1 2018-04-13 Jakub Jelinek <jakub@redhat.com>
3 PR rtl-optimization/85376
4 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
5 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
6 instead of a specific value.
8 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
9 Bin Cheng <bin.cheng@arm.com>
11 PR tree-optimization/82965
12 PR tree-optimization/83991
13 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
14 by_profile_only parameter.
15 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
16 information if the loop was predicted to iterate too many times.
17 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
19 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
22 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
25 2018-04-13 Martin Liska <mliska@suse.cz>
26 Jakub Jelinek <jakub@redhat.com>
29 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
30 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
31 * builtins.c (expand_builtin_memory_copy_args): Use
32 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
33 handle dest_addr == pc_rtx.
35 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
38 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
39 asked to not generate direct moves.
40 (fix_trunc<mode>si2_stfiwx): Similar.
41 (fix_trunc<mode>si2_internal): Similar.
43 2018-04-12 Jakub Jelinek <jakub@redhat.com>
46 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
47 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
48 lookup if dest in some wider mode is known to be const0_rtx and
49 if so, record permanent equivalence for it to be ZERO_EXTEND of
50 the narrower mode destination.
52 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
54 * lto-streamer-out.c (output_function): Revert 259346.
55 * omp-expand.c (expand_oacc_for): Likewise.
57 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
59 PR rtl-optimization/85354
60 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
61 * sel-sched.c (sel_global_init): ... here.
63 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
66 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
67 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
68 mode for PE-COFF targets.
69 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
70 (i386_pe_asm_lto_end): Likewise.
71 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
72 (TARGET_ASM_LTO_END): Likewise.
73 * config/i386/winnt.c (saved_debug_info_level): New static variable.
74 (i386_pe_asm_lto_start): New function.
75 (i386_pe_asm_lto_end): Likewise.
77 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
78 Richard Biener <rguenther@suse.de>
81 * lto-streamer-out.c (output_function): Fix CFG loop state before
83 * omp-expand.c (expand_oacc_for): Handle calls to internal
84 functions like regular functions.
86 2018-04-12 Richard Biener <rguenther@suse.de>
89 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
90 for the early LTO debug to properly generate references to it
91 during DIE emission. Do not re-use that for the skeleton for
93 (dwarf2out_early_finish): Likewise.
95 2018-04-12 Jakub Jelinek <jakub@redhat.com>
99 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
100 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
101 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
102 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
103 and output is a reg, avoid creating invalid lowpart subreg, but
104 instead split into a 512-bit move. Don't split if not AVX512VL,
105 input is xmm16+ reg and output is a mem.
106 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
107 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
108 xmm16+ reg and output is a mem.
110 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
112 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
113 also for flag_dwarf2_cfi_asm.
115 2018-04-12 Jakub Jelinek <jakub@redhat.com>
117 PR rtl-optimization/85342
118 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
119 a bool scalar var inside of the loop instead. Don't try to update
120 recog_data.operand after failed apply_change_group.
122 2018-04-12 Tom de Vries <tom@codesourcery.com>
125 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
126 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
127 array with flexible array member as array without given dimension.
128 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
129 argument for undefined param to true.
131 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
134 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
135 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
136 from PowerPC section.
137 * config/rs6000/sysv4.opt (mcall-): Improve help text.
138 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
139 help text that is too long.
140 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
141 help text that is too long.
142 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
143 help text that is too long.
145 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
147 * config/alpha/alpha.md (stack_probe_internal): Rename
148 from "probe_stack". Update all callers.
150 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
152 PR rtl-optimization/84566
153 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
154 sched_macro_fuse_insns.
156 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
159 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
160 (compute_block_dependences): ... from here.
162 2018-04-11 Jakub Jelinek <jakub@redhat.com>
164 PR tree-optimization/85331
165 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
166 from int to HOST_WIDE_INT.
168 2018-04-11 Martin Jambor <mjambor@suse.cz>
171 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
172 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
173 not the same as the source val.
174 (cgraph_edge_brings_value_p): New parameter.
175 (gather_edges_for_value): Pass destination value to
176 cgraph_edge_brings_value_p.
177 (perhaps_add_new_callers): Likewise.
178 (get_info_about_necessary_edges): Likewise and exclude values brought
179 only by self-recursive edges.
180 (create_specialized_node): Redirect only clones of self-calling edges.
181 (+self_recursive_pass_through_p): New function.
182 (find_more_scalar_values_for_callers_subset): Use it.
183 (find_aggregate_values_for_callers_subset): Likewise.
184 (known_aggs_to_agg_replacement_list): Removed.
185 (decide_whether_version_node): Re-calculate known constants for all
186 remaining context clones.
188 2018-04-11 Richard Biener <rguenther@suse.de>
191 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
192 from early DWARF output.
193 (dwarf2out_early_finish): Output line info unconditionally into
194 early DWARF and add reference to it.
196 2018-04-11 Jakub Jelinek <jakub@redhat.com>
199 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
200 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
201 other than V2DFmode using iptr mode attribute.
202 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
204 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
206 PR rtl-optimization/84659
207 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
209 2018-04-11 Jakub Jelinek <jakub@redhat.com>
212 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
214 (output_loc_list): Pass address of a dummy size variable even in the
215 locview handling loop.
216 (index_location_lists): Add comment on why skip_loc_list_entry can't
219 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
222 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
225 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
228 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
229 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
230 and -mstring-compare-inline-limit.
232 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
235 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
236 for stack clash protection in a register whenever we need it to be in
239 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
241 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
242 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
244 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
247 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
249 (mlong-double-): Ditto.
250 * config/rs6000/sysv4.opt (msdata=): Ditto.
253 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
255 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
256 erroneous entries for
257 "vector int vec_ldl (int, long int *)", and
258 "vector unsigned int vec_ldl (int, unsigned long int *)".
259 Add comments and entries for
260 "vector bool char vec_ldl (int, bool char *)",
261 "vector bool short vec_ldl (int, bool short *)",
262 "vector bool int vec_ldl (int, bool int *)",
263 "vector bool long long vec_ldl (int, bool long long *)",
264 "vector pixel vec_ldl (int, pixel *)",
265 "vector long long vec_ldl (int, long long *)",
266 "vector unsigned long long vec_ldl (int, unsigned long long *)".
267 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
268 type tree bool_long_long_type_node and correct definition of
269 bool_V2DI_type_node to make reference to this new type tree.
270 (rs6000_mangle_type): Replace erroneous reference to
271 bool_long_type_node with bool_long_long_type_node.
272 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
273 comments to emphasize sign distinctions for char and int types and
274 replace RS6000_BTI_bool_long constant with
275 RS6000_BTI_bool_long_long constant. Also add comment to restrict
276 use of RS6000_BTI_pixel.
277 (bool_long_type_node): Remove this macro definition.
278 (bool_long_long_type_node): New macro definition
280 2018-04-10 Jakub Jelinek <jakub@redhat.com>
282 PR rtl-optimization/85300
283 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
284 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
285 simplify_unary_operation fails.
287 2018-04-10 Martin Liska <mliska@suse.cz>
289 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
290 cgraph_edge and ipa_ref.
292 2018-04-10 Jakub Jelinek <jakub@redhat.com>
297 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
298 computation of the VEC_MERGE selector from mask.
299 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
300 Fix decoding of the VEC_MERGE selector into mask.
302 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
304 PR tree-optimization/85286
305 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
307 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
309 * final.c (final_1): Set insn_last_address as well as
310 insn_current_address.
312 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
315 * explow.c (emit_stack_probe): Call validize_mem on memory location
316 before passing it to gen_probe_stack. Create address operand and
317 legitimize it for the probe_stack_address case.
319 2018-04-09 Jan Hubicka <jh@suse.cz>
322 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
323 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
324 * tree.c (free_lang_data_in_type): Fix handling of binfos;
326 (free_lang_data): Rebuild type inheritance graph.
328 2018-04-09 Martin Sebor <msebor@redhat.com>
330 * invoke.texi (-finline-small-functions): Mention other optimization
332 (-findirect-inlining, -fpartial-inlining): Same.
333 (-finline-functions-called-once): Same.
334 (-freorder-blocks-and-partition): Same.
336 2018-04-09 Jan Hubicka <jh@suse.cz>
339 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
340 jumps; choose last target that matches the criteria (i.e.
341 no partition changes for non-crossing jumps).
342 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
343 support for redirecting crossing jumps to non-crossing.
345 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
347 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
348 also for naked functions.
350 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
352 * config/arc/arc.md (add_shift): New pattern.
353 (add_shift2): Likewise.
354 (sub_shift): Likewise.
355 (sub_shift_cmp0_noout): Likewise.
356 (compare_si_ashiftsi): Likewise.
357 (xbfu_cmp0_noout): New combine pattern.
358 (xbfu_cmp0"): Likewise.
359 (movsi_set_cc_insn): Place the predicable variant first.
360 (commutative_binary_cmp0_noout): Remove clobber.
361 (commutative_binary_cmp0): New pattern.
362 (noncommutative_binary_cmp0): Likewise.
363 (noncommutative_binary_cmp0_noout): Likewise.
364 (noncommutative_binary_comparison_result_used): Removed.
365 (rsub_cmp0): New pattern.
366 (rsub_cmp0_noout): Likewise.
367 (extzvsi): Changed, keep only meaningful variants.
368 (SQH, SEZ): New iterators.
369 (SQH_postfix): New mode attribute.
370 (SEZ_prefix): New code attribute.
371 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
372 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
373 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
375 (noncommutative_operator): Check the availability of barrel
378 2018-04-09 Richard Biener <rguenther@suse.de>
380 PR tree-optimization/85284
381 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
382 Only use the niter constraining form of simple_iv when the exit
385 2018-04-09 Tom de Vries <tom@codesourcery.com>
388 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
389 (define_expand "*memory_barrier"): New define_expand.
390 (define_insn "memory_barrier"): New insn.
392 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
394 PR rtl-optimization/80463
395 PR rtl-optimization/83972
396 PR rtl-optimization/83480
398 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
399 correct producer for the insn.
400 (tidy_control_flow): Fixup seqnos in case of debug insns.
402 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
404 PR rtl-optimization/83913
406 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
407 different sched-times when merging exprs.
409 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
411 PR rtl-optimization/83962
413 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
414 tidy_fallthru_edge and tidy_control_flow.
416 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
418 PR rtl-optimization/83530
420 * sel-sched.c (force_next_insn): New global variable.
421 (remove_insn_for_debug): When force_next_insn is true, also leave only
422 next insn in the ready list.
423 (sel_sched_region): When the region wasn't scheduled, make another pass
424 over it with force_next_insn set to 1.
426 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
428 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
430 * config/nds32/constants.md (unspec_volatile_element): Add enum values
431 for interrupt control.
432 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
433 functions for interrupt control.
434 * config/nds32/nds32-intrinsic.md: Likewise.
435 * config/nds32/nds32_intrinsic.h: Likewise.
436 * config/nds32/nds32.h (nds32_builtins): Likewise.
438 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
440 * config/nds32/nds32.c (nds32_init_machine_status,
441 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
442 strict_aligned_p field.
443 (nds32_expand_to_rtl_hook): New function.
444 (TARGET_EXPAND_TO_RTL_HOOK): Define.
445 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
447 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
448 Chung-Ju Wu <jasonwucj@gmail.com>
450 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
451 * config/nds32/nds32-n7.md: New file.
452 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
453 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
455 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
456 * config/nds32/nds32.md (pipeline_model): Add n7.
457 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
458 * config/nds32/pipelines.md: Include n7 settings.
460 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
461 Chung-Ju Wu <jasonwucj@gmail.com>
463 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
464 * config/nds32/nds32-e8.md: New file.
465 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
466 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
468 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
469 * config/nds32/nds32.md (pipeline_model): Add e8.
470 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
471 * config/nds32/pipelines.md: Include e8 settings.
473 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
474 Chung-Ju Wu <jasonwucj@gmail.com>
476 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
477 * config/nds32/nds32-n8.md: New file.
478 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
479 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
481 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
482 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
483 * config/nds32/nds32.md (pipeline_model): Add n8.
484 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
485 * config/nds32/pipelines.md: Include n8 settings.
487 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
488 Chung-Ju Wu <jasonwucj@gmail.com>
490 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
491 * config/nds32/nds32-n9-2r1w.md: New file.
492 * config/nds32/nds32-n9-3r2w.md: New file.
493 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
494 nds32_register_ports): New or modify for cpu n9.
495 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
497 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
498 * config/nds32/nds32-utils.c: New file.
499 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
500 TARGET_MUL_SLOW): Define.
501 * config/nds32/nds32.md (pipeline_model): New attribute.
502 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
503 New options that support cpu n9.
504 * config/nds32/pipelines.md: Include n9 settings.
505 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
507 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
509 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
510 information if necessary.
511 (output_cond_branch_compare_zero): Likewise.
512 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
513 (nds32_target_alignment): Refine for alignment.
514 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
515 (FUNCTION_BOUNDARY): Modify.
516 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
518 * config/nds32/nds32.opt (malways-align, malign-functions): New.
520 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
522 * config/nds32/constants.md (unspec_volatile_element): Add values for
523 TLB operation and data prefetch.
524 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
525 functions for TLB operation and data prefetch.
526 * config/nds32/nds32-intrinsic.md: Likewise.
527 * config/nds32/nds32_intrinsic.h: Likewise.
528 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
529 (nds32_print_operand): Likewise.
530 * config/nds32/nds32.h (nds32_builtins): Likewise.
532 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
533 Andrew Pinski <pinsika@gcc.gnu.org>
536 * match.pd: Use constant_boolean_node of correct type instead of
537 boolean_true_node or boolean_false_node for simplifying
538 pointer comparisons to zero.
540 2018-04-07 Jakub Jelinek <jakub@redhat.com>
542 PR tree-optimization/80021
543 * tree.c (verify_type_variant): Make error call in verify_variant_match
544 translatable and remove final full stop.
546 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
548 * config/nds32/constants.md (unspec_volatile_element): Add
549 UNSPEC_VOLATILE_EH_RETURN.
550 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
551 nds32_output_stack_pop): Support dwarf exception handling process.
552 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
553 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
554 exception handling process.
555 (nds32_compute_stack_frame): Likewise.
556 (nds32_return_addr_rtx): Likewise.
557 (nds32_initial_elimination_offset): Likewise.
558 (nds32_expand_prologue): Likewise.
559 (nds32_expand_epilogue): Likewise.
560 (nds32_dynamic_chain_address): New function.
561 * config/nds32/nds32.h (machine_function): Add fields for dwarf
563 (DYNAMIC_CHAIN_ADDRESS): Define.
564 (EH_RETURN_DATA_REGNO): Define.
565 (EH_RETURN_STACKADJ_RTX): Define.
566 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
567 patterns for dwarf exception handling.
569 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
571 * config/nds32/nds32.h: Clean up obsolete macros.
573 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
575 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
576 Add enum values for particular instructions.
577 * config/nds32/nds32-intrinsic.c: Implementation of expanding
578 particular intrinsic functions.
579 * config/nds32/nds32-intrinsic.md: Likewise.
580 * config/nds32/nds32_intrinsic.h: Likewise.
581 * config/nds32/nds32.h (nds32_builtins): Likewise.
582 * config/nds32/nds32.md (type): Add pbsad and pbsada.
583 (btst, ave): New patterns for particular instructions.
585 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
587 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
588 Add enum values for atomic load/store and memory sync.
589 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
591 * config/nds32/nds32-intrinsic.md: Likewise.
592 * config/nds32/nds32_intrinsic.h: Likewise.
593 * config/nds32/nds32.h (nds32_builtins): Likewise.
595 2018-04-07 Jakub Jelinek <jakub@redhat.com>
597 PR tree-optimization/85257
598 * fold-const.c (native_encode_vector): If not all elts could fit
599 and off is -1, return 0 rather than offset.
600 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
601 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
602 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
603 adjust buffer in native_interpret_expr call.
605 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
607 * config/nds32/constants.md (unspec_volatile_element): Add cache
609 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
610 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
611 * config/nds32/nds32.c (nds32_cctl_names): New.
612 (nds32_print_operand): Handle cache control register names.
613 * config/nds32/nds32.h (nds32_builtins): New enum values.
614 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
616 * config/nds32/nds32.md (type): Add mmu.
617 * config/nds32/pipelines.md (simple_insn): Add mmu.
619 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
621 * config/nds32/nds32.md (type): Remove call.
622 * config/nds32/pipelines.md (simple_insn): Likewise.
624 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
626 * config/nds32/constants.md (unspec_volatile_element): Add
627 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
628 UNSPEC_VOLATILE_FMFCFG.
629 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
630 description for fmfcfg and fmfcsr.
631 (bdesc_1arg): Add fmtcsr.
632 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
633 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
634 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
635 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
636 unspec_fmfcfg): New patterns.
637 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
638 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
639 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
640 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
641 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
642 __nds32__fmfcfg): Define.
644 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
646 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
647 intrinsic register names.
648 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
649 intrinsic register enum values and macros.
651 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
653 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
654 for load/store addressing form.
655 (nds32_print_operand_address): Likewise.
657 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
660 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
661 based on LABEL_REF. Remove useless assertion.
662 (pic_address_needs_scratch): Fix formatting.
663 (sparc_legitimize_pic_address): Minor tweaks.
664 (sparc_delegitimize_address): Adjust assertion accordingly.
665 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
666 into symbolic_operand.
667 (movsi_high_pic_label_ref): Likewise.
668 (movsi_lo_sum_pic_label_ref): Likewise.
669 (movdi_pic_label_ref): Likewise.
670 (movdi_high_pic_label_ref): Likewise.
671 (movdi_lo_sum_pic_label_ref): Likewise.
673 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
675 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
676 custom LIB_SPEC setup.
678 2018-04-06 Ruslan Bukin <br@bsdpad.com>
679 Kito Cheng <kito.cheng@gmail.com>
681 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
682 * config/riscv/freebsd.h: New.
684 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
686 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
687 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
690 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
691 Kito Cheng <kito.cheng@gmail.com>
693 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
694 nds32_output_call, nds32_symbol_binds_local_p): New functions.
695 * config/nds32/nds32-protos.h (nds32_output_call,
696 nds32_output_return): Declare.
697 * config/nds32/nds32.md: Refine all the call and return patterns.
699 2018-04-06 Jakub Jelinek <jakub@redhat.com>
702 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
703 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
705 PR rtl-optimization/84872
706 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
707 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
710 2018-04-06 Tamar Christina <tamar.christina@arm.com>
712 * expr.c (copy_blkmode_to_reg): Revert 254862.
713 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
715 2018-04-06 Richard Biener <rguenther@suse.de>
718 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
719 after seeing a component reference with an adjacent field. Treat
720 refs to arrays at struct end of external decls similar to
721 refs to unconstrained commons.
723 2018-04-06 Jakub Jelinek <jakub@redhat.com>
726 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
727 look through SAVE_EXPRs with non-side-effects argument. Adjust
729 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
732 2018-04-06 Richard Biener <rguenther@suse.de>
735 * alias.c (find_base_term): New wrapper around find_base_term
736 unwinding CSELIB_VAL_PTR changes.
737 (find_base_term): Do not restore CSELIB_VAL_PTR during the
740 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
742 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
744 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
745 constant definitions.
746 ("nop"): lr 0,0 -> nopr r0
747 ("nop_lr0", "nop_lr1"): New insn definitions.
749 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
751 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
752 NDS32_V3PUSH_AVAILABLE_P macro.
754 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
755 Chung-Ju Wu <jasonwucj@gmail.com>
757 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
758 (nds32*-*-*): Add float and fpu_config into supported_defaults.
759 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
760 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
761 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
762 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
763 * config/nds32/constraints.md: New constraints and checking for hard
765 * config/nds32/iterators.md: New mode iterator and attribute for hard
767 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
769 * config/nds32/nds32-fpu.md: New file.
770 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
771 deal with hard float code generation.
772 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
774 (abi_type, float_reg_number): New enum type.
775 * config/nds32/nds32-predicates.c: New predicates for hard float.
776 * config/nds32/nds32-protos.h: Declare functions for hard float.
777 * config/nds32/nds32.c: Implementation for hard float configuration.
778 * config/nds32/nds32.h: Definitions for hard float configuration.
779 * config/nds32/nds32.md: Include hard float machine description and
780 modify patterns for hard float configuration.
781 * config/nds32/nds32.opt: New options for hard float configuration.
782 * config/nds32/predicates.md: New predicates for hard float
785 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
787 * common/config/nds32/nds32-common.c
788 (nds32_option_optimization_table): Enable -mreleax-hint by default.
790 2018-04-05 Jakub Jelinek <jakub@redhat.com>
793 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
794 CONSTRUCTOR_ELT (ctor, ...)->value.
796 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
799 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
801 2018-04-05 Tom de Vries <tom@codesourcery.com>
804 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
807 2018-04-05 Shiva Chen <shiva0217@gmail.com>
808 Kito Cheng <kito.cheng@gmail.com>
810 * config/nds32/constraints.md (U33): Fine-tune checking condition.
811 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
812 * config/nds32/nds32.h (nds32_16bit_address_type): Add
813 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
815 2018-04-05 Shiva Chen <shiva0217@gmail.com>
816 Kito Cheng <kito.cheng@gmail.com>
818 * config/nds32/constraints.md (Ufe): New memory constraint.
819 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
820 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
821 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
823 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
824 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
826 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
828 * config/nds32/nds32.md: Use optimize_size in the condition for
829 alu-shift instructions.
831 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
833 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
835 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
837 * config/nds32/nds32.md (negsi2): Refine pattern.
839 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
840 Chung-Ju Wu <jasonwucj@gmail.com>
842 * config/nds32/iterators.md (shift_rotate): New code iterator.
843 (shift): New code attribute.
844 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
845 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
846 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
847 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
849 (andsi3, *andsi3): Ditto.
850 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
851 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
852 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
853 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
854 nds32_ior_operand, nds32_xor_operand): New predicates.
856 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
858 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
859 (addsi3, subsi3): ... this.
861 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
863 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
865 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
867 * config/nds32/nds32.md: Adjust indention.
869 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
871 * config/nds32/nds32.md (feature): New attribute.
873 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
875 * config/nds32/nds32.md (subtype): New attribute.
877 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
880 * config/arm/arm-builtins.c (arm_expand_builtin): Change
881 expansion to perform a bitwise AND of the argument followed by a
882 boolean negation of the result.
884 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
886 PR rtl-optimization/84878
887 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
888 the basic block. Assert the use reference is not artificial and that
889 it has an associated insn.
891 2018-04-04 Michael Matz <matz@suse.de>
893 * builtins.c (compute_objsize): Pass correct operand
894 to array_at_struct_end_p.
896 2018-04-04 Richard Biener <rguenther@suse.de>
899 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
900 from contexts for DINFO_LEVEL_TERSE and below.
902 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
904 * config/nds32/nds32-doubleword.md (move_<mode>): Require
905 resiter_operand condition.
906 * config/nds32/nds32.md (*move<mode>): Ditto.
908 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
909 Monk Chiang <sh.chiang04@gmail.com>
911 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
913 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
915 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
917 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
918 Kito Cheng <kito.cheng@gmail.com>
920 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
921 nds32_cond_code_str, output_cond_branch,
922 output_cond_branch_compare_zero, nds32_expand_cbranch,
923 nds32_expand_cstore, nds32_expand_movcc,
924 nds32_output_cbranchsi4_equality_zero,
925 nds32_output_cbranchsi4_equality_reg,
926 nds32_output_cbranchsi4_equality_reg_or_const_int,
927 nds32_output_cbranchsi4_greater_less_zero: New functions.
928 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
929 nds32_expand_cstore, nds32_expand_movcc,
930 nds32_output_cbranchsi4_equality_zero,
931 nds32_output_cbranchsi4_equality_reg,
932 nds32_output_cbranchsi4_equality_reg_or_const_int,
933 nds32_output_cbranchsi4_greater_less_zero): Declare.
934 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
935 nds32_rimm11s_operand): New predicates.
936 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
937 * config/nds32/nds32.md: Rewrite all the branch and conditional move
940 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
942 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
943 * config/nds32/nds32.md: Ditto.
944 * config/nds32/pipelines.md: Ditto.
946 2018-04-04 Richard Biener <rguenther@suse.de>
948 PR tree-optimization/85168
949 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
950 propagating abnormals.
952 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
954 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
956 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
957 Kito Cheng <kito.cheng@gmail.com>
959 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
960 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
961 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
962 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
963 * config/nds32/nds32.md (sibcall_internal): New.
964 (sibcall_register): Remove.
965 (sibcall_immediate): Remove.
966 (sibcall_value_internal): New.
967 (sibcall_value_register): Remove.
968 (sibcall_value_immediate): Remove.
969 * config/nds32/predicates.md (nds32_general_register_operand): New.
970 (nds32_call_address_operand): New.
972 2018-04-03 Jakub Jelinek <jakub@redhat.com>
974 PR rtl-optimization/85167
975 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
976 bb_defs if *split_p, instead preinitialize it to NULL.
978 PR tree-optimization/85156
979 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
980 evaluating the argument multiple times.
982 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
984 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
986 (_mm_cvtpd_ps): Likewise.
987 (_mm_cvttpd_epi32): Likewise.
988 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
989 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
990 vector, pixel, and bool following altivec.h include.
992 2018-04-03 Martin Sebor <msebor@redhat.com>
994 * doc/extend.texi (Common Function Attributes): Clarify.
995 (const attribute): Likewise.
996 (pure attribute): Likewise.
998 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1001 * config/i386/i386.c (ix86_expand_vector_set): Use
1002 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
1004 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
1006 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
1007 instructions when changing rounding bits to preserve precision bits
1008 in the x87 control word.
1010 2018-04-03 Martin Liska <mliska@suse.cz>
1012 PR tree-optimization/82491
1013 * rtl.h (strip_offset_and_add): Replace += suboffset with
1014 poly_uint64 () + suboffset.
1016 2018-03-29 Martin Liska <mliska@suse.cz>
1017 Martin Jambor <mjambor@suse.cz>
1020 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
1021 param_type is not an integral or pointer type.
1023 2018-04-03 Richard Biener <rguenther@suse.de>
1025 * sese.h (recompute_all_dominators): Remove.
1027 2018-04-02 Martin Sebor <msebor@redhat.com>
1029 * doc/invoke.texi (-Wrestrict): Fix typos.
1031 2018-04-02 Jim Wilson <jimw@sifive.com>
1033 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
1034 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
1035 (<optab>di3, <optab>si3_extend): Likewise.
1036 (<optab>si3_mask, <optab>si3_mask_1): New.
1037 (<optab>di3_mask, <optab>di3_mask_1): New.
1038 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
1039 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
1040 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
1042 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
1044 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
1047 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
1049 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
1050 (nds32_canonicalize_comparison): New function.
1052 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1053 Kito Cheng <kito.cheng@gmail.com>
1054 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1056 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
1057 * config/nds32/constants.md (unspec_volatile_element): Add
1058 UNSPEC_VOLATILE_RELAX_GROUP.
1059 * config/nds32/nds32-relax-opt.c: New file.
1060 * config/nds32/nds32-predicates.c
1061 (nds32_symbol_load_store_p): New function.
1062 * config/nds32/nds32-protos.h
1063 (nds32_symbol_load_store_p): Declare function.
1064 (make_pass_nds32_relax_opt): Declare new rtl pass function.
1065 * config/nds32/nds32.c
1066 (nds32_register_pass): New function to register pass.
1067 (nds32_register_passes): New function to register passes.
1068 * config/nds32/nds32.md (relax_group): New pattern.
1069 * config/nds32/nds32.opt (mrelax-hint): New option.
1070 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
1072 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
1074 * config/nds32/t-nds32: Modify files dependency.
1076 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1078 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
1079 (PROFILE_HOOK): Define its implementation.
1081 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1083 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
1084 type and 32-bit size.
1086 2018-04-01 Jakub Jelinek <jakub@redhat.com>
1089 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
1090 (V_128_256): New mode iterator.
1091 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
1092 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
1093 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
1095 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
1096 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
1098 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
1101 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
1102 NaN inputs correctly.
1104 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
1107 * config/rs6000/vsx.md (??r): New mode attribute.
1108 (*vsx_mov<mode>_64bit): Use it.
1109 (*vsx_mov<mode>_32bit): Likewise.
1111 2018-03-30 Martin Sebor <msebor@redhat.com>
1113 PR tree-optimization/84818
1114 * builtins.c (check_access): Use warning_n.
1116 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1119 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
1121 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
1124 2018-03-30 Julia Koval <julia.koval@intel.com>
1127 * x86-tune.def (movx, partial_reg_dependency): Enable for
1130 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
1133 * lra-constraints.c (process_alt_operands): Move setting
1134 this_alternative_matches below.
1136 2018-03-29 Martin Liska <mliska@suse.cz>
1139 * doc/invoke.texi: Document how LTO works with debug info.
1140 Describe auto-load support of binutils. Mention 'x86-64'
1141 as valid option value of -march option.
1143 2018-03-29 Jakub Jelinek <jakub@redhat.com>
1145 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
1148 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
1149 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
1150 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
1153 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1156 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
1157 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
1158 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
1159 for RS6000_BTM_POWERPC64.
1160 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
1161 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
1162 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
1167 2018-03-28 Carl Love <cel@us.ibm.com>
1170 2017-09-27 Carl Love <cel@us.ibm.com>
1172 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
1173 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
1174 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
1177 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1179 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
1180 instead of __vector bool.
1181 (_mm_max_pu8): Likewise.
1182 (_mm_min_pi16): Likewise.
1184 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1187 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
1188 (DIVWEUO): Likewise.
1190 (DIVDEUO): Likewise.
1191 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
1192 DIVWEUO and DIVDEUO.
1193 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
1194 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
1195 (div_extend): Likewise.
1196 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
1198 (__builtin_divweuo): Likewise.
1199 (__builtin_divdeo): Likewise.
1200 (__builtin_divdeuo): Likewise.
1202 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1205 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
1206 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
1208 PR tree-optimization/82004
1209 * gimple-match-head.c (optimize_pow_to_exp): New function.
1210 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
1211 Don't fold to exp if optimize_pow_to_exp is false.
1213 2018-03-28 Martin Liska <mliska@suse.cz>
1216 * calls.c (initialize_argument_information): Fix trailing space.
1217 * common.opt: Fix typo and provide better explanation for
1218 -fsanitize-coverage option.
1219 * config/i386/i386.opt: Fix typo.
1221 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1222 Martin Liska <mliska@suse.cz>
1225 * gimplify.c (asan_poison_variable): Don't do the check for
1226 gimplify_omp_ctxp here.
1227 (gimplify_decl_expr): Do it here.
1228 (gimplify_target_expr): Likewise.
1230 2018-03-28 Martin Liska <mliska@suse.cz>
1233 * config/i386/i386.c (ix86_function_arg_advance): Do not call
1234 chkp_type_bounds_count if MPX is not enabled.
1236 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
1238 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
1240 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
1243 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
1244 function to create the function decl for complex long double
1245 multiply and divide for -mabi=ieeelongdouble.
1246 (init_float128_ieee): Call it.
1248 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
1251 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
1252 -fcf-protection=branch -mibt.
1253 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
1255 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1258 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
1260 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
1263 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
1264 extern array declarations.
1266 2018-03-27 Richard Biener <rguenther@suse.de>
1269 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
1270 explicit single_use checks.
1272 2018-03-27 Richard Biener <rguenther@suse.de>
1274 PR tree-optimization/85082
1275 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
1278 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1280 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
1281 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
1282 Turn on fasynchronous-unwind-tables and funwind-tables.
1284 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
1287 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
1288 (*bmi_blsr_<mode>_ccz): Ditto.
1290 2018-03-26 Tom de Vries <tom@codesourcery.com>
1292 PR tree-optimization/85063
1293 * omp-general.c (offloading_function_p): New function. Factor out
1295 * omp-offload.c (pass_omp_target_link::gate): ... here.
1296 * omp-general.h (offloading_function_p): Declare.
1297 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
1298 with attribute omp declare target for offloading functions.
1300 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
1302 PR tree-optimization/84005
1303 * tree-data-ref.h (get_base_for_alignment): Declare.
1304 * tree-data-ref.c (get_base_for_alignment_1): New function.
1305 (get_base_for_alignment): Likewise.
1306 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
1307 get_base_for_alignment to find a suitable base object, instead
1308 of always using drb->base_address.
1310 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1313 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
1314 known size by default.
1316 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
1319 * lra-constraints.c (process_alt_operands): Don't match BLKmode
1320 and non BLKmode operands.
1322 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1325 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
1326 Clean up attributes.
1328 2018-03-23 Richard Biener <rguenther@suse.de>
1331 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
1332 we are going to emit early debug for LTO.
1334 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1337 * function.c (match_asm_constraints_1): Don't optimize if input
1338 doesn't satisfy general_operand predicate for output's mode.
1341 * alias.c (write_dependence_p): Don't require for x_canonicalized
1342 non-VOIDmode if x has VOIDmode.
1345 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
1346 just don't try to optimize it rather than assert it never happens.
1348 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1350 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
1351 macro expansions for definition of ST_INTERNAL_<mode> and
1352 LD_INTERNAL_<mode> builtins.
1353 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
1355 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
1357 (altivec_expand_st_builtin): Likewise.
1358 (altivec_expand_builtin): Remove calls to deleted functions.
1359 (rs6000_address_for_altivec): Delete this function.
1360 * config/rs6000/vector.md: Remove expands for
1361 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
1363 2018-03-22 Sudakshina Das <sudi.das@arm.com>
1366 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
1367 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
1368 re-computing once computed.
1369 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
1370 (arm_init_machine_status): Initialize
1371 machine->static_chain_stack_bytes.
1373 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1376 * doc/extend.texi: Add four new prototypes for vec_ld.
1377 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
1378 definitions for more logical presentation.
1379 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
1380 entries for V1TI variants of __builtin_altivec_ld builtin.
1381 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
1382 handling of V1TI variant of LVX icode pattern.
1383 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
1384 (rs6000_gimple_fold_builtin): Likewise.
1385 (altivec_init_builtins): Add code to define
1386 __builtin_altivec_lvx_v1ti function.
1388 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1391 * function.c (match_asm_constraints_1): Don't do the optimization
1392 if input isn't a REG, SUBREG, MEM or constant.
1394 2018-03-22 Tom de Vries <tom@codesourcery.com>
1396 PR tree-optimization/84956
1397 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
1398 bb_has_abnormal_pred.
1400 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1403 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
1404 DECL_INITIAL (decl) to decl at the end.
1405 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
1408 2018-03-21 Joseph Myers <joseph@codesourcery.com>
1410 * doc/extend.texi (__builtin_tgmath): Document when complex
1411 integer types are treated as _Complex _Float64.
1413 2018-03-21 Tom de Vries <tom@codesourcery.com>
1415 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
1417 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1419 PR tree-optimization/84960
1420 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
1421 if it is ENTRY block, move them into single succ of ENTRY in that case.
1423 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
1425 PR tree-optimization/84811
1426 * poly-int.h (poly_span_traits): Remove the T3 parameter and
1427 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
1428 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
1429 (known_subrange_p): Update accordingly. Cast each value involved
1430 in the size comparison, rather than casting the result of the
1433 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1435 PR tree-optimization/84982
1436 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
1437 by flipping the least significant bit rather than all bits from
1438 bitpos to bitpos + bitsize - 1.
1440 2018-03-21 Nathan Sidwell <nathan@acm.org>
1442 * doc/extend.texi (Deprecated Features): Remove mention of
1443 long-deleted deprecations.
1445 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1448 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
1449 * configure: Regenerate.
1451 2018-03-21 Tom de Vries <tom@codesourcery.com>
1453 PR tree-optimization/83126
1454 * tree-parloops.c (num_phis): New function.
1455 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
1457 2018-03-21 Nathan Sidwell <nathan@acm.org>
1459 * doc/extend.texi (Deprecated Features): Update deprecated flags,
1460 mention anon-struct/union members and trailing attributes.
1462 2018-03-21 Bin Cheng <bin.cheng@arm.com>
1464 PR tree-optimization/84969
1465 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
1466 builtin memset partitions if they set different rhs values.
1468 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1470 PR rtl-optimization/84989
1471 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
1472 VEC_DUPLICATE with scalar result mode.
1474 2018-03-21 Martin Liska <mliska@suse.cz>
1477 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
1478 not intended return statement.
1480 2018-03-21 Martin Liska <mliska@suse.cz>
1483 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
1484 (chkp_find_bound_slots_1): Limit number of iterations.
1486 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
1489 * Minor grammar fixes for x86 options.
1491 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1494 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
1495 holding REG_CFA_RESTORE notes, instead turn them into a USE.
1497 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
1500 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
1501 (altivec_lvx_<mode>_1op): Likewise.
1502 (altivec_stvx_<mode>_2op): Likewise.
1503 (altivec_stvx_<mode>_1op): Likewise.
1504 (altivec_lvx_<VM2:mode>): New define_expand.
1505 (altivec_stvx_<VM2:mode>): Likewise.
1506 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
1507 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1508 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
1509 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1510 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
1511 (rs6000_gen_lvx): Likewise.
1512 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
1513 (altivec_expand_stv_builtin): Likewise.
1514 (altivec_expand_builtin): Likewise.
1515 * config/rs6000/vector.md: Likewise.
1517 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1520 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
1523 2018-03-20 Richard Biener <rguenther@suse.de>
1526 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
1527 sign-conversions as zero, fall back to standard scalar_stmt
1530 2018-03-20 Martin Liska <mliska@suse.cz>
1533 * predict.c (rebuild_frequencies): Handle case when we have
1534 PROFILE_ABSENT, but flag_guess_branch_prob is false.
1536 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1539 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
1540 flag_section_anchors.
1541 * varasm.c (use_blocks_for_decl_p): Remove hack for
1542 dw2_force_const_mem.
1545 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
1547 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
1548 be created, use lowpart_subreg of operands[0] rather than operands[0]
1550 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
1551 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
1552 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
1553 and n constraint instead of aarch64_shift_imm_di and Usd.
1554 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
1555 (*aarch64_<optab>_reg_minus<mode>3): ... this.
1557 2018-03-20 Sudakshina Das <sudi.das@arm.com>
1560 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
1561 to favor GPR over NEON registers.
1562 (<shift>di3_neon): Likewise.
1564 2018-03-20 Tom de Vries <tom@codesourcery.com>
1567 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
1568 (nvptx_process_pars): Emit bar.sync asap and alap.
1570 2018-03-20 Tom de Vries <tom@codesourcery.com>
1573 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
1574 seen_label if seen_label is already set.
1576 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1579 * config/i386/i386.c (fold_builtin_cpu): For features above 31
1580 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
1581 Use 1U instead of 1. Formatting fixes.
1584 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
1585 instead of TREE_TYPE (s1) for the return value.
1587 2018-03-19 Jakub Jelinek <jakub@redhat.com>
1589 PR tree-optimization/84946
1590 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
1591 bitsize + bitsize in poly_uint64 rather than poly_int64.
1594 * dwarf2asm.c: Include fold-const.c.
1595 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
1596 of decl rather than decl itself.
1598 PR rtl-optimization/84643
1599 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
1601 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
1604 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
1605 calling assemble_variable.
1607 2018-03-19 Sudakshina Das <sudi.das@arm.com>
1610 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
1611 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
1613 2018-03-19 Jim Wilson <jimw@sifive.com>
1616 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
1617 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
1618 (riscv_first_stack_step): Likewise.
1619 (riscv_option_override): Use STACK_BOUNDARY instead of
1621 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
1623 (BIGGEST_ALIGNMENT): Set to 128.
1624 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
1625 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
1628 2018-03-19 Richard Biener <rguenther@suse.de>
1630 PR tree-optimization/84933
1631 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
1632 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
1634 2018-03-19 Richard Biener <rguenther@suse.de>
1636 PR tree-optimization/84859
1637 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
1638 (cond_if_else_store_replacement): Perform sinking operation on
1639 single-store BBs regardless of MAX_STORES_TO_SINK setting.
1640 Generalize what a BB with a single eligible store is.
1642 2018-03-19 Richard Biener <rguenther@suse.de>
1644 PR tree-optimization/84929
1645 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
1646 chrec_is_positive against non-chrec arg.
1648 2018-03-19 Tamar Christina <tamar.christina@arm.com>
1651 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
1653 2018-03-18 Martin Liska <mliska@suse.cz>
1655 PR rtl-optimization/84635
1656 * regrename.c (build_def_use): Use matches_mode only when
1659 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
1661 PR tree-optimization/84913
1662 * tree-vect-loop.c (vectorizable_reduction): Don't try to
1663 vectorize chains of COND_EXPRs.
1665 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1667 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
1669 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1671 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
1673 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1675 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
1677 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1678 Kito Cheng <kito.cheng@gmail.com>
1680 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
1681 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
1682 (nds32_adjust_reg_alloc_order): New function.
1683 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
1685 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
1687 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
1688 nds32_print_operand, nds32_print_operand_address): Use
1689 HOST_WIDE_INT_PRINT_DEC instead.
1691 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1693 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
1695 2018-03-17 Jakub Jelinek <jakub@redhat.com>
1698 * config/i386/i386.c (initial_ix86_tune_features,
1699 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
1701 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
1702 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
1703 rather than 1u << ix86_tune. Formatting fix.
1704 (ix86_option_override_internal): Change ix86_arch_mask from
1705 unsigned int to unsigned HOST_WIDE_INT, initialize to
1706 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
1707 (ix86_function_specific_restore): Likewise.
1709 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1712 * postreload.c (reload_combine_recognize_pattern): Perform
1713 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
1714 truncate_int_for_mode the result for the destination's mode.
1717 * hsa-gen.c (mem_type_for_type): Fix comment typo.
1718 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
1720 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1723 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
1726 * lra-assigns.c (lra_split_hard_reg_for): Don't use
1727 regno_allocno_class_array and sorted_pseudos.
1728 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
1729 insns where regno is used.
1731 2018-03-16 Martin Liska <mliska@suse.cz>
1734 * multiple_target.c (create_dispatcher_calls): Redirect
1735 reference in the symbol table.
1737 2018-03-16 Martin Liska <mliska@suse.cz>
1740 * multiple_target.c (create_dispatcher_calls): Redirect also
1743 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1747 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
1748 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
1749 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
1751 2018-03-16 Julia Koval <julia.koval@intel.com>
1753 * doc/invoke.texi (Skylake Server): Add CLWB.
1754 Cannonlake): Remove CLWB.
1756 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1758 PR tree-optimization/84841
1759 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
1761 (FLOAT_ONE_CONST_TYPE): Define.
1762 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
1763 (sort_by_operand_rank): Put entries with higher constant_type last
1764 rather than first to match comments.
1766 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
1768 * config/nios2/nios2.md (movsi_internal): Fix thinko in
1771 2018-03-15 Jakub Jelinek <jakub@redhat.com>
1774 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
1775 check and use address of target always.
1777 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
1780 * config/i386/i386.c (indirect_thunk_needed): Update comments.
1781 (indirect_thunk_bnd_needed): Likewise.
1782 (indirect_thunks_used): Likewise.
1783 (indirect_thunks_bnd_used): Likewise.
1784 (indirect_return_needed): New.
1785 (indirect_return_bnd_needed): Likewise.
1786 (output_indirect_thunk_function): Add a bool argument for
1788 (output_indirect_thunk_function): Don't generate alias for
1789 function return thunk.
1790 (ix86_code_end): Call output_indirect_thunk_function to generate
1791 function return thunks.
1792 (ix86_output_function_return): Set indirect_return_bnd_needed
1793 and indirect_return_needed instead of indirect_thunk_bnd_needed
1794 and indirect_thunk_needed.
1796 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
1798 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
1799 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
1800 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
1802 2018-03-15 David Malcolm <dmalcolm@redhat.com>
1803 Paul Hua <paul.hua.gm@gmail.com>
1806 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
1808 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
1810 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
1811 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
1814 2018-03-15 Tamar Christina <tamar.christina@arm.com>
1817 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
1818 instead of GET_MODE_SIZE when comparing Units.
1820 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1823 * varasm.c (hash_section): Return an unchangeble hash value
1824 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
1825 Return !aarch64_can_use_per_function_literal_pools_p ().
1827 2018-03-15 Jakub Jelinek <jakub@redhat.com>
1830 * optabs.c (emit_conditional_move): Pass address of cmode's copy
1831 rather than address of cmode as last argument to prepare_cmp_insn.
1833 2018-03-15 Julia Koval <julia.koval@intel.com>
1835 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
1836 F_AVX512VNNI, F_AVX512BITALG): New.
1838 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
1841 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
1842 insn for floating-point loads and stores.
1844 2018-03-14 Carl Love <cel@us.ibm.com>
1846 * config/rs6000/rs6000-c.c: Add macro definitions for
1847 ALTIVEC_BUILTIN_VEC_PERMXOR.
1848 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
1849 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
1850 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
1851 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
1853 * config/doc/extend.texi: Add prototypes for vec_permxor.
1855 2018-03-14 David Malcolm <dmalcolm@redhat.com>
1858 * diagnostic-show-locus.c (class layout_point): Convert m_line
1859 from int to linenum_type.
1860 (line_span::comparator): Use linenum "compare" function when
1861 comparing line numbers.
1862 (test_line_span): New function.
1863 (layout_range::contains_point): Convert param "row" from int to
1865 (layout_range::intersects_line_p): Likewise.
1866 (layout::will_show_line_p): Likewise.
1867 (layout::print_source_line): Likewise.
1868 (layout::should_print_annotation_line_p): Likewise.
1869 (layout::print_annotation_line): Likewise.
1870 (layout::print_leading_fixits): Likewise.
1871 (layout::annotation_line_showed_range_p): Likewise.
1872 (struct line_corrections): Likewise for field m_row.
1873 (line_corrections::line_corrections): Likewise for param "row".
1874 (layout::print_trailing_fixits): Likewise.
1875 (layout::get_state_at_point): Likewise.
1876 (layout::get_x_bound_for_row): Likewise.
1877 (layout::print_line): Likewise.
1878 (diagnostic_show_locus): Likewise for locals "last_line" and
1880 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
1881 * input.c (selftest::test_linenum_comparisons): New function.
1882 (selftest::input_c_tests): Call it.
1883 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
1884 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
1885 * selftest.h (ASSERT_GT): New macro.
1886 (ASSERT_GT_AT): New macro.
1887 (ASSERT_LT): New macro.
1888 (ASSERT_LT_AT): New macro.
1890 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
1892 PR rtl-optimization/84780
1893 * combine.c (distribute_links): Don't make a link based on pc_rtx.
1895 2018-03-14 Martin Liska <mliska@suse.cz>
1897 * tree.c (record_node_allocation_statistics): Use
1898 get_stats_node_kind.
1899 (get_stats_node_kind): New function extracted from
1900 record_node_allocation_statistics.
1901 (free_node): Use get_stats_node_kind.
1903 2018-03-14 Richard Biener <rguenther@suse.de>
1905 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
1906 that the value-set of ANTIC_IN doesn't grow.
1909 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1911 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1912 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1914 2018-03-14 Julia Koval <julia.koval@intel.com>
1916 * config.gcc (icelake-client, icelake-server): New.
1918 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
1919 (initial_ix86_arch_features): Ditto.
1920 (PTA_SKYLAKE): Add SGX.
1921 (PTA_ICELAKE): Remove.
1922 (PTA_ICELAKE_CLIENT): New.
1923 (PTA_ICELAKE_SERVER): New.
1924 (ix86_option_override_internal): Split up icelake on icelake client and
1926 (get_builtin_code_for_version): Ditto.
1927 (fold_builtin_cpu): Ditto.
1928 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
1929 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
1930 * config/i386/i386.h (processor_type): Ditto.
1931 * doc/invoke.texi: Ditto.
1933 2018-03-14 Jakub Jelinek <jakub@redhat.com>
1936 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
1937 INTEGER_CST offset, add it together with bitpos / 8 and
1938 sign extend based on POINTER_SIZE.
1942 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
1945 * config/i386/constraints.md (Yc): New register constraint.
1946 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
1947 Use Yc constraint for alternative 2 of operand 0. Remove
1948 preferred_for_speed attribute.
1950 2018-03-14 Richard Biener <rguenther@suse.de>
1952 PR tree-optimization/84830
1953 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
1954 with the old one to avoid oscillations.
1956 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
1959 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
1961 (assign_by_spills): Return a flag of reload assignment failure.
1962 Do not process the reload assignment failures. Do not spill other
1963 reload pseudos if they has the same reg class. Update n if
1965 (lra_assign): Add a return arg. Set up from the result of
1966 assign_by_spills call.
1967 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1968 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1969 usage_insns if it is not NULL.
1970 (spill_hard_reg_in_range): New function.
1971 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1972 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1973 function prototypes.
1974 (lra_assign): Change prototype.
1975 * lra.c (lra): Add code to deal with fails by splitting hard reg
1978 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
1980 * config/riscv/riscv.opt (mrelax): New option.
1981 * config/riscv/riscv.c (riscv_file_start): Emit ".option
1982 "norelax" when riscv_mrelax is disabled.
1983 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
1985 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1988 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
1989 reassociation for int modes.
1991 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1993 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
1994 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
1996 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
1997 * config/aarch64/aarch64-sve.md
1998 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
1999 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
2000 (*extend<mode><Vwide>2): Rename to...
2001 (aarch64_sve_extend<mode><Vwide>2): ...this.
2002 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
2003 renaming the old pattern to...
2004 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
2006 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
2007 define_expand, renaming the old pattern to...
2008 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
2009 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
2010 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
2011 account when deciding which SVE instruction the optab should use.
2012 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
2014 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2016 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
2017 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
2018 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
2019 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
2020 (tlsdesc_small_<mode>): Turn a define_expand and use
2021 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
2022 (tlsdesc_small_advsimd_<mode>): ...this.
2023 (tlsdesc_small_sve_<mode>): New pattern.
2025 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2027 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
2028 (UNSPEC_UMUL_HIGHPART): New constants.
2029 (MUL_HIGHPART): New int iteraor.
2030 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
2031 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
2033 (*<su>mul<mode>3_highpart): New define_insn.
2035 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2038 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2041 2018-03-13 Martin Liska <mliska@suse.cz>
2044 * (sem_item_optimizer::sem_item_optimizer): Initialize new
2046 (sem_item_optimizer::~sem_item_optimizer): Release it.
2047 (sem_item_optimizer::merge_classes): Register variable aliases.
2048 (sem_item_optimizer::fixup_pt_set): New function.
2049 (sem_item_optimizer::fixup_points_to_sets): Likewise.
2050 * ipa-icf.h: Declare new variables and functions.
2052 2018-03-13 Jakub Jelinek <jakub@redhat.com>
2055 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
2056 integer_pow2p@2 and test integer_pow2p in condition.
2057 (A < 0 ? C : 0): Similarly for @1.
2060 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
2061 characters starting at p contain '\0' character, don't look beyond
2065 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
2066 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
2069 * reg-stack.c (change_stack): Change update_end var from int to
2070 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
2071 also call set_block_for_insn on the newly added insns and rescan.
2074 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
2075 on the last operand.
2078 * tree.c (stabilize_reference_1): Return save_expr (e) for
2079 STATEMENT_LIST even if it doesn't have side-effects.
2081 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
2083 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
2085 2018-03-12 Renlin Li <renlin.li@arm.com>
2087 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
2088 aarch64_output_scalar_simd_mov_immediate.
2090 2018-03-12 Martin Sebor <msebor@redhat.com>
2092 PR tree-optimization/83456
2093 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
2094 for perfectly overlapping calls to memcpy.
2095 (gimple_fold_builtin_memory_chk): Same.
2096 (gimple_fold_builtin_strcpy): Handle no-warning.
2097 (gimple_fold_builtin_stxcpy_chk): Same.
2098 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
2100 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2102 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
2103 parameter. Use it for SFmode.
2104 (rs6000_function_arg_advance_1): Adjust.
2105 (rs6000_function_arg): Adjust.
2106 (rs6000_gimplify_va_arg): Pass false for that new parameter.
2108 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2110 PR rtl-optimization/84169
2111 PR rtl-optimization/84780
2112 * combine.c (can_combine_p): Check for a 2-insn combination whether
2113 the destination register is used between the two insns, too.
2115 2018-03-12 Richard Biener <rguenther@suse.de>
2117 PR tree-optimization/84803
2118 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
2119 for refs DR analysis didn't process.
2121 2018-03-12 Richard Biener <rguenther@suse.de>
2123 PR tree-optimization/84777
2124 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
2125 force-vectorize loops ignore whether we are optimizing for size.
2127 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
2129 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
2130 (TARGET_MD_ASM_ADJUST): Define.
2132 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
2133 Kito Cheng <kito.cheng@gmail.com>
2134 Chung-Ju Wu <jasonwucj@gmail.com>
2136 * config/nds32/nds32.c (nds32_compute_stack_frame,
2137 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
2138 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
2139 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
2140 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
2141 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
2142 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
2143 * config/nds32/nds32.md (prologue, epilogue): Use macro
2144 NDS32_V3PUSH_AVAILABLE_P to do checking.
2146 2018-03-11 Jakub Jelinek <jakub@redhat.com>
2149 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
2150 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
2151 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
2152 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
2153 addition of most attributes on !orig_type_die or the attribute not
2154 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
2156 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2157 Chung-Ju Wu <jasonwucj@gmail.com>
2159 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
2161 * config/nds32/nds32.opt (mvh): New option.
2163 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2164 Chung-Ju Wu <jasonwucj@gmail.com>
2166 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
2168 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
2169 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
2172 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2173 Chung-Ju Wu <jasonwucj@gmail.com>
2175 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
2177 * config/nds32/nds32-multiple.md (strlensi): New pattern.
2178 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
2180 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
2181 Kito Cheng <kito.cheng@gmail.com>
2182 Chung-Ju Wu <jasonwucj@gmail.com>
2184 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
2185 UNSPEC_FFMISM and UNSPEC_FLMISM.
2186 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
2187 for ffb, ffmism and flmism.
2188 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
2189 (unspec_ffmism): Ditto.
2190 (unspec_flmism): Ditto.
2191 (nds32_expand_builtin_impl): Check if string extension is available.
2192 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
2193 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
2195 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
2198 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2201 * lra-assigns.c (assign_by_spills): Return a flag of reload
2202 assignment failure. Do not process the reload assignment
2203 failures. Do not spill other reload pseudos if they has the same
2205 (lra_assign): Add a return arg. Set up from the result of
2206 assign_by_spills call.
2207 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2208 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2209 usage_insns if it is not NULL.
2210 (spill_hard_reg_in_range): New function.
2211 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2212 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2213 function prototypes.
2214 (lra_assign): Change prototype.
2215 * lra.c (lra): Add code to deal with fails by splitting hard reg
2218 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
2221 * config/i386/i386.opt: Replace Enforcment with Enforcement.
2223 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
2226 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
2227 (dw_val_node): Add val_symbolic_view.
2228 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
2229 (symview_upper_bound): New.
2230 (new_line_info_table): Initialize symviews_since_reset.
2231 (dwarf2out_source_line): Count symviews_since_reset and set
2232 symview_upper_bound.
2233 (dw_val_equal_p): Handle symview.
2234 (add_AT_symview): New.
2235 (print_dw_val): Handle symview.
2236 (attr_checksum, attr_checksum_ordered): Likewise.
2237 (same_dw_val_p, size_of_die): Likewise.
2238 (value_format, output_die): Likewise.
2239 (add_high_low_attributes): Use add_AT_symview for entry_view.
2240 (dwarf2out_finish): Reset symview_upper_bound, clear
2243 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
2246 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
2247 Add strict argument and use it.
2248 (rs6000_split_multireg_move): Update for new strict argument.
2249 (mem_operand_gpr): Disallow all non-offsettable addresses.
2250 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
2252 2018-03-09 Jakub Jelinek <jakub@redhat.com>
2255 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
2256 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
2257 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
2260 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
2261 decl, use remap_type if we want to use the type.
2263 2018-03-09 Martin Sebor <msebor@redhat.com>
2265 PR tree-optimization/84526
2266 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2268 (builtin_access::generic_overlap): Be prepared to handle non-array
2271 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
2273 PR rtl-optimization/84682
2274 * lra-constraints.c (process_address_1): Check is_address flag
2275 for address constraints.
2276 (process_alt_operands): Likewise.
2277 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
2278 preprocess_constraints.
2279 * recog.h (preprocess_constraints): Add oploc parameter.
2281 * recog.c (preprocess_constraints): Test address_operand for
2282 CT_ADDRESS constraints.
2284 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2287 * lra-assigns.c (assign_by_spills): Return a flag of reload
2288 assignment failure. Do not process the reload assignment
2289 failures. Do not spill other reload pseudos if they has the same
2291 (lra_assign): Add a return arg. Set up from the result of
2292 assign_by_spills call.
2293 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2294 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2295 usage_insns if it is not NULL.
2296 (spill_hard_reg_in_range): New function.
2297 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2298 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2299 function prototypes.
2300 (lra_assign): Change prototype.
2301 * lra.c (lra): Add code to deal with fails by splitting hard reg
2304 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2307 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
2308 Accept complain bool parameter. Only emit errors if it is true.
2309 (arm_parse_cpu_option_name): Likewise.
2310 (arm_target_thumb_only): Adjust callers of the above.
2311 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
2312 prototype to take a default true bool parameter.
2313 (arm_parse_arch_option_name): Likewise.
2315 2018-03-09 David Malcolm <dmalcolm@redhat.com>
2316 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
2320 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
2321 * configure: Regenerate.
2322 * configure.ac ("linker --version-script option"): New.
2323 ("linker soname option"): New.
2325 2018-03-09 Richard Biener <rguenther@suse.de>
2327 PR tree-optimization/84775
2328 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
2329 immediate uses of predicate stmts and mark them modified.
2332 PR tree-optimization/84178
2333 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2335 (version_loop_for_if_conversion): Delay update_ssa call.
2336 (tree_if_conversion): Delay update_ssa until after predicate
2339 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
2342 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
2343 when the function accesses prior frames.
2345 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2348 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
2349 gen_llsym, otherwise call maybe_gen_llsym.
2352 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
2353 has ',' character inside of it.
2355 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2358 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
2359 as clobbering CC_REGNUM.
2361 2018-03-08 Richard Biener <rguenther@suse.de>
2364 * tree-scalar-evolution.c: Include tree-into-ssa.h.
2365 (follow_copies_to_constant): Do not follow SSA names registered
2368 2018-03-08 Richard Biener <rguenther@suse.de>
2370 PR tree-optimization/84178
2371 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2373 (version_loop_for_if_conversion): Delay update_ssa call.
2374 (tree_if_conversion): Delay update_ssa until after predicate
2377 2018-03-08 David Malcolm <dmalcolm@redhat.com>
2379 PR tree-optimization/84178
2380 * tree-if-conv.c (release_bb_predicate): Remove the
2381 the assertion that the stmts have NULL use_ops.
2382 Discard the statements, asserting that they haven't
2383 yet been added to a BB.
2385 2018-03-08 Richard Biener <rguenther@suse.de>
2387 PR tree-optimization/84746
2388 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
2389 (phi_translate): Pass in destination ANTIC_OUT set.
2390 (phi_translate_1): Likewise. For a simplified result lookup
2391 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
2392 (phi_translate_set): Adjust.
2393 (do_pre_regular_insertion): Likewise.
2394 (do_pre_partial_partial_insertion): Likewise.
2396 2018-03-08 Martin Liska <mliska@suse.cz>
2398 PR gcov-profile/84735
2399 * doc/gcov.texi: Document usage of profile files.
2400 * gcov-io.h: Document changes in the format.
2402 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
2406 * dwarf2out.c (struct dw_line_info_table): Update comments for
2408 (FORCE_RESET_NEXT_VIEW): New.
2409 (FORCE_RESETTING_VIEW_P): New.
2410 (RESETTING_VIEW_P): Check for -1 too.
2411 (ZERO_VIEW_P): Likewise.
2412 (new_line_info_table): Force-reset next view.
2413 (dwarf2out_begin_function): Likewise.
2414 (dwarf2out_source_line): Simplify zero_view_p initialization.
2415 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
2416 view directly. Omit view when omitting .loc at line 0.
2418 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2420 PR tree-optimization/84740
2421 * tree-switch-conversion.c (process_switch): Call build_constructors
2422 only if info.phi_count is non-zero.
2424 PR tree-optimization/84739
2425 * tree-tailcall.c (find_tail_calls): Check call arguments against
2426 DECL_ARGUMENTS (current_function_decl) rather than
2427 DECL_ARGUMENTS (func) when checking for tail recursion.
2429 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2431 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
2432 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
2433 Volker Reichelt's entry and add entries for people that perform
2434 GCC fuzzy testing and report numerous bugs.
2436 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
2439 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
2440 readonly data in sdata, if that is disabled.
2441 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
2442 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
2443 -mreadonly-in-sdata option.
2445 2018-03-07 Martin Sebor <msebor@redhat.com>
2447 PR tree-optimization/84468
2448 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
2449 basic block when looking for nul assignment.
2451 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
2454 * except.h (output_function_exception_table): Adjust prototype.
2455 * except.c (output_function_exception_table): Remove FNNAME parameter
2456 and add SECTION parameter. Ouput one part of the table at a time.
2457 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
2458 the first part of the exception table and emit unwind directives.
2459 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
2460 (i386_pe_seh_cold_init): Likewise.
2461 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
2462 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
2463 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
2464 (ix86_output_call_insn): Emit a nop in one more case for SEH.
2465 * config/i386/winnt.c: Include except.h.
2466 (struct seh_frame_state): Add reg_offset, after_prologue and
2467 in_cold_section fields.
2468 (i386_pe_seh_end_prologue): Set seh->after_prologue.
2469 (i386_pe_seh_cold_init): New function.
2470 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
2471 to seh->in_cold_section.
2472 (seh_emit_push): Record the offset of the push.
2473 (seh_emit_save): Record the offet of the save.
2474 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
2475 Test seh->after_prologue to disregard the epilogue.
2476 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
2477 (i386_pe_end_cold_function): New function.
2479 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2482 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
2483 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
2486 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
2488 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
2489 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
2492 * multiple_target.c: Include tree-inline.h and intl.h.
2493 (expand_target_clones): Diagnose and fail if node->definition and
2494 !tree_versionable_function_p (node->decl).
2496 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
2498 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
2500 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
2501 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
2502 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
2504 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2507 * combine.c (try_combine): Use reg_or_subregno instead of handling
2508 just paradoxical SUBREGs and REGs.
2510 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
2512 * config/arc/arc.c (arc_finalize_pic): Remove function.
2513 (arc_must_save_register): We use single base PIC register, remove
2514 checks to save/restore the PIC register.
2515 (arc_expand_prologue): Likewise.
2516 * config/arc/arc-protos.h (arc_set_default_type_attributes):
2518 (arc_verify_short): Likewise.
2519 (arc_attr_type): Likewise.
2520 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
2521 (walk_stores): Likewise.
2522 (arc_address_cost): Make it static.
2523 (arc_verify_short): Likewise.
2524 (branch_dest): Likewise.
2525 (arc_attr_type): Likewise.
2526 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
2527 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
2528 (arc_final_prescan_insn): Remove inserting the nops due to
2529 hardware hazards. It is done in reorg step.
2530 (insn_length_variant_t): Remove.
2531 (insn_length_parameters_t): Likewise.
2532 (arc_insn_length_parameters): Likewise.
2533 (arc_get_insn_variants): Likewise.
2534 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
2536 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2539 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
2542 PR tree-optimization/84687
2543 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
2545 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
2547 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2549 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
2550 Rename to ppc_speculation_barrier.
2551 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
2552 __builtin_ppc_speculation_barrier.
2554 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2557 * combine.c (combine_simplify_rtx): Don't try to simplify if
2558 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
2561 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
2563 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
2564 to 32 bytes when compiling for POWER9.
2566 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2569 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
2570 regparm >= 3 with no arg reg available also for calls with
2571 flag_force_indirect_call. Pass decl to ix86_function_regparm.
2574 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
2576 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
2578 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
2581 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
2583 2018-03-05 Richard Biener <rguenther@suse.de>
2585 PR tree-optimization/84486
2586 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
2587 When inserting a __builtin_assume_aligned call set the LHS
2588 SSA name alignment info accordingly.
2590 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
2592 PR tree-optimization/84114
2593 * config/aarch64/aarch64.c (aarch64_reassociation_width)
2594 Avoid reassociation of FLOAT_MODE addition.
2596 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
2598 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
2599 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
2600 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
2601 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
2602 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
2603 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
2604 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
2606 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
2607 __builtin_ia32_wbinvd): New builtins.
2608 (SPECIAL_ARGS2): New.
2609 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
2610 (SPECIAL_ARGS2): New.
2611 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
2612 (ix86_valid_target_attribute_inner_p): Ditto.
2613 (ix86_init_mmx_sse_builtins): Add special_args2.
2614 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
2615 TARGET_WBNOINVD_P): New.
2616 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
2617 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
2618 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
2619 * config/i386/immintrin.h (_wbinvd): New intrinsic.
2620 * config/i386/pconfigintrin.h: New file.
2621 * config/i386/wbnoinvdintrin.h: Ditto.
2622 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
2623 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
2625 2018-03-05 Richard Biener <rguenther@suse.de>
2627 PR tree-optimization/84670
2628 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2630 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2631 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2632 (compute_antic_aux): Only assert the number of values in ANTIC_IN
2633 doesn't grow if all successors (recursively) were visited at least
2636 2018-03-05 Richard Biener <rguenther@suse.de>
2638 PR tree-optimization/84650
2639 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
2640 if executed in the loop pipeline.
2642 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
2644 * doc/configfiles.texi (Configuration Files): Move info about
2645 conditionalizing $target-protos.h to...
2646 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
2647 differs from $target-protos.h.
2649 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
2650 Chung-Ju Wu <jasonwucj@gmail.com>
2652 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
2653 * config/nds32/nds32-multiple.md (setmemsi): Define.
2654 * config/nds32/nds32-memory-manipulation.c
2655 (nds32_gen_dup_4_byte_to_word_value): New.
2656 (emit_setmem_word_loop): New.
2657 (emit_setmem_byte_loop): New.
2658 (nds32_expand_setmem_loop): New.
2659 (nds32_expand_setmem_loop_v3m): New.
2660 (nds32_expand_setmem_unroll): New.
2661 (nds32_expand_setmem): New.
2663 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2664 Chung-Ju Wu <jasonwucj@gmail.com>
2666 * config/nds32/nds32-memory-manipulation.c
2667 (nds32_emit_load_store): New.
2668 (nds32_emit_post_inc_load_store): New.
2669 (nds32_emit_mem_move): New.
2670 (nds32_emit_mem_move_block): New.
2671 (nds32_expand_movmemsi_loop_unknown_size): New.
2672 (nds32_expand_movmemsi_loop_known_size): New.
2673 (nds32_expand_movmemsi_loop): New.
2674 (nds32_expand_movmemsi_unroll): New.
2675 (nds32_expand_movmemqi): Rename ...
2676 (nds32_expand_movmemsi): ... to this.
2677 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
2678 (movmemsi): ... to this.
2679 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
2680 (nds32_expand_movmemsi): ... to this.
2682 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2683 Monk Chiang <sh.chiang04@gmail.com>
2684 Chung-Ju Wu <jasonwucj@gmail.com>
2686 * config/nds32/nds32-protos.h
2687 (nds32_expand_load_multiple): New arguments.
2688 (nds32_expand_store_multiple): Ditto.
2689 (nds32_valid_multiple_load_store): Rename ...
2690 (nds32_valid_multiple_load_store_p): ... to this.
2691 * config/nds32/nds32-memory-manipulation.c
2692 (nds32_expand_load_multiple): Refine implementation.
2693 (nds32_expand_store_multiple): Ditto.
2694 * config/nds32/nds32-multiple.md
2695 (load_multiple): Update nds32_expand_load_multiple interface.
2696 (store_multiple): Update nds32_expand_store_multiple interface.
2697 * config/nds32/nds32-predicates.c
2698 (nds32_valid_multiple_load_store): Rename ...
2699 (nds32_valid_multiple_load_store_p): ... to this and refine
2701 * config/nds32/predicates.md
2702 (nds32_load_multiple_and_update_address_operation): New predicate.
2703 (nds32_store_multiple_and_update_address_operation): New predicate.
2705 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2706 Chung-Ju Wu <jasonwucj@gmail.com>
2708 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
2709 (combo): New attribute.
2710 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
2712 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
2714 * config/nds32/nds32.opt: Change -mcmodel= default value.
2716 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
2717 Monk Chiang <sh.chiang04@gmail.com>
2718 Chung-Ju Wu <jasonwucj@gmail.com>
2720 * config/nds32/constants.md (unspec_element): New enum.
2721 * config/nds32/constraints.md (Umw): New constraint.
2722 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
2723 * config/nds32/nds32-intrinsic.md: Likewise.
2724 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
2725 (nds32_valid_smw_lwm_base_p): New.
2726 (nds32_output_smw_single_word): New.
2727 (nds32_output_lmw_single_word): New.
2728 (nds32_expand_unaligned_load): New.
2729 (nds32_expand_unaligned_store): New.
2730 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
2731 (nds32_output_smw_single_word): Declare.
2732 (nds32_output_lmw_single_word): Declare.
2733 (nds32_expand_unaligned_load): Declare.
2734 (nds32_expand_unaligned_store): Declare.
2735 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
2736 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
2737 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
2738 NDS32_BUILTIN_UASTORE_DW.
2739 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
2742 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
2743 Kito Cheng <kito.cheng@gmail.com>
2744 Chung-Ju Wu <jasonwucj@gmail.com>
2746 * config/nds32/nds32-intrinsic.c
2747 (nds32_expand_builtin_null_ftype_reg): Delete.
2748 (nds32_expand_builtin_reg_ftype_imm): Ditto.
2749 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
2750 (nds32_read_argument): New.
2751 (nds32_legitimize_target): Ditto.
2752 (nds32_legitimize_argument): Ditto.
2753 (nds32_check_constant_argument): Ditto.
2754 (nds32_expand_unop_builtin): Ditto.
2755 (nds32_expand_unopimm_builtin): Ditto.
2756 (nds32_expand_binop_builtin): Ditto.
2757 (nds32_builtin_decl_impl): Ditto.
2758 (builtin_description): Ditto.
2759 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
2760 (nds32_init_builtins_impl): Ditto.
2761 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
2762 (nds32_builtin_decl): New.
2763 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
2764 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
2766 2018-03-02 Jeff Law <law@redhat.com>
2768 * reorg.c (stop_search_p): Handle DEBUG_INSN.
2769 (redundant_insn, fill_simple_delay_slots): Likewise.
2770 (fill_slots_from_thread): Likewise.
2771 * resource.c (mark_referenced_resources): Likewise.
2772 (mark_set_resources, find_dead_or_set_registers): Likewise.
2774 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2776 * substring-locations.h (format_warning_va): Formatting fix for
2778 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
2780 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
2781 * substring-locations.c: Include intl.h.
2782 (format_warning_va): Turned into small wrapper around
2783 format_warning_n_va, renamed to ...
2784 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
2785 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
2787 (format_warning_at_substring_n): New function.
2788 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
2789 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
2790 format_warning_at_substring with just a shorter name instead of
2791 const function pointer.
2792 (fmtwarn_n): New function.
2793 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
2794 appropriate, get rid of all the fmtstr temporaries, move conditionals
2795 with G_() wrapped string literals directly into fmtwarn arguments,
2796 cast dir.len to (int), formatting fixes.
2798 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
2800 * doc/invoke.texi: Remove "Cilk Plus" references.
2802 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2803 Richard Biener <rguenther@suse.de>
2806 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
2807 for error or warning attributes if CALL_FROM_THUNK_P is set.
2810 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2813 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
2814 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
2817 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
2818 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
2820 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
2821 instead of -1U in last predictors element's probability member.
2823 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2826 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2827 arguments if they are comparable.
2829 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
2831 PR tree-optimization/84634
2832 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
2833 masks and masked_loop_p with a single loop_masks, making sure it's
2834 null for bb vectorization.
2836 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
2838 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
2839 (vect_analyze_data_ref_access): Use loop->safe_len rather than
2840 loop->force_vectorize to check whether there is no alias.
2842 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2845 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
2847 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
2849 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
2850 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
2851 instead of a loop around prev_real_insn.
2852 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
2856 * config/i386/i386.c (ix86_print_operand): Use conditional
2857 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
2860 2018-03-02 Richard Biener <rguenther@suse.de>
2862 PR tree-optimization/84427
2863 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
2864 (bitmap_set_subtract_values): Rewrite to handle multiple
2867 (prune_clobbered_mems): Likewise.
2868 (phi_translate): Take edge instead of pred/phiblock.
2869 (phi_translate_1): Likewise.
2870 (phi_translate_set): Likewise. Insert all translated
2871 exprs for a value into the set, keeping possibly multiple
2872 expressions per value.
2873 (compute_antic_aux): Adjust for phi_translate changes.
2874 When intersecting union the expressions and prune those
2875 not in the final value set, keeping possibly multiple
2876 expressions per value. Do not use value-insertion
2877 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
2878 all expressions. Add verification that the value-sets
2879 only shrink during iteration.
2880 (compute_partial_antic_aux): Adjust for the phi_translate changes.
2881 (do_pre_regular_insertion): Likewise.
2882 (do_pre_partial_partial_insertion): Likewise.
2884 2018-03-02 Richard Biener <rguenther@suse.de>
2887 * config/darwin.c (saved_debug_info_level): New static global.
2888 (darwin_asm_lto_start): Disable debug info generation for LTO out.
2889 (darwin_asm_lto_end): Restore debug info generation settings.
2891 2018-03-01 Martin Liska <mliska@suse.cz>
2894 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
2897 2018-03-01 Richard Biener <rguenther@suse.de>
2900 * dwarf2out.c (gen_variable_die): Properly handle late VLA
2901 type annotation with LTO when debug was disabled at compile-time.
2903 2018-03-01 Matthew Fortune <mfortune@gmail.com>
2905 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
2907 (mips_final_postscan_insn): Likewise.
2909 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
2911 PR rtl-optimization/84528
2912 * alias.c (init_alias_target): Add commentary.
2913 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
2914 a unique base value if the frame pointer is not eliminated
2915 to the stack pointer.
2917 2018-03-01 Tom de Vries <tom@codesourcery.com>
2919 PR rtl-optimization/83327
2920 * lra-int.h (hard_regs_spilled_into): Declare.
2921 * lra.c (hard_regs_spilled_into): Define.
2922 (init_reg_info): Init hard_regs_spilled_into.
2923 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
2924 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
2925 (process_bb_lives): Handle hard_regs_spilled_into.
2926 (lra_create_live_ranges_1): Before doing liveness propagation, clear
2927 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
2929 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
2931 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
2932 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
2933 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
2934 * config/rs6000/aix72.h: New file.
2936 2018-02-28 Jakub Jelinek <jakub@redhat.com>
2938 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
2939 instead of warning_at with conditional singular and plural messages
2943 * stor-layout.c (update_alignment_for_field): For
2944 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
2945 && !DECL_PACKED (field), do the alignment update, just use
2946 only desired_align instead of MAX (type_align, desired_align)
2948 (place_field): Don't do known_align < desired_align handling
2949 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
2950 is non-NULL, instead do it after rli->prev_field handling and
2951 only if not within a bitfield word. For DECL_PACKED (field)
2952 use type_align of BITS_PER_UNIT.
2954 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
2956 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
2957 superfluous parentheses and trailing spaces.
2959 2018-02-28 Richard Biener <rguenther@suse.de>
2961 PR tree-optimization/84584
2962 * graphite-scop-detection.c (scop_detection::add_scop): Discard
2963 SCoPs with fake exit edge.
2965 2018-02-28 Martin Liska <mliska@suse.cz>
2968 * timevar.c (timer::print): Fix format to properly print 100%
2971 2018-02-28 Richard Biener <rguenther@suse.de>
2974 * genmatch.c (capture_info::walk_match): Do not mark
2975 captured expressions without operands as expr_p given
2976 they act more like predicates and should be subject to
2977 "lost tail" side-effect preserving.
2979 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
2981 PR rtl-optimization/81611
2982 * auto-inc-dec.c (attempt_change): Move dead note from
2983 mem_insn if it's the next use of regno
2984 (find_address): Take address use of reg holding
2985 non-incremented value. Add parm to limit search to the named
2987 (merge_in_block): Attempt to use a mem insn that is the next
2988 use of the original regno.
2990 2018-02-27 Martin Sebor <msebor@redhat.com>
2993 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
2994 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
2996 2018-02-27 Martin Sebor <msebor@redhat.com>
2998 PR translation/84207
2999 * diagnostic-core.h (warning_n, error_n, inform_n): Change
3000 n argument to unsigned HOST_WIDE_INT.
3001 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
3002 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
3003 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
3004 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
3006 2018-02-27 Richard Biener <rguenther@suse.de>
3008 PR tree-optimization/84512
3009 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
3010 Do not use the estimate returned from record_stmt_cost for
3011 the scalar iteration cost but sum properly using add_stmt_cost.
3013 2018-02-27 Richard Biener <rguenther@suse.de>
3015 PR tree-optimization/84466
3016 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3017 Adjust last change to less strictly validate use operands.
3019 2018-02-27 Martin Liska <mliska@suse.cz>
3021 PR gcov-profile/84548
3022 * gcov.c (process_file): Allow partial overlap and consider it
3023 also as group functions.
3024 (output_lines): Properly calculate range of lines for a group.
3026 2018-02-27 Martin Liska <mliska@suse.cz>
3028 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
3029 'ggc' suffixes. Change first column width.
3030 (timer::print): Fix formatting of the column.
3032 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
3034 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
3035 preserve inline entry blocks for the sake of debug inline
3036 entry point markers alone.
3037 (remove_unused_locals): Suggest in comments a better place to
3038 force the preservation of inline entry blocks that are
3039 otherwise unused, but do not preserve them.
3041 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3043 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
3045 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3048 * config/i386/constraints.md (Bs): Replace
3049 ix86_indirect_branch_register with
3050 TARGET_INDIRECT_BRANCH_REGISTER.
3052 * config/i386/i386.md (indirect_jump): Likewise.
3053 (tablejump): Likewise.
3054 (*sibcall_memory): Likewise.
3055 (*sibcall_value_memory): Likewise.
3056 Peepholes of indirect call and jump via memory: Likewise.
3057 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
3058 (*sibcall_value_GOT_32): Likewise.
3059 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3060 (GOT_memory_operand): Likewise.
3061 (call_insn_operand): Likewise.
3062 (sibcall_insn_operand): Likewise.
3063 (GOT32_symbol_operand): Likewise.
3064 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
3066 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3068 PR rtl-optimization/83496
3069 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
3070 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
3071 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
3072 redundant insn, if any.
3073 (relax_delay_slots): Likewise.
3074 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
3076 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
3078 PR tree-optimization/83965
3079 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
3080 that grouped statements are part of a reduction chain. Return
3081 true if the statement is not marked as a reduction itself but
3083 (vect_recog_dot_prod_pattern): Don't check whether the statement
3084 is part of a group here.
3085 (vect_recog_sad_pattern): Likewise.
3086 (vect_recog_widen_sum_pattern): Likewise.
3088 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3091 * final.c (rest_of_clean_state): Also look for calls inside sequences.
3093 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3096 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
3098 (ix86_output_indirect_function_return): New prototype.
3099 (ix86_split_simple_return_pop_internal): Likewise.
3100 * config/i386/i386.c (indirect_return_via_cx): New.
3101 (indirect_return_via_cx_bnd): Likewise.
3102 (indirect_thunk_name): Handle return va CX_REG.
3103 (output_indirect_thunk_function): Create alias for
3104 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
3105 (ix86_output_indirect_jmp): Remove the bool argument.
3106 (ix86_output_indirect_function_return): New function.
3107 (ix86_split_simple_return_pop_internal): Likewise.
3108 * config/i386/i386.md (*indirect_jump): Don't pass false
3109 to ix86_output_indirect_jmp.
3110 (*tablejump_1): Likewise.
3111 (simple_return_pop_internal): Change it to define_insn_and_split.
3112 Call ix86_split_simple_return_pop_internal to split it for
3114 (simple_return_indirect_internal): Call
3115 ix86_output_indirect_function_return instead of
3116 ix86_output_indirect_jmp.
3118 2018-02-26 Jakub Jelinek <jakub@redhat.com>
3121 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
3122 memset and value initialization afterwards.
3124 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
3126 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
3128 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3131 * common/config/aarch64/aarch64-common.c
3132 (aarch_option_optimization_table[]): Switch
3133 off fomit-frame-pointer
3135 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3136 Chung-Ju Wu <jasonwucj@gmail.com>
3138 * config/nds32/nds32-multiple.md (load_multiple): Disallow
3140 (store_multiple): Ditto.
3142 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3144 * config.gcc: Add --with-cpu support for nds32 target.
3145 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
3146 * config/nds32/nds32.opt: Add -mcpu= option.
3148 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
3150 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
3151 isel=yes): Warn for these deprecated options.
3153 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
3155 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
3156 ISA_2_5_MASKS_EMBEDDED.
3158 2018-02-23 Jakub Jelinek <jakub@redhat.com>
3160 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
3161 p->max as pointers rather than using iterative_hash_expr.
3163 2018-02-23 Carl Love <cel@us.ibm.com>
3165 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
3166 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
3168 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
3169 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
3170 P8V_BUILTIN_VEC_VUNSIGNED2.
3172 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
3175 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
3176 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
3178 * lra-constraints.c (curr_insn_transform): Set up
3179 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
3181 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
3182 * lra-eliminations.c (spill_pseudos): Ditto.
3183 (process_insn_for_elimination): Ditto.
3184 * lra-lives.c (reg_early_clobber_p): Use the new macros.
3185 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
3186 LRA_NON_CLOBBERED_ALT.
3188 2018-02-22 Martin Sebor <msebor@redhat.com>
3190 PR tree-optimization/84480
3191 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
3192 to maybe_diag_stxncpy_trunc. Call it.
3193 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
3194 from gimple_fold_builtin_strcpy. Print inlining stack.
3195 (handle_builtin_stxncpy): Print inlining stack.
3196 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
3198 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
3201 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
3202 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
3203 and -fcheck-pointer-bounds are used together.
3204 (indirect_thunk_prefix): New enum.
3205 (indirect_thunk_need_prefix): New function.
3206 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
3207 "_nt" instead of "_bnd" for NOTRACK prefix.
3208 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
3209 (output_indirect_thunk_function): Likewise.
3211 (ix86_code_end): Update output_indirect_thunk_function calls.
3212 (ix86_output_indirect_branch_via_reg): Replace
3213 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
3214 (ix86_output_indirect_branch_via_push): Likewise.
3215 (ix86_output_function_return): Likewise.
3216 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
3217 incompatible with -fcf-protection=branch and
3218 -fcheck-pointer-bounds.
3220 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3223 * config/aarch64/aarch64.c (aarch64_print_address_internal):
3224 Change gcc_assert call to output_operand_lossage.
3226 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3228 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
3230 2018-02-22 DJ Delorie <dj@redhat.com>
3231 Sebastian Perta <sebastian.perta@renesas.com>
3232 Oleg Endo <olegendo@gcc.gnu.org>
3234 * config/rx/rx.c (rx_rtx_costs): New function.
3235 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
3237 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
3239 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
3241 2018-02-22 Martin Liska <mliska@suse.cz>
3244 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
3245 Add "native" as a possible value.
3247 2018-02-22 Martin Liska <mliska@suse.cz>
3250 * config/i386/i386.c (ix86_option_override_internal):
3251 Add "native" as a possible value for -march and -mtune.
3253 2018-02-22 Jakub Jelinek <jakub@redhat.com>
3256 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
3257 to all type variants.
3259 PR tree-optimization/84503
3260 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
3261 width as info->bitpos + info->bitsize - start.
3262 (merged_store_group::merge_overlapping): Simplify width computation.
3263 (check_no_overlap): New function.
3264 (imm_store_chain_info::try_coalesce_bswap): Compute expected
3265 start + width and last_order of the group, fail if check_no_overlap
3267 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
3268 to group if check_no_overlap fails.
3270 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3272 * config/rs6000/altivec.md: Delete contraint arguments to
3273 define_expand, define_split, and define_peephole2, and in
3274 define_insn_and_split if always unused.
3275 * config/rs6000/darwin.md: Ditto.
3276 * config/rs6000/dfp.md: Ditto.
3277 * config/rs6000/rs6000.md: Ditto.
3278 * config/rs6000/sync.md: Ditto.
3279 * config/rs6000/vector.md: Ditto.
3280 * config/rs6000/vsx.md: Ditto.
3282 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3284 * config/rs6000/altivec.md: Write output control strings as braced
3285 blocks instead of double-quoted strings.
3286 * config/rs6000/darwin.md: Ditto.
3287 * config/rs6000/rs6000.md: Ditto.
3288 * config/rs6000/vector.md: Ditto.
3289 * config/rs6000/vsx.md: Ditto.
3291 2018-02-21 Jason Merrill <jason@redhat.com>
3293 PR c++/84314 - ICE with templates and fastcall attribute.
3294 * attribs.c (build_type_attribute_qual_variant): Remove assert.
3296 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3298 * ipa-cp.c (determine_versionability): Fix comment typos.
3300 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3303 * ipa-cp.c (determine_versionability): Do not version functions caling
3306 2018-02-21 Martin Liska <mliska@suse.cz>
3309 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
3310 Add "native" as a possible value.
3311 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
3312 the macro when native cpu detection is available.
3314 2018-02-21 Martin Liska <mliska@suse.cz>
3317 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
3318 Add "native" as a possible value.
3319 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
3320 when native cpu detection is available.
3322 2018-02-21 Jakub Jelinek <jakub@redhat.com>
3323 Martin Sebor <msebor@redhat.com>
3325 PR tree-optimization/84478
3326 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
3328 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
3329 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
3330 support which is conservatively correct, for 2 only stay conservative
3331 for maxlen. Formatting and comment capitalization fixes. Add STRICT
3332 argument to the 2 argument get_range_strlen, adjust 6 arg
3333 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
3335 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
3336 (gimple_fold_builtin_strlen): Pass true as last argument to
3339 2018-02-20 Martin Sebor <msebor@redhat.com>
3342 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
3343 (builtin_memref::set_base_and_offset): Same. Handle inner references.
3344 (builtin_memref::builtin_memref): Factor out parts into
3345 set_base_and_offset and call it.
3347 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3350 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
3351 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
3352 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
3353 search at the associated MODE_INT.
3355 2018-02-20 Jeff Law <law@redhat.com>
3358 PR tree-optimization/81592
3360 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
3361 for range data rather than using global data.
3362 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
3363 range data rather than using global data.
3364 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
3365 pass it to children as needed.
3366 (struct directive::fmtresult): Similarly.
3367 (struct directive::set_width): Similarly.
3368 (struct directive::set_precision): Similarly.
3369 (format_integer, format_directive, parse_directive): Similarly.
3370 (format_none): Accept unnamed vr_values parameter.
3371 (format_percent, format_floating, format_character): Similarly.
3372 (format_string, format_plain): Similarly.
3373 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
3374 the EVRP range analyzer for range data rather than using global data.
3375 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
3376 gimple-ssa-evrp-analyze.h
3377 (class sprintf_dom_walker): Add after_dom_children member function.
3378 Add evrp_range_analyzer member.
3379 (sprintf_dom_walker::before_dom_children): Call into the EVRP
3380 range analyzer as needed.
3381 (sprintf_dom_walker::after_dom_children): New member function.
3382 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
3384 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
3385 (evrp_range_analyzer::pop_to_marker): Likewise.
3387 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3389 PR tree-optimization/84419
3390 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
3391 with the required type if its current type is compatible but
3394 2018-02-20 Jakub Jelinek <jakub@redhat.com>
3397 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
3398 after vectorization.
3400 2018-02-20 Martin Liska <mliska@suse.cz>
3403 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
3404 possible values if we don't have a hint.
3406 2018-02-20 Martin Liska <mliska@suse.cz>
3410 * final.c (shorten_branches): Build align_tab array with one
3412 * opts.c (finish_options): Add alignment option limit check.
3413 (MAX_CODE_ALIGN): Likewise.
3414 (MAX_CODE_ALIGN_VALUE): Likewise.
3415 * doc/invoke.texi: Document maximum allowed option value for
3416 all -falign-* options.
3418 2018-02-19 Jakub Jelinek <jakub@redhat.com>
3421 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
3422 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
3423 * var-tracking.c (emit_note_insn_var_location): Remove all references
3424 to NOTE_INSN_CALL_ARG_LOCATION.
3425 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
3426 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
3427 Use copy_rtx_if_shared.
3428 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
3429 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
3430 (dwarf2out_var_location): Remove handling of
3431 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
3433 * final.c (final_scan_insn): Remove all references to
3434 NOTE_INSN_CALL_ARG_LOCATION.
3435 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
3436 before dumping final insns.
3437 * except.c (emit_note_eh_region_end): Remove all references to
3438 NOTE_INSN_CALL_ARG_LOCATION.
3439 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
3440 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
3441 * config/arc/arc.c (hwloop_optimize): Likewise.
3442 * config/arm/arm.c (create_fix_barrier): Likewise.
3443 * config/s390/s390.c (s390_chunkify_start): Likewise.
3444 * config/sh/sh.c (find_barrier): Likewise.
3445 * config/i386/i386.c (rest_of_insert_endbranch,
3446 ix86_seh_fixup_eh_fallthru): Likewise.
3447 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
3448 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
3449 * config/frv/frv.c (frv_function_prologue): Likewise.
3450 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
3452 (note_outside_basic_block_p): Remove all references to
3453 NOTE_INSN_CALL_ARG_LOCATION.
3454 * gengtype.c (adjust_field_rtx_def): Likewise.
3455 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
3457 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
3458 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
3461 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
3464 PR tree-optimization/84452
3465 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
3466 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
3469 2018-02-19 Martin Liska <mliska@suse.cz>
3472 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
3474 2018-02-19 Martin Liska <mliska@suse.cz>
3475 Richard Sandiford <richard.sandiford@linaro.org>
3477 PR tree-optimization/82491
3478 * gimple-fold.c (get_base_constructor): Make earlier bail out
3481 2018-02-19 Carl Love <cel@us.ibm.com>
3483 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
3484 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
3486 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
3487 P8V_BUILTIN_VEC_NEG.
3489 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
3491 * config/rl78/rl78.md (movdf): New define expand.
3493 2018-02-19 Martin Liska <mliska@suse.cz>
3496 * doc/invoke.texi: Fix typo.
3497 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
3499 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
3501 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
3502 handle rs6000_single_float and rs6000_double_float specially for
3505 2018-02-16 Jeff Law <law@redhat.com>
3507 * config/rx/rx.c (add_pop_cfi_notes): New function.;
3510 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3513 * ipa-inline.c (inline_small_functions): Fix a typo.
3515 2018-02-16 Nathan Sidwell <nathan@acm.org>
3517 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
3519 2018-02-16 Carl Love <cel@us.ibm.com>
3521 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
3522 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
3523 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
3524 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
3525 expansion to P8V_BUILTIN_VEC_FLOAT2.
3527 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
3529 PR rtl-optimization/70023
3530 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
3531 src_regno into account.
3533 2018-02-16 Carl Love <cel@us.ibm.com>
3535 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
3536 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
3537 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
3538 * config/rs6000/rs6000.c: Remove case statements for
3539 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
3540 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
3541 and P9V_BUILTIN_VEC_VINSERT4B.
3542 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
3543 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
3544 * config/rs6000/vsx.md:
3545 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
3548 2018-02-16 Carl Love <cel@us.ibm.com>
3550 * config/rs6000/altivec.h: Add builtin names vec_extract4b
3552 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
3554 * config/rs6000/rs6000-c.c: Add the definitions for
3555 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
3556 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
3557 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
3558 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
3559 definition for insert4b and define insn *insert3b_internal.
3560 * doc/extend.texi: Add documentation for vec_extract4b.
3562 2018-02-16 Nathan Sidwell <nathan@acm.org>
3564 * doc/extend.texi (Backwards Compatibility): Mention friend
3565 injection. Note for-scope is deprecated.
3566 * doc/invoke.texi (-ffriend-injection): Deprecate.
3568 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
3570 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
3571 that moved to I2, also allow destinations that are a paradoxical
3572 subreg (instead of a normal reg).
3574 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
3577 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
3580 2018-02-16 Richard Biener <rguenther@suse.de>
3582 PR tree-optimization/84037
3583 PR tree-optimization/84016
3585 * config/i386/i386.c (ix86_builtin_vectorization_cost):
3586 Adjust vec_construct for the fact we need additional higher latency
3587 128bit inserts for AVX256 and AVX512 vector builds.
3588 (ix86_add_stmt_cost): Scale vector construction cost for
3591 2018-02-16 Richard Biener <rguenther@suse.de>
3593 PR tree-optimization/84417
3594 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
3595 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
3596 (non_rewritable_lvalue_p): Likewise, use poly-ints.
3598 2018-02-16 Martin Liska <mliska@suse.cz>
3601 * internal-fn.def (ASAN_CHECK): Set proper flags.
3602 (ASAN_MARK): Likewise.
3604 2018-02-16 Julia Koval <julia.koval@intel.com>
3606 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
3607 from PTA_CANNONLAKE.
3609 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3612 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
3613 Use ++iter rather than iter++ for std::list iterators.
3614 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
3615 defer deleting them until all nodes in the forest are processed. Do
3616 free even leaf nodes. Change to_process into auto_vec.
3619 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
3620 * vec.h (vec_default_construct): Use memset instead of placement new
3621 if BROKEN_VALUE_INITIALIZATION is defined.
3622 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
3623 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
3626 PR rtl-optimization/83723
3627 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
3628 * lra.c (lra_substitute_pseudo): Likewise. If true, use
3629 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
3631 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
3633 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
3635 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
3637 PR rtl-optimization/81443
3638 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
3639 from inner REGs to paradoxical SUBREGs.
3641 2018-02-16 Richard Biener <rguenther@suse.de>
3643 PR tree-optimization/84399
3644 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3645 For operands we can analyze at their definition make sure we can
3646 analyze them at each use as well.
3648 2018-02-16 Richard Biener <rguenther@suse.de>
3650 PR tree-optimization/84190
3651 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
3652 volatile accesses if the decl isn't volatile.
3654 2018-02-15 Jason Merrill <jason@redhat.com>
3656 PR c++/84314 - ICE with templates and fastcall attribute.
3657 * attribs.c (build_type_attribute_qual_variant): Don't clobber
3658 TYPE_CANONICAL on an existing type.
3660 2018-02-15 Jakub Jelinek <jakub@redhat.com>
3662 PR tree-optimization/84383
3663 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
3664 dstoff nor call operand_equal_p if dstbase is NULL.
3666 PR tree-optimization/84334
3667 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
3668 also a CONSTANT_CLASS_P, punt.
3670 2018-02-14 Jim Wilson <jimw@sifive.com>
3672 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
3673 first SMALL_OPERAND check. New local min_second_step. Move assert
3674 to where locals are set. Add TARGET_RVC support.
3675 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
3677 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
3679 * doc/invoke.texi: Correct -Wformat-overflow code sample.
3681 2018-02-14 Martin Sebor <msebor@redhat.com>
3683 PR tree-optimization/83698
3684 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
3685 arrays constrain the offset range to their bounds.
3686 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
3687 (builtin_access::overlap): Avoid setting the size of overlap if it's
3689 (maybe_diag_overlap): Also consider arrays when deciding what values
3690 of offsets to include in diagnostics.
3692 2018-02-14 Martin Sebor <msebor@redhat.com>
3695 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
3696 that correspond to the kind of a declaration.
3698 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
3701 * config/pa/pa.md: Load address of PIC label using the linkage table
3702 if the label is nonlocal.
3704 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
3706 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
3707 warning message if user requests -maltivec=be.
3708 * doc/invoke.texi: Document deprecation of -maltivec=be.
3710 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
3713 * config/rs6000/rs6000-c.c: Update definitions for
3714 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
3715 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
3717 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3720 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
3721 add _get_ssp intrinsics. Remove argument from
3722 __builtin_ia32_rdssp[d|q].
3723 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
3724 * config/i386/i386-builtin.def: Remove argument from
3725 __builtin_ia32_rdssp[d|q].
3726 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
3727 ix86_expand_special_args_builtin for _rdssp[d|q].
3728 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
3729 Clear register before usage.
3730 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
3731 Add documentation for new _get_ssp and _inc_ssp intrinsics.
3733 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
3735 PR tree-optimization/84357
3736 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
3737 operand 1 of an ARRAY_REF too.
3739 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
3742 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
3743 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3745 (set_of_reg): New struct.
3746 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
3747 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
3748 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3750 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
3751 Split into bitclr, bitset, bitinvert patterns if appropriate.
3752 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
3753 use rx_fuse_in_memory_bitop.
3754 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
3755 to named insn, correct maximum insn length.
3757 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
3760 * machmode.def: Define a complex mode for PARTIAL_INT.
3761 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
3763 * doc/rtl.texi: Document CSPImode.
3764 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
3766 (msp430_hard_regno_nregs_with_padding): Likewise.
3768 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
3771 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
3773 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
3775 PR rtl-optimization/84169
3776 * combine.c (try_combine): New variable split_i2i3. Set it to true if
3777 we generated a parallel as new i3 and we split that to new i2 and i3
3778 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
3779 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
3780 those to i2, not i1. Partially rewrite this scan code.
3782 2018-02-13 Jakub Jelinek <jakub@redhat.com>
3785 * stor-layout.c (place_field): For variable length fields, adjust
3786 offset_align afterwards not just based on the field's alignment,
3787 but also on the size.
3790 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
3791 of exps and logs in the use_exp2 case.
3793 2018-02-13 Jeff Law <law@redhat.com>
3795 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
3798 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
3801 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
3805 * common.opt (gas-loc-support, gas-locview-support): New.
3806 (ginline-points, ginternal-reset-location-views): New.
3807 * doc/invoke.texi: Document them. Use @itemx where intended.
3808 (gvariable-location-views): Adjust.
3809 * target.def (reset_location_view): New.
3810 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
3811 (TARGET_RESET_LOCATION_VIEW): New.
3812 * doc/tm.texi: Rebuilt.
3813 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
3814 (dwarf2out_default_as_locview_support): New.
3815 (output_asm_line_debug_info): Use option variables.
3816 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
3817 (output_loc_list): Likewise.
3818 (add_high_low_attributes): Check option variables.
3819 Don't output entry view attribute in strict mode.
3820 (gen_inlined_subroutine_die): Check option variables.
3821 (dwarf2out_inline_entry): Likewise.
3822 (init_sections_and_labels): Likewise.
3823 (dwarf2out_early_finish): Likewise.
3824 (maybe_reset_location_view): New, from...
3825 (dwarf2out_var_location): ... here. Call it.
3826 * debug.h (dwarf2out_default_as_loc_support): Declare.
3827 (dwarf2out_default_as_locview_support): Declare.
3828 * hooks.c (hook_int_rtx_insn_0): New.
3829 * hooks.h (hook_int_rtx_insn_0): Declare.
3830 * toplev.c (process_options): Take -gas-loc-support and
3831 -gas-locview-support from dwarf2out. Enable
3832 -gvariable-location-views by default only with locview
3833 assembler support. Enable -ginternal-reset-location-views by
3834 default only if the target defines the corresponding hook.
3835 Enable -ginline-points by default if location views are
3836 enabled; force it disabled if statement frontiers are
3838 * tree-inline.c (expand_call_inline): Check option variables.
3839 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
3841 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
3843 PR tree-optimization/84321
3844 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
3845 handling. Also check whether the anti-range contains any values
3846 that satisfy the mask; switch to a VR_RANGE if not.
3848 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
3851 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
3853 2018-02-13 Martin Jambor <mjambor@suse.cz>
3856 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
3857 of call statements, also set location of a load to a temporary.
3859 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
3861 * config/rl78/rl78.c (add_vector_labels): New function.
3862 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
3863 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
3864 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
3865 which checks that no arguments are passed.
3866 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
3867 * doc/extend.texi: Documentation for the new attribute.
3869 2018-02-13 Andreas Schwab <schwab@suse.de>
3871 * config/riscv/linux.h (CPP_SPEC): Define.
3873 2018-02-13 Jakub Jelinek <jakub@redhat.com>
3876 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
3877 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
3878 OPTION_MASK_ISA_AES as first argument to def_builtin_const
3879 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
3880 instead of OPTION_MASK_ISA_PCLMUL as first argument to
3881 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
3882 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
3883 temporarily for AES and PCLMUL builtins.
3885 PR tree-optimization/84339
3886 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
3887 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
3891 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
3892 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
3893 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
3895 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
3897 * omp-simd-clone.h: New file.
3898 * omp-simd-clone.c: Include omp-simd-clone.h.
3899 (expand_simd_clones): No longer static.
3900 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
3901 cgraph.h and omp-simd-clone.h.
3902 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
3903 (vect_recog_widen_shift_pattern): Formatting fix.
3904 (vect_pattern_recog_1): Don't check optab for calls.
3907 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
3908 operands[2] into a REG before using gen_lowpart on it.
3910 2018-02-12 Jeff Law <law@redhat.com>
3913 * config/sh/sh.c (find_barrier): Consider a sibling call
3916 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
3917 successfully back substituting a reg.
3919 2018-02-12 Richard Biener <rguenther@suse.de>
3921 PR tree-optimization/84037
3922 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
3923 parameter, move visited init to caller.
3924 (vect_slp_analyze_operations): Separate cost from validity
3925 check, initialize visited once for all instances.
3926 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
3928 * tree-vect-stmts.c (vect_model_simple_cost): Make early
3930 (vect_model_promotion_demotion_cost): Likewise.
3931 (vectorizable_bswap): Guard cost modeling with !slp_node
3932 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
3934 (vectorizable_call): Likewise.
3935 (vectorizable_conversion): Likewise.
3936 (vectorizable_assignment): Likewise.
3937 (vectorizable_shift): Likewise.
3938 (vectorizable_operation): Likewise.
3939 (vectorizable_store): Likewise.
3940 (vectorizable_load): Likewise.
3941 (vectorizable_condition): Likewise.
3942 (vectorizable_comparison): Likewise.
3944 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
3947 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
3948 (ASAN_MARK): Fix fnspec to account for return value, change pointer
3949 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
3951 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3954 * params.def (inline-min-speedup): Increase from 8 to 15.
3955 (max-inline-insns-auto): Decrease from 40 to 30.
3956 * ipa-split.c (consider_split): Add some buffer for function to
3957 be considered inlining candidate.
3958 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
3961 2018-02-12 Richard Biener <rguenther@suse.de>
3963 PR tree-optimization/84037
3964 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
3965 matched stmts if we cannot swap the non-matched ones.
3967 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3969 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
3970 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
3971 _mm_maskz_scalef_round_ss): New intrinsics.
3972 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
3973 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
3974 __builtin_ia32_scalefss_round): Remove.
3975 (__builtin_ia32_scalefsd_mask_round,
3976 __builtin_ia32_scalefss_mask_round): New intrinsics.
3977 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
3978 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
3979 ((match_operand:VF_128 2 "<round_nimm_predicate>"
3980 "<round_constraint>")): Changed to ...
3981 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
3982 "<round_scalar_constraint>")): ... this.
3983 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
3984 %0, %1, %2<round_op3>}"): Changed to ...
3985 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
3986 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
3987 %2<round_scalar_mask_op3>}"): ... this.
3988 * config/i386/subst.md (round_scalar_nimm_predicate): New.
3990 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3992 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
3993 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
3994 (_mm_maskz_sqrt_round_ss): New intrinsics.
3995 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
3996 (__builtin_ia32_sqrtsd_mask_round)
3997 (__builtin_ia32_sqrtss_mask_round): New builtins.
3998 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
3999 (__builtin_ia32_sqrtss_round): Remove.
4000 (__builtin_ia32_sqrtsd_mask_round)
4001 (__builtin_ia32_sqrtss_mask_round): New builtins.
4002 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
4003 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
4004 ((match_operand:VF_128 1 "vector_operand"
4005 "xBm,<round_constraint>")): Changed to ...
4006 ((match_operand:VF_128 1 "vector_operand"
4007 "xBm,<round_scalar_constraint>")): ... this.
4008 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
4009 %0, %2, %<iptr>1<round_op3>}): Changed to ...
4010 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
4011 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
4012 %<iptr>1<round_scalar_mask_op3>}): ... this.
4013 ((set_attr "prefix" "<round_prefix>")): Changed to ...
4014 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
4016 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
4019 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
4020 Cast vec_cmpeq result to correct type.
4021 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
4022 Cast vec_cmpgt result to correct type.
4024 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
4026 * final.c (final_scan_insn_1): Renamed from...
4027 (final_scan_insn): ... this. New wrapper, to recover
4028 seen from the outermost call in recursive ones.
4029 * config/sparc/sparc.c (output_return): Drop seen from call.
4030 (output_sibcall): Likewise.
4031 * config/visium/visium.c (output_branch): Likewise.
4033 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
4035 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
4038 2018-02-10 Alan Modra <amodra@gmail.com>
4041 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
4042 Specify LR as an input.
4044 2018-02-10 Jakub Jelinek <jakub@redhat.com>
4047 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
4048 remove_member_access_dummy_vars): New functions.
4049 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
4050 lower_omp_1, execute_lower_omp): Use them.
4052 PR rtl-optimization/84308
4053 * shrink-wrap.c (spread_components): Release todo vector.
4055 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
4057 PR rtl-optimization/57193
4058 * ira-color.c (struct allocno_color_data): Add member
4059 conflict_allocno_hard_prefs.
4060 (update_conflict_allocno_hard_prefs): New.
4061 (bucket_allocno_compare_func): Add a preference based on
4062 conflict_allocno_hard_prefs.
4063 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
4064 (color_allocnos): Remove a dead code. Initiate
4065 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
4067 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4070 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
4071 constraint from =wa to wa. Avoid a subreg on the output operand,
4072 instead use a pseudo and subreg it in a move.
4073 (p9_xxbrd_<mode>): Changed to ...
4074 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
4075 (p9_xxbrd_v2df): New expander.
4076 (p9_xxbrw_<mode>): Changed to ...
4077 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
4078 (p9_xxbrw_v4sf): New expander.
4080 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4082 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
4084 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
4087 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
4088 multiply in 32-bit mode.
4089 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
4090 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
4093 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4095 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
4096 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
4097 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
4098 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
4100 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
4103 * dwarf2out.c (is_trivial_indirect_ref): New function.
4104 (dwarf2out_late_global_decl): Do not generate a location
4105 attribute for variables that have a non-trivial DECL_VALUE_EXPR
4106 and that are not defined in the current unit.
4108 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4110 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
4111 instead of a libcall for UNORDERED.
4113 2018-02-09 Tamar Christina <tamar.christina@arm.com>
4116 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
4117 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
4119 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4122 * config/s390/s390.c (s390_set_current_function): Invoke
4123 s390_indirect_branch_settings also if fndecl didn't change.
4125 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4127 * config/rs6000/rs6000.md (blockage): Set length to zero.
4129 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4131 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
4133 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4136 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
4137 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
4141 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
4142 PARALLEL incoming that failed vt_get_decl_and_offset check.
4145 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
4146 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
4148 (get_variable_section): For decls in named .bss* sections pass true as
4149 second argument to bss_initializer_p.
4151 2018-02-09 Marek Polacek <polacek@redhat.com>
4152 Jakub Jelinek <jakub@redhat.com>
4155 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
4156 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
4157 Sync some changes from cxx_fold_indirect_ref.
4159 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4161 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
4163 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
4164 (BLOCK_INLINE_ENTRY_LABEL): New.
4165 (dwarf2out_var_location): Disregard inline entry markers.
4166 (inline_entry_data): New struct.
4167 (inline_entry_data_hasher): New hashtable type.
4168 (inline_entry_data_hasher::hash): New.
4169 (inline_entry_data_hasher::equal): New.
4170 (inline_entry_data_table): New variable.
4171 (add_high_low_attributes): Add DW_AT_entry_pc and
4172 DW_AT_GNU_entry_view attributes if a pending entry is found
4173 in inline_entry_data_table. Add old entry_pc attribute only
4174 if debug nonbinding markers are disabled.
4175 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
4176 markers are enabled.
4177 (block_within_block_p, dwarf2out_inline_entry): New.
4178 (dwarf2out_finish): Check that no entries remained in
4179 inline_entry_data_table.
4180 * final.c (reemit_insn_block_notes): Handle inline entry notes.
4181 (final_scan_insn, notice_source_line): Likewise.
4182 (rest_of_clean_state): Skip inline entry markers.
4183 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
4185 * gimple.c (gimple_build_debug_inline_entry): New.
4186 * gimple.h (enum gimple_debug_subcode): Add
4187 GIMPLE_DEBUG_INLINE_ENTRY.
4188 (gimple_build_debug_inline_entry): Declare.
4189 (gimple_debug_inline_entry_p): New.
4190 (gimple_debug_nonbind_marker_p): Adjust.
4191 * insn-notes.def (INLINE_ENTRY): New.
4192 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
4193 inline entry marker notes.
4194 (print_insn): Likewise.
4195 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
4196 (INSN_DEBUG_MARKER_KIND): Likewise.
4197 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
4198 * tree-inline.c (expand_call_inline): Build and insert
4199 debug_inline_entry stmt.
4200 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
4201 inline entry blocks early, if nonbind markers are enabled.
4202 (dump_scope_block): Dump fragment info.
4203 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
4204 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
4205 (gimple_build_debug_inline_entry): New.
4206 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
4207 Enable/disable inline entry points too.
4208 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
4209 (DEBUG_INSN): Describe inline entry markers.
4211 * common.opt (gvariable-location-views): New.
4212 (gvariable-location-views=incompat5): New.
4213 * config.in: Rebuilt.
4214 * configure: Rebuilt.
4215 * configure.ac: Test assembler for view support.
4216 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
4217 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
4218 * dwarf2out.c (var_loc_view): New typedef.
4219 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
4220 (dwarf2out_locviews_in_attribute): New.
4221 (dwarf2out_locviews_in_loclist): New.
4222 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
4223 (enum dw_line_info_opcode): Add LI_adv_address.
4224 (struct dw_line_info_table): Add view.
4225 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
4226 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
4227 (zero_view_p): New variable.
4228 (ZERO_VIEW_P): New macro.
4229 (output_asm_line_debug_info): New.
4230 (struct var_loc_node): Add view.
4231 (add_AT_view_list, AT_loc_list): New.
4232 (add_var_loc_to_decl): Add view param. Test it against last.
4233 (new_loc_list): Add view params. Record them.
4234 (AT_loc_list_ptr): Handle loc and view lists.
4235 (view_list_to_loc_list_val_node): New.
4236 (print_dw_val): Handle dw_val_class_view_list.
4237 (size_of_die): Likewise.
4238 (value_format): Likewise.
4239 (loc_list_has_views): New.
4240 (gen_llsym): Set vl_symbol too.
4241 (maybe_gen_llsym, skip_loc_list_entry): New.
4242 (dwarf2out_maybe_output_loclist_view_pair): New.
4243 (output_loc_list): Output view list or entries too.
4244 (output_view_list_offset): New.
4245 (output_die): Handle dw_val_class_view_list.
4246 (output_dwarf_version): New.
4247 (output_compilation_unit_header): Use it.
4248 (output_skeleton_debug_sections): Likewise.
4249 (output_rnglists, output_line_info): Likewise.
4250 (output_pubnames, output_aranges): Update version comments.
4251 (output_one_line_info_table): Output view numbers in asm comments.
4252 (dw_loc_list): Determine current endview, pass it to new_loc_list.
4253 Call maybe_gen_llsym.
4254 (loc_list_from_tree_1): Adjust.
4255 (add_AT_location_description): Create view list attribute if
4256 needed, check it's absent otherwise.
4257 (convert_cfa_to_fb_loc_list): Adjust.
4258 (maybe_emit_file): Call output_asm_line_debug_info for test.
4259 (dwarf2out_var_location): Reset views as needed. Precompute
4260 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
4261 attribute. Set view.
4262 (new_line_info_table): Reset next view.
4263 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
4264 (dwarf2out_source_line): Likewise. Output view resets and labels to
4265 the assembler, or select appropriate line info opcodes.
4266 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
4267 (optimize_string_length): Catch it. Adjust.
4268 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
4269 dw_val_class_view_list, and remove it if no longer needed.
4270 (hash_loc_list): Hash view numbers.
4271 (loc_list_hasher::equal): Compare them.
4272 (optimize_location_lists): Check whether a view list symbol is
4273 needed, and whether the locview attribute is present, and
4274 whether they match. Remove the locview attribute if no longer
4276 (index_location_lists): Call skip_loc_list_entry for test.
4277 (dwarf2out_finish): Call output_asm_line_debug_info for test.
4278 Use output_dwarf_version.
4279 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
4280 (struct dw_val_node): Add val_view_list.
4281 * final.c (SEEN_NEXT_VIEW): New.
4282 (set_next_view_needed): New.
4283 (clear_next_view_needed): New.
4284 (maybe_output_next_view): New.
4285 (final_start_function): Rename to...
4286 (final_start_function_1): ... this. Take pointer to FIRST,
4287 add SEEN parameter. Emit param bindings in the initial view.
4288 (final_start_function): Reintroduce SEEN-less interface.
4289 (final): Rename to...
4290 (final_1): ... this. Take SEEN parameter. Output final pending
4291 next view at the end.
4292 (final): Reintroduce seen-less interface.
4293 (final_scan_insn): Output pending next view before switching
4294 sections or ending a block. Mark the next view as needed when
4295 outputting variable locations. Notify debug backend of section
4296 changes, and of location view changes.
4297 (rest_of_handle_final): Adjust.
4298 * toplev.c (process_options): Autodetect value for debug variable
4299 location views option. Warn on incompat5 without -gdwarf-5.
4300 * doc/invoke.texi (gvariable-location-views): New.
4301 (gvariable-location-views=incompat5): New.
4302 (gno-variable-location-views): New.
4304 2018-02-08 David Malcolm <dmalcolm@redhat.com>
4306 PR tree-optimization/84136
4307 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
4308 that the result of find_edge is non-NULL.
4310 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4313 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
4314 storing integer register in SImode. Fix cost of 256 and 512
4315 byte aligned SSE register store.
4317 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4319 * config/i386/i386.c (ix86_multiplication_cost): Fix
4320 multiplication cost for TARGET_AVX512DQ.
4322 2018-02-08 Marek Polacek <polacek@redhat.com>
4324 PR tree-optimization/84238
4325 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
4328 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4330 PR tree-optimization/84265
4331 * tree-vect-stmts.c (vectorizable_store): Don't treat
4332 VMAT_CONTIGUOUS accesses as grouped.
4333 (vectorizable_load): Likewise.
4335 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4337 PR tree-optimization/81635
4338 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
4339 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
4340 (test_round_for_mask): New functions.
4341 (wide_int_cc_tests): Call test_round_for_mask.
4342 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
4343 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
4344 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
4345 range returned by get_range_info.
4347 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4350 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
4351 * symtab.c: Include builtins.h
4352 (symtab_node::output_to_lto_symbol_table_p): Move here
4353 from lto-streamer-out.c:output_symbol_p.
4354 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
4355 (output_symbol_p): Move all logic to symtab.c
4356 (produce_symtab): Update.
4358 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4360 * config/s390/s390-opts.h (enum indirect_branch): Define.
4361 * config/s390/s390-protos.h (s390_return_addr_from_memory)
4362 (s390_indirect_branch_via_thunk)
4363 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
4364 (enum s390_indirect_branch_type): Define.
4365 * config/s390/s390.c (struct s390_frame_layout, struct
4366 machine_function): Remove.
4367 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
4368 (indirect_branch_table_label_no, indirect_branch_table_name):
4370 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
4371 (enum s390_indirect_branch_option): Define.
4372 (s390_return_addr_from_memory): New function.
4373 (s390_handle_string_attribute): New function.
4374 (s390_attribute_table): Add new attribute handler.
4375 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
4376 (s390_indirect_branch_via_thunk): New function.
4377 (s390_indirect_branch_via_inline_thunk): New function.
4378 (s390_function_ok_for_sibcall): When jumping via thunk disallow
4379 sibling call optimization for non z10 compiles.
4380 (s390_emit_call): Force indirect branch target to be a single
4381 register. Add r1 clobber for non-z10 compiles.
4382 (s390_emit_epilogue): Emit return jump via return_use expander.
4383 (s390_reorg): Handle JUMP_INSNs as execute targets.
4384 (s390_option_override_internal): Perform validity checks for the
4385 new command line options.
4386 (s390_indirect_branch_attrvalue): New function.
4387 (s390_indirect_branch_settings): New function.
4388 (s390_set_current_function): Invoke s390_indirect_branch_settings.
4389 (s390_output_indirect_thunk_function): New function.
4390 (s390_code_end): Implement target hook.
4391 (s390_case_values_threshold): Implement target hook.
4392 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
4394 * config/s390/s390.h (struct s390_frame_layout)
4395 (struct machine_function): Move here from s390.c.
4396 (TARGET_INDIRECT_BRANCH_NOBP_RET)
4397 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
4398 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
4399 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
4400 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
4401 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
4402 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
4403 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
4404 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
4405 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
4406 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
4407 (mnemonic attribute): Add values which aren't recognized
4409 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
4410 pattern for branch conversion. Fix mnemonic attribute.
4411 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
4412 indirect branch via thunk if requested.
4413 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
4414 ("*indirect_jump"): Disable for branch conversion using out of
4416 ("indirect_jump_via_thunk<mode>_z10")
4417 ("indirect_jump_via_thunk<mode>")
4418 ("indirect_jump_via_inlinethunk<mode>_z10")
4419 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
4420 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
4421 ("casesi_jump_via_inlinethunk<mode>_z10")
4422 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
4423 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
4424 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
4425 ("*indirect2_jump"): Disable for branch conversion.
4426 ("casesi_jump"): Turn into expander and expand patterns for branch
4428 ("return_use"): New expander.
4429 ("*return"): Emit return via thunk and rename it to ...
4430 ("*return<mode>"): ... this one.
4431 * config/s390/s390.opt: Add new options and and enum for the
4434 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4436 * lra-constraints.c (match_reload): Unconditionally use
4437 gen_lowpart_SUBREG, rather than selecting between that
4438 and equivalent gen_rtx_SUBREG code.
4440 2018-02-08 Richard Biener <rguenther@suse.de>
4442 PR tree-optimization/84233
4443 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
4444 changed flag instead of boguously re-using phi_inserted.
4446 2018-02-08 Martin Jambor <mjambor@suse.cz>
4448 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
4449 static local variables.
4451 2018-02-08 Richard Biener <rguenther@suse.de>
4453 PR tree-optimization/84278
4454 * tree-vect-stmts.c (vectorizable_store): When looking for
4455 smaller vector types to perform grouped strided loads/stores
4456 make sure the mode is supported by the target.
4457 (vectorizable_load): Likewise.
4459 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4461 * config/aarch64/aarch64.c (aarch64_components_for_bb):
4462 Increase LDP/STP opportunities by adding adjacent callee-saves.
4464 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4466 PR rtl-optimization/84068
4467 PR rtl-optimization/83459
4468 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
4470 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
4472 PR tree-optimization/84224
4473 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
4474 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
4477 2018-02-07 Iain Sandoe <iain@codesourcery.com>
4480 * config/rs6000/altivec.md (*restore_world): Remove LR use.
4481 * config/rs6000/predicates.md (restore_world_operation): Adjust op
4482 count, remove one USE.
4484 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
4486 * doc/install.texi (Configuration): Document the
4487 --with-long-double-format={ibm,ieee} PowerPC configuration
4491 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
4492 Convert from define_expand to be define_insn_and_split. Rework
4493 float/double/_Float128 conversions to QI/HI/SImode to work with
4494 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
4495 conversions to QI/HImode types did a store and then a load to
4496 truncate the value. For conversions to VSX registers, don't split
4497 the insn, instead emit the code directly. Use the code iterator
4498 any_fix to combine signed and unsigned conversions.
4499 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
4500 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
4501 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4502 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
4503 (fix_<mode>di2_hw): Likewise.
4504 (fixuns_<mode>di2_hw): Likewise.
4505 (fix_<mode>si2_hw): Likewise.
4506 (fixuns_<mode>si2_hw): Likewise.
4507 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
4508 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4509 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
4510 fix<uns>_trunc<SFDF:mode>si2_p8.
4511 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
4513 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
4514 (fix<uns>_<mode>_mem): Likewise.
4515 (fctiw<u>z_<mode>_mem): Likewise.
4516 (fix<uns>_<mode>_mem): Likewise.
4517 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
4518 the register allocator from doing a direct move to the GPRs to do
4519 a store, and instead use the ISA 3.0 store byte/half-word from
4520 vector register instruction. For IEEE 128-bit floating point,
4521 also optimize stores of 32-bit ints.
4522 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
4524 2018-02-07 Alan Hayward <alan.hayward@arm.com>
4526 * genextract.c (push_pathstr_operand): New function to support
4528 (walk_rtx): Call push_pathstr_operand.
4529 (print_path): Support [a-zA-Z].
4531 2018-02-07 Richard Biener <rguenther@suse.de>
4533 PR tree-optimization/84037
4534 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
4535 (cse_and_gimplify_to_preheader): Declare.
4536 (vect_get_place_in_interleaving_chain): Likewise.
4537 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4539 (_loop_vec_info::~_loop_vec_info): Delete it.
4540 (cse_and_gimplify_to_preheader): New function.
4541 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
4542 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
4543 (vectorizable_load): Likewise. For grouped stores always base
4544 the IV on the first element.
4545 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
4546 condition before gimplifying.
4548 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4550 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
4551 *DIV_EXPR and *MOD_EXPR.
4553 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
4556 * config/i386/i386.c (ix86_option_override_internal): Mask out
4557 the CF_SET bit when checking -fcf-protection.
4559 2018-02-07 Tom de Vries <tom@codesourcery.com>
4562 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
4565 2018-02-07 Richard Biener <rguenther@suse.de>
4567 PR tree-optimization/84204
4568 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
4571 PR tree-optimization/84205
4572 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
4573 special-case isl_ast_op_zdiv_r.
4575 PR tree-optimization/84223
4576 * graphite-scop-detection.c (gather_bbs::before_dom_children):
4577 Only add conditions from within the region.
4578 (gather_bbs::after_dom_children): Adjust.
4580 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
4583 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
4584 * config/avr/avr.md: Only post-reload split REG-REG moves if
4585 either register is GENERAL_REG_P.
4587 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4589 PR tree-optimization/84235
4590 * tree-ssa-scopedtables.c
4591 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
4592 if the subtraction is performed in floating point type where NaNs are
4593 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
4594 build 1. Formatting fix.
4596 2018-02-06 Jakub Jelinek <jakub@redhat.com>
4599 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
4600 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
4601 and skip it regardless of bb boundaries. Use CALL_P macro,
4602 don't test INSN_P (insn) together with CALL_P or JUMP_P check
4603 unnecessarily, formatting fix.
4605 2018-02-06 Michael Collison <michael.collison@arm.com>
4607 * config/arm/thumb2.md:
4608 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
4609 (*thumb_mov_notscc): Ditto.
4611 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
4614 * config/rs6000/rs6000.md (su code attribute): Use "u" for
4615 unsigned_fix, not "s".
4617 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4619 * configure.ac (gcc_fn_eh_frame_ro): New function.
4620 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
4621 correct .eh_frame permissions.
4622 * configure: Regenerate.
4624 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
4626 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
4629 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4631 * config/rs6000/rs6000.c (rs6000_option_override_internal):
4632 Display warning message for -mno-speculate-indirect-jumps.
4634 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
4636 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
4638 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
4640 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
4642 PR tree-optimization/84225
4643 * tree-eh.c (find_trapping_overflow): Only call
4644 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
4646 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4649 * config/i386/i386.c: Reimplement the check of possible options
4650 -mibt/-mshstk conbination. Change error messages.
4651 * doc/invoke.texi: Fix a typo: remove extra '='.
4653 2018-02-06 Marek Polacek <polacek@redhat.com>
4655 PR tree-optimization/84228
4656 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
4658 2018-02-06 Tamar Christina <tamar.christina@arm.com>
4661 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
4662 emitted arch directives.
4663 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
4664 __ARM_FEATURE_COPROC before changing architectures.
4666 2018-02-06 Richard Biener <rguenther@suse.de>
4668 * config/i386/i386.c (print_reg): Fix typo.
4669 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
4671 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
4673 * configure: Regenerate.
4675 2018-02-05 Martin Sebor <msebor@redhat.com>
4677 PR tree-optimization/83369
4678 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
4681 2018-02-05 Martin Liska <mliska@suse.cz>
4683 * doc/invoke.texi: Cherry-pick upstream r323995.
4685 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
4687 * ira.c (ira_init_register_move_cost): Adjust comment.
4689 2018-02-05 Martin Liska <mliska@suse.cz>
4691 PR gcov-profile/84137
4692 * doc/gcov.texi: Fix typo in documentation.
4694 2018-02-05 Martin Liska <mliska@suse.cz>
4696 PR gcov-profile/83879
4697 * doc/gcov.texi: Document necessity of --dynamic-list-data when
4698 using dlopen functionality.
4700 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
4702 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
4703 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
4704 _mm_maskz_range_ss, _mm_mask_range_round_ss,
4705 _mm_maskz_range_round_ss): New intrinsics.
4706 (__builtin_ia32_rangesd128_round)
4707 (__builtin_ia32_rangess128_round): Remove.
4708 (__builtin_ia32_rangesd128_mask_round,
4709 __builtin_ia32_rangess128_mask_round): New builtins.
4710 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
4711 __builtin_ia32_rangess128_round): Remove.
4712 (__builtin_ia32_rangesd128_mask_round,
4713 __builtin_ia32_rangess128_mask_round): New builtins.
4714 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
4715 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
4716 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
4717 "<round_saeonly_constraint>")): Changed to ...
4718 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
4719 "<round_saeonly_scalar_constraint>")): ... this.
4720 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
4721 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
4722 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
4723 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
4724 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
4726 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
4728 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
4730 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
4731 Remove all values except native, 8540 and 8548.
4733 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
4735 * config/i386/i386.c (ix86_output_function_return): Pass
4736 INVALID_REGNUM, instead of -1, as invalid register number to
4737 indirect_thunk_name and output_indirect_thunk.
4739 2018-02-02 Julia Koval <julia.koval@intel.com>
4741 * config.gcc: Add -march=icelake.
4742 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
4743 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
4744 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
4745 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
4746 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
4747 (processor_target_table): Add icelake.
4748 (ix86_option_override_internal): Handle new PTAs.
4749 (get_builtin_code_for_version): Handle icelake.
4750 (M_INTEL_COREI7_ICELAKE): New.
4751 (fold_builtin_cpu): Handle icelake.
4752 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
4753 * doc/invoke.texi: Add -march=icelake.
4755 2018-02-02 Julia Koval <julia.koval@intel.com>
4757 * config/i386/i386.c (ix86_option_override_internal): Change flags type
4758 to wide_int_bitmask.
4759 * wide-int-bitmask.h: New.
4761 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4764 * config/i386/i386.md: Replace Pmode with word_mode in
4765 builtin_setjmp_setup and builtin_longjmp to support x32.
4767 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
4771 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
4773 (rs6000_supported_cpu_names): New static variable.
4774 (linux_cpu_translation_table): Likewise.
4775 (elf_platform) <cpu>: Define new static variable and use it.
4776 Translate kernel AT_PLATFORM name to canonical name if needed.
4777 Error if platform name is unknown.
4779 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
4782 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
4784 2018-02-01 Jeff Law <law@redhat.com>
4787 * config/i386/i386.c (release_scratch_register_on_entry): Add new
4788 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
4789 the scratch if RELEASE_VIA_POP is false.
4790 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
4791 If we have to save a temporary register, decrement SIZE appropriately.
4792 Pass new arguments to release_scratch_register_on_entry.
4793 (ix86_adjust_stack_and_probe): Likewise.
4794 (ix86_emit_probe_stack_range): Pass new arguments to
4795 release_scratch_register_on_entry.
4797 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
4799 PR rtl-optimization/84157
4800 * combine.c (change_zero_ext): Use REG_P predicate in
4801 front of HARD_REGISTER_P predicate.
4803 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
4805 * config/avr/avr.c (avr_option_override): Move disabling of
4806 -fdelete-null-pointer-checks to...
4807 * common/config/avr/avr-common.c (avr_option_optimization_table):
4810 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4812 PR tree-optimization/81635
4813 * tree-data-ref.c (split_constant_offset_1): For types that
4814 wrap on overflow, try to use range info to prove that wrapping
4817 2018-02-01 Renlin Li <renlin.li@arm.com>
4820 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
4822 (aarch64_register_move_cost): Likewise.
4823 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
4825 (REG_CLASS_NAMES): Likewise.
4826 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
4827 TAILCALL_ADDR_REGS. Remove IP registers.
4828 * config/aarch64/aarch64.md (Ucs): Update register constraint.
4830 2018-02-01 Richard Biener <rguenther@suse.de>
4832 * domwalk.h (dom_walker::dom_walker): Add additional constructor
4833 for specifying RPO order and allow NULL for that.
4834 * domwalk.c (dom_walker::dom_walker): Likewise.
4835 (dom_walker::walk): Handle NULL RPO order.
4836 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
4838 (rewrite_update_dom_walker): Likewise.
4839 (mark_def_dom_walker): Likewise.
4841 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4843 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
4844 (aarch64_maybe_expand_sve_subreg_move): Declare.
4845 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
4846 * config/aarch64/predicates.md (aarch64_any_register_operand): New
4848 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
4849 that are semantically a reverse operation.
4850 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
4851 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
4852 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
4854 (aarch64_can_change_mode_class): For big-endian, forbid changes
4855 between two SVE modes if they have different element sizes.
4857 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4859 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
4860 the TImode handling for big-endian targets.
4862 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4864 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
4865 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
4867 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
4868 Remove BSWAP handing for big-endian targets and use the form of
4869 LD1RQ appropariate for the mode.
4871 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4873 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
4874 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
4877 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4880 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
4881 check for operands that need to go through aarch64_sve_reload_be.
4883 2018-02-01 Jakub Jelinek <jakub@redhat.com>
4885 PR tree-optimization/81661
4886 PR tree-optimization/84117
4887 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
4888 * tree-eh.c: Include gimplify.h.
4889 (find_trapping_overflow, replace_trapping_overflow,
4890 rewrite_to_non_trapping_overflow): New functions.
4891 * tree-vect-loop.c: Include tree-eh.h.
4892 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
4893 * tree-data-ref.c: Include tree-eh.h.
4894 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
4896 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
4898 PR rtl-optimization/84123
4899 * combine.c (change_zero_ext): Check if hard register satisfies
4900 can_change_dest_mode before calling gen_lowpart_SUBREG.
4902 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
4905 * ira.c (ira_init_register_move_cost): Remove assert.
4907 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4909 PR rtl-optimization/84071
4910 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
4911 * doc/tm.texi: Regenerate.
4913 2018-01-31 Richard Biener <rguenther@suse.de>
4915 PR tree-optimization/84132
4916 * tree-data-ref.c (analyze_miv_subscript): Properly
4917 check whether evolution_function_is_affine_multivariate_p
4918 before calling gcd_of_steps_may_divide_p.
4920 2018-01-31 Julia Koval <julia.koval@intel.com>
4923 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
4924 * config/i386/i386.md (rdpid_rex64) New.
4925 (rdpid): Make 32bit only.
4927 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
4930 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
4931 an IDENTIFIER_NODE for FUNCTION_TYPE's.
4933 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4936 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4938 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
4940 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4942 PR rtl-optimization/84071
4943 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
4944 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
4946 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4948 * config/arc/arc.c (arc_handle_aux_attribute): New function.
4949 (arc_attribute_table): Add 'aux' attribute.
4950 (arc_in_small_data_p): Consider aux like variables.
4951 (arc_is_aux_reg_p): New function.
4952 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
4953 (arc_get_aux_arg): New function.
4954 (prepare_move_operands): Handle aux-register access.
4955 (arc_handle_aux_attribute): New function.
4956 * doc/extend.texi (ARC Variable attributes): Add subsection.
4958 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4960 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
4961 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
4962 (arc_attribute_table): Add 'uncached' attribute.
4963 (arc_print_operand): Print '.di' flag for uncached memory
4965 (arc_in_small_data_p): Do not consider for small data the uncached
4967 (arc_is_uncached_mem_p): New function.
4968 * config/arc/predicates.md (compact_store_memory_operand): Check
4969 for uncached memory accesses.
4970 (nonvol_nonimm_operand): Likewise.
4971 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
4973 2018-01-31 Jakub Jelinek <jakub@redhat.com>
4976 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
4977 falign-loops=): Add Optimization flag.
4979 2018-01-30 Jeff Law <law@redhat.com>
4982 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
4983 INT_REGISTERS_SAVED. Check it prior to calling
4984 get_scratch_register_on_entry.
4985 (ix86_adjust_stack_and_probe): Similarly.
4986 (ix86_emit_probe_stack_range): Similarly.
4987 (ix86_expand_prologue): Corresponding changes.
4989 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4992 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
4993 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
4995 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
4998 * lra-constraints.c (curr_insn_transform): Process AND in the
5001 2018-01-30 Jakub Jelinek <jakub@redhat.com>
5003 PR rtl-optimization/83986
5004 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
5005 dependence against last_pending_memory_flush in addition to
5008 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
5010 PR tree-optimization/81611
5011 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
5014 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5017 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
5020 2018-01-30 Richard Biener <rguenther@suse.de>
5021 Jakub Jelinek <jakub@redhat.com>
5023 PR tree-optimization/84111
5024 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
5025 inner loops added during recursion, as they don't have up-to-date
5028 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5031 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
5032 (can_inline_edge_by_limits_p): ... here.
5033 (can_early_inline_edge_p, check_callers,
5034 update_caller_keys, update_callee_keys, recursive_inlining,
5035 add_new_edges_to_heap, speculation_useful_p,
5036 inline_small_functions,
5037 inline_small_functions, flatten_function,
5038 inline_to_all_callers_1): Update.
5040 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5042 * profile-count.c (profile_count::combine_with_ipa_count): Handle
5045 2018-01-30 Richard Biener <rguenther@suse.de>
5047 PR tree-optimization/83008
5048 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
5049 invariant and constant vector uses in stmts when they need
5052 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5055 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
5056 * configure: Regenerate.
5058 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5060 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
5062 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
5063 Use gen_rtx_REG rather than gen_lowpart.
5065 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5067 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
5068 rather than 0 when creating partial subregs.
5070 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5072 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
5075 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
5078 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
5079 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
5080 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
5081 flags. This restores the settings used before the 2017-07-24.
5082 Turning off pre increment/decrement/modify allows IVOPTS to
5083 optimize DF/SF loops where the index is an int.
5085 2018-01-29 Richard Biener <rguenther@suse.de>
5086 Kelvin Nilsen <kelvin@gcc.gnu.org>
5089 * tree-vect-stmts.c (vectorizable_call): Don't call
5090 targetm.vectorize_builtin_md_vectorized_function if callee is
5093 2018-01-22 Carl Love <cel@us.ibm.com>
5095 * doc/extend.tex: Fix typo in second arg in
5096 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
5098 2018-01-29 Richard Biener <rguenther@suse.de>
5100 PR tree-optimization/84086
5101 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
5102 (flush_ssaname_freelist): When SSA names were released reset
5103 the SCEV hash table.
5105 2018-01-29 Richard Biener <rguenther@suse.de>
5107 PR tree-optimization/84057
5108 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
5109 removed paths when removing edges.
5111 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
5113 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
5114 -mfunction-return=@var{choice}.
5116 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
5119 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
5121 (layout::print_source_line): Likewise.
5122 (test_get_line_width_without_trailing_whitespace): Add test cases.
5124 2018-01-27 Jakub Jelinek <jakub@redhat.com>
5127 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
5130 2018-01-26 Jim Wilson <jimw@sifive.com>
5132 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
5134 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
5137 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5139 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
5140 and CMP + SUB-immediate -> SUBS.
5142 2018-01-26 Martin Sebor <msebor@redhat.com>
5144 PR tree-optimization/83896
5145 * tree-ssa-strlen.c (get_string_len): Rename...
5146 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
5147 Avoid assuming length is constant.
5148 (handle_char_store): Use HOST_WIDE_INT for string length.
5150 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
5153 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
5154 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
5156 2018-01-26 Richard Biener <rguenther@suse.de>
5158 PR rtl-optimization/84003
5159 * dse.c (record_store): Only record redundant stores when
5160 the earlier store aliases at least all accesses the later one does.
5162 2018-01-26 Jakub Jelinek <jakub@redhat.com>
5164 PR rtl-optimization/83985
5165 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
5166 REG_CFA_RESTORE insns.
5167 (delete_unmarked_insns): Don't ignore separate shrink wrapping
5168 REG_CFA_RESTORE insns here.
5171 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
5172 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
5174 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5176 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
5177 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
5178 (arc_init): Likewise.
5179 (arc_override_options): Likewise.
5180 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
5182 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
5184 * config/arc/arc.h (TARGET_DBNZ): Define.
5185 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
5186 properly set the tune attribute.
5187 (dbnz): Use TARGET_DBNZ guard.
5188 * config/arc/arc.opt (mtune): Add core3 option.
5190 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5192 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
5193 recognize new pic like addresses.
5194 (arc_delegitimize_address): Clean up.
5196 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5198 * config/arc/arc-arches.def: Option mrf16 valid for all
5200 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
5201 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
5202 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
5203 * config/arc/arc-tables.opt: Regenerate.
5204 * config/arc/arc.c (arc_conditional_register_usage): Handle
5205 reduced register file case.
5206 (arc_file_start): Set must have build attributes.
5207 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
5209 * config/arc/arc.opt (mrf16): Add new option.
5210 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
5211 * config/arc/genmultilib.awk: Handle new mrf16 option.
5212 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
5213 * config/arc/t-multilib: Regenerate.
5214 * doc/invoke.texi (ARC Options): Document mrf16 option.
5216 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5218 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
5219 * config/arc/arc.c (arc_handle_secure_attribute): New function.
5220 (arc_attribute_table): Add 'secure_call' attribute.
5221 (arc_print_operand): Print secure call operand.
5222 (arc_function_ok_for_sibcall): Don't optimize tail calls when
5224 (arc_is_secure_call_p): New function. * config/arc/arc.md
5225 (call_i): Add support for sjli instruction.
5226 (call_value_i): Likewise.
5227 * config/arc/constraints.md (Csc): New constraint.
5229 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5230 John Eric Martin <John.Martin@emmicro-us.com>
5232 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
5233 * config/arc/arc.c (_arc_jli_section): New struct.
5234 (arc_jli_section): New type.
5235 (rc_jli_sections): New static variable.
5236 (arc_handle_jli_attribute): New function.
5237 (arc_attribute_table): Add jli_always and jli_fixed attribute.
5238 (arc_file_end): New function.
5239 (TARGET_ASM_FILE_END): Define.
5240 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
5241 (arc_add_jli_section): New function.
5242 (jli_call_scan): Likewise.
5243 (arc_reorg): Call jli_call_scan.
5244 (arc_output_addsi): Remove 'S' from printing asm operand.
5245 (arc_is_jli_call_p): New function.
5246 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
5248 (movhi_insn): Likewise.
5249 (movsi_insn): Likewise.
5250 (movsi_set_cc_insn): Likewise.
5251 (loadqi_update): Likewise.
5252 (load_zeroextendqisi_update): Likewise.
5253 (load_signextendqisi_update): Likewise.
5254 (loadhi_update): Likewise.
5255 (load_zeroextendhisi_update): Likewise.
5256 (load_signextendhisi_update): Likewise.
5257 (loadsi_update): Likewise.
5258 (loadsf_update): Likewise.
5259 (movsicc_insn): Likewise.
5260 (bset_insn): Likewise.
5261 (bxor_insn): Likewise.
5262 (bclr_insn): Likewise.
5263 (bmsk_insn): Likewise.
5264 (bicsi3_insn): Likewise.
5265 (cmpsi_cc_c_insn): Likewise.
5266 (movsi_ne): Likewise.
5267 (movsi_cond_exec): Likewise.
5268 (clrsbsi2): Likewise.
5279 (call_i): Remove 'S' asm letter, add jli instruction.
5280 (call_value_i): Likewise.
5281 * config/arc/arc.op (mjli-always): New option.
5282 * config/arc/constraints.md (Cji): New constraint.
5283 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
5285 (subsf3_fpx): Likewise.
5286 (mulsf3_fpx): Likewise.
5287 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
5289 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
5291 * doc/invoke.texi (ARC): Document mjli-always option.
5293 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
5295 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
5296 avoid addition with 0 and use incw and decw where possible.
5298 2018-01-26 Richard Biener <rguenther@suse.de>
5300 PR tree-optimization/81082
5301 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
5302 association if it requires casting to unsigned.
5303 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
5304 from fold_plusminus_mult_expr to catch important cases late when
5305 range info is available.
5307 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5309 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
5310 * configure.ac (hidden_linkonce): New test.
5311 * configure: Regenerate.
5312 * config.in: Regenerate.
5314 2018-01-26 Julia Koval <julia.koval@intel.com>
5316 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
5317 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
5318 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
5319 _mm_mask_bitshuffle_epi64_mask): Fix type.
5320 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
5321 USI_FTYPE_V4DI_V4DI_USI): Remove.
5322 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
5323 __builtin_ia32_vpshufbitqmb256_mask,
5324 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
5325 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
5326 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
5328 2018-01-26 Alan Modra <amodra@gmail.com>
5331 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
5332 UNSPEC_VBPERMQ. Sort other unspecs.
5334 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
5336 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
5338 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
5341 * predict.c (drop_profile): Do not push/pop cfun; update also
5343 (handle_missing_profiles): Fix logic looking for zero profiles.
5345 2018-01-25 Jakub Jelinek <jakub@redhat.com>
5348 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
5349 on functions with #pragma omp declare simd or functions with simd
5351 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
5352 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
5353 Remove trailing \n from warning_at calls.
5355 2018-01-25 Tom de Vries <tom@codesourcery.com>
5358 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5359 for neutered workers.
5361 2018-01-24 Joseph Myers <joseph@codesourcery.com>
5364 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
5365 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
5367 2018-01-24 Jeff Law <law@redhat.com>
5370 * i386.c (get_probe_interval): Move to earlier point.
5371 (ix86_compute_frame_layout): If -fstack-clash-protection and
5372 the frame is larger than the probe interval, then use pushes
5373 to save registers rather than reg->mem moves.
5374 (ix86_expand_prologue): Remove conditional for int_registers_saved
5377 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
5380 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
5381 min/max for never referenced object.
5383 2018-01-24 Jakub Jelinek <jakub@redhat.com>
5386 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
5388 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
5389 attributes from DECL_ATTRIBUTES (decl) without affecting
5390 DECL_ATTRIBUTES (current_function_decl).
5391 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
5392 functions with non-NULL DECL_ABSTRACT_ORIGIN.
5394 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
5396 PR tree-optimization/83979
5397 * fold-const.c (fold_comparison): Use constant_boolean_node
5398 instead of boolean_{true,false}_node.
5400 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
5402 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
5405 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5407 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5408 Simplify the clause that sets the length attribute.
5409 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5410 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
5411 clause that sets the length attribute.
5412 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5414 2018-01-24 Tom de Vries <tom@codesourcery.com>
5417 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
5418 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
5419 Add strict parameter.
5420 (prevent_branch_around_nothing): Insert dummy insn between branch to
5421 label and label with no ptx insn inbetween.
5422 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
5424 2018-01-24 Tom de Vries <tom@codesourcery.com>
5427 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5428 for neutered threads in warp.
5429 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
5431 2018-01-24 Richard Biener <rguenther@suse.de>
5433 PR tree-optimization/83176
5434 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
5437 2018-01-24 Richard Biener <rguenther@suse.de>
5439 PR tree-optimization/82819
5440 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
5441 code generating pluses that are no-ops in the target precision.
5443 2018-01-24 Richard Biener <rguenther@suse.de>
5446 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
5448 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5450 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
5451 to merge probabilities.
5452 * predict.c (probably_never_executed): Also mark as cold functions
5453 with global 0 profile and guessed local profile.
5454 * profile-count.c (profile_probability::combine_with_count): New
5456 * profile-count.h (profile_probability::operator*,
5457 profile_probability::operator*=, profile_probability::operator/,
5458 profile_probability::operator/=): Reduce precision to adjusted
5459 and set value to guessed on contradictory divisions.
5460 (profile_probability::combine_with_freq): Remove.
5461 (profile_probability::combine_wiht_count): Declare.
5462 (profile_count::force_nonzero):: Set to adjusted.
5463 (profile_count::probability_in):: Set quality to adjusted.
5464 * tree-ssa-tail-merge.c (replace_block_by): Use
5467 2018-01-23 Andrew Waterman <andrew@sifive.com>
5468 Jim Wilson <jimw@sifive.com>
5470 * config/riscv/riscv.c (riscv_stack_boundary): New.
5471 (riscv_option_override): Set riscv_stack_boundary. Handle
5472 riscv_preferred_stack_boundary_arg.
5473 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
5474 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
5475 (STACK_BOUNDARY): Set to riscv_stack_boundary.
5476 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
5477 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
5478 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
5480 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
5483 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
5484 of struct ix86_frame.
5485 (ix86_expand_epilogue): Likewise. Add a local variable for
5486 the reg_save_offset field in struct ix86_frame.
5488 2018-01-23 Bin Cheng <bin.cheng@arm.com>
5490 PR tree-optimization/82604
5491 * tree-loop-distribution.c (enum partition_kind): New enum item
5492 PKIND_PARTIAL_MEMSET.
5493 (partition_builtin_p): Support above new enum item.
5494 (generate_code_for_partition): Ditto.
5495 (compute_access_range): Differentiate cases that equality can be
5496 proven at all loops, the innermost loops or no loops.
5497 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
5498 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
5499 (finalize_partitions, distribute_loop): Don't fuse partition of
5500 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
5501 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
5504 2018-01-23 Martin Liska <mliska@suse.cz>
5506 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
5507 order to ignore the predictor.
5508 (PRED_POLYMORPHIC_CALL): Likewise.
5509 (PRED_RECURSIVE_CALL): Likewise.
5511 2018-01-23 Martin Liska <mliska@suse.cz>
5513 * tree-profile.c (tree_profiling): Print function header to
5514 aware reader which function we are working on.
5515 * value-prof.c (gimple_find_values_to_profile): Do not print
5516 not interesting value histograms.
5518 2018-01-23 Martin Liska <mliska@suse.cz>
5520 * profile-count.h (enum profile_quality): Add
5521 profile_uninitialized as the first value. Do not number values
5522 as they are zero based.
5523 (profile_count::verify): Update sanity check.
5524 (profile_probability::verify): Likewise.
5526 2018-01-23 Nathan Sidwell <nathan@acm.org>
5528 * doc/invoke.texi (ffor-scope): Deprecate.
5530 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5532 PR tree-optimization/83510
5533 * domwalk.c (set_all_edges_as_executable): New function.
5534 (dom_walker::dom_walker): Convert bool param
5535 "skip_unreachable_blocks" to enum reachability. Move setup of
5536 edge flags to set_all_edges_as_executable and only do it when
5537 reachability is REACHABLE_BLOCKS.
5538 * domwalk.h (enum dom_walker::reachability): New enum.
5539 (dom_walker::dom_walker): Convert bool param
5540 "skip_unreachable_blocks" to enum reachability.
5541 (set_all_edges_as_executable): New decl.
5542 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
5543 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
5545 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
5546 but converting true to REACHABLE_BLOCKS.
5547 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
5549 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
5550 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
5551 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
5553 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
5554 if check_all_array_refs will be called.
5556 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5558 * tree.c (selftest::test_location_wrappers): Add more test
5561 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5563 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
5564 (selftest::test_bit_in_range): Likewise.
5566 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5569 * doc/sourcebuild.texi (vect_float): Say that the selector
5570 only describes the situation when -funsafe-math-optimizations is on.
5571 (vect_float_strict): Document.
5573 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5575 PR tree-optimization/83965
5576 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
5577 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
5578 instead of checking only for a reduction.
5579 (vect_recog_widen_sum_pattern): Likewise.
5581 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5583 * predict.c (probably_never_executed): Only use precise profile info.
5584 (compute_function_frequency): Skip after inlining hack since we now
5585 have quality checking.
5587 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5589 * profile-count.h (profile_probability::very_unlikely,
5590 profile_probability::unlikely, profile_probability::even): Set
5591 precision to guessed.
5593 2018-01-23 Richard Biener <rguenther@suse.de>
5595 PR tree-optimization/83963
5596 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
5597 Properly terminate dominator walk when crossing the exit edge not
5598 when visiting its source block.
5600 2018-01-23 Jakub Jelinek <jakub@redhat.com>
5603 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
5604 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
5606 2018-01-22 Jakub Jelinek <jakub@redhat.com>
5608 PR tree-optimization/83957
5609 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
5610 semicolon after for body surrounded by braces.
5612 PR tree-optimization/83081
5613 * profile-count.h (profile_probability::split): New method.
5614 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
5615 Use profile_probability::split.
5616 (do_compare_rtx_and_jump): Fix adjustment of probabilities
5617 when splitting a single conditional jump into 2.
5619 2018-01-22 David Malcolm <dmalcolm@redhat.com>
5621 PR tree-optimization/69452
5622 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
5625 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5627 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
5628 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
5629 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
5631 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5633 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
5634 * config/rl78/rl78.md (movdi): New define_expand.
5635 * config/rl78/rl78.c (rl78_split_movdi): New function.
5637 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
5640 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
5642 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
5643 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
5644 128-bit to produce an UNSPEC move to get the double word with the
5645 signbit and then a shift directly to do signbit.
5646 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
5647 implementation with a new version that just does either a direct
5648 move or a regular move. Move memory interface to separate insns.
5649 Move insns so they are next to the expander.
5650 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
5651 with signbit move. Split big and little endian case.
5652 (signbit<mode>2_dm_mem_le): Likewise.
5653 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
5654 (signbit<mode>2_dm2): Likewise.
5656 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5658 * config/rl78/rl78.md (anddi3): New define_expand.
5660 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5662 * config/rl78/rl78.md (umindi3): New define_expand.
5664 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5666 * config/rl78/rl78.md (smindi3): New define_expand.
5668 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5670 * config/rl78/rl78.md (smaxdi3): New define_expand.
5672 2018-01-22 Carl Love <cel@us.ibm.com>
5674 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
5675 LVX_V1TI): Add macro expansion.
5676 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
5677 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
5678 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
5679 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
5680 Change check to determine if the instruction is a byte reversing
5681 entry. Fix typo in comment.
5682 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
5683 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
5684 Add def_builtin calls for new builtins.
5685 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
5686 Add define_insn expansion.
5688 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5690 * config/rl78/rl78.md (umaxdi3): New define_expand.
5692 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5694 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
5695 for non-QImode registers.
5697 2018-01-22 Richard Biener <rguenther@suse.de>
5699 PR tree-optimization/83963
5700 * graphite-scop-detection.c (scop_detection::get_sese): Delay
5701 including the loop exit block.
5702 (scop_detection::merge_sese): Likewise.
5703 (scop_detection::add_scop): Do it here instead.
5705 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5707 * doc/sourcebuild.texi (arm_softfloat): Document.
5709 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
5712 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
5713 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
5714 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
5716 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5717 David Edelsohn <dje.gcc@gmail.com>
5720 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5721 Change "crset eq" to "crset 2".
5722 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5723 (*call_indirect_aix<mode>_nospec): Likewise.
5724 (*call_value_indirect_aix<mode>_nospec): Likewise.
5725 (*call_indirect_elfv2<mode>_nospec): Likewise.
5726 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
5727 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
5728 change assembly output from . to $.
5729 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5730 (indirect_jump<mode>_nospec): Change assembly output from . to $.
5731 (*tablejump<mode>_internal1_nospec): Likewise.
5733 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
5736 * config/sh/sh_optimize_sett_clrt.cc:
5737 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
5739 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
5741 PR tree-optimization/83940
5742 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
5743 offset_dt to vect_constant_def rather than vect_unknown_def_type.
5744 (vect_check_load_store_mask): Add a mask_dt_out parameter and
5745 use it to pass back the definition type.
5746 (vect_check_store_rhs): Likewise rhs_dt_out.
5747 (vect_build_gather_load_calls): Add a mask_dt argument and use
5748 it instead of a call to vect_is_simple_use.
5749 (vectorizable_store): Update calls to vect_check_load_store_mask
5750 and vect_check_store_rhs. Use the dt returned by the latter instead
5751 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
5752 instead of calls to vect_is_simple_use. Pass the scalar rather
5753 than the vector operand to vect_is_simple_use when handling
5754 second and subsequent copies of an rhs value.
5755 (vectorizable_load): Update calls to vect_check_load_store_mask
5756 and vect_build_gather_load_calls. Use the cached mask_dt and
5757 gs_info.offset_dt instead of calls to vect_is_simple_use.
5759 2018-01-20 Jakub Jelinek <jakub@redhat.com>
5762 * tree-emutls.c: Include gimplify.h.
5763 (lower_emutls_2): New function.
5764 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
5765 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
5766 it before further processing.
5769 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
5770 UINTVAL (trueop1) instead of INTVAL (op1).
5772 2018-01-19 Jakub Jelinek <jakub@redhat.com>
5776 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
5777 INCOMING_FRAME_SP_OFFSET if not defined.
5778 (scan_trace): Add ENTRY argument. If true and
5779 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
5780 emit a note to adjust the CFA offset.
5781 (create_cfi_notes): Adjust scan_trace callers.
5782 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
5783 INCOMING_FRAME_SP_OFFSET in the CIE.
5784 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
5785 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
5787 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
5788 * doc/tm.texi: Regenerated.
5790 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5792 PR rtl-optimization/83147
5793 * lra-constraints.c (remove_inheritance_pseudos): Use
5794 lra_substitute_pseudo_within_insn.
5796 2018-01-19 Tom de Vries <tom@codesourcery.com>
5797 Cesar Philippidis <cesar@codesourcery.com>
5800 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
5802 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
5805 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
5806 spaces for function labels.
5808 2018-01-19 Martin Liska <mliska@suse.cz>
5810 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
5811 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
5812 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
5813 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
5814 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
5815 (PRED_CONST_RETURN): Change from 69 to 65.
5816 (PRED_NULL_RETURN): Change from 91 to 71.
5817 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
5818 (PRED_LOOP_GUARD): Change from 66 to 73.
5820 2018-01-19 Martin Liska <mliska@suse.cz>
5822 * predict.c (predict_insn_def): Add new assert.
5823 (struct branch_predictor): Change type to signed integer.
5824 (test_prediction_value_range): Amend test to cover
5826 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
5827 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
5828 (PRED_LOOP_ITERATIONS_MAX): Likewise.
5829 (PRED_LOOP_IV_COMPARE): Likewise.
5830 * predict.h (PROB_UNINITIALIZED): Define new constant.
5832 2018-01-19 Martin Liska <mliska@suse.cz>
5834 * predict.c (dump_prediction): Add new format for
5835 analyze_brprob.py script which is enabled with -details
5837 * profile-count.h (precise_p): New function.
5839 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
5841 PR tree-optimization/83922
5842 * tree-vect-loop.c (vect_verify_full_masking): Return false if
5843 there are no statements that need masking.
5844 (vect_active_double_reduction_p): New function.
5845 (vect_analyze_loop_operations): Use it when handling phis that
5846 are not in the loop header.
5848 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
5850 PR tree-optimization/83914
5851 * tree-vect-loop.c (vectorizable_induction): Don't convert
5852 init_expr or apply the peeling adjustment for inductions
5853 that are nested within the vectorized loop.
5855 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5857 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
5860 2018-01-18 Jakub Jelinek <jakub@redhat.com>
5864 * function.h (gimplify_parameters): Add gimple_seq * argument.
5865 * function.c: Include gimple.h and options.h.
5866 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
5867 for the added local temporaries if needed.
5868 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
5869 if there are any parameter cleanups, wrap whole body into a
5870 try/finally with the cleanups.
5872 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
5875 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
5876 Use GET_MODE_CLASS for scalar floating point.
5878 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
5882 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
5883 Fix call of call_cgraph_insertion_hooks.
5885 2018-01-18 Martin Sebor <msebor@redhat.com>
5887 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
5889 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
5892 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
5895 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
5898 * common.opt: (-ffile-prefix-map): New option.
5899 * opts.c (common_handle_option): Defer it.
5900 * opts-global.c (handle_common_deferred_options): Handle it.
5901 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
5902 * file-prefix-map.h: New file.
5903 (remap_debug_filename, add_debug_prefix_map): ...here.
5904 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
5905 * final.c (debug_prefix_map, add_debug_prefix_map
5906 remap_debug_filename): Move to...
5907 * file-prefix-map.c: New file.
5908 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
5909 generalize, get rid of alloca(), use strrchr() instead of strchr().
5910 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
5911 Implement in terms of add_prefix_map().
5912 (remap_macro_filename, remap_debug_filename): Implement in term of
5914 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
5915 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
5916 * dbxout.c: Include file-prefix-map.h.
5917 * varasm.c: Likewise.
5918 * vmsdbgout.c: Likewise.
5919 * xcoffout.c: Likewise.
5920 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
5921 * doc/cppopts.texi (-fmacro-prefix-map): Document.
5922 * doc/invoke.texi (-ffile-prefix-map): Document.
5923 (-fdebug-prefix-map): Update description.
5925 2018-01-18 Martin Liska <mliska@suse.cz>
5927 * config/i386/i386.c (indirect_thunk_name): Document that also
5929 (output_indirect_thunk): Document why both instructions
5930 (pause and lfence) are generated.
5932 2018-01-18 Richard Biener <rguenther@suse.de>
5934 PR tree-optimization/83887
5935 * graphite-scop-detection.c
5936 (scop_detection::get_nearest_dom_with_single_entry): Remove.
5937 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
5938 (scop_detection::merge_sese): Re-implement with a flood-fill
5939 algorithm that properly finds a SESE region if it exists.
5941 2018-01-18 Jakub Jelinek <jakub@redhat.com>
5944 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
5945 pointer_diff optimizations use view_convert instead of convert.
5947 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5949 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5950 Generate different code for -mno-speculate-indirect-jumps.
5951 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5952 (*call_indirect_aix<mode>): Disable for
5953 -mno-speculate-indirect-jumps.
5954 (*call_indirect_aix<mode>_nospec): New define_insn.
5955 (*call_value_indirect_aix<mode>): Disable for
5956 -mno-speculate-indirect-jumps.
5957 (*call_value_indirect_aix<mode>_nospec): New define_insn.
5958 (*sibcall_nonlocal_sysv<mode>): Generate different code for
5959 -mno-speculate-indirect-jumps.
5960 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5962 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
5964 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
5965 long double type, set the flags for noting the default long double
5966 type, even if we don't pass or return a long double type.
5968 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
5971 * ipa-inline.c (flatten_function): Do not overwrite final inlining
5974 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
5976 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
5977 support for merge[hl].
5978 (fold_mergehl_helper): New helper function.
5979 (tree-vector-builder.h): New #include for tree_vector_builder usage.
5980 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
5981 (altivec_vmrglw_direct): Add xxmrglw insn.
5983 2018-01-17 Andrew Waterman <andrew@sifive.com>
5985 * config/riscv/riscv.c (riscv_conditional_register_usage): If
5986 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
5988 2018-01-17 David Malcolm <dmalcolm@redhat.com>
5991 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
5992 call the lto_location_cache before reading the
5993 DECL_SOURCE_LOCATION of the types.
5995 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
5996 Richard Sandiford <richard.sandiford@linaro.org>
5998 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
5999 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
6000 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
6001 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
6002 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
6004 * config/aarch64/constraints.md (aarch64_movti_operand):
6006 * config/aarch64/predicates.md (Uti): Add new constraint.
6008 2018-01-17 Carl Love <cel@us.ibm.com>
6010 * config/rs6000/vsx.md (define_expand xl_len_r,
6011 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
6012 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
6014 (define_expand, define_insn): Move the shift left from the
6015 define_insn to the define_expand for lxvl and stxvl instructions.
6016 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
6017 and XL_LEN_R definitions to PURE.
6019 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
6021 * config/i386/i386.c (indirect_thunk_name): Declare regno
6022 as unsigned int. Compare regno with INVALID_REGNUM.
6023 (output_indirect_thunk): Ditto.
6024 (output_indirect_thunk_function): Ditto.
6025 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
6026 in the call to output_indirect_thunk_function.
6028 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
6031 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
6032 rather than the size of inner_type to determine the stack slot size
6033 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
6035 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
6038 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
6041 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
6043 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
6044 endian Linux systems to optionally enable multilibs for selecting
6045 the long double type if the user configured an explicit type.
6046 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
6047 have no long double multilibs if not defined.
6048 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
6049 warn if the user used -mabi={ieee,ibm}longdouble and we built
6050 multilibs for long double.
6051 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
6052 appropriate multilib option.
6053 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
6055 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
6056 for building long double multilibs.
6057 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
6059 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
6061 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
6064 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
6066 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
6069 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
6072 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
6075 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6077 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
6078 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
6080 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
6082 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
6083 different rtl trees depending on TARGET_64BIT.
6084 (rs6000_gen_lvx): Likewise.
6086 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6088 * config/visium/visium.md (nop): Tweak comment.
6089 (hazard_nop): Likewise.
6091 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6093 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
6094 -mspeculate-indirect-jumps.
6095 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
6096 for -mno-speculate-indirect-jumps.
6097 (*call_indirect_elfv2<mode>_nospec): New define_insn.
6098 (*call_value_indirect_elfv2<mode>): Disable for
6099 -mno-speculate-indirect-jumps.
6100 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
6101 (indirect_jump): Emit different RTL for
6102 -mno-speculate-indirect-jumps.
6103 (*indirect_jump<mode>): Disable for
6104 -mno-speculate-indirect-jumps.
6105 (*indirect_jump<mode>_nospec): New define_insn.
6106 (tablejump): Emit different RTL for
6107 -mno-speculate-indirect-jumps.
6108 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
6109 (tablejumpsi_nospec): New define_expand.
6110 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
6111 (tablejumpdi_nospec): New define_expand.
6112 (*tablejump<mode>_internal1): Disable for
6113 -mno-speculate-indirect-jumps.
6114 (*tablejump<mode>_internal1_nospec): New define_insn.
6115 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
6118 2018-01-16 Artyom Skrobov tyomitch@gmail.com
6120 * caller-save.c (insert_save): Drop unnecessary parameter. All
6123 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6124 Richard Biener <rguenth@suse.de>
6127 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
6128 return early, inline manually is_gimple_sizepos. Make sure if we
6129 call gimplify_expr we don't end up with a gimple constant.
6130 * tree.c (variably_modified_type_p): Don't return true for
6131 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
6132 * gimplify.h (is_gimple_sizepos): Remove.
6134 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6136 PR tree-optimization/83857
6137 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
6138 vectorizable_live_operation for pure SLP statements.
6139 (vectorizable_live_operation): Handle PHIs.
6141 2018-01-16 Richard Biener <rguenther@suse.de>
6143 PR tree-optimization/83867
6144 * tree-vect-stmts.c (vect_transform_stmt): Precompute
6145 nested_in_vect_loop_p since the scalar stmt may get invalidated.
6147 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6150 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
6151 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
6152 If off is not INTEGER_CST, issue a may not be aligned warning
6153 rather than isn't aligned. Use isn%'t rather than isn't.
6154 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
6156 <case MULT_EXPR>: Improve the case when bottom and one of the
6157 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
6158 operand, in that case check if the other operand is multiple of
6159 bottom divided by the INTEGER_CST operand.
6161 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6164 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
6165 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
6166 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
6167 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
6168 * config/pa/pa.c (pa_function_arg_advance): Likewise.
6169 (pa_function_arg, pa_arg_partial_bytes): Likewise.
6170 (pa_function_arg_size): New function.
6172 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6174 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
6175 in a separate statement.
6177 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6179 PR tree-optimization/83847
6180 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
6181 group gathers and scatters.
6183 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6185 PR rtl-optimization/86620
6186 * params.def (max-sched-ready-insns): Bump minimum value to 1.
6188 PR rtl-optimization/83213
6189 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
6190 to last if both are JUMP_INSNs.
6192 PR tree-optimization/83843
6193 * gimple-ssa-store-merging.c
6194 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
6195 store_immediate_info for bswap/nop orig_stores.
6197 2018-01-15 Andrew Waterman <andrew@sifive.com>
6199 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
6201 <UDIV>: Increase cost if !TARGET_DIV.
6203 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
6205 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
6206 (define_attr "cr_logical_3op"): New.
6207 (cceq_ior_compare): Adjust.
6208 (cceq_ior_compare_complement): Adjust.
6209 (*cceq_rev_compare): Adjust.
6210 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
6211 (is_cracked_insn): Adjust.
6212 (insn_must_be_first_in_group): Adjust.
6213 * config/rs6000/40x.md: Adjust.
6214 * config/rs6000/440.md: Adjust.
6215 * config/rs6000/476.md: Adjust.
6216 * config/rs6000/601.md: Adjust.
6217 * config/rs6000/603.md: Adjust.
6218 * config/rs6000/6xx.md: Adjust.
6219 * config/rs6000/7450.md: Adjust.
6220 * config/rs6000/7xx.md: Adjust.
6221 * config/rs6000/8540.md: Adjust.
6222 * config/rs6000/cell.md: Adjust.
6223 * config/rs6000/e300c2c3.md: Adjust.
6224 * config/rs6000/e500mc.md: Adjust.
6225 * config/rs6000/e500mc64.md: Adjust.
6226 * config/rs6000/e5500.md: Adjust.
6227 * config/rs6000/e6500.md: Adjust.
6228 * config/rs6000/mpc.md: Adjust.
6229 * config/rs6000/power4.md: Adjust.
6230 * config/rs6000/power5.md: Adjust.
6231 * config/rs6000/power6.md: Adjust.
6232 * config/rs6000/power7.md: Adjust.
6233 * config/rs6000/power8.md: Adjust.
6234 * config/rs6000/power9.md: Adjust.
6235 * config/rs6000/rs64.md: Adjust.
6236 * config/rs6000/titan.md: Adjust.
6238 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6240 * config/i386/predicates.md (indirect_branch_operand): Rewrite
6241 ix86_indirect_branch_register logic.
6243 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6245 * config/i386/constraints.md (Bs): Update
6246 ix86_indirect_branch_register check. Don't check
6247 ix86_indirect_branch_register with GOT_memory_operand.
6249 * config/i386/predicates.md (GOT_memory_operand): Don't check
6250 ix86_indirect_branch_register here.
6251 (GOT32_symbol_operand): Likewise.
6253 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6255 * config/i386/predicates.md (constant_call_address_operand):
6256 Rewrite ix86_indirect_branch_register logic.
6257 (sibcall_insn_operand): Likewise.
6259 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6261 * config/i386/constraints.md (Bs): Replace
6262 ix86_indirect_branch_thunk_register with
6263 ix86_indirect_branch_register.
6265 * config/i386/i386.md (indirect_jump): Likewise.
6266 (tablejump): Likewise.
6267 (*sibcall_memory): Likewise.
6268 (*sibcall_value_memory): Likewise.
6269 Peepholes of indirect call and jump via memory: Likewise.
6270 * config/i386/i386.opt: Likewise.
6271 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6272 (GOT_memory_operand): Likewise.
6273 (call_insn_operand): Likewise.
6274 (sibcall_insn_operand): Likewise.
6275 (GOT32_symbol_operand): Likewise.
6277 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6280 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
6281 type rather than type addr's type points to.
6282 (expand_omp_atomic_mutex): Likewise.
6283 (expand_omp_atomic): Likewise.
6285 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6288 * config/i386/i386.c (output_indirect_thunk_function): Use
6289 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
6290 for __x86_return_thunk.
6292 2018-01-15 Richard Biener <rguenther@suse.de>
6295 * expmed.c (extract_bit_field_1): Fix typo.
6297 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6300 * config/arm/iterators.md (VF): New mode iterator.
6301 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
6302 Remove integer-related logic from pattern.
6303 (neon_vabd<mode>_3): Likewise.
6305 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6308 * common.opt (fstrict-overflow): No longer an alias.
6309 (fwrapv-pointer): New option.
6310 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
6311 also for pointer types based on flag_wrapv_pointer.
6312 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
6313 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
6314 opts->x_flag_wrapv got set.
6315 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
6316 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
6317 POINTER_TYPE_OVERFLOW_UNDEFINED.
6318 * match.pd: Likewise in address comparison pattern.
6319 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
6321 2018-01-15 Richard Biener <rguenther@suse.de>
6324 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
6325 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
6326 Reset type names to their identifier if their TYPE_DECL doesn't
6327 have linkage (and thus is used for ODR and devirt).
6328 (save_debug_info_for_decl): Remove.
6329 (save_debug_info_for_type): Likewise.
6330 (add_tree_to_fld_list): Adjust.
6331 * tree-pretty-print.c (dump_generic_node): Make dumping of
6332 type names more robust.
6334 2018-01-15 Richard Biener <rguenther@suse.de>
6336 * BASE-VER: Bump to 8.0.1.
6338 2018-01-14 Martin Sebor <msebor@redhat.com>
6341 * builtins.c (check_access): Avoid warning when the no-warning bit
6344 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
6346 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
6347 * ira-color (allocno_hard_regs_compare): Likewise.
6349 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
6352 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
6353 Use .pushsection/.popsection.
6355 2018-01-14 Martin Sebor <msebor@redhat.com>
6358 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
6360 2018-01-14 Jakub Jelinek <jakub@redhat.com>
6362 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
6363 entry from extra_headers.
6364 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
6365 extra_headers, make the list bitwise identical to the i?86-*-* one.
6367 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6369 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
6370 -mcmodel=large with -mindirect-branch=thunk,
6371 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
6372 -mfunction-return=thunk-extern.
6373 * doc/invoke.texi: Document -mcmodel=large is incompatible with
6374 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
6375 -mfunction-return=thunk and -mfunction-return=thunk-extern.
6377 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6379 * config/i386/i386.c (print_reg): Print the name of the full
6380 integer register without '%'.
6381 (ix86_print_operand): Handle 'V'.
6382 * doc/extend.texi: Document 'V' modifier.
6384 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6386 * config/i386/constraints.md (Bs): Disallow memory operand for
6387 -mindirect-branch-register.
6389 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6390 (GOT_memory_operand): Likewise.
6391 (call_insn_operand): Likewise.
6392 (sibcall_insn_operand): Likewise.
6393 (GOT32_symbol_operand): Likewise.
6394 * config/i386/i386.md (indirect_jump): Call convert_memory_address
6395 for -mindirect-branch-register.
6396 (tablejump): Likewise.
6397 (*sibcall_memory): Likewise.
6398 (*sibcall_value_memory): Likewise.
6399 Disallow peepholes of indirect call and jump via memory for
6400 -mindirect-branch-register.
6401 (*call_pop): Replace m with Bw.
6402 (*call_value_pop): Likewise.
6403 (*sibcall_pop_memory): Replace m with Bs.
6404 * config/i386/i386.opt (mindirect-branch-register): New option.
6405 * doc/invoke.texi: Document -mindirect-branch-register option.
6407 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6409 * config/i386/i386-protos.h (ix86_output_function_return): New.
6410 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
6411 set function_return_type.
6412 (indirect_thunk_name): Add ret_p to indicate thunk for function
6414 (output_indirect_thunk_function): Pass false to
6415 indirect_thunk_name.
6416 (ix86_output_indirect_branch_via_reg): Likewise.
6417 (ix86_output_indirect_branch_via_push): Likewise.
6418 (output_indirect_thunk_function): Create alias for function
6419 return thunk if regno < 0.
6420 (ix86_output_function_return): New function.
6421 (ix86_handle_fndecl_attribute): Handle function_return.
6422 (ix86_attribute_table): Add function_return.
6423 * config/i386/i386.h (machine_function): Add
6424 function_return_type.
6425 * config/i386/i386.md (simple_return_internal): Use
6426 ix86_output_function_return.
6427 (simple_return_internal_long): Likewise.
6428 * config/i386/i386.opt (mfunction-return=): New option.
6429 (indirect_branch): Mention -mfunction-return=.
6430 * doc/extend.texi: Document function_return function attribute.
6431 * doc/invoke.texi: Document -mfunction-return= option.
6433 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6435 * config/i386/i386-opts.h (indirect_branch): New.
6436 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
6437 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
6438 with local indirect jump when converting indirect call and jump.
6439 (ix86_set_indirect_branch_type): New.
6440 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
6441 (indirectlabelno): New.
6442 (indirect_thunk_needed): Likewise.
6443 (indirect_thunk_bnd_needed): Likewise.
6444 (indirect_thunks_used): Likewise.
6445 (indirect_thunks_bnd_used): Likewise.
6446 (INDIRECT_LABEL): Likewise.
6447 (indirect_thunk_name): Likewise.
6448 (output_indirect_thunk): Likewise.
6449 (output_indirect_thunk_function): Likewise.
6450 (ix86_output_indirect_branch_via_reg): Likewise.
6451 (ix86_output_indirect_branch_via_push): Likewise.
6452 (ix86_output_indirect_branch): Likewise.
6453 (ix86_output_indirect_jmp): Likewise.
6454 (ix86_code_end): Call output_indirect_thunk_function if needed.
6455 (ix86_output_call_insn): Call ix86_output_indirect_branch if
6457 (ix86_handle_fndecl_attribute): Handle indirect_branch.
6458 (ix86_attribute_table): Add indirect_branch.
6459 * config/i386/i386.h (machine_function): Add indirect_branch_type
6460 and has_local_indirect_jump.
6461 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
6463 (tablejump): Likewise.
6464 (*indirect_jump): Use ix86_output_indirect_jmp.
6465 (*tablejump_1): Likewise.
6466 (simple_return_indirect_internal): Likewise.
6467 * config/i386/i386.opt (mindirect-branch=): New option.
6468 (indirect_branch): New.
6471 (thunk-inline): Likewise.
6472 (thunk-extern): Likewise.
6473 * doc/extend.texi: Document indirect_branch function attribute.
6474 * doc/invoke.texi: Document -mindirect-branch= option.
6476 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
6479 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
6481 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
6483 * ipa-inline.c (want_inline_small_function_p): Return false if
6484 inlining has already failed with CIF_FINAL_ERROR.
6485 (update_caller_keys): Call want_inline_small_function_p before
6487 (update_callee_keys): Likewise.
6489 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6491 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
6493 (rs6000_quadword_masked_address_p): Likewise.
6494 (quad_aligned_load_p): Likewise.
6495 (quad_aligned_store_p): Likewise.
6496 (const_load_sequence_p): Add comment to describe the outer-most loop.
6497 (mimic_memory_attributes_and_flags): New function.
6498 (rs6000_gen_stvx): Likewise.
6499 (replace_swapped_aligned_store): Likewise.
6500 (rs6000_gen_lvx): Likewise.
6501 (replace_swapped_aligned_load): Likewise.
6502 (replace_swapped_load_constant): Capitalize argument name in
6503 comment describing this function.
6504 (rs6000_analyze_swaps): Add a third pass to search for vector loads
6505 and stores that access quad-word aligned addresses and replace
6506 with stvx or lvx instructions when appropriate.
6507 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
6508 New function prototype.
6509 (rs6000_quadword_masked_address_p): Likewise.
6510 (rs6000_gen_lvx): Likewise.
6511 (rs6000_gen_stvx): Likewise.
6512 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
6513 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
6514 when memory address is aligned.
6515 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
6516 this split to select lvx instruction when memory address is aligned.
6517 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
6518 instruction when memory address is aligned.
6519 (*vsx_le_perm_load_v16qi): Likewise.
6520 (four unnamed splitters): Modify to select the stvx instruction
6521 when memory is aligned.
6523 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
6525 * predict.c (determine_unlikely_bbs): Handle correctly BBs
6526 which appears in the queue multiple times.
6528 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6529 Alan Hayward <alan.hayward@arm.com>
6530 David Sherwood <david.sherwood@arm.com>
6532 * tree-vectorizer.h (vec_lower_bound): New structure.
6533 (_loop_vec_info): Add check_nonzero and lower_bounds.
6534 (LOOP_VINFO_CHECK_NONZERO): New macro.
6535 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
6536 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
6537 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
6538 fields. Make seg_len the distance travelled, not including the
6540 (dr_direction_indicator): Declare.
6541 (dr_zero_step_indicator): Likewise.
6542 (dr_known_forward_stride_p): Likewise.
6543 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
6545 (runtime_alias_check_p): Allow runtime alias checks with
6547 (operator ==): Compare access_size and align.
6548 (prune_runtime_alias_test_list): Rework for new distinction between
6549 the access_size and seg_len.
6550 (create_intersect_range_checks_index): Likewise. Cope with polynomial
6552 (get_segment_min_max): New function.
6553 (create_intersect_range_checks): Use it.
6554 (dr_step_indicator): New function.
6555 (dr_direction_indicator): Likewise.
6556 (dr_zero_step_indicator): Likewise.
6557 (dr_known_forward_stride_p): Likewise.
6558 * tree-loop-distribution.c (data_ref_segment_size): Return
6559 DR_STEP * (niters - 1).
6560 (compute_alias_check_pairs): Update call to the dr_with_seg_len
6562 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
6563 (vect_preserves_scalar_order_p): New function, split out from...
6564 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
6565 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
6566 (vect_vfa_access_size): New function.
6567 (vect_vfa_align): Likewise.
6568 (vect_compile_time_alias): Take access_size_a and access_b arguments.
6569 (dump_lower_bound): New function.
6570 (vect_check_lower_bound): Likewise.
6571 (vect_small_gap_p): Likewise.
6572 (vectorizable_with_step_bound_p): Likewise.
6573 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
6574 depencies if the vectorization factor is 1. Convert the checks
6575 for nonzero steps into checks on the bounds of DR_STEP. Try using
6576 a bunds check for variable steps if the minimum required step is
6577 relatively small. Update calls to the dr_with_seg_len
6578 constructor and to vect_compile_time_alias.
6579 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
6581 (vect_loop_versioning): Call it.
6582 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
6584 (vect_estimate_min_profitable_iters): Account for any bounds checks.
6586 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6587 Alan Hayward <alan.hayward@arm.com>
6588 David Sherwood <david.sherwood@arm.com>
6590 * doc/sourcebuild.texi (vect_scatter_store): Document.
6591 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
6593 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
6595 * genopinit.c (main): Add supports_vec_scatter_store and
6596 supports_vec_scatter_store_cached to target_optabs.
6597 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
6598 IFN_MASK_SCATTER_STORE.
6599 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
6601 * internal-fn.h (internal_store_fn_p): Declare.
6602 (internal_fn_stored_value_index): Likewise.
6603 * internal-fn.c (scatter_store_direct): New macro.
6604 (expand_scatter_store_optab_fn): New function.
6605 (direct_scatter_store_optab_supported_p): New macro.
6606 (internal_store_fn_p): New function.
6607 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
6608 IFN_MASK_SCATTER_STORE.
6609 (internal_fn_mask_index): Likewise.
6610 (internal_fn_stored_value_index): New function.
6611 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
6613 * optabs-query.h (supports_vec_scatter_store_p): Declare.
6614 * optabs-query.c (supports_vec_scatter_store_p): New function.
6615 * tree-vectorizer.h (vect_get_store_rhs): Declare.
6616 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
6617 true for scatter stores.
6618 (vect_gather_scatter_fn_p): Handle scatter stores too.
6619 (vect_check_gather_scatter): Consider using scatter stores if
6620 supports_vec_scatter_store_p.
6621 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
6623 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6624 internal_fn_stored_value_index.
6625 (check_load_store_masking): Handle scatter stores too.
6626 (vect_get_store_rhs): Make public.
6627 (vectorizable_call): Use internal_store_fn_p.
6628 (vectorizable_store): Handle scatter store internal functions.
6629 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
6630 when deciding whether the end of the group has been reached.
6631 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
6632 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
6633 (mask_scatter_store<mode>): New insns.
6635 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6636 Alan Hayward <alan.hayward@arm.com>
6637 David Sherwood <david.sherwood@arm.com>
6639 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
6640 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
6641 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
6643 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
6644 Use vect_truncate_gather_scatter_offset if we can't treat the
6645 operation as a normal gather load or scatter store.
6646 (get_group_load_store_type): Take the gather_scatter_info
6647 as argument. Try using a gather load or scatter store for
6648 single-element groups.
6649 (get_load_store_type): Update calls to get_group_load_store_type
6650 and vect_use_strided_gather_scatters_p.
6652 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6653 Alan Hayward <alan.hayward@arm.com>
6654 David Sherwood <david.sherwood@arm.com>
6656 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
6657 optional tree argument.
6658 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
6660 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
6661 but continue to use the current value as a fallback.
6662 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
6663 to compare the updates.
6664 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
6665 (get_load_store_type): Use it when handling a strided access.
6666 (vect_get_strided_load_store_ops): New function.
6667 (vect_get_data_ptr_increment): Likewise.
6668 (vectorizable_load): Handle strided gather loads. Always pass
6669 a step to vect_create_data_ref_ptr and bump_vector_ptr.
6671 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6672 Alan Hayward <alan.hayward@arm.com>
6673 David Sherwood <david.sherwood@arm.com>
6675 * doc/md.texi (gather_load@var{m}): Document.
6676 (mask_gather_load@var{m}): Likewise.
6677 * genopinit.c (main): Add supports_vec_gather_load and
6678 supports_vec_gather_load_cached to target_optabs.
6679 * optabs-tree.c (init_tree_optimization_optabs): Use
6680 ggc_cleared_alloc to allocate target_optabs.
6681 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
6682 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
6684 * internal-fn.h (internal_load_fn_p): Declare.
6685 (internal_gather_scatter_fn_p): Likewise.
6686 (internal_fn_mask_index): Likewise.
6687 (internal_gather_scatter_fn_supported_p): Likewise.
6688 * internal-fn.c (gather_load_direct): New macro.
6689 (expand_gather_load_optab_fn): New function.
6690 (direct_gather_load_optab_supported_p): New macro.
6691 (direct_internal_fn_optab): New function.
6692 (internal_load_fn_p): Likewise.
6693 (internal_gather_scatter_fn_p): Likewise.
6694 (internal_fn_mask_index): Likewise.
6695 (internal_gather_scatter_fn_supported_p): Likewise.
6696 * optabs-query.c (supports_at_least_one_mode_p): New function.
6697 (supports_vec_gather_load_p): Likewise.
6698 * optabs-query.h (supports_vec_gather_load_p): Declare.
6699 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
6700 and memory_type field.
6701 (NUM_PATTERNS): Bump to 15.
6702 * tree-vect-data-refs.c: Include internal-fn.h.
6703 (vect_gather_scatter_fn_p): New function.
6704 (vect_describe_gather_scatter_call): Likewise.
6705 (vect_check_gather_scatter): Try using internal functions for
6706 gather loads. Recognize existing calls to a gather load function.
6707 (vect_analyze_data_refs): Consider using gather loads if
6708 supports_vec_gather_load_p.
6709 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
6710 (vect_get_gather_scatter_offset_type): Likewise.
6711 (vect_convert_mask_for_vectype): Likewise.
6712 (vect_add_conversion_to_patterm): Likewise.
6713 (vect_try_gather_scatter_pattern): Likewise.
6714 (vect_recog_gather_scatter_pattern): New pattern recognizer.
6715 (vect_vect_recog_func_ptrs): Add it.
6716 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6717 internal_fn_mask_index and internal_gather_scatter_fn_p.
6718 (check_load_store_masking): Take the gather_scatter_info as an
6719 argument and handle gather loads.
6720 (vect_get_gather_scatter_ops): New function.
6721 (vectorizable_call): Check internal_load_fn_p.
6722 (vectorizable_load): Likewise. Handle gather load internal
6724 (vectorizable_store): Update call to check_load_store_masking.
6725 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
6726 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
6727 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
6728 (aarch64_gather_scale_operand_d): New predicates.
6729 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
6730 (mask_gather_load<mode>): New insns.
6732 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6733 Alan Hayward <alan.hayward@arm.com>
6734 David Sherwood <david.sherwood@arm.com>
6736 * optabs.def (fold_left_plus_optab): New optab.
6737 * doc/md.texi (fold_left_plus_@var{m}): Document.
6738 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
6739 * internal-fn.c (fold_left_direct): Define.
6740 (expand_fold_left_optab_fn): Likewise.
6741 (direct_fold_left_optab_supported_p): Likewise.
6742 * fold-const-call.c (fold_const_fold_left): New function.
6743 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
6744 * tree-parloops.c (valid_reduction_p): New function.
6745 (gather_scalar_reductions): Use it.
6746 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
6747 (vect_finish_replace_stmt): Declare.
6748 * tree-vect-loop.c (fold_left_reduction_fn): New function.
6749 (needs_fold_left_reduction_p): New function, split out from...
6750 (vect_is_simple_reduction): ...here. Accept reductions that
6751 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
6752 (vect_force_simple_reduction): Also store the reduction type in
6753 the assignment's STMT_VINFO_REDUC_TYPE.
6754 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
6755 (merge_with_identity): New function.
6756 (vect_expand_fold_left): Likewise.
6757 (vectorize_fold_left_reduction): Likewise.
6758 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
6759 scalar phi in place for it. Check for target support and reject
6760 cases that would reassociate the operation. Defer the transform
6761 phase to vectorize_fold_left_reduction.
6762 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
6763 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
6764 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
6766 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6768 * tree-if-conv.c (predicate_mem_writes): Remove redundant
6769 call to ifc_temp_var.
6771 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6772 Alan Hayward <alan.hayward@arm.com>
6773 David Sherwood <david.sherwood@arm.com>
6775 * target.def (legitimize_address_displacement): Take the original
6776 offset as a poly_int.
6777 * targhooks.h (default_legitimize_address_displacement): Update
6779 * targhooks.c (default_legitimize_address_displacement): Likewise.
6780 * doc/tm.texi: Regenerate.
6781 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
6782 as an argument, moving assert of ad->disp == ad->disp_term to...
6783 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
6784 Try calling targetm.legitimize_address_displacement before expanding
6785 the address rather than afterwards, and adjust for the new interface.
6786 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
6787 Match the new hook interface. Handle SVE addresses.
6788 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
6791 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6793 * Makefile.in (OBJS): Add early-remat.o.
6794 * target.def (select_early_remat_modes): New hook.
6795 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
6796 * doc/tm.texi: Regenerate.
6797 * targhooks.h (default_select_early_remat_modes): Declare.
6798 * targhooks.c (default_select_early_remat_modes): New function.
6799 * timevar.def (TV_EARLY_REMAT): New timevar.
6800 * passes.def (pass_early_remat): New pass.
6801 * tree-pass.h (make_pass_early_remat): Declare.
6802 * early-remat.c: New file.
6803 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
6805 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
6807 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6808 Alan Hayward <alan.hayward@arm.com>
6809 David Sherwood <david.sherwood@arm.com>
6811 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
6812 vfm1 with a bound_epilog parameter.
6813 (vect_do_peeling): Update calls accordingly, and move the prologue
6814 call earlier in the function. Treat the base bound_epilog as 0 for
6815 fully-masked loops and retain vf - 1 for other loops. Add 1 to
6816 this base when peeling for gaps.
6817 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
6818 with fully-masked loops.
6819 (vect_estimate_min_profitable_iters): Handle the single peeled
6820 iteration in that case.
6822 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6823 Alan Hayward <alan.hayward@arm.com>
6824 David Sherwood <david.sherwood@arm.com>
6826 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
6827 single-element interleaving even if the size is not a power of 2.
6828 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
6829 accesses for single-element interleaving if the group size is
6832 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6833 Alan Hayward <alan.hayward@arm.com>
6834 David Sherwood <david.sherwood@arm.com>
6836 * doc/md.texi (fold_extract_last_@var{m}): Document.
6837 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
6838 * optabs.def (fold_extract_last_optab): New optab.
6839 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
6840 * internal-fn.c (fold_extract_direct): New macro.
6841 (expand_fold_extract_optab_fn): Likewise.
6842 (direct_fold_extract_optab_supported_p): Likewise.
6843 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
6844 * tree-vect-loop.c (vect_model_reduction_cost): Handle
6845 EXTRACT_LAST_REDUCTION.
6846 (get_initial_def_for_reduction): Do not create an initial vector
6847 for EXTRACT_LAST_REDUCTION reductions.
6848 (vectorizable_reduction): Leave the scalar phi in place for
6849 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
6850 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
6851 epilogue code for EXTRACT_LAST_REDUCTION and defer the
6852 transform phase to vectorizable_condition.
6853 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
6855 (vect_finish_stmt_generation): ...here.
6856 (vect_finish_replace_stmt): New function.
6857 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
6858 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
6860 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
6862 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6863 Alan Hayward <alan.hayward@arm.com>
6864 David Sherwood <david.sherwood@arm.com>
6866 * doc/md.texi (extract_last_@var{m}): Document.
6867 * optabs.def (extract_last_optab): New optab.
6868 * internal-fn.def (EXTRACT_LAST): New internal function.
6869 * internal-fn.c (cond_unary_direct): New macro.
6870 (expand_cond_unary_optab_fn): Likewise.
6871 (direct_cond_unary_optab_supported_p): Likewise.
6872 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
6873 loops using EXTRACT_LAST.
6874 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
6875 (extract_last_<mode>): ...this optab.
6876 (vec_extract<mode><Vel>): Update accordingly.
6878 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6879 Alan Hayward <alan.hayward@arm.com>
6880 David Sherwood <david.sherwood@arm.com>
6882 * target.def (empty_mask_is_expensive): New hook.
6883 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
6884 * doc/tm.texi: Regenerate.
6885 * targhooks.h (default_empty_mask_is_expensive): Declare.
6886 * targhooks.c (default_empty_mask_is_expensive): New function.
6887 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
6888 if the target says that empty masks are expensive.
6889 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
6891 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
6893 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6894 Alan Hayward <alan.hayward@arm.com>
6895 David Sherwood <david.sherwood@arm.com>
6897 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
6898 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
6899 (vect_use_loop_mask_for_alignment_p): New function.
6900 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
6901 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
6902 niters_skip argument. Make sure that the first niters_skip elements
6903 of the first iteration are inactive.
6904 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
6905 Update call to vect_set_loop_masks_directly.
6906 (get_misalign_in_elems): New function, split out from...
6907 (vect_gen_prolog_loop_niters): ...here.
6908 (vect_update_init_of_dr): Take a code argument that specifies whether
6909 the adjustment should be added or subtracted.
6910 (vect_update_init_of_drs): Likewise.
6911 (vect_prepare_for_masked_peels): New function.
6912 (vect_do_peeling): Skip prologue peeling if we're using a mask
6913 instead. Update call to vect_update_inits_of_drs.
6914 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6916 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
6917 alignment. Do not include the number of peeled iterations in
6918 the minimum threshold in that case.
6919 (vectorizable_induction): Adjust the start value down by
6920 LOOP_VINFO_MASK_SKIP_NITERS iterations.
6921 (vect_transform_loop): Call vect_prepare_for_masked_peels.
6922 Take the number of skipped iterations into account when calculating
6924 * tree-vect-stmts.c (vect_gen_while_not): New function.
6926 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6927 Alan Hayward <alan.hayward@arm.com>
6928 David Sherwood <david.sherwood@arm.com>
6930 * doc/sourcebuild.texi (vect_fully_masked): Document.
6931 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
6933 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
6935 (vect_analyze_loop_2): ...here. Don't check the vectorization
6936 factor against the number of loop iterations if the loop is
6939 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6940 Alan Hayward <alan.hayward@arm.com>
6941 David Sherwood <david.sherwood@arm.com>
6943 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
6944 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
6945 (dump_groups): Update accordingly.
6946 (iv_use::mem_type): New member variable.
6947 (address_p): New function.
6948 (record_use): Add a mem_type argument and initialize the new
6950 (record_group_use): Add a mem_type argument. Use address_p.
6951 Remove obsolete null checks of base_object. Update call to record_use.
6952 (find_interesting_uses_op): Update call to record_group_use.
6953 (find_interesting_uses_cond): Likewise.
6954 (find_interesting_uses_address): Likewise.
6955 (get_mem_type_for_internal_fn): New function.
6956 (find_address_like_use): Likewise.
6957 (find_interesting_uses_stmt): Try find_address_like_use before
6958 calling find_interesting_uses_op.
6959 (addr_offset_valid_p): Use the iv mem_type field as the type
6960 of the addressed memory.
6961 (add_autoinc_candidates): Likewise.
6962 (get_address_cost): Likewise.
6963 (split_small_address_groups_p): Use address_p.
6964 (split_address_groups): Likewise.
6965 (add_iv_candidate_for_use): Likewise.
6966 (autoinc_possible_for_pair): Likewise.
6967 (rewrite_groups): Likewise.
6968 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
6969 (determine_group_iv_cost): Update after split of USE_ADDRESS.
6970 (get_alias_ptr_type_for_ptr_address): New function.
6971 (rewrite_use_address): Rewrite address uses in calls that were
6972 identified by find_address_like_use.
6974 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6975 Alan Hayward <alan.hayward@arm.com>
6976 David Sherwood <david.sherwood@arm.com>
6978 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
6980 * gimple-expr.h (is_gimple_addressable: Likewise.
6981 * gimple-expr.c (is_gimple_address): Likewise.
6982 * internal-fn.c (expand_call_mem_ref): New function.
6983 (expand_mask_load_optab_fn): Use it.
6984 (expand_mask_store_optab_fn): Likewise.
6986 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6987 Alan Hayward <alan.hayward@arm.com>
6988 David Sherwood <david.sherwood@arm.com>
6990 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
6991 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
6992 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
6993 (cond_umax@var{mode}): Document.
6994 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
6995 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
6996 (cond_umin_optab, cond_umax_optab): New optabs.
6997 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
6998 (COND_IOR, COND_XOR): New internal functions.
6999 * internal-fn.h (get_conditional_internal_fn): Declare.
7000 * internal-fn.c (cond_binary_direct): New macro.
7001 (expand_cond_binary_optab_fn): Likewise.
7002 (direct_cond_binary_optab_supported_p): Likewise.
7003 (get_conditional_internal_fn): New function.
7004 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
7005 Cope with reduction statements that are vectorized as calls rather
7007 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
7008 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
7009 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
7010 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
7011 (UNSPEC_COND_EOR): New unspecs.
7012 (optab): Add mappings for them.
7013 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
7014 (sve_int_op, sve_fp_op): New int attributes.
7016 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7017 Alan Hayward <alan.hayward@arm.com>
7018 David Sherwood <david.sherwood@arm.com>
7020 * optabs.def (while_ult_optab): New optab.
7021 * doc/md.texi (while_ult@var{m}@var{n}): Document.
7022 * internal-fn.def (WHILE_ULT): New internal function.
7023 * internal-fn.h (direct_internal_fn_supported_p): New override
7024 that takes two types as argument.
7025 * internal-fn.c (while_direct): New macro.
7026 (expand_while_optab_fn): New function.
7027 (convert_optab_supported_p): Likewise.
7028 (direct_while_optab_supported_p): New macro.
7029 * wide-int.h (wi::udiv_ceil): New function.
7030 * tree-vectorizer.h (rgroup_masks): New structure.
7031 (vec_loop_masks): New typedef.
7032 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
7034 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
7035 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
7036 (vect_max_vf): New function.
7037 (slpeel_make_loop_iterate_ntimes): Delete.
7038 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
7039 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
7040 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
7041 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
7042 internal-fn.h, stor-layout.h and optabs-query.h.
7043 (vect_set_loop_mask): New function.
7044 (add_preheader_seq): Likewise.
7045 (add_header_seq): Likewise.
7046 (interleave_supported_p): Likewise.
7047 (vect_maybe_permute_loop_masks): Likewise.
7048 (vect_set_loop_masks_directly): Likewise.
7049 (vect_set_loop_condition_masked): Likewise.
7050 (vect_set_loop_condition_unmasked): New function, split out from
7051 slpeel_make_loop_iterate_ntimes.
7052 (slpeel_make_loop_iterate_ntimes): Rename to..
7053 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
7054 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
7055 (vect_do_peeling): Update call accordingly.
7056 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
7058 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7059 mask_compare_type, can_fully_mask_p and fully_masked_p.
7060 (release_vec_loop_masks): New function.
7061 (_loop_vec_info): Use it to free the loop masks.
7062 (can_produce_all_loop_masks_p): New function.
7063 (vect_get_max_nscalars_per_iter): Likewise.
7064 (vect_verify_full_masking): Likewise.
7065 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
7066 retries, and free the mask rgroups before retrying. Check loop-wide
7067 reasons for disallowing fully-masked loops. Make the final decision
7068 about whether use a fully-masked loop or not.
7069 (vect_estimate_min_profitable_iters): Do not assume that peeling
7070 for the number of iterations will be needed for fully-masked loops.
7071 (vectorizable_reduction): Disable fully-masked loops.
7072 (vectorizable_live_operation): Likewise.
7073 (vect_halve_mask_nunits): New function.
7074 (vect_double_mask_nunits): Likewise.
7075 (vect_record_loop_mask): Likewise.
7076 (vect_get_loop_mask): Likewise.
7077 (vect_transform_loop): Handle the case in which the final loop
7078 iteration might handle a partial vector. Call vect_set_loop_condition
7079 instead of slpeel_make_loop_iterate_ntimes.
7080 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
7081 (check_load_store_masking): New function.
7082 (prepare_load_store_mask): Likewise.
7083 (vectorizable_store): Handle fully-masked loops.
7084 (vectorizable_load): Likewise.
7085 (supportable_widening_operation): Use vect_halve_mask_nunits for
7087 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
7088 (vect_gen_while): New function.
7089 * config/aarch64/aarch64.md (umax<mode>3): New expander.
7090 (aarch64_uqdec<mode>): New insn.
7092 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7093 Alan Hayward <alan.hayward@arm.com>
7094 David Sherwood <david.sherwood@arm.com>
7096 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
7097 (reduc_xor_scal_optab): New optabs.
7098 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
7099 (reduc_xor_scal_@var{m}): Document.
7100 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
7101 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
7103 * fold-const-call.c (fold_const_call): Handle them.
7104 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
7105 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
7106 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
7107 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
7108 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
7109 (UNSPEC_XORV): New unspecs.
7110 (optab): Add entries for them.
7111 (BITWISEV): New int iterator.
7112 (bit_reduc_op): New int attributes.
7114 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7115 Alan Hayward <alan.hayward@arm.com>
7116 David Sherwood <david.sherwood@arm.com>
7118 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
7119 * internal-fn.def (VEC_SHL_INSERT): New internal function.
7120 * optabs.def (vec_shl_insert_optab): New optab.
7121 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
7122 (duplicate_and_interleave): Likewise.
7123 * tree-vect-loop.c: Include internal-fn.h.
7124 (neutral_op_for_slp_reduction): New function, split out from
7125 get_initial_defs_for_reduction.
7126 (get_initial_def_for_reduction): Handle option 2 for variable-length
7127 vectors by loading the neutral value into a vector and then shifting
7128 the initial value into element 0.
7129 (get_initial_defs_for_reduction): Replace the code argument with
7130 the neutral value calculated by neutral_op_for_slp_reduction.
7131 Use gimple_build_vector for constant-length vectors.
7132 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
7133 but the first group_size elements have a neutral value.
7134 Use duplicate_and_interleave otherwise.
7135 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
7136 Update call to get_initial_defs_for_reduction. Handle SLP
7137 reductions for variable-length vectors by creating one vector
7138 result for each scalar result, with the elements associated
7139 with other scalar results stubbed out with the neutral value.
7140 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
7141 Require IFN_VEC_SHL_INSERT for double reductions on
7142 variable-length vectors, or SLP reductions that have
7143 a neutral value. Require can_duplicate_and_interleave_p
7144 support for variable-length unchained SLP reductions if there
7145 is no neutral value, such as for MIN/MAX reductions. Also require
7146 the number of vector elements to be a multiple of the number of
7147 SLP statements when doing variable-length unchained SLP reductions.
7148 Update call to vect_create_epilog_for_reduction.
7149 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
7150 and remove initial values.
7151 (duplicate_and_interleave): Make public.
7152 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
7153 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
7155 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7156 Alan Hayward <alan.hayward@arm.com>
7157 David Sherwood <david.sherwood@arm.com>
7159 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
7160 (can_duplicate_and_interleave_p): New function.
7161 (vect_get_and_check_slp_defs): Take the vector of statements
7162 rather than just the current one. Remove excess parentheses.
7163 Restriction rejectinon of vect_constant_def and vect_external_def
7164 for variable-length vectors to boolean types, or types for which
7165 can_duplicate_and_interleave_p is false.
7166 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
7167 (duplicate_and_interleave): New function.
7168 (vect_get_constant_vectors): Use gimple_build_vector for
7169 constant-length vectors and suitable variable-length constant
7170 vectors. Use duplicate_and_interleave for other variable-length
7171 vectors. Don't defer the update when inserting new statements.
7173 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7174 Alan Hayward <alan.hayward@arm.com>
7175 David Sherwood <david.sherwood@arm.com>
7177 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
7178 min_profitable_iters doesn't go negative.
7180 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7181 Alan Hayward <alan.hayward@arm.com>
7182 David Sherwood <david.sherwood@arm.com>
7184 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
7185 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
7186 * optabs.def (vec_mask_load_lanes_optab): New optab.
7187 (vec_mask_store_lanes_optab): Likewise.
7188 * internal-fn.def (MASK_LOAD_LANES): New internal function.
7189 (MASK_STORE_LANES): Likewise.
7190 * internal-fn.c (mask_load_lanes_direct): New macro.
7191 (mask_store_lanes_direct): Likewise.
7192 (expand_mask_load_optab_fn): Handle masked operations.
7193 (expand_mask_load_lanes_optab_fn): New macro.
7194 (expand_mask_store_optab_fn): Handle masked operations.
7195 (expand_mask_store_lanes_optab_fn): New macro.
7196 (direct_mask_load_lanes_optab_supported_p): Likewise.
7197 (direct_mask_store_lanes_optab_supported_p): Likewise.
7198 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
7200 (vect_load_lanes_supported): Likewise.
7201 * tree-vect-data-refs.c (strip_conversion): New function.
7202 (can_group_stmts_p): Likewise.
7203 (vect_analyze_data_ref_accesses): Use it instead of checking
7204 for a pair of assignments.
7205 (vect_store_lanes_supported): Take a masked_p parameter.
7206 (vect_load_lanes_supported): Likewise.
7207 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
7208 vect_store_lanes_supported and vect_load_lanes_supported.
7209 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7210 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
7211 parameter. Don't allow gaps for masked accesses.
7212 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
7213 and vect_load_lanes_supported.
7214 (get_load_store_type): Take a masked_p parameter and update
7215 call to get_group_load_store_type.
7216 (vectorizable_store): Update call to get_load_store_type.
7217 Handle IFN_MASK_STORE_LANES.
7218 (vectorizable_load): Update call to get_load_store_type.
7219 Handle IFN_MASK_LOAD_LANES.
7221 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7222 Alan Hayward <alan.hayward@arm.com>
7223 David Sherwood <david.sherwood@arm.com>
7225 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
7227 * config/aarch64/aarch64-protos.h
7228 (aarch64_sve_struct_memory_operand_p): Declare.
7229 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
7230 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
7231 (VPRED, vpred): Handle SVE structure modes.
7232 * config/aarch64/constraints.md (Utx): New constraint.
7233 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
7234 (aarch64_sve_struct_nonimmediate_operand): New predicates.
7235 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
7236 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
7237 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
7238 structure modes. Split into pieces after RA.
7239 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
7240 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
7242 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
7243 SVE structure modes.
7244 (aarch64_classify_address): Likewise.
7245 (sizetochar): Move earlier in file.
7246 (aarch64_print_operand): Handle SVE register lists.
7247 (aarch64_array_mode): New function.
7248 (aarch64_sve_struct_memory_operand_p): Likewise.
7249 (TARGET_ARRAY_MODE): Redefine.
7251 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7252 Alan Hayward <alan.hayward@arm.com>
7253 David Sherwood <david.sherwood@arm.com>
7255 * target.def (array_mode): New target hook.
7256 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
7257 * doc/tm.texi: Regenerate.
7258 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
7259 * hooks.c (hook_optmode_mode_uhwi_none): New function.
7260 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
7262 * stor-layout.c (mode_for_array): Likewise. Support polynomial
7265 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7266 Alan Hayward <alan.hayward@arm.com>
7267 David Sherwood <david.sherwood@arm.com>
7269 * fold-const.c (fold_binary_loc): Check the argument types
7270 rather than the result type when testing for a vector operation.
7272 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7274 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
7275 * doc/tm.texi: Regenerate.
7277 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7278 Alan Hayward <alan.hayward@arm.com>
7279 David Sherwood <david.sherwood@arm.com>
7281 * doc/invoke.texi (-msve-vector-bits=): Document new option.
7282 (sve): Document new AArch64 extension.
7283 * doc/md.texi (w): Extend the description of the AArch64
7284 constraint to include SVE vectors.
7285 (Upl, Upa): Document new AArch64 predicate constraints.
7286 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
7288 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
7289 (msve-vector-bits=): New option.
7290 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
7291 SVE when these are disabled.
7292 (sve): New extension.
7293 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
7294 modes. Adjust their number of units based on aarch64_sve_vg.
7295 (MAX_BITSIZE_MODE_ANY_MODE): Define.
7296 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
7297 aarch64_addr_query_type.
7298 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
7299 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
7300 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
7301 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
7302 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
7303 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
7304 (aarch64_simd_imm_zero_p): Delete.
7305 (aarch64_check_zero_based_sve_index_immediate): Declare.
7306 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7307 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7308 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7309 (aarch64_sve_float_mul_immediate_p): Likewise.
7310 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7312 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
7313 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
7314 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
7315 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
7316 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
7317 (aarch64_regmode_natural_size): Likewise.
7318 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
7319 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
7321 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
7322 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
7323 for VG and the SVE predicate registers.
7324 (V_ALIASES): Add a "z"-prefixed alias.
7325 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
7326 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
7327 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
7328 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
7329 (REG_CLASS_NAMES): Add entries for them.
7330 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
7331 and the predicate registers.
7332 (aarch64_sve_vg): Declare.
7333 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
7334 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
7335 (REGMODE_NATURAL_SIZE): Define.
7336 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
7338 * config/aarch64/aarch64.c: Include cfgrtl.h.
7339 (simd_immediate_info): Add a constructor for series vectors,
7340 and an associated step field.
7341 (aarch64_sve_vg): New variable.
7342 (aarch64_dbx_register_number): Handle VG and the predicate registers.
7343 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
7344 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
7345 (VEC_ANY_DATA, VEC_STRUCT): New constants.
7346 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
7347 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
7348 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
7349 (aarch64_get_mask_mode): New functions.
7350 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
7351 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
7352 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
7353 predicate modes and predicate registers. Explicitly restrict
7354 GPRs to modes of 16 bytes or smaller. Only allow FP registers
7355 to store a vector mode if it is recognized by
7356 aarch64_classify_vector_mode.
7357 (aarch64_regmode_natural_size): New function.
7358 (aarch64_hard_regno_caller_save_mode): Return the original mode
7360 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
7361 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
7362 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
7363 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
7365 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
7366 does not overlap dest if the function is frame-related. Handle
7368 (aarch64_split_add_offset): New function.
7369 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
7370 them aarch64_add_offset.
7371 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
7372 and update call to aarch64_sub_sp.
7373 (aarch64_add_cfa_expression): New function.
7374 (aarch64_expand_prologue): Pass extra temporary registers to the
7375 functions above. Handle the case in which we need to emit new
7376 DW_CFA_expressions for registers that were originally saved
7377 relative to the stack pointer, but now have to be expressed
7378 relative to the frame pointer.
7379 (aarch64_output_mi_thunk): Pass extra temporary registers to the
7381 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
7382 IP0 and IP1 values for SVE frames.
7383 (aarch64_expand_vec_series): New function.
7384 (aarch64_expand_sve_widened_duplicate): Likewise.
7385 (aarch64_expand_sve_const_vector): Likewise.
7386 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
7387 Handle SVE constants. Use emit_move_insn to move a force_const_mem
7388 into the register, rather than emitting a SET directly.
7389 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
7390 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
7391 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
7392 (offset_9bit_signed_scaled_p): New functions.
7393 (aarch64_replicate_bitmask_imm): New function.
7394 (aarch64_bitmask_imm): Use it.
7395 (aarch64_cannot_force_const_mem): Reject expressions involving
7396 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
7397 (aarch64_classify_index): Handle SVE indices, by requiring
7398 a plain register index with a scale that matches the element size.
7399 (aarch64_classify_address): Handle SVE addresses. Assert that
7400 the mode of the address is VOIDmode or an integer mode.
7401 Update call to aarch64_classify_symbol.
7402 (aarch64_classify_symbolic_expression): Update call to
7403 aarch64_classify_symbol.
7404 (aarch64_const_vec_all_in_range_p): New function.
7405 (aarch64_print_vector_float_operand): Likewise.
7406 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
7407 "vN" for FP registers with SVE modes. Handle (const ...) vectors
7408 and the FP immediates 1.0 and 0.5.
7409 (aarch64_print_address_internal): Handle SVE addresses.
7410 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
7411 (aarch64_regno_regclass): Handle predicate registers.
7412 (aarch64_secondary_reload): Handle big-endian reloads of SVE
7414 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
7415 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
7416 (aarch64_convert_sve_vector_bits): New function.
7417 (aarch64_override_options): Use it to handle -msve-vector-bits=.
7418 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7420 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
7421 Handle SVE vector and predicate modes. Accept VL-based constants
7422 that need only one temporary register, and VL offsets that require
7423 no temporary registers.
7424 (aarch64_conditional_register_usage): Mark the predicate registers
7425 as fixed if SVE isn't available.
7426 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
7427 Return true for SVE vector and predicate modes.
7428 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
7429 rather than an unsigned int. Handle SVE modes.
7430 (aarch64_preferred_simd_mode): Update call accordingly. Handle
7432 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
7434 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7435 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7436 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7437 (aarch64_sve_float_mul_immediate_p): New functions.
7438 (aarch64_sve_valid_immediate): New function.
7439 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
7440 Explicitly reject structure modes. Check for INDEX constants.
7441 Handle PTRUE and PFALSE constants.
7442 (aarch64_check_zero_based_sve_index_immediate): New function.
7443 (aarch64_simd_imm_zero_p): Delete.
7444 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
7445 vector modes. Accept constants in the range of CNT[BHWD].
7446 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
7447 ask for an Advanced SIMD mode.
7448 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
7449 (aarch64_simd_vector_alignment): Handle SVE predicates.
7450 (aarch64_vectorize_preferred_vector_alignment): New function.
7451 (aarch64_simd_vector_alignment_reachable): Use it instead of
7453 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
7454 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
7456 (MAX_VECT_LEN): Delete.
7457 (expand_vec_perm_d): Add a vec_flags field.
7458 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
7459 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
7460 (aarch64_evpc_ext): Don't apply a big-endian lane correction
7462 (aarch64_evpc_rev): Rename to...
7463 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
7464 (aarch64_evpc_rev_global): New function.
7465 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
7466 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
7468 (aarch64_evpc_sve_tbl): New function.
7469 (aarch64_expand_vec_perm_const_1): Update after rename of
7470 aarch64_evpc_rev. Handle SVE permutes too, trying
7471 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
7472 than aarch64_evpc_tbl.
7473 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
7474 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
7475 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
7476 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
7477 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
7478 (aarch64_expand_sve_vcond): New functions.
7479 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
7480 of aarch64_vector_mode_p.
7481 (aarch64_dwarf_poly_indeterminate_value): New function.
7482 (aarch64_compute_pressure_classes): Likewise.
7483 (aarch64_can_change_mode_class): Likewise.
7484 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
7485 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
7486 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
7487 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
7488 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
7489 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
7490 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
7491 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
7493 (Dn, Dl, Dr): Accept const as well as const_vector.
7494 (Dz): Likewise. Compare against CONST0_RTX.
7495 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
7496 of "vector" where appropriate.
7497 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
7498 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
7499 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
7500 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
7501 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
7502 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
7503 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
7504 (v_int_equiv): Extend to SVE modes.
7505 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
7507 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
7508 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
7509 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
7510 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
7511 (SVE_COND_FP_CMP): New int iterators.
7512 (perm_hilo): Handle the new unpack unspecs.
7513 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
7515 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
7516 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
7517 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
7518 (aarch64_equality_operator, aarch64_constant_vector_operand)
7519 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
7520 (aarch64_sve_nonimmediate_operand): Likewise.
7521 (aarch64_sve_general_operand): Likewise.
7522 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
7523 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
7524 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
7525 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
7526 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
7527 (aarch64_sve_float_arith_immediate): Likewise.
7528 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
7529 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
7530 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
7531 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
7532 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
7533 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
7534 (aarch64_sve_float_arith_operand): Likewise.
7535 (aarch64_sve_float_arith_with_sub_operand): Likewise.
7536 (aarch64_sve_float_mul_operand): Likewise.
7537 (aarch64_sve_vec_perm_operand): Likewise.
7538 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
7539 (aarch64_mov_operand): Accept const_poly_int and const_vector.
7540 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
7541 as well as const_vector.
7542 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
7543 in file. Use CONST0_RTX and CONSTM1_RTX.
7544 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
7545 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
7546 Use aarch64_simd_imm_zero.
7547 * config/aarch64/aarch64-sve.md: New file.
7548 * config/aarch64/aarch64.md: Include it.
7549 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
7550 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
7551 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
7552 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
7553 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
7554 (sve): New attribute.
7555 (enabled): Disable instructions with the sve attribute unless
7557 (movqi, movhi): Pass CONST_POLY_INT operaneds through
7558 aarch64_expand_mov_immediate.
7559 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
7560 CNT[BHSD] immediates.
7561 (movti): Split CONST_POLY_INT moves into two halves.
7562 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
7563 Split additions that need a temporary here if the destination
7564 is the stack pointer.
7565 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
7566 (*add<mode>3_poly_1): New instruction.
7567 (set_clobber_cc): New expander.
7569 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7571 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
7572 parameter and use it instead of GET_MODE_SIZE (innermode). Use
7573 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
7574 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
7575 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
7576 Change innermode from fixed_mode_size to machine_mode.
7577 (simplify_subreg): Update call accordingly. Handle a constant-sized
7578 subreg of a variable-length CONST_VECTOR.
7580 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7581 Alan Hayward <alan.hayward@arm.com>
7582 David Sherwood <david.sherwood@arm.com>
7584 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
7585 (add_offset_to_base): New function, split out from...
7586 (create_mem_ref): ...here. When handling a scale other than 1,
7587 check first whether the address is valid without the offset.
7588 Add it into the base if so, leaving the index and scale as-is.
7590 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7593 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
7594 fold_for_warn before checking if arg2 is INTEGER_CST.
7596 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
7598 * config/rs6000/predicates.md (load_multiple_operation): Delete.
7599 (store_multiple_operation): Delete.
7600 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
7601 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
7602 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
7603 guarded by TARGET_STRING.
7604 (rs6000_output_load_multiple): Delete.
7605 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
7606 OPTION_MASK_STRING / TARGET_STRING handling.
7607 (print_operand) <'N', 'O'>: Add comment that these are unused now.
7608 (const rs6000_opt_masks) <"string">: Change mask to 0.
7609 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
7610 (MASK_STRING): Delete.
7611 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
7613 (load_multiple): Delete.
7620 (store_multiple): Delete.
7627 (movmemsi_8reg): Delete.
7628 (corresponding unnamed define_insn): Delete.
7629 (movmemsi_6reg): Delete.
7630 (corresponding unnamed define_insn): Delete.
7631 (movmemsi_4reg): Delete.
7632 (corresponding unnamed define_insn): Delete.
7633 (movmemsi_2reg): Delete.
7634 (corresponding unnamed define_insn): Delete.
7635 (movmemsi_1reg): Delete.
7636 (corresponding unnamed define_insn): Delete.
7637 * config/rs6000/rs6000.opt (mno-string): New.
7638 (mstring): Replace by deprecation warning stub.
7639 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
7641 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7643 * regrename.c (regrename_do_replace): If replacing the same
7644 reg multiple times, try to reuse last created gen_raw_REG.
7647 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
7648 main to workaround a bug in GDB.
7650 2018-01-12 Tom de Vries <tom@codesourcery.com>
7653 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
7655 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
7657 PR rtl-optimization/80481
7658 * ira-color.c (get_cap_member): New function.
7659 (allocnos_conflict_by_live_ranges_p): Use it.
7660 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
7661 (setup_slot_coalesced_allocno_live_ranges): Ditto.
7663 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
7666 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
7667 (*saddl_se_1): Ditto.
7669 (*ssubl_se_1): Ditto.
7671 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7673 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
7674 rather than wi::to_widest for DR_INITs.
7675 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
7676 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
7677 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
7679 (vect_analyze_group_access_1): Note that here.
7681 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7683 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
7684 polynomial type sizes.
7686 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7688 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
7689 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
7690 (gimple_add_tmp_var): Likewise.
7692 2018-01-12 Martin Liska <mliska@suse.cz>
7694 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
7695 (gimple_alloc_sizes): Likewise.
7696 (dump_gimple_statistics): Use PRIu64 in printf format.
7697 * gimple.h: Change uint64_t to int.
7699 2018-01-12 Martin Liska <mliska@suse.cz>
7701 * tree-core.h: Use uint64_t instead of int.
7702 * tree.c (tree_node_counts): Likewise.
7703 (tree_node_sizes): Likewise.
7704 (dump_tree_statistics): Use PRIu64 in printf format.
7706 2018-01-12 Martin Liska <mliska@suse.cz>
7708 * Makefile.in: As qsort_chk is implemented in vec.c, add
7709 vec.o to linkage of gencfn-macros.
7710 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
7711 passing the info to record_node_allocation_statistics.
7712 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
7714 * ggc-common.c (struct ggc_usage): Add operator== and use
7715 it in operator< and compare function.
7716 * mem-stats.h (struct mem_usage): Likewise.
7717 * vec.c (struct vec_usage): Remove operator< and compare
7718 function. Can be simply inherited.
7720 2018-01-12 Martin Jambor <mjambor@suse.cz>
7723 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
7724 * tree-ssa-math-opts.c: Include domwalk.h.
7725 (convert_mult_to_fma_1): New function.
7726 (fma_transformation_info): New type.
7727 (fma_deferring_state): Likewise.
7728 (cancel_fma_deferring): New function.
7729 (result_of_phi): Likewise.
7730 (last_fma_candidate_feeds_initial_phi): Likewise.
7731 (convert_mult_to_fma): Added deferring logic, split actual
7732 transformation to convert_mult_to_fma_1.
7733 (math_opts_dom_walker): New type.
7734 (math_opts_dom_walker::after_dom_children): New method, body moved
7735 here from pass_optimize_widening_mul::execute, added deferring logic
7737 (pass_optimize_widening_mul::execute): Moved most of code to
7738 math_opts_dom_walker::after_dom_children.
7739 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
7740 * config/i386/i386.c (ix86_option_override_internal): Added
7741 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
7743 2018-01-12 Richard Biener <rguenther@suse.de>
7746 * dwarf2out.c (gen_variable_die): Do not reset old_die for
7747 inline instance vars.
7749 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
7752 * config/rx/rx.c (rx_is_restricted_memory_address):
7755 2018-01-12 Richard Biener <rguenther@suse.de>
7757 PR tree-optimization/80846
7758 * target.def (split_reduction): New target hook.
7759 * targhooks.c (default_split_reduction): New function.
7760 * targhooks.h (default_split_reduction): Declare.
7761 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
7762 target requests first reduce vectors by combining low and high
7764 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
7765 (get_vectype_for_scalar_type_and_size): Export.
7766 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
7767 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
7768 * doc/tm.texi: Regenerate.
7769 * config/i386/i386.c (ix86_split_reduction): Implement
7770 TARGET_VECTORIZE_SPLIT_REDUCTION.
7772 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7775 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
7776 in PIC mode except for TARGET_VXWORKS_RTP.
7777 * config/sparc/sparc.c: Include cfgrtl.h.
7778 (TARGET_INIT_PIC_REG): Define.
7779 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
7780 (sparc_pic_register_p): New predicate.
7781 (sparc_legitimate_address_p): Use it.
7782 (sparc_legitimize_pic_address): Likewise.
7783 (sparc_delegitimize_address): Likewise.
7784 (sparc_mode_dependent_address_p): Likewise.
7785 (gen_load_pcrel_sym): Remove 4th parameter.
7786 (load_got_register): Adjust call to above. Remove obsolete stuff.
7787 (sparc_expand_prologue): Do not call load_got_register here.
7788 (sparc_flat_expand_prologue): Likewise.
7789 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
7790 (sparc_use_pseudo_pic_reg): New function.
7791 (sparc_init_pic_reg): Likewise.
7792 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7793 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
7795 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
7797 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
7798 Add item for branch_cost.
7800 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7802 PR rtl-optimization/83565
7803 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
7804 not extend the result to a larger mode for rotate operations.
7805 (num_sign_bit_copies1): Likewise.
7807 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7810 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
7812 Use values-Xc.o for -pedantic.
7813 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
7815 2018-01-12 Martin Liska <mliska@suse.cz>
7818 * ipa-devirt.c (final_warning_record::grow_type_warnings):
7820 (possible_polymorphic_call_targets): Use it.
7821 (ipa_devirt): Likewise.
7823 2018-01-12 Martin Liska <mliska@suse.cz>
7825 * profile-count.h (enum profile_quality): Use 0 as invalid
7826 enum value of profile_quality.
7828 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
7830 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
7831 -mext-string options.
7833 2018-01-12 Richard Biener <rguenther@suse.de>
7835 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
7836 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
7837 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
7839 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
7841 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
7843 * configure.ac (--with-long-double-format): Add support for the
7844 configuration option to change the default long double format on
7846 * config.gcc (powerpc*-linux*-*): Likewise.
7847 * configure: Regenerate.
7848 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
7849 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
7850 used without modification.
7852 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7854 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
7855 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
7856 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
7857 MISC_BUILTIN_SPEC_BARRIER.
7858 (rs6000_init_builtins): Likewise.
7859 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
7861 (speculation_barrier): New define_insn.
7862 * doc/extend.texi: Document __builtin_speculation_barrier.
7864 2018-01-11 Jakub Jelinek <jakub@redhat.com>
7867 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
7868 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
7869 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
7871 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
7872 integral modes instead of "ss" and "sd".
7873 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
7874 vectors with 32-bit and 64-bit elements.
7875 (vecdupssescalarmodesuffix): New mode attribute.
7876 (vec_dup<mode>): Use it.
7878 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
7881 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
7882 frame if argument is passed on stack.
7884 2018-01-11 Jakub Jelinek <jakub@redhat.com>
7887 * ree.c (combine_reaching_defs): Optimize also
7888 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
7889 reg2=any_extend(exp); reg1=reg2;, formatting fix.
7891 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
7894 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
7896 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
7899 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
7900 after they are computed.
7902 2018-01-11 Bin Cheng <bin.cheng@arm.com>
7904 PR tree-optimization/83695
7905 * gimple-loop-linterchange.cc
7906 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
7907 reset cached scev information after interchange.
7908 (pass_linterchange::execute): Remove call to scev_reset_htab.
7910 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7912 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
7913 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
7914 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
7915 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
7916 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
7917 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
7918 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
7919 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
7920 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
7921 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
7922 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
7923 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
7924 (V_lane_reg): Likewise.
7925 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
7927 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
7928 (vfmal_lane_low<mode>_intrinsic,
7929 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
7930 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
7931 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
7932 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
7933 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
7934 vfmsl_lane_high<mode>_intrinsic): New define_insns.
7936 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7938 * config/arm/arm-cpus.in (fp16fml): New feature.
7939 (ALL_SIMD): Add fp16fml.
7940 (armv8.2-a): Add fp16fml as an option.
7941 (armv8.3-a): Likewise.
7942 (armv8.4-a): Add fp16fml as part of fp16.
7943 * config/arm/arm.h (TARGET_FP16FML): Define.
7944 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
7946 * config/arm/arm-modes.def (V2HF): Define.
7947 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
7948 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
7949 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
7950 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
7951 vfmsl_low, vfmsl_high): New set of builtins.
7952 * config/arm/iterators.md (PLUSMINUS): New code iterator.
7953 (vfml_op): New code attribute.
7954 (VFMLHALVES): New int iterator.
7955 (VFML, VFMLSEL): New mode attributes.
7956 (V_reg): Define mapping for V2HF.
7957 (V_hi, V_lo): New mode attributes.
7958 (VF_constraint): Likewise.
7959 (vfml_half, vfml_half_selector): New int attributes.
7960 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
7962 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
7963 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
7965 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
7966 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
7967 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
7968 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
7970 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
7971 Document new effective target and option set.
7973 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7975 * config/arm/arm-cpus.in (armv8_4): New feature.
7976 (ARMv8_4a): New fgroup.
7977 (armv8.4-a): New arch.
7978 * config/arm/arm-tables.opt: Regenerate.
7979 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
7980 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
7981 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
7982 Add matching rules for -march=armv8.4-a and extensions.
7983 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
7985 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
7988 * config/rx/rx.md (BW): New mode attribute.
7989 (sync_lock_test_and_setsi): Add mode suffix to insn output.
7991 2018-01-11 Richard Biener <rguenther@suse.de>
7993 PR tree-optimization/83435
7994 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
7995 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
7996 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
7998 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7999 Alan Hayward <alan.hayward@arm.com>
8000 David Sherwood <david.sherwood@arm.com>
8002 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
8004 (aarch64_classify_address): Initialize it. Track polynomial offsets.
8005 (aarch64_print_address_internal): Use it to check for a zero offset.
8007 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8008 Alan Hayward <alan.hayward@arm.com>
8009 David Sherwood <david.sherwood@arm.com>
8011 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
8012 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
8013 Return a poly_int64 rather than a HOST_WIDE_INT.
8014 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
8015 rather than a HOST_WIDE_INT.
8016 * config/aarch64/aarch64.h (aarch64_frame): Protect with
8017 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
8018 hard_fp_offset, frame_size, initial_adjust, callee_offset and
8019 final_offset from HOST_WIDE_INT to poly_int64.
8020 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
8021 to_constant when getting the number of units in an Advanced SIMD
8023 (aarch64_builtin_vectorized_function): Check for a constant number
8025 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
8027 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
8028 attribute instead of GET_MODE_NUNITS.
8029 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
8030 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
8031 GET_MODE_SIZE for fixed-size registers.
8032 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
8033 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
8034 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
8035 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
8036 (aarch64_print_operand, aarch64_print_address_internal)
8037 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
8038 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
8039 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
8040 Handle polynomial GET_MODE_SIZE.
8041 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
8042 wider than SImode without modification.
8043 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
8044 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
8045 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
8046 passing and returning SVE modes.
8047 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
8048 rather than GEN_INT.
8049 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
8050 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
8051 (aarch64_allocate_and_probe_stack_space): Likewise.
8052 (aarch64_layout_frame): Cope with polynomial offsets.
8053 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
8054 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
8056 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
8057 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
8058 poly_int64 rather than a HOST_WIDE_INT.
8059 (aarch64_get_separate_components, aarch64_process_components)
8060 (aarch64_expand_prologue, aarch64_expand_epilogue)
8061 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
8062 (aarch64_anchor_offset): New function, split out from...
8063 (aarch64_legitimize_address): ...here.
8064 (aarch64_builtin_vectorization_cost): Handle polynomial
8065 TYPE_VECTOR_SUBPARTS.
8066 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
8068 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
8069 number of elements from the PARALLEL rather than the mode.
8070 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
8071 rather than GET_MODE_BITSIZE.
8072 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
8073 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
8074 (aarch64_expand_vec_perm_const_1): Handle polynomial
8075 d->perm.length () and d->perm elements.
8076 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
8077 Apply to_constant to d->perm elements.
8078 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
8079 polynomial CONST_VECTOR_NUNITS.
8080 (aarch64_move_pointer): Take amount as a poly_int64 rather
8082 (aarch64_progress_pointer): Avoid temporary variable.
8083 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
8084 the mode attribute instead of GET_MODE.
8086 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8087 Alan Hayward <alan.hayward@arm.com>
8088 David Sherwood <david.sherwood@arm.com>
8090 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
8091 x exists before using it.
8092 (aarch64_add_constant_internal): Rename to...
8093 (aarch64_add_offset_1): ...this. Replace regnum with separate
8094 src and dest rtxes. Handle the case in which they're different,
8095 including when the offset is zero. Replace scratchreg with an rtx.
8096 Use 2 additions if there is no spare register into which we can
8097 move a 16-bit constant.
8098 (aarch64_add_constant): Delete.
8099 (aarch64_add_offset): Replace reg with separate src and dest
8100 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
8101 Use aarch64_add_offset_1.
8102 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
8103 an rtx rather than an int. Take the delta as a poly_int64
8104 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
8105 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
8106 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
8107 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
8108 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
8110 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
8111 aarch64_add_constant.
8113 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8115 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
8116 Use scalar_float_mode.
8118 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8120 * config/aarch64/aarch64-simd.md
8121 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
8122 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
8123 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
8124 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
8125 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
8126 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
8127 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
8128 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
8129 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
8130 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
8132 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8135 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
8136 targ_options->x_arm_arch_string is non NULL.
8138 2018-01-11 Tamar Christina <tamar.christina@arm.com>
8140 * config/aarch64/aarch64.h
8141 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
8143 2018-01-11 Sudakshina Das <sudi.das@arm.com>
8146 * expmed.c (emit_store_flag_force): Swap if const op0
8147 and change VOIDmode to mode of op0.
8149 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8151 PR rtl-optimization/83761
8152 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
8153 than bytes to mode_for_size.
8155 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8158 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
8159 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
8162 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8165 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
8166 when in layout mode.
8167 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
8168 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
8171 2018-01-10 Michael Collison <michael.collison@arm.com>
8173 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
8174 * config/aarch64/aarch64-option-extension.def: Add
8175 AARCH64_OPT_EXTENSION of 'fp16fml'.
8176 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8177 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
8178 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
8179 * config/aarch64/constraints.md (Ui7): New constraint.
8180 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
8181 (VFMLA_SEL_W): Ditto.
8184 (VFMLA16_LOW): New int iterator.
8185 (VFMLA16_HIGH): Ditto.
8186 (UNSPEC_FMLAL): New unspec.
8187 (UNSPEC_FMLSL): Ditto.
8188 (UNSPEC_FMLAL2): Ditto.
8189 (UNSPEC_FMLSL2): Ditto.
8190 (f16mac): New code attribute.
8191 * config/aarch64/aarch64-simd-builtins.def
8192 (aarch64_fmlal_lowv2sf): Ditto.
8193 (aarch64_fmlsl_lowv2sf): Ditto.
8194 (aarch64_fmlalq_lowv4sf): Ditto.
8195 (aarch64_fmlslq_lowv4sf): Ditto.
8196 (aarch64_fmlal_highv2sf): Ditto.
8197 (aarch64_fmlsl_highv2sf): Ditto.
8198 (aarch64_fmlalq_highv4sf): Ditto.
8199 (aarch64_fmlslq_highv4sf): Ditto.
8200 (aarch64_fmlal_lane_lowv2sf): Ditto.
8201 (aarch64_fmlsl_lane_lowv2sf): Ditto.
8202 (aarch64_fmlal_laneq_lowv2sf): Ditto.
8203 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
8204 (aarch64_fmlalq_lane_lowv4sf): Ditto.
8205 (aarch64_fmlsl_lane_lowv4sf): Ditto.
8206 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
8207 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
8208 (aarch64_fmlal_lane_highv2sf): Ditto.
8209 (aarch64_fmlsl_lane_highv2sf): Ditto.
8210 (aarch64_fmlal_laneq_highv2sf): Ditto.
8211 (aarch64_fmlsl_laneq_highv2sf): Ditto.
8212 (aarch64_fmlalq_lane_highv4sf): Ditto.
8213 (aarch64_fmlsl_lane_highv4sf): Ditto.
8214 (aarch64_fmlalq_laneq_highv4sf): Ditto.
8215 (aarch64_fmlsl_laneq_highv4sf): Ditto.
8216 * config/aarch64/aarch64-simd.md:
8217 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
8218 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8219 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
8220 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8221 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
8222 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
8223 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
8224 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
8225 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
8226 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
8227 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
8228 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
8229 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
8230 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
8231 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
8232 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
8233 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
8234 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
8235 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
8236 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
8237 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
8238 (vfmlsl_low_u32): Ditto.
8239 (vfmlalq_low_u32): Ditto.
8240 (vfmlslq_low_u32): Ditto.
8241 (vfmlal_high_u32): Ditto.
8242 (vfmlsl_high_u32): Ditto.
8243 (vfmlalq_high_u32): Ditto.
8244 (vfmlslq_high_u32): Ditto.
8245 (vfmlal_lane_low_u32): Ditto.
8246 (vfmlsl_lane_low_u32): Ditto.
8247 (vfmlal_laneq_low_u32): Ditto.
8248 (vfmlsl_laneq_low_u32): Ditto.
8249 (vfmlalq_lane_low_u32): Ditto.
8250 (vfmlslq_lane_low_u32): Ditto.
8251 (vfmlalq_laneq_low_u32): Ditto.
8252 (vfmlslq_laneq_low_u32): Ditto.
8253 (vfmlal_lane_high_u32): Ditto.
8254 (vfmlsl_lane_high_u32): Ditto.
8255 (vfmlal_laneq_high_u32): Ditto.
8256 (vfmlsl_laneq_high_u32): Ditto.
8257 (vfmlalq_lane_high_u32): Ditto.
8258 (vfmlslq_lane_high_u32): Ditto.
8259 (vfmlalq_laneq_high_u32): Ditto.
8260 (vfmlslq_laneq_high_u32): Ditto.
8261 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
8262 (AARCH64_FL_FOR_ARCH8_4): New.
8263 (AARCH64_ISA_F16FML): New ISA flag.
8264 (TARGET_F16FML): New feature flag for fp16fml.
8265 (doc/invoke.texi): Document new fp16fml option.
8267 2018-01-10 Michael Collison <michael.collison@arm.com>
8269 * config/aarch64/aarch64-builtins.c:
8270 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
8271 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8272 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
8273 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
8274 (AARCH64_ISA_SHA3): New ISA flag.
8275 (TARGET_SHA3): New feature flag for sha3.
8276 * config/aarch64/iterators.md (sha512_op): New int attribute.
8277 (CRYPTO_SHA512): New int iterator.
8278 (UNSPEC_SHA512H): New unspec.
8279 (UNSPEC_SHA512H2): Ditto.
8280 (UNSPEC_SHA512SU0): Ditto.
8281 (UNSPEC_SHA512SU1): Ditto.
8282 * config/aarch64/aarch64-simd-builtins.def
8283 (aarch64_crypto_sha512hqv2di): New builtin.
8284 (aarch64_crypto_sha512h2qv2di): Ditto.
8285 (aarch64_crypto_sha512su0qv2di): Ditto.
8286 (aarch64_crypto_sha512su1qv2di): Ditto.
8287 (aarch64_eor3qv8hi): Ditto.
8288 (aarch64_rax1qv2di): Ditto.
8289 (aarch64_xarqv2di): Ditto.
8290 (aarch64_bcaxqv8hi): Ditto.
8291 * config/aarch64/aarch64-simd.md:
8292 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
8293 (aarch64_crypto_sha512su0qv2di): Ditto.
8294 (aarch64_crypto_sha512su1qv2di): Ditto.
8295 (aarch64_eor3qv8hi): Ditto.
8296 (aarch64_rax1qv2di): Ditto.
8297 (aarch64_xarqv2di): Ditto.
8298 (aarch64_bcaxqv8hi): Ditto.
8299 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
8300 (vsha512h2q_u64): Ditto.
8301 (vsha512su0q_u64): Ditto.
8302 (vsha512su1q_u64): Ditto.
8303 (veor3q_u16): Ditto.
8304 (vrax1q_u64): Ditto.
8306 (vbcaxq_u16): Ditto.
8307 * config/arm/types.md (crypto_sha512): New type attribute.
8308 (crypto_sha3): Ditto.
8309 (doc/invoke.texi): Document new sha3 option.
8311 2018-01-10 Michael Collison <michael.collison@arm.com>
8313 * config/aarch64/aarch64-builtins.c:
8314 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
8315 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8316 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
8317 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
8318 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
8319 (AARCH64_ISA_SM4): New ISA flag.
8320 (TARGET_SM4): New feature flag for sm4.
8321 * config/aarch64/aarch64-simd-builtins.def
8322 (aarch64_sm3ss1qv4si): Ditto.
8323 (aarch64_sm3tt1aq4si): Ditto.
8324 (aarch64_sm3tt1bq4si): Ditto.
8325 (aarch64_sm3tt2aq4si): Ditto.
8326 (aarch64_sm3tt2bq4si): Ditto.
8327 (aarch64_sm3partw1qv4si): Ditto.
8328 (aarch64_sm3partw2qv4si): Ditto.
8329 (aarch64_sm4eqv4si): Ditto.
8330 (aarch64_sm4ekeyqv4si): Ditto.
8331 * config/aarch64/aarch64-simd.md:
8332 (aarch64_sm3ss1qv4si): Ditto.
8333 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
8334 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
8335 (aarch64_sm4eqv4si): Ditto.
8336 (aarch64_sm4ekeyqv4si): Ditto.
8337 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
8338 (sm3part_op): Ditto.
8339 (CRYPTO_SM3TT): Ditto.
8340 (CRYPTO_SM3PART): Ditto.
8341 (UNSPEC_SM3SS1): New unspec.
8342 (UNSPEC_SM3TT1A): Ditto.
8343 (UNSPEC_SM3TT1B): Ditto.
8344 (UNSPEC_SM3TT2A): Ditto.
8345 (UNSPEC_SM3TT2B): Ditto.
8346 (UNSPEC_SM3PARTW1): Ditto.
8347 (UNSPEC_SM3PARTW2): Ditto.
8348 (UNSPEC_SM4E): Ditto.
8349 (UNSPEC_SM4EKEY): Ditto.
8350 * config/aarch64/constraints.md (Ui2): New constraint.
8351 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
8352 * config/arm/types.md (crypto_sm3): New type attribute.
8353 (crypto_sm4): Ditto.
8354 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
8355 (vsm3tt1aq_u32): Ditto.
8356 (vsm3tt1bq_u32): Ditto.
8357 (vsm3tt2aq_u32): Ditto.
8358 (vsm3tt2bq_u32): Ditto.
8359 (vsm3partw1q_u32): Ditto.
8360 (vsm3partw2q_u32): Ditto.
8361 (vsm4eq_u32): Ditto.
8362 (vsm4ekeyq_u32): Ditto.
8363 (doc/invoke.texi): Document new sm4 option.
8365 2018-01-10 Michael Collison <michael.collison@arm.com>
8367 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
8368 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
8369 (AARCH64_FL_FOR_ARCH8_4): New.
8370 (AARCH64_FL_V8_4): New flag.
8371 (doc/invoke.texi): Document new armv8.4-a option.
8373 2018-01-10 Michael Collison <michael.collison@arm.com>
8375 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8376 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
8377 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
8378 * config/aarch64/aarch64-option-extension.def: Add
8379 AARCH64_OPT_EXTENSION of 'sha2'.
8380 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
8381 (crypto): Disable sha2 and aes if crypto disabled.
8382 (crypto): Enable aes and sha2 if enabled.
8383 (simd): Disable sha2 and aes if simd disabled.
8384 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
8386 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
8387 (TARGET_SHA2): New feature flag for sha2.
8388 (TARGET_AES): New feature flag for aes.
8389 * config/aarch64/aarch64-simd.md:
8390 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
8391 conditional on TARGET_AES.
8392 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
8393 (aarch64_crypto_sha1hsi): Make pattern conditional
8395 (aarch64_crypto_sha1hv4si): Ditto.
8396 (aarch64_be_crypto_sha1hv4si): Ditto.
8397 (aarch64_crypto_sha1su1v4si): Ditto.
8398 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
8399 (aarch64_crypto_sha1su0v4si): Ditto.
8400 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
8401 (aarch64_crypto_sha256su0v4si): Ditto.
8402 (aarch64_crypto_sha256su1v4si): Ditto.
8403 (doc/invoke.texi): Document new aes and sha2 options.
8405 2018-01-10 Martin Sebor <msebor@redhat.com>
8407 PR tree-optimization/83781
8408 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
8411 2018-01-11 Martin Sebor <msebor@gmail.com>
8412 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8414 PR tree-optimization/83501
8415 PR tree-optimization/81703
8417 * tree-ssa-strlen.c (get_string_cst): Rename...
8418 (get_string_len): ...to this. Handle global constants.
8419 (handle_char_store): Adjust.
8421 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
8422 Jim Wilson <jimw@sifive.com>
8424 * config/riscv/riscv-protos.h (riscv_output_return): New.
8425 * config/riscv/riscv.c (struct machine_function): New naked_p field.
8426 (riscv_attribute_table, riscv_output_return),
8427 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
8428 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
8429 (riscv_compute_frame_info): Only compute frame->mask if not a naked
8431 (riscv_expand_prologue): Add early return for naked function.
8432 (riscv_expand_epilogue): Likewise.
8433 (riscv_function_ok_for_sibcall): Return false for naked function.
8434 (riscv_set_current_function): New.
8435 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
8436 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
8437 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
8438 * doc/extend.texi (RISC-V Function Attributes): New.
8440 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
8442 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
8443 check for 128-bit long double before checking TCmode.
8444 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
8445 128-bit long doubles before checking TFmode or TCmode.
8446 (FLOAT128_IBM_P): Likewise.
8448 2018-01-10 Martin Sebor <msebor@redhat.com>
8450 PR tree-optimization/83671
8451 * builtins.c (c_strlen): Unconditionally return zero for the empty
8453 Use -Warray-bounds for warnings.
8454 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
8455 for non-constant array indices with COMPONENT_REF, arrays of
8456 arrays, and pointers to arrays.
8457 (gimple_fold_builtin_strlen): Determine and set length range for
8458 non-constant character arrays.
8460 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
8463 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
8466 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
8468 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
8470 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8473 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
8474 VECTOR_MEM_ALTIVEC_OR_VSX_P.
8475 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
8476 indexed_or_indirect_operand predicate.
8477 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
8478 (*vsx_le_perm_load_v8hi): Likewise.
8479 (*vsx_le_perm_load_v16qi): Likewise.
8480 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
8481 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
8482 (*vsx_le_perm_store_v8hi): Likewise.
8483 (*vsx_le_perm_store_v16qi): Likewise.
8484 (eight unnamed splitters): Likewise.
8486 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8488 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
8489 * config/rs6000/emmintrin.h: Likewise.
8490 * config/rs6000/mmintrin.h: Likewise.
8491 * config/rs6000/xmmintrin.h: Likewise.
8493 2018-01-10 David Malcolm <dmalcolm@redhat.com>
8496 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
8498 * tree.c (tree_nop_conversion): Return true for location wrapper
8500 (maybe_wrap_with_location): New function.
8501 (selftest::check_strip_nops): New function.
8502 (selftest::test_location_wrappers): New function.
8503 (selftest::tree_c_tests): Call it.
8504 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
8505 (maybe_wrap_with_location): New decl.
8506 (EXPR_LOCATION_WRAPPER_P): New macro.
8507 (location_wrapper_p): New inline function.
8508 (tree_strip_any_location_wrapper): New inline function.
8510 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
8513 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
8514 stack_realign_offset for the largest alignment of stack slot
8516 (ix86_find_max_used_stack_alignment): New function.
8517 (ix86_finalize_stack_frame_flags): Use it. Set
8518 max_used_stack_alignment if we don't realign stack.
8519 * config/i386/i386.h (machine_function): Add
8520 max_used_stack_alignment.
8522 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
8524 * config/arm/arm.opt (-mbranch-cost): New option.
8525 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
8528 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
8531 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
8532 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
8534 2018-01-10 Richard Biener <rguenther@suse.de>
8537 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
8538 early out so it also covers the case where we have a non-NULL
8541 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8543 PR tree-optimization/83753
8544 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
8545 for non-strided grouped accesses if the number of elements is 1.
8547 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8550 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
8551 * i386.h (TARGET_USE_GATHER): Define.
8552 * x86-tune.def (X86_TUNE_USE_GATHER): New.
8554 2018-01-10 Martin Liska <mliska@suse.cz>
8557 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
8558 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
8560 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
8561 CLEANUP_NO_PARTITIONING is not set.
8563 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8565 * doc/rtl.texi: Remove documentation of (const ...) wrappers
8566 for vectors, as a partial revert of r254296.
8567 * rtl.h (const_vec_p): Delete.
8568 (const_vec_duplicate_p): Don't test for vector CONSTs.
8569 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
8570 * expmed.c (make_tree): Likewise.
8573 * common.md (E, F): Use CONSTANT_P instead of checking for
8575 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
8576 checking for CONST_VECTOR.
8578 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8581 * predict.c (force_edge_cold): Handle in more sane way edges
8584 2018-01-09 Carl Love <cel@us.ibm.com>
8586 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
8588 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
8589 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
8590 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
8591 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
8592 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
8593 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
8594 * config/rs6000/rs6000-protos.h: Add extern defition for
8595 rs6000_generate_float2_double_code.
8596 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
8598 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
8599 (float2_v2df): Add define_expand.
8601 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
8604 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
8605 op_mode in the force_to_mode call.
8607 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8609 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
8610 instead of checking each element individually.
8611 (aarch64_evpc_uzp): Likewise.
8612 (aarch64_evpc_zip): Likewise.
8613 (aarch64_evpc_ext): Likewise.
8614 (aarch64_evpc_rev): Likewise.
8615 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
8616 instead of checking each element individually. Return true without
8618 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
8619 whether all selected elements come from the same input, instead of
8620 checking each element individually. Remove calls to gen_rtx_REG,
8621 start_sequence and end_sequence and instead assert that no rtl is
8624 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8626 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
8627 order of HIGH and CONST checks.
8629 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8631 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
8632 if the destination isn't an SSA_NAME.
8634 2018-01-09 Richard Biener <rguenther@suse.de>
8636 PR tree-optimization/83668
8637 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
8639 (canonicalize_loop_form): ... here, renamed from ...
8640 (canonicalize_loop_closed_ssa_form): ... this and amended to
8641 swap successor edges for loop exit blocks to make us use
8642 the RPO order we need for initial schedule generation.
8644 2018-01-09 Joseph Myers <joseph@codesourcery.com>
8646 PR tree-optimization/64811
8647 * match.pd: When optimizing comparisons with Inf, avoid
8648 introducing or losing exceptions from comparisons with NaN.
8650 2018-01-09 Martin Liska <mliska@suse.cz>
8653 * asan.c (shadow_mem_size): Add gcc_assert.
8655 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
8657 Don't save registers in main().
8660 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
8661 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
8662 * config/avr/avr.c (avr_set_current_function): Don't error if
8663 naked, OS_task or OS_main are specified at the same time.
8664 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
8666 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
8668 * common/config/avr/avr-common.c (avr_option_optimization_table):
8669 Switch on -mmain-is-OS_task for optimizing compilations.
8671 2018-01-09 Richard Biener <rguenther@suse.de>
8673 PR tree-optimization/83572
8674 * graphite.c: Include cfganal.h.
8675 (graphite_transform_loops): Connect infinite loops to exit
8676 and remove fake edges at the end.
8678 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8680 * ipa-inline.c (edge_badness): Revert accidental checkin.
8682 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8685 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
8686 symbols; not inline clones.
8688 2018-01-09 Jakub Jelinek <jakub@redhat.com>
8691 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
8692 hard registers. Formatting fixes.
8694 PR preprocessor/83722
8695 * gcc.c (try_generate_repro): Pass
8696 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
8697 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
8700 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
8701 Kito Cheng <kito.cheng@gmail.com>
8703 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
8704 (riscv_leaf_function_p): Delete.
8705 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
8707 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8709 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
8711 (do_ifelse): New function.
8712 (do_isel): New function.
8713 (do_sub3): New function.
8714 (do_add3): New function.
8715 (do_load_mask_compare): New function.
8716 (do_overlap_load_compare): New function.
8717 (expand_compare_loop): New function.
8718 (expand_block_compare): Call expand_compare_loop() when appropriate.
8719 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
8721 (-mblock-compare-inline-loop-limit): New option.
8723 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8726 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
8727 Reverse order of second and third operands in first alternative.
8728 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
8729 of first and second elements in UNSPEC_VPERMR vector.
8730 (altivec_expand_vec_perm_le): Likewise.
8732 2018-01-08 Jeff Law <law@redhat.com>
8734 PR rtl-optimizatin/81308
8735 * tree-switch-conversion.c (cfg_altered): New file scoped static.
8736 (process_switch): If group_case_labels makes a change, then set
8738 (pass_convert_switch::execute): If a switch is converted, then
8739 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
8741 PR rtl-optimization/81308
8742 * recog.c (split_all_insns): Conditionally cleanup the CFG after
8745 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
8747 PR target/83663 - Revert r255946
8748 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
8749 generation for cases where splatting a value is not useful.
8750 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
8751 across a vec_duplicate and a paradoxical subreg forming a vector
8752 mode to a vec_concat.
8754 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8756 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
8757 -march=armv8.3-a variants.
8758 * config/arm/t-multilib: Likewise.
8759 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
8761 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8763 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
8765 (cceq_ior_compare_complement): Give it a name so I can use it, and
8766 change boolean_or_operator predicate to boolean_operator so it can
8767 be used to generate a crand.
8768 (eqne): New code iterator.
8769 (bd/bd_neg): New code_attrs.
8770 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
8771 a single define_insn.
8772 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
8773 decrement (bdnzt/bdnzf/bdzt/bdzf).
8774 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
8775 with the new names of the branch decrement patterns, and added the
8776 names of the branch decrement conditional patterns.
8778 2018-01-08 Richard Biener <rguenther@suse.de>
8780 PR tree-optimization/83563
8781 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
8784 2018-01-08 Richard Biener <rguenther@suse.de>
8787 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
8789 2018-01-08 Richard Biener <rguenther@suse.de>
8791 PR tree-optimization/83685
8792 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
8793 references to abnormals.
8795 2018-01-08 Richard Biener <rguenther@suse.de>
8798 * dwarf2out.c (output_indirect_strings): Handle empty
8799 skeleton_debug_str_hash.
8800 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
8802 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
8804 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
8805 (emit_store_direct): Likewise.
8806 (arc_trampoline_adjust_address): Likewise.
8807 (arc_asm_trampoline_template): New function.
8808 (arc_initialize_trampoline): Use asm_trampoline_template.
8809 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
8810 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
8811 * config/arc/arc.md (flush_icache): Delete pattern.
8813 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
8815 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
8816 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
8819 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
8822 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
8823 by not USED_FOR_TARGET.
8824 (make_pass_resolve_sw_modes): Likewise.
8826 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
8828 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
8831 2018-01-08 Richard Biener <rguenther@suse.de>
8834 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
8836 2018-01-08 Richard Biener <rguenther@suse.de>
8839 * match.pd ((t * 2) / 2) -> t): Add missing :c.
8841 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
8844 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
8845 basic blocks with a small number of successors.
8846 (convert_control_dep_chain_into_preds): Improve handling of
8848 (dump_predicates): Split apart into...
8849 (dump_pred_chain): ...here...
8850 (dump_pred_info): ...and here.
8851 (can_one_predicate_be_invalidated_p): Add debugging printfs.
8852 (can_chain_union_be_invalidated_p): Improve check for invalidation
8854 (uninit_uses_cannot_happen): Avoid unnecessary if
8855 convert_control_dep_chain_into_preds yielded nothing.
8857 2018-01-06 Martin Sebor <msebor@redhat.com>
8859 PR tree-optimization/83640
8860 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
8861 subtracting negative offset from size.
8862 (builtin_access::overlap): Adjust offset bounds of the access to fall
8863 within the size of the object if possible.
8865 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
8867 PR rtl-optimization/83699
8868 * expmed.c (extract_bit_field_1): Restrict the vector usage of
8869 extract_bit_field_as_subreg to cases in which the extracted
8870 value is also a vector.
8872 * lra-constraints.c (process_alt_operands): Test for the equivalence
8873 substitutions when detecting a possible reload cycle.
8875 2018-01-06 Jakub Jelinek <jakub@redhat.com>
8878 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
8879 by default if flag_selective_schedling{,2}. Formatting fixes.
8881 PR rtl-optimization/83682
8882 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
8883 if it has non-VECTOR_MODE element mode.
8884 (vec_duplicate_p): Likewise.
8887 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
8888 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
8890 2018-01-05 Jakub Jelinek <jakub@redhat.com>
8893 * config/i386/i386-builtin.def
8894 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
8895 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
8896 Require also OPTION_MASK_ISA_AVX512F in addition to
8897 OPTION_MASK_ISA_GFNI.
8898 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
8899 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
8900 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
8901 to OPTION_MASK_ISA_GFNI.
8902 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
8903 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
8904 OPTION_MASK_ISA_AVX512BW.
8905 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
8906 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
8907 addition to OPTION_MASK_ISA_GFNI.
8908 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
8909 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
8910 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
8911 to OPTION_MASK_ISA_GFNI.
8912 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
8913 a requirement for all ISAs rather than any of them with a few
8915 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
8917 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
8918 bitmasks to be enabled with 3 exceptions, instead of requiring any
8919 enabled ISA with lots of exceptions.
8920 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
8921 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
8922 Change avx512bw in isa attribute to avx512f.
8923 * config/i386/sgxintrin.h: Add license boilerplate.
8924 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
8925 to __AVX512F__ and __AVX512VL to __AVX512VL__.
8926 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
8927 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
8929 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
8930 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
8931 temporarily sse2 rather than sse if not enabled already.
8934 * config/i386/sse.md (VI248_VLBW): Rename to ...
8935 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
8936 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
8937 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
8938 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
8939 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
8940 mode iterator instead of VI248_VLBW.
8942 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
8944 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
8945 (record_modified): Skip clobbers; add debug output.
8946 (param_change_prob): Use sreal frequencies.
8948 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8950 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
8951 punt for user-aligned variables.
8953 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8955 * tree-chrec.c (chrec_contains_symbols): Return true for
8958 2018-01-05 Sudakshina Das <sudi.das@arm.com>
8961 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
8962 of (x|y) == x for BICS pattern.
8964 2018-01-05 Jakub Jelinek <jakub@redhat.com>
8966 PR tree-optimization/83605
8967 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
8968 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
8971 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
8973 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
8974 * config/epiphany/rtems.h: New file.
8976 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8977 Uros Bizjak <ubizjak@gmail.com>
8980 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
8981 QIreg_operand instead of register_operand predicate.
8982 * config/i386/i386.c (ix86_rop_should_change_byte_p,
8983 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
8984 comments instead of -fmitigate[-_]rop.
8986 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8989 * cgraphunit.c (symbol_table::compile): Switch to text_section
8990 before calling assembly_start debug hook.
8991 * run-rtl-passes.c (run_rtl_passes): Likewise.
8994 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8996 * tree-vrp.c (extract_range_from_binary_expr_1): Check
8997 range_int_cst_p rather than !symbolic_range_p before calling
8998 extract_range_from_multiplicative_op_1.
9000 2018-01-04 Jeff Law <law@redhat.com>
9002 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
9003 redundant test in assertion.
9005 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9007 * doc/rtl.texi: Document machine_mode wrapper classes.
9009 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9011 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
9014 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9016 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
9017 the VEC_PERM_EXPR fold to fail.
9019 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9022 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
9023 to switched_sections.
9025 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9028 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
9031 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
9034 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
9035 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
9037 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9040 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
9041 is BLKmode and bitpos not zero or mode change is needed.
9043 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9046 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
9049 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
9052 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
9053 instead of MULT rtx. Update all corresponding splitters.
9055 (*ssub<modesuffix>): Ditto.
9057 (*cmp_sadd_di): Update split patterns.
9058 (*cmp_sadd_si): Ditto.
9059 (*cmp_sadd_sidi): Ditto.
9060 (*cmp_ssub_di): Ditto.
9061 (*cmp_ssub_si): Ditto.
9062 (*cmp_ssub_sidi): Ditto.
9063 * config/alpha/predicates.md (const23_operand): New predicate.
9064 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
9065 Look for ASHIFT, not MULT inner operand.
9066 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
9068 2018-01-04 Martin Liska <mliska@suse.cz>
9070 PR gcov-profile/83669
9071 * gcov.c (output_intermediate_file): Add version to intermediate
9073 * doc/gcov.texi: Document new field 'version' in intermediate
9074 file format. Fix location of '-k' option of gcov command.
9076 2018-01-04 Martin Liska <mliska@suse.cz>
9079 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
9081 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9083 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
9085 2018-01-03 Martin Sebor <msebor@redhat.com>
9087 PR tree-optimization/83655
9088 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
9089 checking calls with invalid arguments.
9091 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9093 * tree-vect-stmts.c (vect_get_store_rhs): New function.
9094 (vectorizable_mask_load_store): Delete.
9095 (vectorizable_call): Return false for masked loads and stores.
9096 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
9097 instead of gimple_assign_rhs1.
9098 (vectorizable_load): Handle IFN_MASK_LOAD.
9099 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
9101 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9103 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
9105 (vectorizable_mask_load_store): ...here.
9106 (vectorizable_load): ...and here.
9108 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9110 * tree-vect-stmts.c (vect_build_all_ones_mask)
9111 (vect_build_zero_merge_argument): New functions, split out from...
9112 (vectorizable_load): ...here.
9114 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9116 * tree-vect-stmts.c (vect_check_store_rhs): New function,
9118 (vectorizable_mask_load_store): ...here.
9119 (vectorizable_store): ...and here.
9121 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9123 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
9125 (vectorizable_mask_load_store): ...here.
9127 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9129 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
9130 (vect_model_store_cost): Take a vec_load_store_type instead of a
9132 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
9133 (vect_model_store_cost): Take a vec_load_store_type instead of a
9135 (vectorizable_mask_load_store): Update accordingly.
9136 (vectorizable_store): Likewise.
9137 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
9139 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9141 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
9142 IFN_MASK_LOAD calls here rather than...
9143 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
9145 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9146 Alan Hayward <alan.hayward@arm.com>
9147 David Sherwood <david.sherwood@arm.com>
9149 * expmed.c (extract_bit_field_1): For vector extracts,
9150 fall back to extract_bit_field_as_subreg if vec_extract
9153 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9154 Alan Hayward <alan.hayward@arm.com>
9155 David Sherwood <david.sherwood@arm.com>
9157 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
9158 they are variable or constant sized.
9159 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
9160 slots for constant-sized data.
9162 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9163 Alan Hayward <alan.hayward@arm.com>
9164 David Sherwood <david.sherwood@arm.com>
9166 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
9167 handling COND_EXPRs with boolean comparisons, try to find a better
9168 basis for the mask type than the boolean itself.
9170 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9172 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
9173 is calculated and how it can be overridden.
9174 * genmodes.c (max_bitsize_mode_any_mode): New variable.
9175 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
9177 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
9180 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9181 Alan Hayward <alan.hayward@arm.com>
9182 David Sherwood <david.sherwood@arm.com>
9184 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
9185 Remove the mode argument.
9186 (aarch64_simd_valid_immediate): Remove the mode and inverse
9188 * config/aarch64/iterators.md (bitsize): New iterator.
9189 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
9190 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
9191 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
9192 aarch64_simd_valid_immediate.
9193 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
9194 (aarch64_reg_or_bic_imm): Likewise.
9195 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
9196 with an insn_type enum and msl with a modifier_type enum.
9197 Replace element_width with a scalar_mode. Change the shift
9198 to unsigned int. Add constructors for scalar_float_mode and
9199 scalar_int_mode elements.
9200 (aarch64_vect_float_const_representable_p): Delete.
9201 (aarch64_can_const_movi_rtx_p)
9202 (aarch64_simd_scalar_immediate_valid_for_move)
9203 (aarch64_simd_make_constant): Update call to
9204 aarch64_simd_valid_immediate.
9205 (aarch64_advsimd_valid_immediate_hs): New function.
9206 (aarch64_advsimd_valid_immediate): Likewise.
9207 (aarch64_simd_valid_immediate): Remove mode and inverse
9208 arguments. Rewrite to use the above. Use const_vec_duplicate_p
9209 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
9210 and aarch64_float_const_representable_p on the result.
9211 (aarch64_output_simd_mov_immediate): Remove mode argument.
9212 Update call to aarch64_simd_valid_immediate and use of
9213 simd_immediate_info.
9214 (aarch64_output_scalar_simd_mov_immediate): Update call
9217 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9218 Alan Hayward <alan.hayward@arm.com>
9219 David Sherwood <david.sherwood@arm.com>
9221 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
9222 (mode_nunits): Likewise CONST_MODE_NUNITS.
9223 * machmode.def (ADJUST_NUNITS): Document.
9224 * genmodes.c (mode_data::need_nunits_adj): New field.
9225 (blank_mode): Update accordingly.
9226 (adj_nunits): New variable.
9227 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
9229 (emit_mode_size_inline): Set need_bytesize_adj for all modes
9230 listed in adj_nunits.
9231 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
9232 listed in adj_nunits. Don't emit case statements for such modes.
9233 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
9234 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
9235 nothing if adj_nunits is nonnull.
9236 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
9237 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
9238 (emit_mode_fbit): Update use of print_maybe_const_decl.
9239 (emit_move_size): Likewise. Treat the array as non-const
9241 (emit_mode_adjustments): Handle adj_nunits.
9243 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9245 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
9246 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
9247 (VECTOR_MODES): Use it.
9248 (make_vector_modes): Take the prefix as an argument.
9250 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9251 Alan Hayward <alan.hayward@arm.com>
9252 David Sherwood <david.sherwood@arm.com>
9254 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
9255 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
9256 for MODE_VECTOR_BOOL.
9257 * machmode.def (VECTOR_BOOL_MODE): Document.
9258 * genmodes.c (VECTOR_BOOL_MODE): New macro.
9259 (make_vector_bool_mode): New function.
9260 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
9262 * lto-streamer-in.c (lto_input_mode_table): Likewise.
9263 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
9265 * stor-layout.c (int_mode_for_mode): Likewise.
9266 * tree.c (build_vector_type_for_mode): Likewise.
9267 * varasm.c (output_constant_pool_2): Likewise.
9268 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
9269 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
9270 for MODE_VECTOR_BOOL.
9271 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
9272 of mode class checks.
9273 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
9274 instead of a list of mode class checks.
9275 (expand_vector_scalar_condition): Likewise.
9276 (type_for_widest_vector_mode): Handle BImode as an inner mode.
9278 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9279 Alan Hayward <alan.hayward@arm.com>
9280 David Sherwood <david.sherwood@arm.com>
9282 * machmode.h (mode_size): Change from unsigned short to
9284 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
9285 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9286 or if measurement_type is not polynomial.
9287 (fixed_size_mode::includes_p): Check for constant-sized modes.
9288 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
9289 return a poly_uint16 rather than an unsigned short.
9290 (emit_mode_size): Change the type of mode_size from unsigned short
9291 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
9292 (emit_mode_adjustments): Cope with polynomial vector sizes.
9293 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9295 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9297 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
9298 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
9299 * caller-save.c (setup_save_areas): Likewise.
9300 (replace_reg_with_saved_mem): Likewise.
9301 * calls.c (emit_library_call_value_1): Likewise.
9302 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
9303 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
9304 (gen_lowpart_for_combine): Likewise.
9305 * convert.c (convert_to_integer_1): Likewise.
9306 * cse.c (equiv_constant, cse_insn): Likewise.
9307 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
9308 (cselib_subst_to_values): Likewise.
9309 * dce.c (word_dce_process_block): Likewise.
9310 * df-problems.c (df_word_lr_mark_ref): Likewise.
9311 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
9312 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
9313 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
9314 (rtl_for_decl_location): Likewise.
9315 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
9316 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
9317 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
9318 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
9319 (expand_expr_real_1): Likewise.
9320 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
9321 (pad_below): Likewise.
9322 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9323 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
9324 * ira.c (get_subreg_tracking_sizes): Likewise.
9325 * ira-build.c (ira_create_allocno_objects): Likewise.
9326 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
9327 (ira_sort_regnos_for_alter_reg): Likewise.
9328 * ira-costs.c (record_operand_costs): Likewise.
9329 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
9330 (resolve_simple_move): Likewise.
9331 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
9332 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
9333 (lra_constraints): Likewise.
9334 (CONST_POOL_OK_P): Reject variable-sized modes.
9335 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
9336 (add_pseudo_to_slot, lra_spill): Likewise.
9337 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9338 * optabs-query.c (get_best_extraction_insn): Likewise.
9339 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9340 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
9341 (expand_mult_highpart, valid_multiword_target_p): Likewise.
9342 * recog.c (offsettable_address_addr_space_p): Likewise.
9343 * regcprop.c (maybe_mode_change): Likewise.
9344 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
9345 * regrename.c (build_def_use): Likewise.
9346 * regstat.c (dump_reg_info): Likewise.
9347 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
9348 (find_reloads, find_reloads_subreg_address): Likewise.
9349 * reload1.c (eliminate_regs_1): Likewise.
9350 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
9351 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
9352 (simplify_binary_operation_1, simplify_subreg): Likewise.
9353 * targhooks.c (default_function_arg_padding): Likewise.
9354 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
9355 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
9356 (verify_gimple_assign_ternary): Likewise.
9357 * tree-inline.c (estimate_move_cost): Likewise.
9358 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9359 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
9360 (get_address_cost_ainc): Likewise.
9361 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
9362 (vect_supportable_dr_alignment): Likewise.
9363 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9364 (vectorizable_reduction): Likewise.
9365 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
9366 (vectorizable_operation, vectorizable_load): Likewise.
9367 * tree.c (build_same_sized_truth_vector_type): Likewise.
9368 * valtrack.c (cleanup_auto_inc_dec): Likewise.
9369 * var-tracking.c (emit_note_insn_var_location): Likewise.
9370 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
9371 (ADDR_VEC_ALIGN): Likewise.
9373 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9374 Alan Hayward <alan.hayward@arm.com>
9375 David Sherwood <david.sherwood@arm.com>
9377 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
9379 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9380 or if measurement_type is polynomial.
9381 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
9382 * combine.c (make_extraction): Likewise.
9383 * dse.c (find_shift_sequence): Likewise.
9384 * dwarf2out.c (mem_loc_descriptor): Likewise.
9385 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
9386 (extract_bit_field, extract_low_bits): Likewise.
9387 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
9388 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
9389 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
9390 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
9391 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9392 * reload.c (find_reloads): Likewise.
9393 * reload1.c (alter_reg): Likewise.
9394 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
9395 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
9396 * tree-if-conv.c (predicate_mem_writes): Likewise.
9397 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
9398 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
9399 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
9400 * valtrack.c (dead_debug_insert_temp): Likewise.
9401 * varasm.c (mergeable_constant_section): Likewise.
9402 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
9404 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9405 Alan Hayward <alan.hayward@arm.com>
9406 David Sherwood <david.sherwood@arm.com>
9408 * expr.c (expand_assignment): Cope with polynomial mode sizes
9409 when assigning to a CONCAT.
9411 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9412 Alan Hayward <alan.hayward@arm.com>
9413 David Sherwood <david.sherwood@arm.com>
9415 * machmode.h (mode_precision): Change from unsigned short to
9417 (mode_to_precision): Return a poly_uint16 rather than an unsigned
9419 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
9420 or if measurement_type is not polynomial.
9421 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
9422 in which the mode is already known to be a scalar_int_mode.
9423 * genmodes.c (emit_mode_precision): Change the type of mode_precision
9424 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
9426 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9427 for GET_MODE_PRECISION.
9428 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9429 for GET_MODE_PRECISION.
9430 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
9432 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
9433 (expand_field_assignment, make_extraction): Likewise.
9434 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
9435 (get_last_value): Likewise.
9436 * convert.c (convert_to_integer_1): Likewise.
9437 * cse.c (cse_insn): Likewise.
9438 * expr.c (expand_expr_real_1): Likewise.
9439 * lra-constraints.c (simplify_operand_subreg): Likewise.
9440 * optabs-query.c (can_atomic_load_p): Likewise.
9441 * optabs.c (expand_atomic_load): Likewise.
9442 (expand_atomic_store): Likewise.
9443 * ree.c (combine_reaching_defs): Likewise.
9444 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
9445 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
9446 * tree.h (type_has_mode_precision_p): Likewise.
9447 * ubsan.c (instrument_si_overflow): Likewise.
9449 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9450 Alan Hayward <alan.hayward@arm.com>
9451 David Sherwood <david.sherwood@arm.com>
9453 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
9454 polynomial numbers of units.
9455 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
9456 (valid_vector_subparts_p): New function.
9457 (build_vector_type): Remove temporary shim and take the number
9458 of units as a poly_uint64 rather than an int.
9459 (build_opaque_vector_type): Take the number of units as a
9460 poly_uint64 rather than an int.
9461 * tree.c (build_vector_from_ctor): Handle polynomial
9462 TYPE_VECTOR_SUBPARTS.
9463 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
9464 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
9465 (build_vector_from_val): If the number of units is variable,
9466 use build_vec_duplicate_cst for constant operands and
9467 VEC_DUPLICATE_EXPR otherwise.
9468 (make_vector_type): Remove temporary is_constant ().
9469 (build_vector_type, build_opaque_vector_type): Take the number of
9470 units as a poly_uint64 rather than an int.
9471 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
9473 * cfgexpand.c (expand_debug_expr): Likewise.
9474 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
9475 (store_constructor, expand_expr_real_1): Likewise.
9476 (const_scalar_mask_from_tree): Likewise.
9477 * fold-const-call.c (fold_const_reduction): Likewise.
9478 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
9479 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
9480 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
9481 (fold_relational_const): Likewise.
9482 (native_interpret_vector): Likewise. Change the size from an
9483 int to an unsigned int.
9484 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
9485 TYPE_VECTOR_SUBPARTS.
9486 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
9487 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
9488 duplicating a non-constant operand into a variable-length vector.
9489 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
9490 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
9491 * ipa-icf.c (sem_variable::equals): Likewise.
9492 * match.pd: Likewise.
9493 * omp-simd-clone.c (simd_clone_subparts): Likewise.
9494 * print-tree.c (print_node): Likewise.
9495 * stor-layout.c (layout_type): Likewise.
9496 * targhooks.c (default_builtin_vectorization_cost): Likewise.
9497 * tree-cfg.c (verify_gimple_comparison): Likewise.
9498 (verify_gimple_assign_binary): Likewise.
9499 (verify_gimple_assign_ternary): Likewise.
9500 (verify_gimple_assign_single): Likewise.
9501 * tree-pretty-print.c (dump_generic_node): Likewise.
9502 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9503 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
9504 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
9505 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
9506 (vect_shift_permute_load_chain): Likewise.
9507 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
9508 (expand_vector_condition, optimize_vector_constructor): Likewise.
9509 (lower_vec_perm, get_compute_type): Likewise.
9510 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9511 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
9512 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
9513 (vect_recog_mask_conversion_pattern): Likewise.
9514 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
9515 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
9516 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9517 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
9518 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
9519 (vectorizable_shift, vectorizable_operation, vectorizable_store)
9520 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
9521 (supportable_widening_operation): Likewise.
9522 (supportable_narrowing_operation): Likewise.
9523 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
9525 * varasm.c (output_constant): Likewise.
9527 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9528 Alan Hayward <alan.hayward@arm.com>
9529 David Sherwood <david.sherwood@arm.com>
9531 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
9532 so that both the length == 3 and length != 3 cases set up their
9533 own permute vectors. Add comments explaining why we know the
9534 number of elements is constant.
9535 (vect_permute_load_chain): Likewise.
9537 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9538 Alan Hayward <alan.hayward@arm.com>
9539 David Sherwood <david.sherwood@arm.com>
9541 * machmode.h (mode_nunits): Change from unsigned char to
9543 (ONLY_FIXED_SIZE_MODES): New macro.
9544 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
9545 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
9546 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
9548 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
9549 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
9550 or if measurement_type is not polynomial.
9551 * genmodes.c (ZERO_COEFFS): New macro.
9552 (emit_mode_nunits_inline): Make mode_nunits_inline return a
9554 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
9555 Use ZERO_COEFFS when emitting initializers.
9556 * data-streamer.h (bp_pack_poly_value): New function.
9557 (bp_unpack_poly_value): Likewise.
9558 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9559 for GET_MODE_NUNITS.
9560 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9561 for GET_MODE_NUNITS.
9562 * tree.c (make_vector_type): Remove temporary shim and make
9563 the real function take the number of units as a poly_uint64
9565 (build_vector_type_for_mode): Handle polynomial nunits.
9566 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
9567 * emit-rtl.c (const_vec_series_p_1): Likewise.
9568 (gen_rtx_CONST_VECTOR): Likewise.
9569 * fold-const.c (test_vec_duplicate_folding): Likewise.
9570 * genrecog.c (validate_pattern): Likewise.
9571 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
9572 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9573 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
9574 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
9575 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
9576 * rtlanal.c (subreg_get_info): Likewise.
9577 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9578 (vect_grouped_load_supported): Likewise.
9579 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
9580 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
9581 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
9582 (simplify_const_unary_operation, simplify_binary_operation_1)
9583 (simplify_const_binary_operation, simplify_ternary_operation)
9584 (test_vector_ops_duplicate, test_vector_ops): Likewise.
9585 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
9586 instead of CONST_VECTOR_NUNITS.
9587 * varasm.c (output_constant_pool_2): Likewise.
9588 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
9589 explicit-encoded elements in the XVEC for variable-length vectors.
9591 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9593 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
9595 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9596 Alan Hayward <alan.hayward@arm.com>
9597 David Sherwood <david.sherwood@arm.com>
9599 * coretypes.h (fixed_size_mode): Declare.
9600 (fixed_size_mode_pod): New typedef.
9601 * builtins.h (target_builtins::x_apply_args_mode)
9602 (target_builtins::x_apply_result_mode): Change type to
9603 fixed_size_mode_pod.
9604 * builtins.c (apply_args_size, apply_result_size, result_vector)
9605 (expand_builtin_apply_args_1, expand_builtin_apply)
9606 (expand_builtin_return): Update accordingly.
9608 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9610 * cse.c (hash_rtx_cb): Hash only the encoded elements.
9611 * cselib.c (cselib_hash_rtx): Likewise.
9612 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
9613 CONST_VECTOR encoding.
9615 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9616 Jeff Law <law@redhat.com>
9619 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
9620 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
9621 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
9622 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
9625 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
9626 explicitly probe *sp in a noreturn function if there were any callee
9627 register saves or frame pointer is needed.
9629 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9632 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
9633 BLKmode for ternary, binary or unary expressions.
9636 * var-tracking.c (delete_vta_debug_insn): New inline function.
9637 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
9638 insns from get_insns () to NULL instead of each bb separately.
9639 Use delete_vta_debug_insn. No longer static.
9640 (vt_debug_insns_local, variable_tracking_main_1): Adjust
9641 delete_vta_debug_insns callers.
9642 * rtl.h (delete_vta_debug_insns): Declare.
9643 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
9644 instead of variable_tracking_main.
9646 2018-01-03 Martin Sebor <msebor@redhat.com>
9648 PR tree-optimization/83603
9649 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
9650 arguments past the endof the argument list in functions declared
9651 without a prototype.
9652 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
9653 Avoid checking when arguments are null.
9655 2018-01-03 Martin Sebor <msebor@redhat.com>
9658 * doc/extend.texi (attribute const): Fix a typo.
9659 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
9660 issuing -Wsuggest-attribute for void functions.
9662 2018-01-03 Martin Sebor <msebor@redhat.com>
9664 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
9665 offset_int::from instead of wide_int::to_shwi.
9666 (maybe_diag_overlap): Remove assertion.
9667 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
9668 * gimple-ssa-sprintf.c (format_directive): Same.
9669 (parse_directive): Same.
9670 (sprintf_dom_walker::compute_format_length): Same.
9671 (try_substitute_return_value): Same.
9673 2018-01-03 Jeff Law <law@redhat.com>
9676 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
9677 non-constant residual for zero at runtime and avoid probing in
9678 that case. Reorganize code for trailing problem to mirror handling
9681 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9683 PR tree-optimization/83501
9684 * tree-ssa-strlen.c (get_string_cst): New.
9685 (handle_char_store): Call get_string_cst.
9687 2018-01-03 Martin Liska <mliska@suse.cz>
9689 PR tree-optimization/83593
9690 * tree-ssa-strlen.c: Include tree-cfg.h.
9691 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
9692 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
9693 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
9695 (strlen_dom_walker::before_dom_children): Call
9696 gimple_purge_dead_eh_edges. Dump tranformation with details
9698 (strlen_dom_walker::before_dom_children): Update call by adding
9699 new argument cleanup_eh.
9700 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
9702 2018-01-03 Martin Liska <mliska@suse.cz>
9705 * cif-code.def (VARIADIC_THUNK): New enum value.
9706 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
9709 2018-01-03 Jan Beulich <jbeulich@suse.com>
9711 * sse.md (mov<mode>_internal): Tighten condition for when to use
9712 vmovdqu<ssescalarsize> for TI and OI modes.
9714 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9716 Update copyright years.
9718 2018-01-03 Martin Liska <mliska@suse.cz>
9721 * ipa-visibility.c (function_and_variable_visibility): Skip
9722 functions with noipa attribure.
9724 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9726 * gcc.c (process_command): Update copyright notice dates.
9727 * gcov-dump.c (print_version): Ditto.
9728 * gcov.c (print_version): Ditto.
9729 * gcov-tool.c (print_version): Ditto.
9730 * gengtype.c (create_file): Ditto.
9731 * doc/cpp.texi: Bump @copying's copyright year.
9732 * doc/cppinternals.texi: Ditto.
9733 * doc/gcc.texi: Ditto.
9734 * doc/gccint.texi: Ditto.
9735 * doc/gcov.texi: Ditto.
9736 * doc/install.texi: Ditto.
9737 * doc/invoke.texi: Ditto.
9739 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9741 * vector-builder.h (vector_builder::m_full_nelts): Change from
9742 unsigned int to poly_uint64.
9743 (vector_builder::full_nelts): Update prototype accordingly.
9744 (vector_builder::new_vector): Likewise.
9745 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
9746 (vector_builder::operator ==): Likewise.
9747 (vector_builder::finalize): Likewise.
9748 * int-vector-builder.h (int_vector_builder::int_vector_builder):
9749 Take the number of elements as a poly_uint64 rather than an
9751 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
9752 from unsigned int to poly_uint64.
9753 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
9754 (vec_perm_indices::new_vector): Likewise.
9755 (vec_perm_indices::length): Likewise.
9756 (vec_perm_indices::nelts_per_input): Likewise.
9757 (vec_perm_indices::input_nelts): Likewise.
9758 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
9759 number of elements per input as a poly_uint64 rather than an
9760 unsigned int. Use the original encoding for variable-length
9761 vectors, rather than clamping each individual element.
9762 For the second and subsequent elements in each pattern,
9763 clamp the step and base before clamping their sum.
9764 (vec_perm_indices::series_p): Handle polynomial element counts.
9765 (vec_perm_indices::all_in_range_p): Likewise.
9766 (vec_perm_indices_to_tree): Likewise.
9767 (vec_perm_indices_to_rtx): Likewise.
9768 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
9769 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
9770 (tree_vector_builder::new_binary_operation): Handle polynomial
9771 element counts. Return false if we need to know the number
9772 of elements at compile time.
9773 * fold-const.c (fold_vec_perm): Punt if the number of elements
9774 isn't known at compile time.
9776 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9778 * vec-perm-indices.h (vec_perm_builder): Change element type
9779 from HOST_WIDE_INT to poly_int64.
9780 (vec_perm_indices::element_type): Update accordingly.
9781 (vec_perm_indices::clamp): Handle polynomial element_types.
9782 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9783 (vec_perm_indices::all_in_range_p): Likewise.
9784 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
9786 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
9787 polynomial vec_perm_indices element types.
9788 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
9789 * fold-const.c (fold_vec_perm): Likewise.
9790 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
9791 * tree-vect-generic.c (lower_vec_perm): Likewise.
9792 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9793 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
9794 element type to HOST_WIDE_INT.
9796 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9797 Alan Hayward <alan.hayward@arm.com>
9798 David Sherwood <david.sherwood@arm.com>
9800 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
9801 rather than an int. Use plus_constant.
9802 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
9803 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
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 * calls.c (emit_call_1, expand_call): Change struct_value_size from
9810 a HOST_WIDE_INT to a poly_int64.
9812 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9813 Alan Hayward <alan.hayward@arm.com>
9814 David Sherwood <david.sherwood@arm.com>
9816 * calls.c (load_register_parameters): Cope with polynomial
9817 mode sizes. Require a constant size for BLKmode parameters
9818 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
9819 forces a parameter to be padded at the lsb end in order to
9820 fill a complete number of words, require the parameter size
9821 to be ordered wrt UNITS_PER_WORD.
9823 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9824 Alan Hayward <alan.hayward@arm.com>
9825 David Sherwood <david.sherwood@arm.com>
9827 * reload1.c (spill_stack_slot_width): Change element type
9828 from unsigned int to poly_uint64_pod.
9829 (alter_reg): Treat mode sizes as polynomial.
9831 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9832 Alan Hayward <alan.hayward@arm.com>
9833 David Sherwood <david.sherwood@arm.com>
9835 * reload.c (complex_word_subreg_p): New function.
9836 (reload_inner_reg_of_subreg, push_reload): Use it.
9838 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9839 Alan Hayward <alan.hayward@arm.com>
9840 David Sherwood <david.sherwood@arm.com>
9842 * lra-constraints.c (process_alt_operands): Reject matched
9843 operands whose sizes aren't ordered.
9844 (match_reload): Refer to this check here.
9846 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9847 Alan Hayward <alan.hayward@arm.com>
9848 David Sherwood <david.sherwood@arm.com>
9850 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
9851 that the mode size is in the set {1, 2, 4, 8, 16}.
9853 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9854 Alan Hayward <alan.hayward@arm.com>
9855 David Sherwood <david.sherwood@arm.com>
9857 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
9858 Use plus_constant instead of gen_rtx_PLUS.
9860 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9861 Alan Hayward <alan.hayward@arm.com>
9862 David Sherwood <david.sherwood@arm.com>
9864 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
9865 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
9866 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
9867 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
9868 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
9869 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
9870 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
9871 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
9872 * config/i386/i386.c (ix86_push_rounding): ...this new function.
9873 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
9875 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
9876 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
9877 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
9878 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
9879 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
9880 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
9881 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
9882 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
9883 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
9884 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
9886 * expr.c (emit_move_resolve_push): Treat the input and result
9887 of PUSH_ROUNDING as a poly_int64.
9888 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
9889 (emit_push_insn): Likewise.
9890 * lra-eliminations.c (mark_not_eliminable): Likewise.
9891 * recog.c (push_operand): Likewise.
9892 * reload1.c (elimination_effects): Likewise.
9893 * rtlanal.c (nonzero_bits1): Likewise.
9894 * calls.c (store_one_arg): Likewise. Require the padding to be
9895 known at compile time.
9897 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9898 Alan Hayward <alan.hayward@arm.com>
9899 David Sherwood <david.sherwood@arm.com>
9901 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
9902 Use plus_constant instead of gen_rtx_PLUS.
9904 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9905 Alan Hayward <alan.hayward@arm.com>
9906 David Sherwood <david.sherwood@arm.com>
9908 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
9911 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9912 Alan Hayward <alan.hayward@arm.com>
9913 David Sherwood <david.sherwood@arm.com>
9915 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
9916 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
9917 via stack temporaries. Treat the mode size as polynomial too.
9919 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9920 Alan Hayward <alan.hayward@arm.com>
9921 David Sherwood <david.sherwood@arm.com>
9923 * expr.c (expand_expr_real_2): When handling conversions involving
9924 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
9925 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
9926 as a poly_uint64 too.
9928 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9929 Alan Hayward <alan.hayward@arm.com>
9930 David Sherwood <david.sherwood@arm.com>
9932 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
9934 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9935 Alan Hayward <alan.hayward@arm.com>
9936 David Sherwood <david.sherwood@arm.com>
9938 * combine.c (can_change_dest_mode): Handle polynomial
9939 REGMODE_NATURAL_SIZE.
9940 * expmed.c (store_bit_field_1): Likewise.
9941 * expr.c (store_constructor): Likewise.
9942 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
9943 and polynomial REGMODE_NATURAL_SIZE.
9944 (gen_lowpart_common): Likewise.
9945 * reginfo.c (record_subregs_of_mode): Likewise.
9946 * rtlanal.c (read_modify_subreg_p): Likewise.
9948 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9949 Alan Hayward <alan.hayward@arm.com>
9950 David Sherwood <david.sherwood@arm.com>
9952 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
9953 numbers of elements.
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 * match.pd: Cope with polynomial numbers of vector elements.
9961 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9962 Alan Hayward <alan.hayward@arm.com>
9963 David Sherwood <david.sherwood@arm.com>
9965 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
9966 in a POINTER_PLUS_EXPR.
9968 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9969 Alan Hayward <alan.hayward@arm.com>
9970 David Sherwood <david.sherwood@arm.com>
9972 * omp-simd-clone.c (simd_clone_subparts): New function.
9973 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
9974 (ipa_simd_modify_function_body): Likewise.
9976 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9977 Alan Hayward <alan.hayward@arm.com>
9978 David Sherwood <david.sherwood@arm.com>
9980 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
9981 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
9982 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
9983 (expand_vector_condition, vector_element): Likewise.
9984 (subparts_gt): New function.
9985 (get_compute_type): Use subparts_gt.
9986 (count_type_subparts): Delete.
9987 (expand_vector_operations_1): Use subparts_gt instead of
9988 count_type_subparts.
9990 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9991 Alan Hayward <alan.hayward@arm.com>
9992 David Sherwood <david.sherwood@arm.com>
9994 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
9995 (vect_compile_time_alias): ...this new function. Do the calculation
9996 on poly_ints rather than trees.
9997 (vect_prune_runtime_alias_test_list): Update call accordingly.
9999 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10000 Alan Hayward <alan.hayward@arm.com>
10001 David Sherwood <david.sherwood@arm.com>
10003 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
10005 (vect_schedule_slp_instance): Likewise.
10007 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10008 Alan Hayward <alan.hayward@arm.com>
10009 David Sherwood <david.sherwood@arm.com>
10011 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
10012 constant and extern definitions for variable-length vectors.
10013 (vect_get_constant_vectors): Note that the number of units
10014 is known to be constant.
10016 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10017 Alan Hayward <alan.hayward@arm.com>
10018 David Sherwood <david.sherwood@arm.com>
10020 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
10021 of units as polynomial. Choose between WIDE and NARROW based
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 * tree-vect-stmts.c (simd_clone_subparts): New function.
10029 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
10031 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10032 Alan Hayward <alan.hayward@arm.com>
10033 David Sherwood <david.sherwood@arm.com>
10035 * tree-vect-stmts.c (vectorizable_call): Treat the number of
10036 vectors as polynomial. Use build_index_vector for
10037 IFN_GOMP_SIMD_LANE.
10039 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10040 Alan Hayward <alan.hayward@arm.com>
10041 David Sherwood <david.sherwood@arm.com>
10043 * tree-vect-stmts.c (get_load_store_type): Treat the number of
10044 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
10045 for variable-length vectors.
10046 (vectorizable_mask_load_store): Treat the number of units as
10047 polynomial, asserting that it is constant if the condition has
10048 already been enforced.
10049 (vectorizable_store, vectorizable_load): Likewise.
10051 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10052 Alan Hayward <alan.hayward@arm.com>
10053 David Sherwood <david.sherwood@arm.com>
10055 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
10056 of units as polynomial. Punt if we can't tell at compile time
10057 which vector contains the final result.
10059 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10060 Alan Hayward <alan.hayward@arm.com>
10061 David Sherwood <david.sherwood@arm.com>
10063 * tree-vect-loop.c (vectorizable_induction): Treat the number
10064 of units as polynomial. Punt on SLP inductions. Use an integer
10065 VEC_SERIES_EXPR for variable-length integer reductions. Use a
10066 cast of such a series for variable-length floating-point
10069 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10070 Alan Hayward <alan.hayward@arm.com>
10071 David Sherwood <david.sherwood@arm.com>
10073 * tree.h (build_index_vector): Declare.
10074 * tree.c (build_index_vector): New function.
10075 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
10076 of units as polynomial, forcibly converting it to a constant if
10077 vectorizable_reduction has already enforced the condition.
10078 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
10079 to create a {1,2,3,...} vector.
10080 (vectorizable_reduction): Treat the number of units as polynomial.
10081 Choose vectype_in based on the largest scalar element size rather
10082 than the smallest number of units. Enforce the restrictions
10085 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10086 Alan Hayward <alan.hayward@arm.com>
10087 David Sherwood <david.sherwood@arm.com>
10089 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
10090 number of units as polynomial.
10092 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10093 Alan Hayward <alan.hayward@arm.com>
10094 David Sherwood <david.sherwood@arm.com>
10096 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
10097 * target.def (autovectorize_vector_sizes): Return the vector sizes
10098 by pointer, using vector_sizes rather than a bitmask.
10099 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
10100 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
10101 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
10103 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
10104 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
10105 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
10106 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
10107 * omp-general.c (omp_max_vf): Likewise.
10108 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10109 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
10110 * tree-vect-loop.c (vect_analyze_loop): Likewise.
10111 * tree-vect-slp.c (vect_slp_bb): Likewise.
10112 * doc/tm.texi: Regenerate.
10113 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
10115 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
10116 the vector size as a poly_uint64 rather than an unsigned int.
10117 (current_vector_size): Change from an unsigned int to a poly_uint64.
10118 (get_vectype_for_scalar_type): Update accordingly.
10119 * tree.h (build_truth_vector_type): Take the size and number of
10120 units as a poly_uint64 rather than an unsigned int.
10121 (build_vector_type): Add a temporary overload that takes
10122 the number of units as a poly_uint64 rather than an unsigned int.
10123 * tree.c (make_vector_type): Likewise.
10124 (build_truth_vector_type): Take the number of units as a poly_uint64
10125 rather than an unsigned int.
10127 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10128 Alan Hayward <alan.hayward@arm.com>
10129 David Sherwood <david.sherwood@arm.com>
10131 * target.def (get_mask_mode): Take the number of units and length
10132 as poly_uint64s rather than unsigned ints.
10133 * targhooks.h (default_get_mask_mode): Update accordingly.
10134 * targhooks.c (default_get_mask_mode): Likewise.
10135 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
10136 * doc/tm.texi: Regenerate.
10138 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10139 Alan Hayward <alan.hayward@arm.com>
10140 David Sherwood <david.sherwood@arm.com>
10142 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
10143 * omp-general.c (omp_max_vf): Likewise.
10144 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
10145 (expand_omp_simd): Handle polynomial safelen.
10146 * omp-low.c (omplow_simd_context): Add a default constructor.
10147 (omplow_simd_context::max_vf): Change from int to poly_uint64.
10148 (lower_rec_simd_input_clauses): Update accordingly.
10149 (lower_rec_input_clauses): Likewise.
10151 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10152 Alan Hayward <alan.hayward@arm.com>
10153 David Sherwood <david.sherwood@arm.com>
10155 * tree-vectorizer.h (vect_nunits_for_cost): New function.
10156 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
10157 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
10158 (vect_analyze_slp_cost): Likewise.
10159 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
10160 (vect_model_load_cost): Likewise.
10162 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10163 Alan Hayward <alan.hayward@arm.com>
10164 David Sherwood <david.sherwood@arm.com>
10166 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
10167 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
10168 from an unsigned int * to a poly_uint64_pod *.
10169 (calculate_unrolling_factor): New function.
10170 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
10172 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10173 Alan Hayward <alan.hayward@arm.com>
10174 David Sherwood <david.sherwood@arm.com>
10176 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
10177 from an unsigned int to a poly_uint64.
10178 (_loop_vec_info::slp_unrolling_factor): Likewise.
10179 (_loop_vec_info::vectorization_factor): Change from an int
10181 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
10182 (vect_get_num_vectors): New function.
10183 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
10184 (vect_get_num_copies): Use vect_get_num_vectors.
10185 (vect_analyze_data_ref_dependences): Change max_vf from an int *
10186 to an unsigned int *.
10187 (vect_analyze_data_refs): Change min_vf from an int * to a
10189 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10190 than an unsigned HOST_WIDE_INT.
10191 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
10192 (vect_analyze_data_ref_dependence): Change max_vf from an int *
10193 to an unsigned int *.
10194 (vect_analyze_data_ref_dependences): Likewise.
10195 (vect_compute_data_ref_alignment): Handle polynomial vf.
10196 (vect_enhance_data_refs_alignment): Likewise.
10197 (vect_prune_runtime_alias_test_list): Likewise.
10198 (vect_shift_permute_load_chain): Likewise.
10199 (vect_supportable_dr_alignment): Likewise.
10200 (dependence_distance_ge_vf): Take the vectorization factor as a
10201 poly_uint64 rather than an unsigned HOST_WIDE_INT.
10202 (vect_analyze_data_refs): Change min_vf from an int * to a
10204 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
10205 vfm1 as a poly_uint64 rather than an int. Make the same change
10206 for the returned bound_scalar.
10207 (vect_gen_vector_loop_niters): Handle polynomial vf.
10208 (vect_do_peeling): Likewise. Update call to
10209 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
10210 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
10212 * tree-vect-loop.c (vect_determine_vectorization_factor)
10213 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
10214 (vect_get_known_peeling_cost): Likewise.
10215 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
10216 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
10217 (vect_transform_loop): Likewise. Use the lowest possible VF when
10218 updating the upper bounds of the loop.
10219 (vect_min_worthwhile_factor): Make static. Return an unsigned int
10220 rather than an int.
10221 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
10222 polynomial unroll factors.
10223 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
10224 (vect_make_slp_decision): Likewise.
10225 (vect_supported_load_permutation_p): Likewise, and polynomial
10227 (vect_analyze_slp_cost): Handle polynomial vf.
10228 (vect_slp_analyze_node_operations): Likewise.
10229 (vect_slp_analyze_bb_1): Likewise.
10230 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10231 than an unsigned HOST_WIDE_INT.
10232 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
10233 (vectorizable_load): Handle polynomial vf.
10234 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
10236 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
10238 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10239 Alan Hayward <alan.hayward@arm.com>
10240 David Sherwood <david.sherwood@arm.com>
10242 * match.pd: Handle bit operations involving three constants
10243 and try to fold one pair.
10245 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10247 * tree-vect-loop-manip.c: Include gimple-fold.h.
10248 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
10249 niters_maybe_zero parameters. Handle other cases besides a step of 1.
10250 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
10251 Add a path that uses a step of VF instead of 1, but disable it
10253 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
10254 and niters_no_overflow parameters. Update calls to
10255 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
10256 Create a new SSA name if the latter choses to use a ste other
10257 than zero, and return it via niters_vector_mult_vf_var.
10258 * tree-vect-loop.c (vect_transform_loop): Update calls to
10259 vect_do_peeling, vect_gen_vector_loop_niters and
10260 slpeel_make_loop_iterate_ntimes.
10261 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
10262 (vect_gen_vector_loop_niters): Update declarations after above changes.
10264 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
10266 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
10267 128-bit round to integer instructions.
10268 (ceil<mode>2): Likewise.
10269 (btrunc<mode>2): Likewise.
10270 (round<mode>2): Likewise.
10272 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10274 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
10275 unaligned VSX load/store on P8/P9.
10276 (expand_block_clear): Allow the use of unaligned VSX
10277 load/store on P8/P9.
10279 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10281 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
10283 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
10284 swap associated with both a load and a store.
10286 2018-01-02 Andrew Waterman <andrew@sifive.com>
10288 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
10289 * config/riscv/riscv.md (clear_cache): Use it.
10291 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
10293 * web.c: Remove out-of-date comment.
10295 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10297 * expr.c (fixup_args_size_notes): Check that any existing
10298 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
10299 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
10300 (emit_single_push_insn): ...here.
10302 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10304 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
10305 (const_vector_encoded_nelts): New function.
10306 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
10307 (const_vector_int_elt, const_vector_elt): Declare.
10308 * emit-rtl.c (const_vector_int_elt_1): New function.
10309 (const_vector_elt): Likewise.
10310 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
10311 of CONST_VECTOR_ELT.
10313 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10315 * expr.c: Include rtx-vector-builder.h.
10316 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
10317 directly on the tree encoding.
10318 (const_vector_from_tree): Likewise.
10319 * optabs.c: Include rtx-vector-builder.h.
10320 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
10321 sequence of "u" values.
10322 * vec-perm-indices.c: Include rtx-vector-builder.h.
10323 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
10324 directly on the vec_perm_indices encoding.
10326 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10328 * doc/rtl.texi (const_vector): Describe new encoding scheme.
10329 * Makefile.in (OBJS): Add rtx-vector-builder.o.
10330 * rtx-vector-builder.h: New file.
10331 * rtx-vector-builder.c: Likewise.
10332 * rtl.h (rtx_def::u2): Add a const_vector field.
10333 (CONST_VECTOR_NPATTERNS): New macro.
10334 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
10335 (CONST_VECTOR_DUPLICATE_P): Likewise.
10336 (CONST_VECTOR_STEPPED_P): Likewise.
10337 (CONST_VECTOR_ENCODED_ELT): Likewise.
10338 (const_vec_duplicate_p): Check for a duplicated vector encoding.
10339 (unwrap_const_vec_duplicate): Likewise.
10340 (const_vec_series_p): Check for a non-duplicated vector encoding.
10341 Say that the function only returns true for integer vectors.
10342 * emit-rtl.c: Include rtx-vector-builder.h.
10343 (gen_const_vec_duplicate_1): Delete.
10344 (gen_const_vector): Call gen_const_vec_duplicate instead of
10345 gen_const_vec_duplicate_1.
10346 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
10347 (gen_const_vec_duplicate): Use rtx_vector_builder.
10348 (gen_const_vec_series): Likewise.
10349 (gen_rtx_CONST_VECTOR): Likewise.
10350 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
10351 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10352 Build a new vector rather than modifying a CONST_VECTOR in-place.
10353 (handle_special_swappables): Update call accordingly.
10354 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
10355 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10356 Build a new vector rather than modifying a CONST_VECTOR in-place.
10357 (handle_special_swappables): Update call accordingly.
10359 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10361 * simplify-rtx.c (simplify_const_binary_operation): Use
10362 CONST_VECTOR_ELT instead of XVECEXP.
10364 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10366 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
10367 the selector elements to be different from the data elements
10368 if the selector is a VECTOR_CST.
10369 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
10370 ssizetype for the selector.
10372 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10374 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
10375 before testing each element individually.
10376 * tree-vect-generic.c (lower_vec_perm): Likewise.
10378 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10380 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
10381 * selftest-run-tests.c (selftest::run_tests): Call it.
10382 * vector-builder.h (vector_builder::operator ==): New function.
10383 (vector_builder::operator !=): Likewise.
10384 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
10385 (vec_perm_indices::all_from_input_p): New function.
10386 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10387 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
10388 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
10389 instead of reading the VECTOR_CST directly. Detect whether both
10390 vector inputs are the same before constructing the vec_perm_indices,
10391 and update the number of inputs argument accordingly. Use the
10392 utility functions added above. Only construct sel2 if we need to.
10394 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10396 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
10397 the broadcast of the low byte.
10398 (expand_mult_highpart): Use an explicit encoding for the permutes.
10399 * optabs-query.c (can_mult_highpart_p): Likewise.
10400 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
10401 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10402 (vectorizable_bswap): Likewise.
10403 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
10404 explicit encoding for the power-of-2 permutes.
10405 (vect_permute_store_chain): Likewise.
10406 (vect_grouped_load_supported): Likewise.
10407 (vect_permute_load_chain): Likewise.
10409 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10411 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
10412 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
10413 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
10414 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10415 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
10416 (vect_gen_perm_mask_any): Likewise.
10418 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10420 * int-vector-builder.h: New file.
10421 * vec-perm-indices.h: Include int-vector-builder.h.
10422 (vec_perm_indices): Redefine as an int_vector_builder.
10423 (auto_vec_perm_indices): Delete.
10424 (vec_perm_builder): Redefine as a stand-alone class.
10425 (vec_perm_indices::vec_perm_indices): New function.
10426 (vec_perm_indices::clamp): Likewise.
10427 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
10428 (vec_perm_indices::new_vector): New function.
10429 (vec_perm_indices::new_expanded_vector): Update for new
10430 vec_perm_indices class.
10431 (vec_perm_indices::rotate_inputs): New function.
10432 (vec_perm_indices::all_in_range_p): Operate directly on the
10433 encoded form, without computing elided elements.
10434 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
10435 encoding. Update for new vec_perm_indices class.
10436 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
10437 the given vec_perm_builder.
10438 (expand_vec_perm_var): Update vec_perm_builder constructor.
10439 (expand_mult_highpart): Use vec_perm_builder instead of
10440 auto_vec_perm_indices.
10441 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
10442 vec_perm_indices instead of auto_vec_perm_indices. Use a single
10443 or double series encoding as appropriate.
10444 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
10445 vec_perm_indices instead of auto_vec_perm_indices.
10446 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10447 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10448 (vect_permute_store_chain): Likewise.
10449 (vect_grouped_load_supported): Likewise.
10450 (vect_permute_load_chain): Likewise.
10451 (vect_shift_permute_load_chain): Likewise.
10452 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10453 (vect_transform_slp_perm_load): Likewise.
10454 (vect_schedule_slp_instance): Likewise.
10455 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10456 (vectorizable_mask_load_store): Likewise.
10457 (vectorizable_bswap): Likewise.
10458 (vectorizable_store): Likewise.
10459 (vectorizable_load): Likewise.
10460 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
10461 vec_perm_indices instead of auto_vec_perm_indices. Use
10462 tree_to_vec_perm_builder to read the vector from a tree.
10463 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
10464 vec_perm_builder instead of a vec_perm_indices.
10465 (have_whole_vector_shift): Use vec_perm_builder and
10466 vec_perm_indices instead of auto_vec_perm_indices. Leave the
10467 truncation to calc_vec_perm_mask_for_shift.
10468 (vect_create_epilog_for_reduction): Likewise.
10469 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
10470 from auto_vec_perm_indices to vec_perm_indices.
10471 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10472 instead of changing individual elements.
10473 (aarch64_vectorize_vec_perm_const): Use new_vector to install
10474 the vector in d.perm.
10475 * config/arm/arm.c (expand_vec_perm_d::perm): Change
10476 from auto_vec_perm_indices to vec_perm_indices.
10477 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10478 instead of changing individual elements.
10479 (arm_vectorize_vec_perm_const): Use new_vector to install
10480 the vector in d.perm.
10481 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
10482 Update vec_perm_builder constructor.
10483 (rs6000_expand_interleave): Likewise.
10484 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
10485 (rs6000_expand_interleave): Likewise.
10487 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10489 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
10490 to qimode could truncate the indices.
10491 * optabs.c (expand_vec_perm_var): Likewise.
10493 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10495 * Makefile.in (OBJS): Add vec-perm-indices.o.
10496 * vec-perm-indices.h: New file.
10497 * vec-perm-indices.c: Likewise.
10498 * target.h (vec_perm_indices): Replace with a forward class
10500 (auto_vec_perm_indices): Move to vec-perm-indices.h.
10501 * optabs.h: Include vec-perm-indices.h.
10502 (expand_vec_perm): Delete.
10503 (selector_fits_mode_p, expand_vec_perm_var): Declare.
10504 (expand_vec_perm_const): Declare.
10505 * target.def (vec_perm_const_ok): Replace with...
10506 (vec_perm_const): ...this new hook.
10507 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
10508 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
10509 * doc/tm.texi: Regenerate.
10510 * optabs.def (vec_perm_const): Delete.
10511 * doc/md.texi (vec_perm_const): Likewise.
10512 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
10513 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
10514 expand_vec_perm for constant permutation vectors. Assert that
10515 the mode of variable permutation vectors is the integer equivalent
10516 of the mode that is being permuted.
10517 * optabs-query.h (selector_fits_mode_p): Declare.
10518 * optabs-query.c: Include vec-perm-indices.h.
10519 (selector_fits_mode_p): New function.
10520 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
10521 is defined, instead of checking whether the vec_perm_const_optab
10522 exists. Use targetm.vectorize.vec_perm_const instead of
10523 targetm.vectorize.vec_perm_const_ok. Check whether the indices
10524 fit in the vector mode before using a variable permute.
10525 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
10526 vec_perm_indices instead of an rtx.
10527 (expand_vec_perm): Replace with...
10528 (expand_vec_perm_const): ...this new function. Take the selector
10529 as a vec_perm_indices rather than an rtx. Also take the mode of
10530 the selector. Update call to shift_amt_for_vec_perm_mask.
10531 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
10532 Use vec_perm_indices::new_expanded_vector to expand the original
10533 selector into bytes. Check whether the indices fit in the vector
10534 mode before using a variable permute.
10535 (expand_vec_perm_var): Make global.
10536 (expand_mult_highpart): Use expand_vec_perm_const.
10537 * fold-const.c: Includes vec-perm-indices.h.
10538 * tree-ssa-forwprop.c: Likewise.
10539 * tree-vect-data-refs.c: Likewise.
10540 * tree-vect-generic.c: Likewise.
10541 * tree-vect-loop.c: Likewise.
10542 * tree-vect-slp.c: Likewise.
10543 * tree-vect-stmts.c: Likewise.
10544 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
10546 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
10547 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
10548 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
10549 (aarch64_vectorize_vec_perm_const): ...this new function.
10550 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10551 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10552 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
10553 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
10554 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10555 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10556 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
10558 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
10559 check for NEON modes.
10560 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
10561 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
10562 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
10563 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
10565 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
10566 the old VEC_PERM_CONST conditions.
10567 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
10568 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
10569 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
10570 (ia64_vectorize_vec_perm_const_ok): Merge into...
10571 (ia64_vectorize_vec_perm_const): ...this new function.
10572 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
10573 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
10574 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
10575 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
10576 * config/mips/mips.c (mips_expand_vec_perm_const)
10577 (mips_vectorize_vec_perm_const_ok): Merge into...
10578 (mips_vectorize_vec_perm_const): ...this new function.
10579 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
10580 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
10581 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
10582 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
10583 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
10584 (rs6000_expand_vec_perm_const): Delete.
10585 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
10587 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10588 (altivec_expand_vec_perm_const_le): Take each operand individually.
10589 Operate on constant selectors rather than rtxes.
10590 (altivec_expand_vec_perm_const): Likewise. Update call to
10591 altivec_expand_vec_perm_const_le.
10592 (rs6000_expand_vec_perm_const): Delete.
10593 (rs6000_vectorize_vec_perm_const_ok): Delete.
10594 (rs6000_vectorize_vec_perm_const): New function.
10595 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10596 an element count and rtx array.
10597 (rs6000_expand_extract_even): Update call accordingly.
10598 (rs6000_expand_interleave): Likewise.
10599 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
10600 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
10601 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
10602 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
10603 (rs6000_expand_vec_perm_const): Delete.
10604 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10605 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10606 (altivec_expand_vec_perm_const_le): Take each operand individually.
10607 Operate on constant selectors rather than rtxes.
10608 (altivec_expand_vec_perm_const): Likewise. Update call to
10609 altivec_expand_vec_perm_const_le.
10610 (rs6000_expand_vec_perm_const): Delete.
10611 (rs6000_vectorize_vec_perm_const_ok): Delete.
10612 (rs6000_vectorize_vec_perm_const): New function. Remove stray
10613 reference to the SPE evmerge intructions.
10614 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10615 an element count and rtx array.
10616 (rs6000_expand_extract_even): Update call accordingly.
10617 (rs6000_expand_interleave): Likewise.
10618 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
10619 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
10621 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10623 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10625 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
10626 vector mode and that that mode matches the mode of the data
10628 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
10629 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
10630 directly using expand_vec_perm_1 when forcing selectors into
10632 (expand_vec_perm_var): New function, split out from expand_vec_perm.
10634 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10636 * optabs-query.h (can_vec_perm_p): Delete.
10637 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
10638 * optabs-query.c (can_vec_perm_p): Split into...
10639 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
10640 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
10641 particular selector is valid.
10642 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10643 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10644 (vect_grouped_load_supported): Likewise.
10645 (vect_shift_permute_load_chain): Likewise.
10646 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10647 (vect_transform_slp_perm_load): Likewise.
10648 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10649 (vectorizable_bswap): Likewise.
10650 (vect_gen_perm_mask_checked): Likewise.
10651 * fold-const.c (fold_ternary_loc): Likewise. Don't take
10652 implementations of variable permutation vectors into account
10653 when deciding which selector to use.
10654 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
10655 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
10656 with a false third argument.
10657 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
10658 to test whether the constant selector is valid and can_vec_perm_var_p
10659 to test whether a variable selector is valid.
10661 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10663 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
10664 * optabs-query.c (can_vec_perm_p): Likewise.
10665 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
10666 instead of vec_perm_indices.
10667 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
10668 (vect_gen_perm_mask_checked): Likewise,
10669 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
10670 (vect_gen_perm_mask_checked): Likewise,
10672 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10674 * optabs-query.h (qimode_for_vec_perm): Declare.
10675 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
10676 (qimode_for_vec_perm): ...this new function.
10677 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
10679 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10681 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
10682 does not have a conditional at the top.
10684 2018-01-02 Richard Biener <rguenther@suse.de>
10686 * ipa-inline.c (big_speedup_p): Fix expression.
10688 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10691 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
10694 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10698 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
10699 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
10700 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
10701 cond_taken_branch_cost 3->4.
10703 2018-01-01 Jakub Jelinek <jakub@redhat.com>
10705 PR tree-optimization/83581
10706 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
10707 TODO_cleanup_cfg if any changes have been made.
10709 PR middle-end/83608
10710 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
10711 convert_modes if target mode has the right side, but different mode
10714 PR middle-end/83609
10715 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
10716 last argument when extracting from CONCAT. If either from_real or
10717 from_imag is NULL, use expansion through memory. If result is not
10718 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
10719 the parts directly to inner mode, if even that fails, use expansion
10722 PR middle-end/83623
10723 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
10724 check for bswap in mode rather than HImode and use that in expand_unop
10727 Copyright (C) 2018 Free Software Foundation, Inc.
10729 Copying and distribution of this file, with or without modification,
10730 are permitted in any medium without royalty provided the copyright
10731 notice and this notice are preserved.