vec_duplicate and a paradoxical subreg forming a vector
8638 mode to a vec_concat.
8640 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8642 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
8643 -march=armv8.3-a variants.
8644 * config/arm/t-multilib: Likewise.
8645 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
8647 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8649 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
8651 (cceq_ior_compare_complement): Give it a name so I can use it, and
8652 change boolean_or_operator predicate to boolean_operator so it can
8653 be used to generate a crand.
8654 (eqne): New code iterator.
8655 (bd/bd_neg): New code_attrs.
8656 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
8657 a single define_insn.
8658 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
8659 decrement (bdnzt/bdnzf/bdzt/bdzf).
8660 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
8661 with the new names of the branch decrement patterns, and added the
8662 names of the branch decrement conditional patterns.
8664 2018-01-08 Richard Biener <rguenther@suse.de>
8666 PR tree-optimization/83563
8667 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
8670 2018-01-08 Richard Biener <rguenther@suse.de>
8673 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
8675 2018-01-08 Richard Biener <rguenther@suse.de>
8677 PR tree-optimization/83685
8678 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
8679 references to abnormals.
8681 2018-01-08 Richard Biener <rguenther@suse.de>
8684 * dwarf2out.c (output_indirect_strings): Handle empty
8685 skeleton_debug_str_hash.
8686 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
8688 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
8690 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
8691 (emit_store_direct): Likewise.
8692 (arc_trampoline_adjust_address): Likewise.
8693 (arc_asm_trampoline_template): New function.
8694 (arc_initialize_trampoline): Use asm_trampoline_template.
8695 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
8696 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
8697 * config/arc/arc.md (flush_icache): Delete pattern.
8699 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
8701 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
8702 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
8705 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
8708 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
8709 by not USED_FOR_TARGET.
8710 (make_pass_resolve_sw_modes): Likewise.
8712 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
8714 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
8717 2018-01-08 Richard Biener <rguenther@suse.de>
8720 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
8722 2018-01-08 Richard Biener <rguenther@suse.de>
8725 * match.pd ((t * 2) / 2) -> t): Add missing :c.
8727 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
8730 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
8731 basic blocks with a small number of successors.
8732 (convert_control_dep_chain_into_preds): Improve handling of
8734 (dump_predicates): Split apart into...
8735 (dump_pred_chain): ...here...
8736 (dump_pred_info): ...and here.
8737 (can_one_predicate_be_invalidated_p): Add debugging printfs.
8738 (can_chain_union_be_invalidated_p): Improve check for invalidation
8740 (uninit_uses_cannot_happen): Avoid unnecessary if
8741 convert_control_dep_chain_into_preds yielded nothing.
8743 2018-01-06 Martin Sebor <msebor@redhat.com>
8745 PR tree-optimization/83640
8746 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
8747 subtracting negative offset from size.
8748 (builtin_access::overlap): Adjust offset bounds of the access to fall
8749 within the size of the object if possible.
8751 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
8753 PR rtl-optimization/83699
8754 * expmed.c (extract_bit_field_1): Restrict the vector usage of
8755 extract_bit_field_as_subreg to cases in which the extracted
8756 value is also a vector.
8758 * lra-constraints.c (process_alt_operands): Test for the equivalence
8759 substitutions when detecting a possible reload cycle.
8761 2018-01-06 Jakub Jelinek <jakub@redhat.com>
8764 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
8765 by default if flag_selective_schedling{,2}. Formatting fixes.
8767 PR rtl-optimization/83682
8768 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
8769 if it has non-VECTOR_MODE element mode.
8770 (vec_duplicate_p): Likewise.
8773 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
8774 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
8776 2018-01-05 Jakub Jelinek <jakub@redhat.com>
8779 * config/i386/i386-builtin.def
8780 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
8781 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
8782 Require also OPTION_MASK_ISA_AVX512F in addition to
8783 OPTION_MASK_ISA_GFNI.
8784 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
8785 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
8786 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
8787 to OPTION_MASK_ISA_GFNI.
8788 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
8789 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
8790 OPTION_MASK_ISA_AVX512BW.
8791 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
8792 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
8793 addition to OPTION_MASK_ISA_GFNI.
8794 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
8795 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
8796 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
8797 to OPTION_MASK_ISA_GFNI.
8798 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
8799 a requirement for all ISAs rather than any of them with a few
8801 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
8803 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
8804 bitmasks to be enabled with 3 exceptions, instead of requiring any
8805 enabled ISA with lots of exceptions.
8806 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
8807 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
8808 Change avx512bw in isa attribute to avx512f.
8809 * config/i386/sgxintrin.h: Add license boilerplate.
8810 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
8811 to __AVX512F__ and __AVX512VL to __AVX512VL__.
8812 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
8813 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
8815 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
8816 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
8817 temporarily sse2 rather than sse if not enabled already.
8820 * config/i386/sse.md (VI248_VLBW): Rename to ...
8821 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
8822 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
8823 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
8824 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
8825 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
8826 mode iterator instead of VI248_VLBW.
8828 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
8830 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
8831 (record_modified): Skip clobbers; add debug output.
8832 (param_change_prob): Use sreal frequencies.
8834 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8836 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
8837 punt for user-aligned variables.
8839 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8841 * tree-chrec.c (chrec_contains_symbols): Return true for
8844 2018-01-05 Sudakshina Das <sudi.das@arm.com>
8847 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
8848 of (x|y) == x for BICS pattern.
8850 2018-01-05 Jakub Jelinek <jakub@redhat.com>
8852 PR tree-optimization/83605
8853 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
8854 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
8857 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
8859 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
8860 * config/epiphany/rtems.h: New file.
8862 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8863 Uros Bizjak <ubizjak@gmail.com>
8866 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
8867 QIreg_operand instead of register_operand predicate.
8868 * config/i386/i386.c (ix86_rop_should_change_byte_p,
8869 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
8870 comments instead of -fmitigate[-_]rop.
8872 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8875 * cgraphunit.c (symbol_table::compile): Switch to text_section
8876 before calling assembly_start debug hook.
8877 * run-rtl-passes.c (run_rtl_passes): Likewise.
8880 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8882 * tree-vrp.c (extract_range_from_binary_expr_1): Check
8883 range_int_cst_p rather than !symbolic_range_p before calling
8884 extract_range_from_multiplicative_op_1.
8886 2018-01-04 Jeff Law <law@redhat.com>
8888 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
8889 redundant test in assertion.
8891 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8893 * doc/rtl.texi: Document machine_mode wrapper classes.
8895 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8897 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
8900 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8902 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
8903 the VEC_PERM_EXPR fold to fail.
8905 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8908 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
8909 to switched_sections.
8911 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8914 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
8917 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
8920 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
8921 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
8923 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8926 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
8927 is BLKmode and bitpos not zero or mode change is needed.
8929 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8932 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
8935 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
8938 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
8939 instead of MULT rtx. Update all corresponding splitters.
8941 (*ssub<modesuffix>): Ditto.
8943 (*cmp_sadd_di): Update split patterns.
8944 (*cmp_sadd_si): Ditto.
8945 (*cmp_sadd_sidi): Ditto.
8946 (*cmp_ssub_di): Ditto.
8947 (*cmp_ssub_si): Ditto.
8948 (*cmp_ssub_sidi): Ditto.
8949 * config/alpha/predicates.md (const23_operand): New predicate.
8950 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
8951 Look for ASHIFT, not MULT inner operand.
8952 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
8954 2018-01-04 Martin Liska <mliska@suse.cz>
8956 PR gcov-profile/83669
8957 * gcov.c (output_intermediate_file): Add version to intermediate
8959 * doc/gcov.texi: Document new field 'version' in intermediate
8960 file format. Fix location of '-k' option of gcov command.
8962 2018-01-04 Martin Liska <mliska@suse.cz>
8965 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
8967 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8969 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
8971 2018-01-03 Martin Sebor <msebor@redhat.com>
8973 PR tree-optimization/83655
8974 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
8975 checking calls with invalid arguments.
8977 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8979 * tree-vect-stmts.c (vect_get_store_rhs): New function.
8980 (vectorizable_mask_load_store): Delete.
8981 (vectorizable_call): Return false for masked loads and stores.
8982 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
8983 instead of gimple_assign_rhs1.
8984 (vectorizable_load): Handle IFN_MASK_LOAD.
8985 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
8987 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8989 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
8991 (vectorizable_mask_load_store): ...here.
8992 (vectorizable_load): ...and here.
8994 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8996 * tree-vect-stmts.c (vect_build_all_ones_mask)
8997 (vect_build_zero_merge_argument): New functions, split out from...
8998 (vectorizable_load): ...here.
9000 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9002 * tree-vect-stmts.c (vect_check_store_rhs): New function,
9004 (vectorizable_mask_load_store): ...here.
9005 (vectorizable_store): ...and here.
9007 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9009 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
9011 (vectorizable_mask_load_store): ...here.
9013 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9015 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
9016 (vect_model_store_cost): Take a vec_load_store_type instead of a
9018 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
9019 (vect_model_store_cost): Take a vec_load_store_type instead of a
9021 (vectorizable_mask_load_store): Update accordingly.
9022 (vectorizable_store): Likewise.
9023 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
9025 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9027 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
9028 IFN_MASK_LOAD calls here rather than...
9029 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
9031 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9032 Alan Hayward <alan.hayward@arm.com>
9033 David Sherwood <david.sherwood@arm.com>
9035 * expmed.c (extract_bit_field_1): For vector extracts,
9036 fall back to extract_bit_field_as_subreg if vec_extract
9039 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9040 Alan Hayward <alan.hayward@arm.com>
9041 David Sherwood <david.sherwood@arm.com>
9043 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
9044 they are variable or constant sized.
9045 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
9046 slots for constant-sized data.
9048 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9049 Alan Hayward <alan.hayward@arm.com>
9050 David Sherwood <david.sherwood@arm.com>
9052 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
9053 handling COND_EXPRs with boolean comparisons, try to find a better
9054 basis for the mask type than the boolean itself.
9056 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9058 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
9059 is calculated and how it can be overridden.
9060 * genmodes.c (max_bitsize_mode_any_mode): New variable.
9061 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
9063 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
9066 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9067 Alan Hayward <alan.hayward@arm.com>
9068 David Sherwood <david.sherwood@arm.com>
9070 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
9071 Remove the mode argument.
9072 (aarch64_simd_valid_immediate): Remove the mode and inverse
9074 * config/aarch64/iterators.md (bitsize): New iterator.
9075 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
9076 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
9077 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
9078 aarch64_simd_valid_immediate.
9079 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
9080 (aarch64_reg_or_bic_imm): Likewise.
9081 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
9082 with an insn_type enum and msl with a modifier_type enum.
9083 Replace element_width with a scalar_mode. Change the shift
9084 to unsigned int. Add constructors for scalar_float_mode and
9085 scalar_int_mode elements.
9086 (aarch64_vect_float_const_representable_p): Delete.
9087 (aarch64_can_const_movi_rtx_p)
9088 (aarch64_simd_scalar_immediate_valid_for_move)
9089 (aarch64_simd_make_constant): Update call to
9090 aarch64_simd_valid_immediate.
9091 (aarch64_advsimd_valid_immediate_hs): New function.
9092 (aarch64_advsimd_valid_immediate): Likewise.
9093 (aarch64_simd_valid_immediate): Remove mode and inverse
9094 arguments. Rewrite to use the above. Use const_vec_duplicate_p
9095 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
9096 and aarch64_float_const_representable_p on the result.
9097 (aarch64_output_simd_mov_immediate): Remove mode argument.
9098 Update call to aarch64_simd_valid_immediate and use of
9099 simd_immediate_info.
9100 (aarch64_output_scalar_simd_mov_immediate): Update call
9103 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9104 Alan Hayward <alan.hayward@arm.com>
9105 David Sherwood <david.sherwood@arm.com>
9107 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
9108 (mode_nunits): Likewise CONST_MODE_NUNITS.
9109 * machmode.def (ADJUST_NUNITS): Document.
9110 * genmodes.c (mode_data::need_nunits_adj): New field.
9111 (blank_mode): Update accordingly.
9112 (adj_nunits): New variable.
9113 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
9115 (emit_mode_size_inline): Set need_bytesize_adj for all modes
9116 listed in adj_nunits.
9117 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
9118 listed in adj_nunits. Don't emit case statements for such modes.
9119 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
9120 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
9121 nothing if adj_nunits is nonnull.
9122 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
9123 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
9124 (emit_mode_fbit): Update use of print_maybe_const_decl.
9125 (emit_move_size): Likewise. Treat the array as non-const
9127 (emit_mode_adjustments): Handle adj_nunits.
9129 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9131 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
9132 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
9133 (VECTOR_MODES): Use it.
9134 (make_vector_modes): Take the prefix as an argument.
9136 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9137 Alan Hayward <alan.hayward@arm.com>
9138 David Sherwood <david.sherwood@arm.com>
9140 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
9141 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
9142 for MODE_VECTOR_BOOL.
9143 * machmode.def (VECTOR_BOOL_MODE): Document.
9144 * genmodes.c (VECTOR_BOOL_MODE): New macro.
9145 (make_vector_bool_mode): New function.
9146 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
9148 * lto-streamer-in.c (lto_input_mode_table): Likewise.
9149 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
9151 * stor-layout.c (int_mode_for_mode): Likewise.
9152 * tree.c (build_vector_type_for_mode): Likewise.
9153 * varasm.c (output_constant_pool_2): Likewise.
9154 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
9155 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
9156 for MODE_VECTOR_BOOL.
9157 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
9158 of mode class checks.
9159 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
9160 instead of a list of mode class checks.
9161 (expand_vector_scalar_condition): Likewise.
9162 (type_for_widest_vector_mode): Handle BImode as an inner mode.
9164 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9165 Alan Hayward <alan.hayward@arm.com>
9166 David Sherwood <david.sherwood@arm.com>
9168 * machmode.h (mode_size): Change from unsigned short to
9170 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
9171 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9172 or if measurement_type is not polynomial.
9173 (fixed_size_mode::includes_p): Check for constant-sized modes.
9174 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
9175 return a poly_uint16 rather than an unsigned short.
9176 (emit_mode_size): Change the type of mode_size from unsigned short
9177 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
9178 (emit_mode_adjustments): Cope with polynomial vector sizes.
9179 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9181 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9183 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
9184 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
9185 * caller-save.c (setup_save_areas): Likewise.
9186 (replace_reg_with_saved_mem): Likewise.
9187 * calls.c (emit_library_call_value_1): Likewise.
9188 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
9189 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
9190 (gen_lowpart_for_combine): Likewise.
9191 * convert.c (convert_to_integer_1): Likewise.
9192 * cse.c (equiv_constant, cse_insn): Likewise.
9193 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
9194 (cselib_subst_to_values): Likewise.
9195 * dce.c (word_dce_process_block): Likewise.
9196 * df-problems.c (df_word_lr_mark_ref): Likewise.
9197 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
9198 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
9199 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
9200 (rtl_for_decl_location): Likewise.
9201 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
9202 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
9203 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
9204 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
9205 (expand_expr_real_1): Likewise.
9206 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
9207 (pad_below): Likewise.
9208 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9209 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
9210 * ira.c (get_subreg_tracking_sizes): Likewise.
9211 * ira-build.c (ira_create_allocno_objects): Likewise.
9212 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
9213 (ira_sort_regnos_for_alter_reg): Likewise.
9214 * ira-costs.c (record_operand_costs): Likewise.
9215 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
9216 (resolve_simple_move): Likewise.
9217 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
9218 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
9219 (lra_constraints): Likewise.
9220 (CONST_POOL_OK_P): Reject variable-sized modes.
9221 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
9222 (add_pseudo_to_slot, lra_spill): Likewise.
9223 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9224 * optabs-query.c (get_best_extraction_insn): Likewise.
9225 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9226 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
9227 (expand_mult_highpart, valid_multiword_target_p): Likewise.
9228 * recog.c (offsettable_address_addr_space_p): Likewise.
9229 * regcprop.c (maybe_mode_change): Likewise.
9230 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
9231 * regrename.c (build_def_use): Likewise.
9232 * regstat.c (dump_reg_info): Likewise.
9233 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
9234 (find_reloads, find_reloads_subreg_address): Likewise.
9235 * reload1.c (eliminate_regs_1): Likewise.
9236 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
9237 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
9238 (simplify_binary_operation_1, simplify_subreg): Likewise.
9239 * targhooks.c (default_function_arg_padding): Likewise.
9240 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
9241 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
9242 (verify_gimple_assign_ternary): Likewise.
9243 * tree-inline.c (estimate_move_cost): Likewise.
9244 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9245 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
9246 (get_address_cost_ainc): Likewise.
9247 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
9248 (vect_supportable_dr_alignment): Likewise.
9249 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9250 (vectorizable_reduction): Likewise.
9251 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
9252 (vectorizable_operation, vectorizable_load): Likewise.
9253 * tree.c (build_same_sized_truth_vector_type): Likewise.
9254 * valtrack.c (cleanup_auto_inc_dec): Likewise.
9255 * var-tracking.c (emit_note_insn_var_location): Likewise.
9256 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
9257 (ADDR_VEC_ALIGN): Likewise.
9259 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9260 Alan Hayward <alan.hayward@arm.com>
9261 David Sherwood <david.sherwood@arm.com>
9263 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
9265 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9266 or if measurement_type is polynomial.
9267 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
9268 * combine.c (make_extraction): Likewise.
9269 * dse.c (find_shift_sequence): Likewise.
9270 * dwarf2out.c (mem_loc_descriptor): Likewise.
9271 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
9272 (extract_bit_field, extract_low_bits): Likewise.
9273 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
9274 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
9275 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
9276 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
9277 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9278 * reload.c (find_reloads): Likewise.
9279 * reload1.c (alter_reg): Likewise.
9280 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
9281 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
9282 * tree-if-conv.c (predicate_mem_writes): Likewise.
9283 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
9284 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
9285 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
9286 * valtrack.c (dead_debug_insert_temp): Likewise.
9287 * varasm.c (mergeable_constant_section): Likewise.
9288 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
9290 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9291 Alan Hayward <alan.hayward@arm.com>
9292 David Sherwood <david.sherwood@arm.com>
9294 * expr.c (expand_assignment): Cope with polynomial mode sizes
9295 when assigning to a CONCAT.
9297 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9298 Alan Hayward <alan.hayward@arm.com>
9299 David Sherwood <david.sherwood@arm.com>
9301 * machmode.h (mode_precision): Change from unsigned short to
9303 (mode_to_precision): Return a poly_uint16 rather than an unsigned
9305 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
9306 or if measurement_type is not polynomial.
9307 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
9308 in which the mode is already known to be a scalar_int_mode.
9309 * genmodes.c (emit_mode_precision): Change the type of mode_precision
9310 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
9312 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9313 for GET_MODE_PRECISION.
9314 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9315 for GET_MODE_PRECISION.
9316 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
9318 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
9319 (expand_field_assignment, make_extraction): Likewise.
9320 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
9321 (get_last_value): Likewise.
9322 * convert.c (convert_to_integer_1): Likewise.
9323 * cse.c (cse_insn): Likewise.
9324 * expr.c (expand_expr_real_1): Likewise.
9325 * lra-constraints.c (simplify_operand_subreg): Likewise.
9326 * optabs-query.c (can_atomic_load_p): Likewise.
9327 * optabs.c (expand_atomic_load): Likewise.
9328 (expand_atomic_store): Likewise.
9329 * ree.c (combine_reaching_defs): Likewise.
9330 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
9331 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
9332 * tree.h (type_has_mode_precision_p): Likewise.
9333 * ubsan.c (instrument_si_overflow): Likewise.
9335 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9336 Alan Hayward <alan.hayward@arm.com>
9337 David Sherwood <david.sherwood@arm.com>
9339 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
9340 polynomial numbers of units.
9341 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
9342 (valid_vector_subparts_p): New function.
9343 (build_vector_type): Remove temporary shim and take the number
9344 of units as a poly_uint64 rather than an int.
9345 (build_opaque_vector_type): Take the number of units as a
9346 poly_uint64 rather than an int.
9347 * tree.c (build_vector_from_ctor): Handle polynomial
9348 TYPE_VECTOR_SUBPARTS.
9349 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
9350 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
9351 (build_vector_from_val): If the number of units is variable,
9352 use build_vec_duplicate_cst for constant operands and
9353 VEC_DUPLICATE_EXPR otherwise.
9354 (make_vector_type): Remove temporary is_constant ().
9355 (build_vector_type, build_opaque_vector_type): Take the number of
9356 units as a poly_uint64 rather than an int.
9357 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
9359 * cfgexpand.c (expand_debug_expr): Likewise.
9360 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
9361 (store_constructor, expand_expr_real_1): Likewise.
9362 (const_scalar_mask_from_tree): Likewise.
9363 * fold-const-call.c (fold_const_reduction): Likewise.
9364 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
9365 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
9366 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
9367 (fold_relational_const): Likewise.
9368 (native_interpret_vector): Likewise. Change the size from an
9369 int to an unsigned int.
9370 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
9371 TYPE_VECTOR_SUBPARTS.
9372 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
9373 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
9374 duplicating a non-constant operand into a variable-length vector.
9375 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
9376 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
9377 * ipa-icf.c (sem_variable::equals): Likewise.
9378 * match.pd: Likewise.
9379 * omp-simd-clone.c (simd_clone_subparts): Likewise.
9380 * print-tree.c (print_node): Likewise.
9381 * stor-layout.c (layout_type): Likewise.
9382 * targhooks.c (default_builtin_vectorization_cost): Likewise.
9383 * tree-cfg.c (verify_gimple_comparison): Likewise.
9384 (verify_gimple_assign_binary): Likewise.
9385 (verify_gimple_assign_ternary): Likewise.
9386 (verify_gimple_assign_single): Likewise.
9387 * tree-pretty-print.c (dump_generic_node): Likewise.
9388 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9389 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
9390 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
9391 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
9392 (vect_shift_permute_load_chain): Likewise.
9393 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
9394 (expand_vector_condition, optimize_vector_constructor): Likewise.
9395 (lower_vec_perm, get_compute_type): Likewise.
9396 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9397 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
9398 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
9399 (vect_recog_mask_conversion_pattern): Likewise.
9400 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
9401 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
9402 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9403 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
9404 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
9405 (vectorizable_shift, vectorizable_operation, vectorizable_store)
9406 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
9407 (supportable_widening_operation): Likewise.
9408 (supportable_narrowing_operation): Likewise.
9409 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
9411 * varasm.c (output_constant): Likewise.
9413 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9414 Alan Hayward <alan.hayward@arm.com>
9415 David Sherwood <david.sherwood@arm.com>
9417 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
9418 so that both the length == 3 and length != 3 cases set up their
9419 own permute vectors. Add comments explaining why we know the
9420 number of elements is constant.
9421 (vect_permute_load_chain): Likewise.
9423 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9424 Alan Hayward <alan.hayward@arm.com>
9425 David Sherwood <david.sherwood@arm.com>
9427 * machmode.h (mode_nunits): Change from unsigned char to
9429 (ONLY_FIXED_SIZE_MODES): New macro.
9430 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
9431 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
9432 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
9434 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
9435 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
9436 or if measurement_type is not polynomial.
9437 * genmodes.c (ZERO_COEFFS): New macro.
9438 (emit_mode_nunits_inline): Make mode_nunits_inline return a
9440 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
9441 Use ZERO_COEFFS when emitting initializers.
9442 * data-streamer.h (bp_pack_poly_value): New function.
9443 (bp_unpack_poly_value): Likewise.
9444 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9445 for GET_MODE_NUNITS.
9446 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9447 for GET_MODE_NUNITS.
9448 * tree.c (make_vector_type): Remove temporary shim and make
9449 the real function take the number of units as a poly_uint64
9451 (build_vector_type_for_mode): Handle polynomial nunits.
9452 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
9453 * emit-rtl.c (const_vec_series_p_1): Likewise.
9454 (gen_rtx_CONST_VECTOR): Likewise.
9455 * fold-const.c (test_vec_duplicate_folding): Likewise.
9456 * genrecog.c (validate_pattern): Likewise.
9457 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
9458 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9459 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
9460 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
9461 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
9462 * rtlanal.c (subreg_get_info): Likewise.
9463 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9464 (vect_grouped_load_supported): Likewise.
9465 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
9466 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
9467 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
9468 (simplify_const_unary_operation, simplify_binary_operation_1)
9469 (simplify_const_binary_operation, simplify_ternary_operation)
9470 (test_vector_ops_duplicate, test_vector_ops): Likewise.
9471 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
9472 instead of CONST_VECTOR_NUNITS.
9473 * varasm.c (output_constant_pool_2): Likewise.
9474 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
9475 explicit-encoded elements in the XVEC for variable-length vectors.
9477 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9479 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
9481 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9482 Alan Hayward <alan.hayward@arm.com>
9483 David Sherwood <david.sherwood@arm.com>
9485 * coretypes.h (fixed_size_mode): Declare.
9486 (fixed_size_mode_pod): New typedef.
9487 * builtins.h (target_builtins::x_apply_args_mode)
9488 (target_builtins::x_apply_result_mode): Change type to
9489 fixed_size_mode_pod.
9490 * builtins.c (apply_args_size, apply_result_size, result_vector)
9491 (expand_builtin_apply_args_1, expand_builtin_apply)
9492 (expand_builtin_return): Update accordingly.
9494 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9496 * cse.c (hash_rtx_cb): Hash only the encoded elements.
9497 * cselib.c (cselib_hash_rtx): Likewise.
9498 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
9499 CONST_VECTOR encoding.
9501 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9502 Jeff Law <law@redhat.com>
9505 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
9506 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
9507 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
9508 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
9511 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
9512 explicitly probe *sp in a noreturn function if there were any callee
9513 register saves or frame pointer is needed.
9515 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9518 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
9519 BLKmode for ternary, binary or unary expressions.
9522 * var-tracking.c (delete_vta_debug_insn): New inline function.
9523 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
9524 insns from get_insns () to NULL instead of each bb separately.
9525 Use delete_vta_debug_insn. No longer static.
9526 (vt_debug_insns_local, variable_tracking_main_1): Adjust
9527 delete_vta_debug_insns callers.
9528 * rtl.h (delete_vta_debug_insns): Declare.
9529 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
9530 instead of variable_tracking_main.
9532 2018-01-03 Martin Sebor <msebor@redhat.com>
9534 PR tree-optimization/83603
9535 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
9536 arguments past the endof the argument list in functions declared
9537 without a prototype.
9538 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
9539 Avoid checking when arguments are null.
9541 2018-01-03 Martin Sebor <msebor@redhat.com>
9544 * doc/extend.texi (attribute const): Fix a typo.
9545 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
9546 issuing -Wsuggest-attribute for void functions.
9548 2018-01-03 Martin Sebor <msebor@redhat.com>
9550 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
9551 offset_int::from instead of wide_int::to_shwi.
9552 (maybe_diag_overlap): Remove assertion.
9553 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
9554 * gimple-ssa-sprintf.c (format_directive): Same.
9555 (parse_directive): Same.
9556 (sprintf_dom_walker::compute_format_length): Same.
9557 (try_substitute_return_value): Same.
9559 2018-01-03 Jeff Law <law@redhat.com>
9562 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
9563 non-constant residual for zero at runtime and avoid probing in
9564 that case. Reorganize code for trailing problem to mirror handling
9567 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9569 PR tree-optimization/83501
9570 * tree-ssa-strlen.c (get_string_cst): New.
9571 (handle_char_store): Call get_string_cst.
9573 2018-01-03 Martin Liska <mliska@suse.cz>
9575 PR tree-optimization/83593
9576 * tree-ssa-strlen.c: Include tree-cfg.h.
9577 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
9578 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
9579 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
9581 (strlen_dom_walker::before_dom_children): Call
9582 gimple_purge_dead_eh_edges. Dump tranformation with details
9584 (strlen_dom_walker::before_dom_children): Update call by adding
9585 new argument cleanup_eh.
9586 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
9588 2018-01-03 Martin Liska <mliska@suse.cz>
9591 * cif-code.def (VARIADIC_THUNK): New enum value.
9592 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
9595 2018-01-03 Jan Beulich <jbeulich@suse.com>
9597 * sse.md (mov<mode>_internal): Tighten condition for when to use
9598 vmovdqu<ssescalarsize> for TI and OI modes.
9600 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9602 Update copyright years.
9604 2018-01-03 Martin Liska <mliska@suse.cz>
9607 * ipa-visibility.c (function_and_variable_visibility): Skip
9608 functions with noipa attribure.
9610 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9612 * gcc.c (process_command): Update copyright notice dates.
9613 * gcov-dump.c (print_version): Ditto.
9614 * gcov.c (print_version): Ditto.
9615 * gcov-tool.c (print_version): Ditto.
9616 * gengtype.c (create_file): Ditto.
9617 * doc/cpp.texi: Bump @copying's copyright year.
9618 * doc/cppinternals.texi: Ditto.
9619 * doc/gcc.texi: Ditto.
9620 * doc/gccint.texi: Ditto.
9621 * doc/gcov.texi: Ditto.
9622 * doc/install.texi: Ditto.
9623 * doc/invoke.texi: Ditto.
9625 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9627 * vector-builder.h (vector_builder::m_full_nelts): Change from
9628 unsigned int to poly_uint64.
9629 (vector_builder::full_nelts): Update prototype accordingly.
9630 (vector_builder::new_vector): Likewise.
9631 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
9632 (vector_builder::operator ==): Likewise.
9633 (vector_builder::finalize): Likewise.
9634 * int-vector-builder.h (int_vector_builder::int_vector_builder):
9635 Take the number of elements as a poly_uint64 rather than an
9637 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
9638 from unsigned int to poly_uint64.
9639 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
9640 (vec_perm_indices::new_vector): Likewise.
9641 (vec_perm_indices::length): Likewise.
9642 (vec_perm_indices::nelts_per_input): Likewise.
9643 (vec_perm_indices::input_nelts): Likewise.
9644 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
9645 number of elements per input as a poly_uint64 rather than an
9646 unsigned int. Use the original encoding for variable-length
9647 vectors, rather than clamping each individual element.
9648 For the second and subsequent elements in each pattern,
9649 clamp the step and base before clamping their sum.
9650 (vec_perm_indices::series_p): Handle polynomial element counts.
9651 (vec_perm_indices::all_in_range_p): Likewise.
9652 (vec_perm_indices_to_tree): Likewise.
9653 (vec_perm_indices_to_rtx): Likewise.
9654 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
9655 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
9656 (tree_vector_builder::new_binary_operation): Handle polynomial
9657 element counts. Return false if we need to know the number
9658 of elements at compile time.
9659 * fold-const.c (fold_vec_perm): Punt if the number of elements
9660 isn't known at compile time.
9662 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9664 * vec-perm-indices.h (vec_perm_builder): Change element type
9665 from HOST_WIDE_INT to poly_int64.
9666 (vec_perm_indices::element_type): Update accordingly.
9667 (vec_perm_indices::clamp): Handle polynomial element_types.
9668 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9669 (vec_perm_indices::all_in_range_p): Likewise.
9670 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
9672 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
9673 polynomial vec_perm_indices element types.
9674 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
9675 * fold-const.c (fold_vec_perm): Likewise.
9676 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
9677 * tree-vect-generic.c (lower_vec_perm): Likewise.
9678 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9679 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
9680 element type to HOST_WIDE_INT.
9682 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9683 Alan Hayward <alan.hayward@arm.com>
9684 David Sherwood <david.sherwood@arm.com>
9686 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
9687 rather than an int. Use plus_constant.
9688 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
9689 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
9691 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9692 Alan Hayward <alan.hayward@arm.com>
9693 David Sherwood <david.sherwood@arm.com>
9695 * calls.c (emit_call_1, expand_call): Change struct_value_size from
9696 a HOST_WIDE_INT to a poly_int64.
9698 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9699 Alan Hayward <alan.hayward@arm.com>
9700 David Sherwood <david.sherwood@arm.com>
9702 * calls.c (load_register_parameters): Cope with polynomial
9703 mode sizes. Require a constant size for BLKmode parameters
9704 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
9705 forces a parameter to be padded at the lsb end in order to
9706 fill a complete number of words, require the parameter size
9707 to be ordered wrt UNITS_PER_WORD.
9709 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9710 Alan Hayward <alan.hayward@arm.com>
9711 David Sherwood <david.sherwood@arm.com>
9713 * reload1.c (spill_stack_slot_width): Change element type
9714 from unsigned int to poly_uint64_pod.
9715 (alter_reg): Treat mode sizes as polynomial.
9717 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9718 Alan Hayward <alan.hayward@arm.com>
9719 David Sherwood <david.sherwood@arm.com>
9721 * reload.c (complex_word_subreg_p): New function.
9722 (reload_inner_reg_of_subreg, push_reload): Use it.
9724 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9725 Alan Hayward <alan.hayward@arm.com>
9726 David Sherwood <david.sherwood@arm.com>
9728 * lra-constraints.c (process_alt_operands): Reject matched
9729 operands whose sizes aren't ordered.
9730 (match_reload): Refer to this check here.
9732 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9733 Alan Hayward <alan.hayward@arm.com>
9734 David Sherwood <david.sherwood@arm.com>
9736 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
9737 that the mode size is in the set {1, 2, 4, 8, 16}.
9739 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9740 Alan Hayward <alan.hayward@arm.com>
9741 David Sherwood <david.sherwood@arm.com>
9743 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
9744 Use plus_constant instead of gen_rtx_PLUS.
9746 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9747 Alan Hayward <alan.hayward@arm.com>
9748 David Sherwood <david.sherwood@arm.com>
9750 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
9751 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
9752 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
9753 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
9754 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
9755 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
9756 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
9757 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
9758 * config/i386/i386.c (ix86_push_rounding): ...this new function.
9759 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
9761 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
9762 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
9763 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
9764 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
9765 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
9766 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
9767 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
9768 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
9769 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
9770 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
9772 * expr.c (emit_move_resolve_push): Treat the input and result
9773 of PUSH_ROUNDING as a poly_int64.
9774 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
9775 (emit_push_insn): Likewise.
9776 * lra-eliminations.c (mark_not_eliminable): Likewise.
9777 * recog.c (push_operand): Likewise.
9778 * reload1.c (elimination_effects): Likewise.
9779 * rtlanal.c (nonzero_bits1): Likewise.
9780 * calls.c (store_one_arg): Likewise. Require the padding to be
9781 known at compile time.
9783 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9784 Alan Hayward <alan.hayward@arm.com>
9785 David Sherwood <david.sherwood@arm.com>
9787 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
9788 Use plus_constant instead of gen_rtx_PLUS.
9790 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9791 Alan Hayward <alan.hayward@arm.com>
9792 David Sherwood <david.sherwood@arm.com>
9794 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
9797 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9798 Alan Hayward <alan.hayward@arm.com>
9799 David Sherwood <david.sherwood@arm.com>
9801 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
9802 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
9803 via stack temporaries. Treat the mode size as polynomial too.
9805 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9806 Alan Hayward <alan.hayward@arm.com>
9807 David Sherwood <david.sherwood@arm.com>
9809 * expr.c (expand_expr_real_2): When handling conversions involving
9810 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
9811 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
9812 as a poly_uint64 too.
9814 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9815 Alan Hayward <alan.hayward@arm.com>
9816 David Sherwood <david.sherwood@arm.com>
9818 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
9820 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9821 Alan Hayward <alan.hayward@arm.com>
9822 David Sherwood <david.sherwood@arm.com>
9824 * combine.c (can_change_dest_mode): Handle polynomial
9825 REGMODE_NATURAL_SIZE.
9826 * expmed.c (store_bit_field_1): Likewise.
9827 * expr.c (store_constructor): Likewise.
9828 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
9829 and polynomial REGMODE_NATURAL_SIZE.
9830 (gen_lowpart_common): Likewise.
9831 * reginfo.c (record_subregs_of_mode): Likewise.
9832 * rtlanal.c (read_modify_subreg_p): Likewise.
9834 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9835 Alan Hayward <alan.hayward@arm.com>
9836 David Sherwood <david.sherwood@arm.com>
9838 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
9839 numbers of elements.
9841 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9842 Alan Hayward <alan.hayward@arm.com>
9843 David Sherwood <david.sherwood@arm.com>
9845 * match.pd: Cope with polynomial numbers of vector elements.
9847 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9848 Alan Hayward <alan.hayward@arm.com>
9849 David Sherwood <david.sherwood@arm.com>
9851 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
9852 in a POINTER_PLUS_EXPR.
9854 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9855 Alan Hayward <alan.hayward@arm.com>
9856 David Sherwood <david.sherwood@arm.com>
9858 * omp-simd-clone.c (simd_clone_subparts): New function.
9859 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
9860 (ipa_simd_modify_function_body): Likewise.
9862 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9863 Alan Hayward <alan.hayward@arm.com>
9864 David Sherwood <david.sherwood@arm.com>
9866 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
9867 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
9868 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
9869 (expand_vector_condition, vector_element): Likewise.
9870 (subparts_gt): New function.
9871 (get_compute_type): Use subparts_gt.
9872 (count_type_subparts): Delete.
9873 (expand_vector_operations_1): Use subparts_gt instead of
9874 count_type_subparts.
9876 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9877 Alan Hayward <alan.hayward@arm.com>
9878 David Sherwood <david.sherwood@arm.com>
9880 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
9881 (vect_compile_time_alias): ...this new function. Do the calculation
9882 on poly_ints rather than trees.
9883 (vect_prune_runtime_alias_test_list): Update call accordingly.
9885 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9886 Alan Hayward <alan.hayward@arm.com>
9887 David Sherwood <david.sherwood@arm.com>
9889 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
9891 (vect_schedule_slp_instance): Likewise.
9893 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9894 Alan Hayward <alan.hayward@arm.com>
9895 David Sherwood <david.sherwood@arm.com>
9897 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
9898 constant and extern definitions for variable-length vectors.
9899 (vect_get_constant_vectors): Note that the number of units
9900 is known to be constant.
9902 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9903 Alan Hayward <alan.hayward@arm.com>
9904 David Sherwood <david.sherwood@arm.com>
9906 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
9907 of units as polynomial. Choose between WIDE and NARROW based
9910 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9911 Alan Hayward <alan.hayward@arm.com>
9912 David Sherwood <david.sherwood@arm.com>
9914 * tree-vect-stmts.c (simd_clone_subparts): New function.
9915 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
9917 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9918 Alan Hayward <alan.hayward@arm.com>
9919 David Sherwood <david.sherwood@arm.com>
9921 * tree-vect-stmts.c (vectorizable_call): Treat the number of
9922 vectors as polynomial. Use build_index_vector for
9925 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9926 Alan Hayward <alan.hayward@arm.com>
9927 David Sherwood <david.sherwood@arm.com>
9929 * tree-vect-stmts.c (get_load_store_type): Treat the number of
9930 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
9931 for variable-length vectors.
9932 (vectorizable_mask_load_store): Treat the number of units as
9933 polynomial, asserting that it is constant if the condition has
9934 already been enforced.
9935 (vectorizable_store, vectorizable_load): Likewise.
9937 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9938 Alan Hayward <alan.hayward@arm.com>
9939 David Sherwood <david.sherwood@arm.com>
9941 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
9942 of units as polynomial. Punt if we can't tell at compile time
9943 which vector contains the final result.
9945 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9946 Alan Hayward <alan.hayward@arm.com>
9947 David Sherwood <david.sherwood@arm.com>
9949 * tree-vect-loop.c (vectorizable_induction): Treat the number
9950 of units as polynomial. Punt on SLP inductions. Use an integer
9951 VEC_SERIES_EXPR for variable-length integer reductions. Use a
9952 cast of such a series for variable-length floating-point
9955 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9956 Alan Hayward <alan.hayward@arm.com>
9957 David Sherwood <david.sherwood@arm.com>
9959 * tree.h (build_index_vector): Declare.
9960 * tree.c (build_index_vector): New function.
9961 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
9962 of units as polynomial, forcibly converting it to a constant if
9963 vectorizable_reduction has already enforced the condition.
9964 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
9965 to create a {1,2,3,...} vector.
9966 (vectorizable_reduction): Treat the number of units as polynomial.
9967 Choose vectype_in based on the largest scalar element size rather
9968 than the smallest number of units. Enforce the restrictions
9971 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9972 Alan Hayward <alan.hayward@arm.com>
9973 David Sherwood <david.sherwood@arm.com>
9975 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
9976 number of units as polynomial.
9978 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9979 Alan Hayward <alan.hayward@arm.com>
9980 David Sherwood <david.sherwood@arm.com>
9982 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
9983 * target.def (autovectorize_vector_sizes): Return the vector sizes
9984 by pointer, using vector_sizes rather than a bitmask.
9985 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
9986 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
9987 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
9989 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
9990 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
9991 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
9992 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
9993 * omp-general.c (omp_max_vf): Likewise.
9994 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9995 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
9996 * tree-vect-loop.c (vect_analyze_loop): Likewise.
9997 * tree-vect-slp.c (vect_slp_bb): Likewise.
9998 * doc/tm.texi: Regenerate.
9999 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
10001 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
10002 the vector size as a poly_uint64 rather than an unsigned int.
10003 (current_vector_size): Change from an unsigned int to a poly_uint64.
10004 (get_vectype_for_scalar_type): Update accordingly.
10005 * tree.h (build_truth_vector_type): Take the size and number of
10006 units as a poly_uint64 rather than an unsigned int.
10007 (build_vector_type): Add a temporary overload that takes
10008 the number of units as a poly_uint64 rather than an unsigned int.
10009 * tree.c (make_vector_type): Likewise.
10010 (build_truth_vector_type): Take the number of units as a poly_uint64
10011 rather than an unsigned int.
10013 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10014 Alan Hayward <alan.hayward@arm.com>
10015 David Sherwood <david.sherwood@arm.com>
10017 * target.def (get_mask_mode): Take the number of units and length
10018 as poly_uint64s rather than unsigned ints.
10019 * targhooks.h (default_get_mask_mode): Update accordingly.
10020 * targhooks.c (default_get_mask_mode): Likewise.
10021 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
10022 * doc/tm.texi: Regenerate.
10024 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10025 Alan Hayward <alan.hayward@arm.com>
10026 David Sherwood <david.sherwood@arm.com>
10028 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
10029 * omp-general.c (omp_max_vf): Likewise.
10030 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
10031 (expand_omp_simd): Handle polynomial safelen.
10032 * omp-low.c (omplow_simd_context): Add a default constructor.
10033 (omplow_simd_context::max_vf): Change from int to poly_uint64.
10034 (lower_rec_simd_input_clauses): Update accordingly.
10035 (lower_rec_input_clauses): Likewise.
10037 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10038 Alan Hayward <alan.hayward@arm.com>
10039 David Sherwood <david.sherwood@arm.com>
10041 * tree-vectorizer.h (vect_nunits_for_cost): New function.
10042 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
10043 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
10044 (vect_analyze_slp_cost): Likewise.
10045 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
10046 (vect_model_load_cost): Likewise.
10048 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10049 Alan Hayward <alan.hayward@arm.com>
10050 David Sherwood <david.sherwood@arm.com>
10052 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
10053 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
10054 from an unsigned int * to a poly_uint64_pod *.
10055 (calculate_unrolling_factor): New function.
10056 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
10058 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10059 Alan Hayward <alan.hayward@arm.com>
10060 David Sherwood <david.sherwood@arm.com>
10062 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
10063 from an unsigned int to a poly_uint64.
10064 (_loop_vec_info::slp_unrolling_factor): Likewise.
10065 (_loop_vec_info::vectorization_factor): Change from an int
10067 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
10068 (vect_get_num_vectors): New function.
10069 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
10070 (vect_get_num_copies): Use vect_get_num_vectors.
10071 (vect_analyze_data_ref_dependences): Change max_vf from an int *
10072 to an unsigned int *.
10073 (vect_analyze_data_refs): Change min_vf from an int * to a
10075 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10076 than an unsigned HOST_WIDE_INT.
10077 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
10078 (vect_analyze_data_ref_dependence): Change max_vf from an int *
10079 to an unsigned int *.
10080 (vect_analyze_data_ref_dependences): Likewise.
10081 (vect_compute_data_ref_alignment): Handle polynomial vf.
10082 (vect_enhance_data_refs_alignment): Likewise.
10083 (vect_prune_runtime_alias_test_list): Likewise.
10084 (vect_shift_permute_load_chain): Likewise.
10085 (vect_supportable_dr_alignment): Likewise.
10086 (dependence_distance_ge_vf): Take the vectorization factor as a
10087 poly_uint64 rather than an unsigned HOST_WIDE_INT.
10088 (vect_analyze_data_refs): Change min_vf from an int * to a
10090 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
10091 vfm1 as a poly_uint64 rather than an int. Make the same change
10092 for the returned bound_scalar.
10093 (vect_gen_vector_loop_niters): Handle polynomial vf.
10094 (vect_do_peeling): Likewise. Update call to
10095 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
10096 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
10098 * tree-vect-loop.c (vect_determine_vectorization_factor)
10099 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
10100 (vect_get_known_peeling_cost): Likewise.
10101 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
10102 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
10103 (vect_transform_loop): Likewise. Use the lowest possible VF when
10104 updating the upper bounds of the loop.
10105 (vect_min_worthwhile_factor): Make static. Return an unsigned int
10106 rather than an int.
10107 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
10108 polynomial unroll factors.
10109 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
10110 (vect_make_slp_decision): Likewise.
10111 (vect_supported_load_permutation_p): Likewise, and polynomial
10113 (vect_analyze_slp_cost): Handle polynomial vf.
10114 (vect_slp_analyze_node_operations): Likewise.
10115 (vect_slp_analyze_bb_1): Likewise.
10116 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10117 than an unsigned HOST_WIDE_INT.
10118 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
10119 (vectorizable_load): Handle polynomial vf.
10120 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
10122 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
10124 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10125 Alan Hayward <alan.hayward@arm.com>
10126 David Sherwood <david.sherwood@arm.com>
10128 * match.pd: Handle bit operations involving three constants
10129 and try to fold one pair.
10131 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10133 * tree-vect-loop-manip.c: Include gimple-fold.h.
10134 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
10135 niters_maybe_zero parameters. Handle other cases besides a step of 1.
10136 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
10137 Add a path that uses a step of VF instead of 1, but disable it
10139 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
10140 and niters_no_overflow parameters. Update calls to
10141 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
10142 Create a new SSA name if the latter choses to use a ste other
10143 than zero, and return it via niters_vector_mult_vf_var.
10144 * tree-vect-loop.c (vect_transform_loop): Update calls to
10145 vect_do_peeling, vect_gen_vector_loop_niters and
10146 slpeel_make_loop_iterate_ntimes.
10147 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
10148 (vect_gen_vector_loop_niters): Update declarations after above changes.
10150 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
10152 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
10153 128-bit round to integer instructions.
10154 (ceil<mode>2): Likewise.
10155 (btrunc<mode>2): Likewise.
10156 (round<mode>2): Likewise.
10158 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10160 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
10161 unaligned VSX load/store on P8/P9.
10162 (expand_block_clear): Allow the use of unaligned VSX
10163 load/store on P8/P9.
10165 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10167 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
10169 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
10170 swap associated with both a load and a store.
10172 2018-01-02 Andrew Waterman <andrew@sifive.com>
10174 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
10175 * config/riscv/riscv.md (clear_cache): Use it.
10177 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
10179 * web.c: Remove out-of-date comment.
10181 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10183 * expr.c (fixup_args_size_notes): Check that any existing
10184 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
10185 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
10186 (emit_single_push_insn): ...here.
10188 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10190 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
10191 (const_vector_encoded_nelts): New function.
10192 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
10193 (const_vector_int_elt, const_vector_elt): Declare.
10194 * emit-rtl.c (const_vector_int_elt_1): New function.
10195 (const_vector_elt): Likewise.
10196 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
10197 of CONST_VECTOR_ELT.
10199 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10201 * expr.c: Include rtx-vector-builder.h.
10202 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
10203 directly on the tree encoding.
10204 (const_vector_from_tree): Likewise.
10205 * optabs.c: Include rtx-vector-builder.h.
10206 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
10207 sequence of "u" values.
10208 * vec-perm-indices.c: Include rtx-vector-builder.h.
10209 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
10210 directly on the vec_perm_indices encoding.
10212 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10214 * doc/rtl.texi (const_vector): Describe new encoding scheme.
10215 * Makefile.in (OBJS): Add rtx-vector-builder.o.
10216 * rtx-vector-builder.h: New file.
10217 * rtx-vector-builder.c: Likewise.
10218 * rtl.h (rtx_def::u2): Add a const_vector field.
10219 (CONST_VECTOR_NPATTERNS): New macro.
10220 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
10221 (CONST_VECTOR_DUPLICATE_P): Likewise.
10222 (CONST_VECTOR_STEPPED_P): Likewise.
10223 (CONST_VECTOR_ENCODED_ELT): Likewise.
10224 (const_vec_duplicate_p): Check for a duplicated vector encoding.
10225 (unwrap_const_vec_duplicate): Likewise.
10226 (const_vec_series_p): Check for a non-duplicated vector encoding.
10227 Say that the function only returns true for integer vectors.
10228 * emit-rtl.c: Include rtx-vector-builder.h.
10229 (gen_const_vec_duplicate_1): Delete.
10230 (gen_const_vector): Call gen_const_vec_duplicate instead of
10231 gen_const_vec_duplicate_1.
10232 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
10233 (gen_const_vec_duplicate): Use rtx_vector_builder.
10234 (gen_const_vec_series): Likewise.
10235 (gen_rtx_CONST_VECTOR): Likewise.
10236 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
10237 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10238 Build a new vector rather than modifying a CONST_VECTOR in-place.
10239 (handle_special_swappables): Update call accordingly.
10240 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
10241 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10242 Build a new vector rather than modifying a CONST_VECTOR in-place.
10243 (handle_special_swappables): Update call accordingly.
10245 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10247 * simplify-rtx.c (simplify_const_binary_operation): Use
10248 CONST_VECTOR_ELT instead of XVECEXP.
10250 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10252 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
10253 the selector elements to be different from the data elements
10254 if the selector is a VECTOR_CST.
10255 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
10256 ssizetype for the selector.
10258 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10260 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
10261 before testing each element individually.
10262 * tree-vect-generic.c (lower_vec_perm): Likewise.
10264 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10266 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
10267 * selftest-run-tests.c (selftest::run_tests): Call it.
10268 * vector-builder.h (vector_builder::operator ==): New function.
10269 (vector_builder::operator !=): Likewise.
10270 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
10271 (vec_perm_indices::all_from_input_p): New function.
10272 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10273 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
10274 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
10275 instead of reading the VECTOR_CST directly. Detect whether both
10276 vector inputs are the same before constructing the vec_perm_indices,
10277 and update the number of inputs argument accordingly. Use the
10278 utility functions added above. Only construct sel2 if we need to.
10280 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10282 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
10283 the broadcast of the low byte.
10284 (expand_mult_highpart): Use an explicit encoding for the permutes.
10285 * optabs-query.c (can_mult_highpart_p): Likewise.
10286 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
10287 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10288 (vectorizable_bswap): Likewise.
10289 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
10290 explicit encoding for the power-of-2 permutes.
10291 (vect_permute_store_chain): Likewise.
10292 (vect_grouped_load_supported): Likewise.
10293 (vect_permute_load_chain): Likewise.
10295 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10297 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
10298 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
10299 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
10300 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10301 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
10302 (vect_gen_perm_mask_any): Likewise.
10304 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10306 * int-vector-builder.h: New file.
10307 * vec-perm-indices.h: Include int-vector-builder.h.
10308 (vec_perm_indices): Redefine as an int_vector_builder.
10309 (auto_vec_perm_indices): Delete.
10310 (vec_perm_builder): Redefine as a stand-alone class.
10311 (vec_perm_indices::vec_perm_indices): New function.
10312 (vec_perm_indices::clamp): Likewise.
10313 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
10314 (vec_perm_indices::new_vector): New function.
10315 (vec_perm_indices::new_expanded_vector): Update for new
10316 vec_perm_indices class.
10317 (vec_perm_indices::rotate_inputs): New function.
10318 (vec_perm_indices::all_in_range_p): Operate directly on the
10319 encoded form, without computing elided elements.
10320 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
10321 encoding. Update for new vec_perm_indices class.
10322 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
10323 the given vec_perm_builder.
10324 (expand_vec_perm_var): Update vec_perm_builder constructor.
10325 (expand_mult_highpart): Use vec_perm_builder instead of
10326 auto_vec_perm_indices.
10327 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
10328 vec_perm_indices instead of auto_vec_perm_indices. Use a single
10329 or double series encoding as appropriate.
10330 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
10331 vec_perm_indices instead of auto_vec_perm_indices.
10332 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10333 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10334 (vect_permute_store_chain): Likewise.
10335 (vect_grouped_load_supported): Likewise.
10336 (vect_permute_load_chain): Likewise.
10337 (vect_shift_permute_load_chain): Likewise.
10338 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10339 (vect_transform_slp_perm_load): Likewise.
10340 (vect_schedule_slp_instance): Likewise.
10341 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10342 (vectorizable_mask_load_store): Likewise.
10343 (vectorizable_bswap): Likewise.
10344 (vectorizable_store): Likewise.
10345 (vectorizable_load): Likewise.
10346 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
10347 vec_perm_indices instead of auto_vec_perm_indices. Use
10348 tree_to_vec_perm_builder to read the vector from a tree.
10349 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
10350 vec_perm_builder instead of a vec_perm_indices.
10351 (have_whole_vector_shift): Use vec_perm_builder and
10352 vec_perm_indices instead of auto_vec_perm_indices. Leave the
10353 truncation to calc_vec_perm_mask_for_shift.
10354 (vect_create_epilog_for_reduction): Likewise.
10355 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
10356 from auto_vec_perm_indices to vec_perm_indices.
10357 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10358 instead of changing individual elements.
10359 (aarch64_vectorize_vec_perm_const): Use new_vector to install
10360 the vector in d.perm.
10361 * config/arm/arm.c (expand_vec_perm_d::perm): Change
10362 from auto_vec_perm_indices to vec_perm_indices.
10363 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10364 instead of changing individual elements.
10365 (arm_vectorize_vec_perm_const): Use new_vector to install
10366 the vector in d.perm.
10367 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
10368 Update vec_perm_builder constructor.
10369 (rs6000_expand_interleave): Likewise.
10370 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
10371 (rs6000_expand_interleave): Likewise.
10373 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10375 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
10376 to qimode could truncate the indices.
10377 * optabs.c (expand_vec_perm_var): Likewise.
10379 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10381 * Makefile.in (OBJS): Add vec-perm-indices.o.
10382 * vec-perm-indices.h: New file.
10383 * vec-perm-indices.c: Likewise.
10384 * target.h (vec_perm_indices): Replace with a forward class
10386 (auto_vec_perm_indices): Move to vec-perm-indices.h.
10387 * optabs.h: Include vec-perm-indices.h.
10388 (expand_vec_perm): Delete.
10389 (selector_fits_mode_p, expand_vec_perm_var): Declare.
10390 (expand_vec_perm_const): Declare.
10391 * target.def (vec_perm_const_ok): Replace with...
10392 (vec_perm_const): ...this new hook.
10393 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
10394 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
10395 * doc/tm.texi: Regenerate.
10396 * optabs.def (vec_perm_const): Delete.
10397 * doc/md.texi (vec_perm_const): Likewise.
10398 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
10399 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
10400 expand_vec_perm for constant permutation vectors. Assert that
10401 the mode of variable permutation vectors is the integer equivalent
10402 of the mode that is being permuted.
10403 * optabs-query.h (selector_fits_mode_p): Declare.
10404 * optabs-query.c: Include vec-perm-indices.h.
10405 (selector_fits_mode_p): New function.
10406 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
10407 is defined, instead of checking whether the vec_perm_const_optab
10408 exists. Use targetm.vectorize.vec_perm_const instead of
10409 targetm.vectorize.vec_perm_const_ok. Check whether the indices
10410 fit in the vector mode before using a variable permute.
10411 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
10412 vec_perm_indices instead of an rtx.
10413 (expand_vec_perm): Replace with...
10414 (expand_vec_perm_const): ...this new function. Take the selector
10415 as a vec_perm_indices rather than an rtx. Also take the mode of
10416 the selector. Update call to shift_amt_for_vec_perm_mask.
10417 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
10418 Use vec_perm_indices::new_expanded_vector to expand the original
10419 selector into bytes. Check whether the indices fit in the vector
10420 mode before using a variable permute.
10421 (expand_vec_perm_var): Make global.
10422 (expand_mult_highpart): Use expand_vec_perm_const.
10423 * fold-const.c: Includes vec-perm-indices.h.
10424 * tree-ssa-forwprop.c: Likewise.
10425 * tree-vect-data-refs.c: Likewise.
10426 * tree-vect-generic.c: Likewise.
10427 * tree-vect-loop.c: Likewise.
10428 * tree-vect-slp.c: Likewise.
10429 * tree-vect-stmts.c: Likewise.
10430 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
10432 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
10433 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
10434 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
10435 (aarch64_vectorize_vec_perm_const): ...this new function.
10436 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10437 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10438 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
10439 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
10440 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10441 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10442 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
10444 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
10445 check for NEON modes.
10446 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
10447 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
10448 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
10449 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
10451 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
10452 the old VEC_PERM_CONST conditions.
10453 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
10454 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
10455 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
10456 (ia64_vectorize_vec_perm_const_ok): Merge into...
10457 (ia64_vectorize_vec_perm_const): ...this new function.
10458 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
10459 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
10460 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
10461 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
10462 * config/mips/mips.c (mips_expand_vec_perm_const)
10463 (mips_vectorize_vec_perm_const_ok): Merge into...
10464 (mips_vectorize_vec_perm_const): ...this new function.
10465 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
10466 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
10467 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
10468 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
10469 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
10470 (rs6000_expand_vec_perm_const): Delete.
10471 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
10473 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10474 (altivec_expand_vec_perm_const_le): Take each operand individually.
10475 Operate on constant selectors rather than rtxes.
10476 (altivec_expand_vec_perm_const): Likewise. Update call to
10477 altivec_expand_vec_perm_const_le.
10478 (rs6000_expand_vec_perm_const): Delete.
10479 (rs6000_vectorize_vec_perm_const_ok): Delete.
10480 (rs6000_vectorize_vec_perm_const): New function.
10481 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10482 an element count and rtx array.
10483 (rs6000_expand_extract_even): Update call accordingly.
10484 (rs6000_expand_interleave): Likewise.
10485 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
10486 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
10487 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
10488 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
10489 (rs6000_expand_vec_perm_const): Delete.
10490 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10491 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10492 (altivec_expand_vec_perm_const_le): Take each operand individually.
10493 Operate on constant selectors rather than rtxes.
10494 (altivec_expand_vec_perm_const): Likewise. Update call to
10495 altivec_expand_vec_perm_const_le.
10496 (rs6000_expand_vec_perm_const): Delete.
10497 (rs6000_vectorize_vec_perm_const_ok): Delete.
10498 (rs6000_vectorize_vec_perm_const): New function. Remove stray
10499 reference to the SPE evmerge intructions.
10500 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10501 an element count and rtx array.
10502 (rs6000_expand_extract_even): Update call accordingly.
10503 (rs6000_expand_interleave): Likewise.
10504 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
10505 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
10507 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10509 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10511 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
10512 vector mode and that that mode matches the mode of the data
10514 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
10515 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
10516 directly using expand_vec_perm_1 when forcing selectors into
10518 (expand_vec_perm_var): New function, split out from expand_vec_perm.
10520 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10522 * optabs-query.h (can_vec_perm_p): Delete.
10523 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
10524 * optabs-query.c (can_vec_perm_p): Split into...
10525 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
10526 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
10527 particular selector is valid.
10528 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10529 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10530 (vect_grouped_load_supported): Likewise.
10531 (vect_shift_permute_load_chain): Likewise.
10532 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10533 (vect_transform_slp_perm_load): Likewise.
10534 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10535 (vectorizable_bswap): Likewise.
10536 (vect_gen_perm_mask_checked): Likewise.
10537 * fold-const.c (fold_ternary_loc): Likewise. Don't take
10538 implementations of variable permutation vectors into account
10539 when deciding which selector to use.
10540 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
10541 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
10542 with a false third argument.
10543 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
10544 to test whether the constant selector is valid and can_vec_perm_var_p
10545 to test whether a variable selector is valid.
10547 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10549 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
10550 * optabs-query.c (can_vec_perm_p): Likewise.
10551 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
10552 instead of vec_perm_indices.
10553 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
10554 (vect_gen_perm_mask_checked): Likewise,
10555 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
10556 (vect_gen_perm_mask_checked): Likewise,
10558 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10560 * optabs-query.h (qimode_for_vec_perm): Declare.
10561 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
10562 (qimode_for_vec_perm): ...this new function.
10563 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
10565 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10567 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
10568 does not have a conditional at the top.
10570 2018-01-02 Richard Biener <rguenther@suse.de>
10572 * ipa-inline.c (big_speedup_p): Fix expression.
10574 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10577 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
10580 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10584 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
10585 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
10586 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
10587 cond_taken_branch_cost 3->4.
10589 2018-01-01 Jakub Jelinek <jakub@redhat.com>
10591 PR tree-optimization/83581
10592 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
10593 TODO_cleanup_cfg if any changes have been made.
10595 PR middle-end/83608
10596 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
10597 convert_modes if target mode has the right side, but different mode
10600 PR middle-end/83609
10601 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
10602 last argument when extracting from CONCAT. If either from_real or
10603 from_imag is NULL, use expansion through memory. If result is not
10604 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
10605 the parts directly to inner mode, if even that fails, use expansion
10608 PR middle-end/83623
10609 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
10610 check for bswap in mode rather than HImode and use that in expand_unop
10613 Copyright (C) 2018 Free Software Foundation, Inc.
10615 Copying and distribution of this file, with or without modification,
10616 are permitted in any medium without royalty provided the copyright
10617 notice and this notice are preserved.