1 2018-04-13 Martin Liska <mliska@suse.cz>
2 Jakub Jelinek <jakub@redhat.com>
5 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
6 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
7 * builtins.c (expand_builtin_memory_copy_args): Use
8 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
9 handle dest_addr == pc_rtx.
11 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
14 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
15 asked to not generate direct moves.
16 (fix_trunc<mode>si2_stfiwx): Similar.
17 (fix_trunc<mode>si2_internal): Similar.
19 2018-04-12 Jakub Jelinek <jakub@redhat.com>
22 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
23 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
24 lookup if dest in some wider mode is known to be const0_rtx and
25 if so, record permanent equivalence for it to be ZERO_EXTEND of
26 the narrower mode destination.
28 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
30 * lto-streamer-out.c (output_function): Revert 259346.
31 * omp-expand.c (expand_oacc_for): Likewise.
33 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
35 PR rtl-optimization/85354
36 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
37 * sel-sched.c (sel_global_init): ... here.
39 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
42 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
43 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
44 mode for PE-COFF targets.
45 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
46 (i386_pe_asm_lto_end): Likewise.
47 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
48 (TARGET_ASM_LTO_END): Likewise.
49 * config/i386/winnt.c (saved_debug_info_level): New static variable.
50 (i386_pe_asm_lto_start): New function.
51 (i386_pe_asm_lto_end): Likewise.
53 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
54 Richard Biener <rguenther@suse.de>
57 * lto-streamer-out.c (output_function): Fix CFG loop state before
59 * omp-expand.c (expand_oacc_for): Handle calls to internal
60 functions like regular functions.
62 2018-04-12 Richard Biener <rguenther@suse.de>
65 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
66 for the early LTO debug to properly generate references to it
67 during DIE emission. Do not re-use that for the skeleton for
69 (dwarf2out_early_finish): Likewise.
71 2018-04-12 Jakub Jelinek <jakub@redhat.com>
75 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
76 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
77 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
78 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
79 and output is a reg, avoid creating invalid lowpart subreg, but
80 instead split into a 512-bit move. Don't split if not AVX512VL,
81 input is xmm16+ reg and output is a mem.
82 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
83 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
84 xmm16+ reg and output is a mem.
86 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
88 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
89 also for flag_dwarf2_cfi_asm.
91 2018-04-12 Jakub Jelinek <jakub@redhat.com>
93 PR rtl-optimization/85342
94 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
95 a bool scalar var inside of the loop instead. Don't try to update
96 recog_data.operand after failed apply_change_group.
98 2018-04-12 Tom de Vries <tom@codesourcery.com>
101 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
102 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
103 array with flexible array member as array without given dimension.
104 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
105 argument for undefined param to true.
107 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
110 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
111 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
112 from PowerPC section.
113 * config/rs6000/sysv4.opt (mcall-): Improve help text.
114 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
115 help text that is too long.
116 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
117 help text that is too long.
118 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
119 help text that is too long.
121 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
123 * config/alpha/alpha.md (stack_probe_internal): Rename
124 from "probe_stack". Update all callers.
126 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
128 PR rtl-optimization/84566
129 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
130 sched_macro_fuse_insns.
132 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
135 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
136 (compute_block_dependences): ... from here.
138 2018-04-11 Jakub Jelinek <jakub@redhat.com>
140 PR tree-optimization/85331
141 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
142 from int to HOST_WIDE_INT.
144 2018-04-11 Martin Jambor <mjambor@suse.cz>
147 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
148 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
149 not the same as the source val.
150 (cgraph_edge_brings_value_p): New parameter.
151 (gather_edges_for_value): Pass destination value to
152 cgraph_edge_brings_value_p.
153 (perhaps_add_new_callers): Likewise.
154 (get_info_about_necessary_edges): Likewise and exclude values brought
155 only by self-recursive edges.
156 (create_specialized_node): Redirect only clones of self-calling edges.
157 (+self_recursive_pass_through_p): New function.
158 (find_more_scalar_values_for_callers_subset): Use it.
159 (find_aggregate_values_for_callers_subset): Likewise.
160 (known_aggs_to_agg_replacement_list): Removed.
161 (decide_whether_version_node): Re-calculate known constants for all
162 remaining context clones.
164 2018-04-11 Richard Biener <rguenther@suse.de>
167 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
168 from early DWARF output.
169 (dwarf2out_early_finish): Output line info unconditionally into
170 early DWARF and add reference to it.
172 2018-04-11 Jakub Jelinek <jakub@redhat.com>
175 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
176 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
177 other than V2DFmode using iptr mode attribute.
178 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
180 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
182 PR rtl-optimization/84659
183 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
185 2018-04-11 Jakub Jelinek <jakub@redhat.com>
188 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
190 (output_loc_list): Pass address of a dummy size variable even in the
191 locview handling loop.
192 (index_location_lists): Add comment on why skip_loc_list_entry can't
195 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
198 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
201 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
204 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
205 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
206 and -mstring-compare-inline-limit.
208 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
211 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
212 for stack clash protection in a register whenever we need it to be in
215 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
217 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
218 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
220 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
223 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
225 (mlong-double-): Ditto.
226 * config/rs6000/sysv4.opt (msdata=): Ditto.
229 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
231 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
232 erroneous entries for
233 "vector int vec_ldl (int, long int *)", and
234 "vector unsigned int vec_ldl (int, unsigned long int *)".
235 Add comments and entries for
236 "vector bool char vec_ldl (int, bool char *)",
237 "vector bool short vec_ldl (int, bool short *)",
238 "vector bool int vec_ldl (int, bool int *)",
239 "vector bool long long vec_ldl (int, bool long long *)",
240 "vector pixel vec_ldl (int, pixel *)",
241 "vector long long vec_ldl (int, long long *)",
242 "vector unsigned long long vec_ldl (int, unsigned long long *)".
243 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
244 type tree bool_long_long_type_node and correct definition of
245 bool_V2DI_type_node to make reference to this new type tree.
246 (rs6000_mangle_type): Replace erroneous reference to
247 bool_long_type_node with bool_long_long_type_node.
248 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
249 comments to emphasize sign distinctions for char and int types and
250 replace RS6000_BTI_bool_long constant with
251 RS6000_BTI_bool_long_long constant. Also add comment to restrict
252 use of RS6000_BTI_pixel.
253 (bool_long_type_node): Remove this macro definition.
254 (bool_long_long_type_node): New macro definition
256 2018-04-10 Jakub Jelinek <jakub@redhat.com>
258 PR rtl-optimization/85300
259 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
260 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
261 simplify_unary_operation fails.
263 2018-04-10 Martin Liska <mliska@suse.cz>
265 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
266 cgraph_edge and ipa_ref.
268 2018-04-10 Jakub Jelinek <jakub@redhat.com>
273 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
274 computation of the VEC_MERGE selector from mask.
275 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
276 Fix decoding of the VEC_MERGE selector into mask.
278 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
280 PR tree-optimization/85286
281 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
283 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
285 * final.c (final_1): Set insn_last_address as well as
286 insn_current_address.
288 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
291 * explow.c (emit_stack_probe): Call validize_mem on memory location
292 before passing it to gen_probe_stack. Create address operand and
293 legitimize it for the probe_stack_address case.
295 2018-04-09 Jan Hubicka <jh@suse.cz>
298 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
299 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
300 * tree.c (free_lang_data_in_type): Fix handling of binfos;
302 (free_lang_data): Rebuild type inheritance graph.
304 2018-04-09 Martin Sebor <msebor@redhat.com>
306 * invoke.texi (-finline-small-functions): Mention other optimization
308 (-findirect-inlining, -fpartial-inlining): Same.
309 (-finline-functions-called-once): Same.
310 (-freorder-blocks-and-partition): Same.
312 2018-04-09 Jan Hubicka <jh@suse.cz>
315 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
316 jumps; choose last target that matches the criteria (i.e.
317 no partition changes for non-crossing jumps).
318 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
319 support for redirecting crossing jumps to non-crossing.
321 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
323 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
324 also for naked functions.
326 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
328 * config/arc/arc.md (add_shift): New pattern.
329 (add_shift2): Likewise.
330 (sub_shift): Likewise.
331 (sub_shift_cmp0_noout): Likewise.
332 (compare_si_ashiftsi): Likewise.
333 (xbfu_cmp0_noout): New combine pattern.
334 (xbfu_cmp0"): Likewise.
335 (movsi_set_cc_insn): Place the predicable variant first.
336 (commutative_binary_cmp0_noout): Remove clobber.
337 (commutative_binary_cmp0): New pattern.
338 (noncommutative_binary_cmp0): Likewise.
339 (noncommutative_binary_cmp0_noout): Likewise.
340 (noncommutative_binary_comparison_result_used): Removed.
341 (rsub_cmp0): New pattern.
342 (rsub_cmp0_noout): Likewise.
343 (extzvsi): Changed, keep only meaningful variants.
344 (SQH, SEZ): New iterators.
345 (SQH_postfix): New mode attribute.
346 (SEZ_prefix): New code attribute.
347 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
348 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
349 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
351 (noncommutative_operator): Check the availability of barrel
354 2018-04-09 Richard Biener <rguenther@suse.de>
356 PR tree-optimization/85284
357 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
358 Only use the niter constraining form of simple_iv when the exit
361 2018-04-09 Tom de Vries <tom@codesourcery.com>
364 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
365 (define_expand "*memory_barrier"): New define_expand.
366 (define_insn "memory_barrier"): New insn.
368 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
370 PR rtl-optimization/80463
371 PR rtl-optimization/83972
372 PR rtl-optimization/83480
374 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
375 correct producer for the insn.
376 (tidy_control_flow): Fixup seqnos in case of debug insns.
378 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
380 PR rtl-optimization/83913
382 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
383 different sched-times when merging exprs.
385 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
387 PR rtl-optimization/83962
389 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
390 tidy_fallthru_edge and tidy_control_flow.
392 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
394 PR rtl-optimization/83530
396 * sel-sched.c (force_next_insn): New global variable.
397 (remove_insn_for_debug): When force_next_insn is true, also leave only
398 next insn in the ready list.
399 (sel_sched_region): When the region wasn't scheduled, make another pass
400 over it with force_next_insn set to 1.
402 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
404 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
406 * config/nds32/constants.md (unspec_volatile_element): Add enum values
407 for interrupt control.
408 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
409 functions for interrupt control.
410 * config/nds32/nds32-intrinsic.md: Likewise.
411 * config/nds32/nds32_intrinsic.h: Likewise.
412 * config/nds32/nds32.h (nds32_builtins): Likewise.
414 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
416 * config/nds32/nds32.c (nds32_init_machine_status,
417 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
418 strict_aligned_p field.
419 (nds32_expand_to_rtl_hook): New function.
420 (TARGET_EXPAND_TO_RTL_HOOK): Define.
421 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
423 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
424 Chung-Ju Wu <jasonwucj@gmail.com>
426 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
427 * config/nds32/nds32-n7.md: New file.
428 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
429 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
431 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
432 * config/nds32/nds32.md (pipeline_model): Add n7.
433 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
434 * config/nds32/pipelines.md: Include n7 settings.
436 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
437 Chung-Ju Wu <jasonwucj@gmail.com>
439 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
440 * config/nds32/nds32-e8.md: New file.
441 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
442 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
444 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
445 * config/nds32/nds32.md (pipeline_model): Add e8.
446 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
447 * config/nds32/pipelines.md: Include e8 settings.
449 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
450 Chung-Ju Wu <jasonwucj@gmail.com>
452 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
453 * config/nds32/nds32-n8.md: New file.
454 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
455 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
457 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
458 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
459 * config/nds32/nds32.md (pipeline_model): Add n8.
460 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
461 * config/nds32/pipelines.md: Include n8 settings.
463 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
464 Chung-Ju Wu <jasonwucj@gmail.com>
466 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
467 * config/nds32/nds32-n9-2r1w.md: New file.
468 * config/nds32/nds32-n9-3r2w.md: New file.
469 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
470 nds32_register_ports): New or modify for cpu n9.
471 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
473 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
474 * config/nds32/nds32-utils.c: New file.
475 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
476 TARGET_MUL_SLOW): Define.
477 * config/nds32/nds32.md (pipeline_model): New attribute.
478 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
479 New options that support cpu n9.
480 * config/nds32/pipelines.md: Include n9 settings.
481 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
483 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
485 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
486 information if necessary.
487 (output_cond_branch_compare_zero): Likewise.
488 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
489 (nds32_target_alignment): Refine for alignment.
490 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
491 (FUNCTION_BOUNDARY): Modify.
492 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
494 * config/nds32/nds32.opt (malways-align, malign-functions): New.
496 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
498 * config/nds32/constants.md (unspec_volatile_element): Add values for
499 TLB operation and data prefetch.
500 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
501 functions for TLB operation and data prefetch.
502 * config/nds32/nds32-intrinsic.md: Likewise.
503 * config/nds32/nds32_intrinsic.h: Likewise.
504 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
505 (nds32_print_operand): Likewise.
506 * config/nds32/nds32.h (nds32_builtins): Likewise.
508 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
509 Andrew Pinski <pinsika@gcc.gnu.org>
512 * match.pd: Use constant_boolean_node of correct type instead of
513 boolean_true_node or boolean_false_node for simplifying
514 pointer comparisons to zero.
516 2018-04-07 Jakub Jelinek <jakub@redhat.com>
518 PR tree-optimization/80021
519 * tree.c (verify_type_variant): Make error call in verify_variant_match
520 translatable and remove final full stop.
522 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
524 * config/nds32/constants.md (unspec_volatile_element): Add
525 UNSPEC_VOLATILE_EH_RETURN.
526 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
527 nds32_output_stack_pop): Support dwarf exception handling process.
528 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
529 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
530 exception handling process.
531 (nds32_compute_stack_frame): Likewise.
532 (nds32_return_addr_rtx): Likewise.
533 (nds32_initial_elimination_offset): Likewise.
534 (nds32_expand_prologue): Likewise.
535 (nds32_expand_epilogue): Likewise.
536 (nds32_dynamic_chain_address): New function.
537 * config/nds32/nds32.h (machine_function): Add fields for dwarf
539 (DYNAMIC_CHAIN_ADDRESS): Define.
540 (EH_RETURN_DATA_REGNO): Define.
541 (EH_RETURN_STACKADJ_RTX): Define.
542 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
543 patterns for dwarf exception handling.
545 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
547 * config/nds32/nds32.h: Clean up obsolete macros.
549 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
551 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
552 Add enum values for particular instructions.
553 * config/nds32/nds32-intrinsic.c: Implementation of expanding
554 particular intrinsic functions.
555 * config/nds32/nds32-intrinsic.md: Likewise.
556 * config/nds32/nds32_intrinsic.h: Likewise.
557 * config/nds32/nds32.h (nds32_builtins): Likewise.
558 * config/nds32/nds32.md (type): Add pbsad and pbsada.
559 (btst, ave): New patterns for particular instructions.
561 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
563 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
564 Add enum values for atomic load/store and memory sync.
565 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
567 * config/nds32/nds32-intrinsic.md: Likewise.
568 * config/nds32/nds32_intrinsic.h: Likewise.
569 * config/nds32/nds32.h (nds32_builtins): Likewise.
571 2018-04-07 Jakub Jelinek <jakub@redhat.com>
573 PR tree-optimization/85257
574 * fold-const.c (native_encode_vector): If not all elts could fit
575 and off is -1, return 0 rather than offset.
576 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
577 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
578 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
579 adjust buffer in native_interpret_expr call.
581 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
583 * config/nds32/constants.md (unspec_volatile_element): Add cache
585 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
586 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
587 * config/nds32/nds32.c (nds32_cctl_names): New.
588 (nds32_print_operand): Handle cache control register names.
589 * config/nds32/nds32.h (nds32_builtins): New enum values.
590 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
592 * config/nds32/nds32.md (type): Add mmu.
593 * config/nds32/pipelines.md (simple_insn): Add mmu.
595 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
597 * config/nds32/nds32.md (type): Remove call.
598 * config/nds32/pipelines.md (simple_insn): Likewise.
600 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
602 * config/nds32/constants.md (unspec_volatile_element): Add
603 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
604 UNSPEC_VOLATILE_FMFCFG.
605 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
606 description for fmfcfg and fmfcsr.
607 (bdesc_1arg): Add fmtcsr.
608 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
609 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
610 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
611 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
612 unspec_fmfcfg): New patterns.
613 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
614 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
615 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
616 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
617 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
618 __nds32__fmfcfg): Define.
620 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
622 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
623 intrinsic register names.
624 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
625 intrinsic register enum values and macros.
627 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
629 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
630 for load/store addressing form.
631 (nds32_print_operand_address): Likewise.
633 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
636 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
637 based on LABEL_REF. Remove useless assertion.
638 (pic_address_needs_scratch): Fix formatting.
639 (sparc_legitimize_pic_address): Minor tweaks.
640 (sparc_delegitimize_address): Adjust assertion accordingly.
641 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
642 into symbolic_operand.
643 (movsi_high_pic_label_ref): Likewise.
644 (movsi_lo_sum_pic_label_ref): Likewise.
645 (movdi_pic_label_ref): Likewise.
646 (movdi_high_pic_label_ref): Likewise.
647 (movdi_lo_sum_pic_label_ref): Likewise.
649 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
651 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
652 custom LIB_SPEC setup.
654 2018-04-06 Ruslan Bukin <br@bsdpad.com>
655 Kito Cheng <kito.cheng@gmail.com>
657 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
658 * config/riscv/freebsd.h: New.
660 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
662 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
663 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
666 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
667 Kito Cheng <kito.cheng@gmail.com>
669 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
670 nds32_output_call, nds32_symbol_binds_local_p): New functions.
671 * config/nds32/nds32-protos.h (nds32_output_call,
672 nds32_output_return): Declare.
673 * config/nds32/nds32.md: Refine all the call and return patterns.
675 2018-04-06 Jakub Jelinek <jakub@redhat.com>
678 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
679 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
681 PR rtl-optimization/84872
682 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
683 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
686 2018-04-06 Tamar Christina <tamar.christina@arm.com>
688 * expr.c (copy_blkmode_to_reg): Revert 254862.
689 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
691 2018-04-06 Richard Biener <rguenther@suse.de>
694 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
695 after seeing a component reference with an adjacent field. Treat
696 refs to arrays at struct end of external decls similar to
697 refs to unconstrained commons.
699 2018-04-06 Jakub Jelinek <jakub@redhat.com>
702 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
703 look through SAVE_EXPRs with non-side-effects argument. Adjust
705 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
708 2018-04-06 Richard Biener <rguenther@suse.de>
711 * alias.c (find_base_term): New wrapper around find_base_term
712 unwinding CSELIB_VAL_PTR changes.
713 (find_base_term): Do not restore CSELIB_VAL_PTR during the
716 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
718 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
720 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
721 constant definitions.
722 ("nop"): lr 0,0 -> nopr r0
723 ("nop_lr0", "nop_lr1"): New insn definitions.
725 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
727 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
728 NDS32_V3PUSH_AVAILABLE_P macro.
730 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
731 Chung-Ju Wu <jasonwucj@gmail.com>
733 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
734 (nds32*-*-*): Add float and fpu_config into supported_defaults.
735 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
736 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
737 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
738 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
739 * config/nds32/constraints.md: New constraints and checking for hard
741 * config/nds32/iterators.md: New mode iterator and attribute for hard
743 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
745 * config/nds32/nds32-fpu.md: New file.
746 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
747 deal with hard float code generation.
748 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
750 (abi_type, float_reg_number): New enum type.
751 * config/nds32/nds32-predicates.c: New predicates for hard float.
752 * config/nds32/nds32-protos.h: Declare functions for hard float.
753 * config/nds32/nds32.c: Implementation for hard float configuration.
754 * config/nds32/nds32.h: Definitions for hard float configuration.
755 * config/nds32/nds32.md: Include hard float machine description and
756 modify patterns for hard float configuration.
757 * config/nds32/nds32.opt: New options for hard float configuration.
758 * config/nds32/predicates.md: New predicates for hard float
761 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
763 * common/config/nds32/nds32-common.c
764 (nds32_option_optimization_table): Enable -mreleax-hint by default.
766 2018-04-05 Jakub Jelinek <jakub@redhat.com>
769 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
770 CONSTRUCTOR_ELT (ctor, ...)->value.
772 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
775 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
777 2018-04-05 Tom de Vries <tom@codesourcery.com>
780 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
783 2018-04-05 Shiva Chen <shiva0217@gmail.com>
784 Kito Cheng <kito.cheng@gmail.com>
786 * config/nds32/constraints.md (U33): Fine-tune checking condition.
787 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
788 * config/nds32/nds32.h (nds32_16bit_address_type): Add
789 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
791 2018-04-05 Shiva Chen <shiva0217@gmail.com>
792 Kito Cheng <kito.cheng@gmail.com>
794 * config/nds32/constraints.md (Ufe): New memory constraint.
795 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
796 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
797 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
799 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
800 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
802 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
804 * config/nds32/nds32.md: Use optimize_size in the condition for
805 alu-shift instructions.
807 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
809 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
811 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
813 * config/nds32/nds32.md (negsi2): Refine pattern.
815 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
816 Chung-Ju Wu <jasonwucj@gmail.com>
818 * config/nds32/iterators.md (shift_rotate): New code iterator.
819 (shift): New code attribute.
820 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
821 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
822 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
823 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
825 (andsi3, *andsi3): Ditto.
826 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
827 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
828 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
829 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
830 nds32_ior_operand, nds32_xor_operand): New predicates.
832 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
834 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
835 (addsi3, subsi3): ... this.
837 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
839 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
841 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
843 * config/nds32/nds32.md: Adjust indention.
845 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
847 * config/nds32/nds32.md (feature): New attribute.
849 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
851 * config/nds32/nds32.md (subtype): New attribute.
853 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
856 * config/arm/arm-builtins.c (arm_expand_builtin): Change
857 expansion to perform a bitwise AND of the argument followed by a
858 boolean negation of the result.
860 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
862 PR rtl-optimization/84878
863 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
864 the basic block. Assert the use reference is not artificial and that
865 it has an associated insn.
867 2018-04-04 Michael Matz <matz@suse.de>
869 * builtins.c (compute_objsize): Pass correct operand
870 to array_at_struct_end_p.
872 2018-04-04 Richard Biener <rguenther@suse.de>
875 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
876 from contexts for DINFO_LEVEL_TERSE and below.
878 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
880 * config/nds32/nds32-doubleword.md (move_<mode>): Require
881 resiter_operand condition.
882 * config/nds32/nds32.md (*move<mode>): Ditto.
884 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
885 Monk Chiang <sh.chiang04@gmail.com>
887 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
889 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
891 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
893 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
894 Kito Cheng <kito.cheng@gmail.com>
896 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
897 nds32_cond_code_str, output_cond_branch,
898 output_cond_branch_compare_zero, nds32_expand_cbranch,
899 nds32_expand_cstore, nds32_expand_movcc,
900 nds32_output_cbranchsi4_equality_zero,
901 nds32_output_cbranchsi4_equality_reg,
902 nds32_output_cbranchsi4_equality_reg_or_const_int,
903 nds32_output_cbranchsi4_greater_less_zero: New functions.
904 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
905 nds32_expand_cstore, nds32_expand_movcc,
906 nds32_output_cbranchsi4_equality_zero,
907 nds32_output_cbranchsi4_equality_reg,
908 nds32_output_cbranchsi4_equality_reg_or_const_int,
909 nds32_output_cbranchsi4_greater_less_zero): Declare.
910 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
911 nds32_rimm11s_operand): New predicates.
912 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
913 * config/nds32/nds32.md: Rewrite all the branch and conditional move
916 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
918 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
919 * config/nds32/nds32.md: Ditto.
920 * config/nds32/pipelines.md: Ditto.
922 2018-04-04 Richard Biener <rguenther@suse.de>
924 PR tree-optimization/85168
925 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
926 propagating abnormals.
928 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
930 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
932 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
933 Kito Cheng <kito.cheng@gmail.com>
935 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
936 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
937 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
938 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
939 * config/nds32/nds32.md (sibcall_internal): New.
940 (sibcall_register): Remove.
941 (sibcall_immediate): Remove.
942 (sibcall_value_internal): New.
943 (sibcall_value_register): Remove.
944 (sibcall_value_immediate): Remove.
945 * config/nds32/predicates.md (nds32_general_register_operand): New.
946 (nds32_call_address_operand): New.
948 2018-04-03 Jakub Jelinek <jakub@redhat.com>
950 PR rtl-optimization/85167
951 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
952 bb_defs if *split_p, instead preinitialize it to NULL.
954 PR tree-optimization/85156
955 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
956 evaluating the argument multiple times.
958 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
960 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
962 (_mm_cvtpd_ps): Likewise.
963 (_mm_cvttpd_epi32): Likewise.
964 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
965 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
966 vector, pixel, and bool following altivec.h include.
968 2018-04-03 Martin Sebor <msebor@redhat.com>
970 * doc/extend.texi (Common Function Attributes): Clarify.
971 (const attribute): Likewise.
972 (pure attribute): Likewise.
974 2018-04-03 Jakub Jelinek <jakub@redhat.com>
977 * config/i386/i386.c (ix86_expand_vector_set): Use
978 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
980 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
982 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
983 instructions when changing rounding bits to preserve precision bits
984 in the x87 control word.
986 2018-04-03 Martin Liska <mliska@suse.cz>
988 PR tree-optimization/82491
989 * rtl.h (strip_offset_and_add): Replace += suboffset with
990 poly_uint64 () + suboffset.
992 2018-03-29 Martin Liska <mliska@suse.cz>
993 Martin Jambor <mjambor@suse.cz>
996 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
997 param_type is not an integral or pointer type.
999 2018-04-03 Richard Biener <rguenther@suse.de>
1001 * sese.h (recompute_all_dominators): Remove.
1003 2018-04-02 Martin Sebor <msebor@redhat.com>
1005 * doc/invoke.texi (-Wrestrict): Fix typos.
1007 2018-04-02 Jim Wilson <jimw@sifive.com>
1009 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
1010 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
1011 (<optab>di3, <optab>si3_extend): Likewise.
1012 (<optab>si3_mask, <optab>si3_mask_1): New.
1013 (<optab>di3_mask, <optab>di3_mask_1): New.
1014 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
1015 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
1016 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
1018 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
1020 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
1023 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
1025 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
1026 (nds32_canonicalize_comparison): New function.
1028 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1029 Kito Cheng <kito.cheng@gmail.com>
1030 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1032 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
1033 * config/nds32/constants.md (unspec_volatile_element): Add
1034 UNSPEC_VOLATILE_RELAX_GROUP.
1035 * config/nds32/nds32-relax-opt.c: New file.
1036 * config/nds32/nds32-predicates.c
1037 (nds32_symbol_load_store_p): New function.
1038 * config/nds32/nds32-protos.h
1039 (nds32_symbol_load_store_p): Declare function.
1040 (make_pass_nds32_relax_opt): Declare new rtl pass function.
1041 * config/nds32/nds32.c
1042 (nds32_register_pass): New function to register pass.
1043 (nds32_register_passes): New function to register passes.
1044 * config/nds32/nds32.md (relax_group): New pattern.
1045 * config/nds32/nds32.opt (mrelax-hint): New option.
1046 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
1048 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
1050 * config/nds32/t-nds32: Modify files dependency.
1052 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1054 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
1055 (PROFILE_HOOK): Define its implementation.
1057 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1059 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
1060 type and 32-bit size.
1062 2018-04-01 Jakub Jelinek <jakub@redhat.com>
1065 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
1066 (V_128_256): New mode iterator.
1067 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
1068 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
1069 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
1071 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
1072 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
1074 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
1077 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
1078 NaN inputs correctly.
1080 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
1083 * config/rs6000/vsx.md (??r): New mode attribute.
1084 (*vsx_mov<mode>_64bit): Use it.
1085 (*vsx_mov<mode>_32bit): Likewise.
1087 2018-03-30 Martin Sebor <msebor@redhat.com>
1089 PR tree-optimization/84818
1090 * builtins.c (check_access): Use warning_n.
1092 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1095 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
1097 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
1100 2018-03-30 Julia Koval <julia.koval@intel.com>
1103 * x86-tune.def (movx, partial_reg_dependency): Enable for
1106 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
1109 * lra-constraints.c (process_alt_operands): Move setting
1110 this_alternative_matches below.
1112 2018-03-29 Martin Liska <mliska@suse.cz>
1115 * doc/invoke.texi: Document how LTO works with debug info.
1116 Describe auto-load support of binutils. Mention 'x86-64'
1117 as valid option value of -march option.
1119 2018-03-29 Jakub Jelinek <jakub@redhat.com>
1121 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
1124 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
1125 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
1126 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
1129 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1132 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
1133 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
1134 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
1135 for RS6000_BTM_POWERPC64.
1136 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
1137 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
1138 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
1143 2018-03-28 Carl Love <cel@us.ibm.com>
1146 2017-09-27 Carl Love <cel@us.ibm.com>
1148 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
1149 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
1150 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
1153 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1155 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
1156 instead of __vector bool.
1157 (_mm_max_pu8): Likewise.
1158 (_mm_min_pi16): Likewise.
1160 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1163 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
1164 (DIVWEUO): Likewise.
1166 (DIVDEUO): Likewise.
1167 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
1168 DIVWEUO and DIVDEUO.
1169 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
1170 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
1171 (div_extend): Likewise.
1172 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
1174 (__builtin_divweuo): Likewise.
1175 (__builtin_divdeo): Likewise.
1176 (__builtin_divdeuo): Likewise.
1178 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1181 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
1182 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
1184 PR tree-optimization/82004
1185 * gimple-match-head.c (optimize_pow_to_exp): New function.
1186 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
1187 Don't fold to exp if optimize_pow_to_exp is false.
1189 2018-03-28 Martin Liska <mliska@suse.cz>
1192 * calls.c (initialize_argument_information): Fix trailing space.
1193 * common.opt: Fix typo and provide better explanation for
1194 -fsanitize-coverage option.
1195 * config/i386/i386.opt: Fix typo.
1197 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1198 Martin Liska <mliska@suse.cz>
1201 * gimplify.c (asan_poison_variable): Don't do the check for
1202 gimplify_omp_ctxp here.
1203 (gimplify_decl_expr): Do it here.
1204 (gimplify_target_expr): Likewise.
1206 2018-03-28 Martin Liska <mliska@suse.cz>
1209 * config/i386/i386.c (ix86_function_arg_advance): Do not call
1210 chkp_type_bounds_count if MPX is not enabled.
1212 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
1214 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
1216 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
1219 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
1220 function to create the function decl for complex long double
1221 multiply and divide for -mabi=ieeelongdouble.
1222 (init_float128_ieee): Call it.
1224 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
1227 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
1228 -fcf-protection=branch -mibt.
1229 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
1231 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1234 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
1236 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
1239 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
1240 extern array declarations.
1242 2018-03-27 Richard Biener <rguenther@suse.de>
1245 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
1246 explicit single_use checks.
1248 2018-03-27 Richard Biener <rguenther@suse.de>
1250 PR tree-optimization/85082
1251 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
1254 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1256 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
1257 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
1258 Turn on fasynchronous-unwind-tables and funwind-tables.
1260 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
1263 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
1264 (*bmi_blsr_<mode>_ccz): Ditto.
1266 2018-03-26 Tom de Vries <tom@codesourcery.com>
1268 PR tree-optimization/85063
1269 * omp-general.c (offloading_function_p): New function. Factor out
1271 * omp-offload.c (pass_omp_target_link::gate): ... here.
1272 * omp-general.h (offloading_function_p): Declare.
1273 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
1274 with attribute omp declare target for offloading functions.
1276 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
1278 PR tree-optimization/84005
1279 * tree-data-ref.h (get_base_for_alignment): Declare.
1280 * tree-data-ref.c (get_base_for_alignment_1): New function.
1281 (get_base_for_alignment): Likewise.
1282 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
1283 get_base_for_alignment to find a suitable base object, instead
1284 of always using drb->base_address.
1286 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1289 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
1290 known size by default.
1292 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
1295 * lra-constraints.c (process_alt_operands): Don't match BLKmode
1296 and non BLKmode operands.
1298 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1301 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
1302 Clean up attributes.
1304 2018-03-23 Richard Biener <rguenther@suse.de>
1307 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
1308 we are going to emit early debug for LTO.
1310 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1313 * function.c (match_asm_constraints_1): Don't optimize if input
1314 doesn't satisfy general_operand predicate for output's mode.
1317 * alias.c (write_dependence_p): Don't require for x_canonicalized
1318 non-VOIDmode if x has VOIDmode.
1321 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
1322 just don't try to optimize it rather than assert it never happens.
1324 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1326 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
1327 macro expansions for definition of ST_INTERNAL_<mode> and
1328 LD_INTERNAL_<mode> builtins.
1329 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
1331 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
1333 (altivec_expand_st_builtin): Likewise.
1334 (altivec_expand_builtin): Remove calls to deleted functions.
1335 (rs6000_address_for_altivec): Delete this function.
1336 * config/rs6000/vector.md: Remove expands for
1337 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
1339 2018-03-22 Sudakshina Das <sudi.das@arm.com>
1342 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
1343 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
1344 re-computing once computed.
1345 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
1346 (arm_init_machine_status): Initialize
1347 machine->static_chain_stack_bytes.
1349 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1352 * doc/extend.texi: Add four new prototypes for vec_ld.
1353 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
1354 definitions for more logical presentation.
1355 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
1356 entries for V1TI variants of __builtin_altivec_ld builtin.
1357 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
1358 handling of V1TI variant of LVX icode pattern.
1359 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
1360 (rs6000_gimple_fold_builtin): Likewise.
1361 (altivec_init_builtins): Add code to define
1362 __builtin_altivec_lvx_v1ti function.
1364 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1367 * function.c (match_asm_constraints_1): Don't do the optimization
1368 if input isn't a REG, SUBREG, MEM or constant.
1370 2018-03-22 Tom de Vries <tom@codesourcery.com>
1372 PR tree-optimization/84956
1373 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
1374 bb_has_abnormal_pred.
1376 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1379 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
1380 DECL_INITIAL (decl) to decl at the end.
1381 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
1384 2018-03-21 Joseph Myers <joseph@codesourcery.com>
1386 * doc/extend.texi (__builtin_tgmath): Document when complex
1387 integer types are treated as _Complex _Float64.
1389 2018-03-21 Tom de Vries <tom@codesourcery.com>
1391 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
1393 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1395 PR tree-optimization/84960
1396 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
1397 if it is ENTRY block, move them into single succ of ENTRY in that case.
1399 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
1401 PR tree-optimization/84811
1402 * poly-int.h (poly_span_traits): Remove the T3 parameter and
1403 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
1404 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
1405 (known_subrange_p): Update accordingly. Cast each value involved
1406 in the size comparison, rather than casting the result of the
1409 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1411 PR tree-optimization/84982
1412 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
1413 by flipping the least significant bit rather than all bits from
1414 bitpos to bitpos + bitsize - 1.
1416 2018-03-21 Nathan Sidwell <nathan@acm.org>
1418 * doc/extend.texi (Deprecated Features): Remove mention of
1419 long-deleted deprecations.
1421 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1424 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
1425 * configure: Regenerate.
1427 2018-03-21 Tom de Vries <tom@codesourcery.com>
1429 PR tree-optimization/83126
1430 * tree-parloops.c (num_phis): New function.
1431 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
1433 2018-03-21 Nathan Sidwell <nathan@acm.org>
1435 * doc/extend.texi (Deprecated Features): Update deprecated flags,
1436 mention anon-struct/union members and trailing attributes.
1438 2018-03-21 Bin Cheng <bin.cheng@arm.com>
1440 PR tree-optimization/84969
1441 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
1442 builtin memset partitions if they set different rhs values.
1444 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1446 PR rtl-optimization/84989
1447 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
1448 VEC_DUPLICATE with scalar result mode.
1450 2018-03-21 Martin Liska <mliska@suse.cz>
1453 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
1454 not intended return statement.
1456 2018-03-21 Martin Liska <mliska@suse.cz>
1459 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
1460 (chkp_find_bound_slots_1): Limit number of iterations.
1462 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
1465 * Minor grammar fixes for x86 options.
1467 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1470 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
1471 holding REG_CFA_RESTORE notes, instead turn them into a USE.
1473 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
1476 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
1477 (altivec_lvx_<mode>_1op): Likewise.
1478 (altivec_stvx_<mode>_2op): Likewise.
1479 (altivec_stvx_<mode>_1op): Likewise.
1480 (altivec_lvx_<VM2:mode>): New define_expand.
1481 (altivec_stvx_<VM2:mode>): Likewise.
1482 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
1483 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1484 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
1485 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1486 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
1487 (rs6000_gen_lvx): Likewise.
1488 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
1489 (altivec_expand_stv_builtin): Likewise.
1490 (altivec_expand_builtin): Likewise.
1491 * config/rs6000/vector.md: Likewise.
1493 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1496 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
1499 2018-03-20 Richard Biener <rguenther@suse.de>
1502 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
1503 sign-conversions as zero, fall back to standard scalar_stmt
1506 2018-03-20 Martin Liska <mliska@suse.cz>
1509 * predict.c (rebuild_frequencies): Handle case when we have
1510 PROFILE_ABSENT, but flag_guess_branch_prob is false.
1512 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1515 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
1516 flag_section_anchors.
1517 * varasm.c (use_blocks_for_decl_p): Remove hack for
1518 dw2_force_const_mem.
1521 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
1523 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
1524 be created, use lowpart_subreg of operands[0] rather than operands[0]
1526 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
1527 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
1528 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
1529 and n constraint instead of aarch64_shift_imm_di and Usd.
1530 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
1531 (*aarch64_<optab>_reg_minus<mode>3): ... this.
1533 2018-03-20 Sudakshina Das <sudi.das@arm.com>
1536 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
1537 to favor GPR over NEON registers.
1538 (<shift>di3_neon): Likewise.
1540 2018-03-20 Tom de Vries <tom@codesourcery.com>
1543 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
1544 (nvptx_process_pars): Emit bar.sync asap and alap.
1546 2018-03-20 Tom de Vries <tom@codesourcery.com>
1549 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
1550 seen_label if seen_label is already set.
1552 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1555 * config/i386/i386.c (fold_builtin_cpu): For features above 31
1556 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
1557 Use 1U instead of 1. Formatting fixes.
1560 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
1561 instead of TREE_TYPE (s1) for the return value.
1563 2018-03-19 Jakub Jelinek <jakub@redhat.com>
1565 PR tree-optimization/84946
1566 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
1567 bitsize + bitsize in poly_uint64 rather than poly_int64.
1570 * dwarf2asm.c: Include fold-const.c.
1571 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
1572 of decl rather than decl itself.
1574 PR rtl-optimization/84643
1575 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
1577 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
1580 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
1581 calling assemble_variable.
1583 2018-03-19 Sudakshina Das <sudi.das@arm.com>
1586 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
1587 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
1589 2018-03-19 Jim Wilson <jimw@sifive.com>
1592 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
1593 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
1594 (riscv_first_stack_step): Likewise.
1595 (riscv_option_override): Use STACK_BOUNDARY instead of
1597 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
1599 (BIGGEST_ALIGNMENT): Set to 128.
1600 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
1601 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
1604 2018-03-19 Richard Biener <rguenther@suse.de>
1606 PR tree-optimization/84933
1607 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
1608 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
1610 2018-03-19 Richard Biener <rguenther@suse.de>
1612 PR tree-optimization/84859
1613 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
1614 (cond_if_else_store_replacement): Perform sinking operation on
1615 single-store BBs regardless of MAX_STORES_TO_SINK setting.
1616 Generalize what a BB with a single eligible store is.
1618 2018-03-19 Richard Biener <rguenther@suse.de>
1620 PR tree-optimization/84929
1621 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
1622 chrec_is_positive against non-chrec arg.
1624 2018-03-19 Tamar Christina <tamar.christina@arm.com>
1627 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
1629 2018-03-18 Martin Liska <mliska@suse.cz>
1631 PR rtl-optimization/84635
1632 * regrename.c (build_def_use): Use matches_mode only when
1635 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
1637 PR tree-optimization/84913
1638 * tree-vect-loop.c (vectorizable_reduction): Don't try to
1639 vectorize chains of COND_EXPRs.
1641 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1643 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
1645 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1647 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
1649 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1651 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
1653 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1654 Kito Cheng <kito.cheng@gmail.com>
1656 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
1657 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
1658 (nds32_adjust_reg_alloc_order): New function.
1659 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
1661 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
1663 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
1664 nds32_print_operand, nds32_print_operand_address): Use
1665 HOST_WIDE_INT_PRINT_DEC instead.
1667 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1669 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
1671 2018-03-17 Jakub Jelinek <jakub@redhat.com>
1674 * config/i386/i386.c (initial_ix86_tune_features,
1675 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
1677 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
1678 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
1679 rather than 1u << ix86_tune. Formatting fix.
1680 (ix86_option_override_internal): Change ix86_arch_mask from
1681 unsigned int to unsigned HOST_WIDE_INT, initialize to
1682 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
1683 (ix86_function_specific_restore): Likewise.
1685 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1688 * postreload.c (reload_combine_recognize_pattern): Perform
1689 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
1690 truncate_int_for_mode the result for the destination's mode.
1693 * hsa-gen.c (mem_type_for_type): Fix comment typo.
1694 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
1696 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1699 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
1702 * lra-assigns.c (lra_split_hard_reg_for): Don't use
1703 regno_allocno_class_array and sorted_pseudos.
1704 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
1705 insns where regno is used.
1707 2018-03-16 Martin Liska <mliska@suse.cz>
1710 * multiple_target.c (create_dispatcher_calls): Redirect
1711 reference in the symbol table.
1713 2018-03-16 Martin Liska <mliska@suse.cz>
1716 * multiple_target.c (create_dispatcher_calls): Redirect also
1719 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1723 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
1724 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
1725 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
1727 2018-03-16 Julia Koval <julia.koval@intel.com>
1729 * doc/invoke.texi (Skylake Server): Add CLWB.
1730 Cannonlake): Remove CLWB.
1732 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1734 PR tree-optimization/84841
1735 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
1737 (FLOAT_ONE_CONST_TYPE): Define.
1738 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
1739 (sort_by_operand_rank): Put entries with higher constant_type last
1740 rather than first to match comments.
1742 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
1744 * config/nios2/nios2.md (movsi_internal): Fix thinko in
1747 2018-03-15 Jakub Jelinek <jakub@redhat.com>
1750 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
1751 check and use address of target always.
1753 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
1756 * config/i386/i386.c (indirect_thunk_needed): Update comments.
1757 (indirect_thunk_bnd_needed): Likewise.
1758 (indirect_thunks_used): Likewise.
1759 (indirect_thunks_bnd_used): Likewise.
1760 (indirect_return_needed): New.
1761 (indirect_return_bnd_needed): Likewise.
1762 (output_indirect_thunk_function): Add a bool argument for
1764 (output_indirect_thunk_function): Don't generate alias for
1765 function return thunk.
1766 (ix86_code_end): Call output_indirect_thunk_function to generate
1767 function return thunks.
1768 (ix86_output_function_return): Set indirect_return_bnd_needed
1769 and indirect_return_needed instead of indirect_thunk_bnd_needed
1770 and indirect_thunk_needed.
1772 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
1774 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
1775 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
1776 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
1778 2018-03-15 David Malcolm <dmalcolm@redhat.com>
1779 Paul Hua <paul.hua.gm@gmail.com>
1782 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
1784 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
1786 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
1787 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
1790 2018-03-15 Tamar Christina <tamar.christina@arm.com>
1793 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
1794 instead of GET_MODE_SIZE when comparing Units.
1796 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1799 * varasm.c (hash_section): Return an unchangeble hash value
1800 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
1801 Return !aarch64_can_use_per_function_literal_pools_p ().
1803 2018-03-15 Jakub Jelinek <jakub@redhat.com>
1806 * optabs.c (emit_conditional_move): Pass address of cmode's copy
1807 rather than address of cmode as last argument to prepare_cmp_insn.
1809 2018-03-15 Julia Koval <julia.koval@intel.com>
1811 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
1812 F_AVX512VNNI, F_AVX512BITALG): New.
1814 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
1817 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
1818 insn for floating-point loads and stores.
1820 2018-03-14 Carl Love <cel@us.ibm.com>
1822 * config/rs6000/rs6000-c.c: Add macro definitions for
1823 ALTIVEC_BUILTIN_VEC_PERMXOR.
1824 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
1825 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
1826 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
1827 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
1829 * config/doc/extend.texi: Add prototypes for vec_permxor.
1831 2018-03-14 David Malcolm <dmalcolm@redhat.com>
1834 * diagnostic-show-locus.c (class layout_point): Convert m_line
1835 from int to linenum_type.
1836 (line_span::comparator): Use linenum "compare" function when
1837 comparing line numbers.
1838 (test_line_span): New function.
1839 (layout_range::contains_point): Convert param "row" from int to
1841 (layout_range::intersects_line_p): Likewise.
1842 (layout::will_show_line_p): Likewise.
1843 (layout::print_source_line): Likewise.
1844 (layout::should_print_annotation_line_p): Likewise.
1845 (layout::print_annotation_line): Likewise.
1846 (layout::print_leading_fixits): Likewise.
1847 (layout::annotation_line_showed_range_p): Likewise.
1848 (struct line_corrections): Likewise for field m_row.
1849 (line_corrections::line_corrections): Likewise for param "row".
1850 (layout::print_trailing_fixits): Likewise.
1851 (layout::get_state_at_point): Likewise.
1852 (layout::get_x_bound_for_row): Likewise.
1853 (layout::print_line): Likewise.
1854 (diagnostic_show_locus): Likewise for locals "last_line" and
1856 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
1857 * input.c (selftest::test_linenum_comparisons): New function.
1858 (selftest::input_c_tests): Call it.
1859 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
1860 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
1861 * selftest.h (ASSERT_GT): New macro.
1862 (ASSERT_GT_AT): New macro.
1863 (ASSERT_LT): New macro.
1864 (ASSERT_LT_AT): New macro.
1866 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
1868 PR rtl-optimization/84780
1869 * combine.c (distribute_links): Don't make a link based on pc_rtx.
1871 2018-03-14 Martin Liska <mliska@suse.cz>
1873 * tree.c (record_node_allocation_statistics): Use
1874 get_stats_node_kind.
1875 (get_stats_node_kind): New function extracted from
1876 record_node_allocation_statistics.
1877 (free_node): Use get_stats_node_kind.
1879 2018-03-14 Richard Biener <rguenther@suse.de>
1881 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
1882 that the value-set of ANTIC_IN doesn't grow.
1885 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1887 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1888 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1890 2018-03-14 Julia Koval <julia.koval@intel.com>
1892 * config.gcc (icelake-client, icelake-server): New.
1894 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
1895 (initial_ix86_arch_features): Ditto.
1896 (PTA_SKYLAKE): Add SGX.
1897 (PTA_ICELAKE): Remove.
1898 (PTA_ICELAKE_CLIENT): New.
1899 (PTA_ICELAKE_SERVER): New.
1900 (ix86_option_override_internal): Split up icelake on icelake client and
1902 (get_builtin_code_for_version): Ditto.
1903 (fold_builtin_cpu): Ditto.
1904 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
1905 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
1906 * config/i386/i386.h (processor_type): Ditto.
1907 * doc/invoke.texi: Ditto.
1909 2018-03-14 Jakub Jelinek <jakub@redhat.com>
1912 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
1913 INTEGER_CST offset, add it together with bitpos / 8 and
1914 sign extend based on POINTER_SIZE.
1918 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
1921 * config/i386/constraints.md (Yc): New register constraint.
1922 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
1923 Use Yc constraint for alternative 2 of operand 0. Remove
1924 preferred_for_speed attribute.
1926 2018-03-14 Richard Biener <rguenther@suse.de>
1928 PR tree-optimization/84830
1929 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
1930 with the old one to avoid oscillations.
1932 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
1935 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
1937 (assign_by_spills): Return a flag of reload assignment failure.
1938 Do not process the reload assignment failures. Do not spill other
1939 reload pseudos if they has the same reg class. Update n if
1941 (lra_assign): Add a return arg. Set up from the result of
1942 assign_by_spills call.
1943 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1944 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1945 usage_insns if it is not NULL.
1946 (spill_hard_reg_in_range): New function.
1947 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1948 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1949 function prototypes.
1950 (lra_assign): Change prototype.
1951 * lra.c (lra): Add code to deal with fails by splitting hard reg
1954 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
1956 * config/riscv/riscv.opt (mrelax): New option.
1957 * config/riscv/riscv.c (riscv_file_start): Emit ".option
1958 "norelax" when riscv_mrelax is disabled.
1959 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
1961 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1964 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
1965 reassociation for int modes.
1967 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1969 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
1970 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
1972 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
1973 * config/aarch64/aarch64-sve.md
1974 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
1975 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
1976 (*extend<mode><Vwide>2): Rename to...
1977 (aarch64_sve_extend<mode><Vwide>2): ...this.
1978 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
1979 renaming the old pattern to...
1980 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
1982 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
1983 define_expand, renaming the old pattern to...
1984 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
1985 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
1986 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
1987 account when deciding which SVE instruction the optab should use.
1988 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
1990 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1992 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
1993 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
1994 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
1995 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
1996 (tlsdesc_small_<mode>): Turn a define_expand and use
1997 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
1998 (tlsdesc_small_advsimd_<mode>): ...this.
1999 (tlsdesc_small_sve_<mode>): New pattern.
2001 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2003 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
2004 (UNSPEC_UMUL_HIGHPART): New constants.
2005 (MUL_HIGHPART): New int iteraor.
2006 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
2007 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
2009 (*<su>mul<mode>3_highpart): New define_insn.
2011 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2014 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2017 2018-03-13 Martin Liska <mliska@suse.cz>
2020 * (sem_item_optimizer::sem_item_optimizer): Initialize new
2022 (sem_item_optimizer::~sem_item_optimizer): Release it.
2023 (sem_item_optimizer::merge_classes): Register variable aliases.
2024 (sem_item_optimizer::fixup_pt_set): New function.
2025 (sem_item_optimizer::fixup_points_to_sets): Likewise.
2026 * ipa-icf.h: Declare new variables and functions.
2028 2018-03-13 Jakub Jelinek <jakub@redhat.com>
2031 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
2032 integer_pow2p@2 and test integer_pow2p in condition.
2033 (A < 0 ? C : 0): Similarly for @1.
2036 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
2037 characters starting at p contain '\0' character, don't look beyond
2041 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
2042 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
2045 * reg-stack.c (change_stack): Change update_end var from int to
2046 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
2047 also call set_block_for_insn on the newly added insns and rescan.
2050 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
2051 on the last operand.
2054 * tree.c (stabilize_reference_1): Return save_expr (e) for
2055 STATEMENT_LIST even if it doesn't have side-effects.
2057 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
2059 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
2061 2018-03-12 Renlin Li <renlin.li@arm.com>
2063 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
2064 aarch64_output_scalar_simd_mov_immediate.
2066 2018-03-12 Martin Sebor <msebor@redhat.com>
2068 PR tree-optimization/83456
2069 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
2070 for perfectly overlapping calls to memcpy.
2071 (gimple_fold_builtin_memory_chk): Same.
2072 (gimple_fold_builtin_strcpy): Handle no-warning.
2073 (gimple_fold_builtin_stxcpy_chk): Same.
2074 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
2076 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2078 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
2079 parameter. Use it for SFmode.
2080 (rs6000_function_arg_advance_1): Adjust.
2081 (rs6000_function_arg): Adjust.
2082 (rs6000_gimplify_va_arg): Pass false for that new parameter.
2084 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2086 PR rtl-optimization/84169
2087 PR rtl-optimization/84780
2088 * combine.c (can_combine_p): Check for a 2-insn combination whether
2089 the destination register is used between the two insns, too.
2091 2018-03-12 Richard Biener <rguenther@suse.de>
2093 PR tree-optimization/84803
2094 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
2095 for refs DR analysis didn't process.
2097 2018-03-12 Richard Biener <rguenther@suse.de>
2099 PR tree-optimization/84777
2100 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
2101 force-vectorize loops ignore whether we are optimizing for size.
2103 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
2105 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
2106 (TARGET_MD_ASM_ADJUST): Define.
2108 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
2109 Kito Cheng <kito.cheng@gmail.com>
2110 Chung-Ju Wu <jasonwucj@gmail.com>
2112 * config/nds32/nds32.c (nds32_compute_stack_frame,
2113 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
2114 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
2115 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
2116 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
2117 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
2118 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
2119 * config/nds32/nds32.md (prologue, epilogue): Use macro
2120 NDS32_V3PUSH_AVAILABLE_P to do checking.
2122 2018-03-11 Jakub Jelinek <jakub@redhat.com>
2125 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
2126 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
2127 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
2128 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
2129 addition of most attributes on !orig_type_die or the attribute not
2130 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
2132 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2133 Chung-Ju Wu <jasonwucj@gmail.com>
2135 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
2137 * config/nds32/nds32.opt (mvh): New option.
2139 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2140 Chung-Ju Wu <jasonwucj@gmail.com>
2142 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
2144 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
2145 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
2148 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2149 Chung-Ju Wu <jasonwucj@gmail.com>
2151 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
2153 * config/nds32/nds32-multiple.md (strlensi): New pattern.
2154 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
2156 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
2157 Kito Cheng <kito.cheng@gmail.com>
2158 Chung-Ju Wu <jasonwucj@gmail.com>
2160 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
2161 UNSPEC_FFMISM and UNSPEC_FLMISM.
2162 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
2163 for ffb, ffmism and flmism.
2164 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
2165 (unspec_ffmism): Ditto.
2166 (unspec_flmism): Ditto.
2167 (nds32_expand_builtin_impl): Check if string extension is available.
2168 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
2169 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
2171 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
2174 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2177 * lra-assigns.c (assign_by_spills): Return a flag of reload
2178 assignment failure. Do not process the reload assignment
2179 failures. Do not spill other reload pseudos if they has the same
2181 (lra_assign): Add a return arg. Set up from the result of
2182 assign_by_spills call.
2183 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2184 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2185 usage_insns if it is not NULL.
2186 (spill_hard_reg_in_range): New function.
2187 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2188 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2189 function prototypes.
2190 (lra_assign): Change prototype.
2191 * lra.c (lra): Add code to deal with fails by splitting hard reg
2194 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
2197 * config/i386/i386.opt: Replace Enforcment with Enforcement.
2199 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
2202 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
2203 (dw_val_node): Add val_symbolic_view.
2204 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
2205 (symview_upper_bound): New.
2206 (new_line_info_table): Initialize symviews_since_reset.
2207 (dwarf2out_source_line): Count symviews_since_reset and set
2208 symview_upper_bound.
2209 (dw_val_equal_p): Handle symview.
2210 (add_AT_symview): New.
2211 (print_dw_val): Handle symview.
2212 (attr_checksum, attr_checksum_ordered): Likewise.
2213 (same_dw_val_p, size_of_die): Likewise.
2214 (value_format, output_die): Likewise.
2215 (add_high_low_attributes): Use add_AT_symview for entry_view.
2216 (dwarf2out_finish): Reset symview_upper_bound, clear
2219 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
2222 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
2223 Add strict argument and use it.
2224 (rs6000_split_multireg_move): Update for new strict argument.
2225 (mem_operand_gpr): Disallow all non-offsettable addresses.
2226 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
2228 2018-03-09 Jakub Jelinek <jakub@redhat.com>
2231 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
2232 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
2233 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
2236 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
2237 decl, use remap_type if we want to use the type.
2239 2018-03-09 Martin Sebor <msebor@redhat.com>
2241 PR tree-optimization/84526
2242 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2244 (builtin_access::generic_overlap): Be prepared to handle non-array
2247 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
2249 PR rtl-optimization/84682
2250 * lra-constraints.c (process_address_1): Check is_address flag
2251 for address constraints.
2252 (process_alt_operands): Likewise.
2253 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
2254 preprocess_constraints.
2255 * recog.h (preprocess_constraints): Add oploc parameter.
2257 * recog.c (preprocess_constraints): Test address_operand for
2258 CT_ADDRESS constraints.
2260 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2263 * lra-assigns.c (assign_by_spills): Return a flag of reload
2264 assignment failure. Do not process the reload assignment
2265 failures. Do not spill other reload pseudos if they has the same
2267 (lra_assign): Add a return arg. Set up from the result of
2268 assign_by_spills call.
2269 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2270 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2271 usage_insns if it is not NULL.
2272 (spill_hard_reg_in_range): New function.
2273 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2274 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2275 function prototypes.
2276 (lra_assign): Change prototype.
2277 * lra.c (lra): Add code to deal with fails by splitting hard reg
2280 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2283 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
2284 Accept complain bool parameter. Only emit errors if it is true.
2285 (arm_parse_cpu_option_name): Likewise.
2286 (arm_target_thumb_only): Adjust callers of the above.
2287 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
2288 prototype to take a default true bool parameter.
2289 (arm_parse_arch_option_name): Likewise.
2291 2018-03-09 David Malcolm <dmalcolm@redhat.com>
2292 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
2296 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
2297 * configure: Regenerate.
2298 * configure.ac ("linker --version-script option"): New.
2299 ("linker soname option"): New.
2301 2018-03-09 Richard Biener <rguenther@suse.de>
2303 PR tree-optimization/84775
2304 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
2305 immediate uses of predicate stmts and mark them modified.
2308 PR tree-optimization/84178
2309 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2311 (version_loop_for_if_conversion): Delay update_ssa call.
2312 (tree_if_conversion): Delay update_ssa until after predicate
2315 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
2318 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
2319 when the function accesses prior frames.
2321 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2324 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
2325 gen_llsym, otherwise call maybe_gen_llsym.
2328 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
2329 has ',' character inside of it.
2331 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2334 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
2335 as clobbering CC_REGNUM.
2337 2018-03-08 Richard Biener <rguenther@suse.de>
2340 * tree-scalar-evolution.c: Include tree-into-ssa.h.
2341 (follow_copies_to_constant): Do not follow SSA names registered
2344 2018-03-08 Richard Biener <rguenther@suse.de>
2346 PR tree-optimization/84178
2347 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2349 (version_loop_for_if_conversion): Delay update_ssa call.
2350 (tree_if_conversion): Delay update_ssa until after predicate
2353 2018-03-08 David Malcolm <dmalcolm@redhat.com>
2355 PR tree-optimization/84178
2356 * tree-if-conv.c (release_bb_predicate): Remove the
2357 the assertion that the stmts have NULL use_ops.
2358 Discard the statements, asserting that they haven't
2359 yet been added to a BB.
2361 2018-03-08 Richard Biener <rguenther@suse.de>
2363 PR tree-optimization/84746
2364 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
2365 (phi_translate): Pass in destination ANTIC_OUT set.
2366 (phi_translate_1): Likewise. For a simplified result lookup
2367 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
2368 (phi_translate_set): Adjust.
2369 (do_pre_regular_insertion): Likewise.
2370 (do_pre_partial_partial_insertion): Likewise.
2372 2018-03-08 Martin Liska <mliska@suse.cz>
2374 PR gcov-profile/84735
2375 * doc/gcov.texi: Document usage of profile files.
2376 * gcov-io.h: Document changes in the format.
2378 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
2382 * dwarf2out.c (struct dw_line_info_table): Update comments for
2384 (FORCE_RESET_NEXT_VIEW): New.
2385 (FORCE_RESETTING_VIEW_P): New.
2386 (RESETTING_VIEW_P): Check for -1 too.
2387 (ZERO_VIEW_P): Likewise.
2388 (new_line_info_table): Force-reset next view.
2389 (dwarf2out_begin_function): Likewise.
2390 (dwarf2out_source_line): Simplify zero_view_p initialization.
2391 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
2392 view directly. Omit view when omitting .loc at line 0.
2394 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2396 PR tree-optimization/84740
2397 * tree-switch-conversion.c (process_switch): Call build_constructors
2398 only if info.phi_count is non-zero.
2400 PR tree-optimization/84739
2401 * tree-tailcall.c (find_tail_calls): Check call arguments against
2402 DECL_ARGUMENTS (current_function_decl) rather than
2403 DECL_ARGUMENTS (func) when checking for tail recursion.
2405 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2407 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
2408 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
2409 Volker Reichelt's entry and add entries for people that perform
2410 GCC fuzzy testing and report numerous bugs.
2412 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
2415 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
2416 readonly data in sdata, if that is disabled.
2417 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
2418 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
2419 -mreadonly-in-sdata option.
2421 2018-03-07 Martin Sebor <msebor@redhat.com>
2423 PR tree-optimization/84468
2424 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
2425 basic block when looking for nul assignment.
2427 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
2430 * except.h (output_function_exception_table): Adjust prototype.
2431 * except.c (output_function_exception_table): Remove FNNAME parameter
2432 and add SECTION parameter. Ouput one part of the table at a time.
2433 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
2434 the first part of the exception table and emit unwind directives.
2435 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
2436 (i386_pe_seh_cold_init): Likewise.
2437 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
2438 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
2439 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
2440 (ix86_output_call_insn): Emit a nop in one more case for SEH.
2441 * config/i386/winnt.c: Include except.h.
2442 (struct seh_frame_state): Add reg_offset, after_prologue and
2443 in_cold_section fields.
2444 (i386_pe_seh_end_prologue): Set seh->after_prologue.
2445 (i386_pe_seh_cold_init): New function.
2446 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
2447 to seh->in_cold_section.
2448 (seh_emit_push): Record the offset of the push.
2449 (seh_emit_save): Record the offet of the save.
2450 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
2451 Test seh->after_prologue to disregard the epilogue.
2452 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
2453 (i386_pe_end_cold_function): New function.
2455 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2458 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
2459 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
2462 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
2464 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
2465 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
2468 * multiple_target.c: Include tree-inline.h and intl.h.
2469 (expand_target_clones): Diagnose and fail if node->definition and
2470 !tree_versionable_function_p (node->decl).
2472 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
2474 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
2476 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
2477 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
2478 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
2480 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2483 * combine.c (try_combine): Use reg_or_subregno instead of handling
2484 just paradoxical SUBREGs and REGs.
2486 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
2488 * config/arc/arc.c (arc_finalize_pic): Remove function.
2489 (arc_must_save_register): We use single base PIC register, remove
2490 checks to save/restore the PIC register.
2491 (arc_expand_prologue): Likewise.
2492 * config/arc/arc-protos.h (arc_set_default_type_attributes):
2494 (arc_verify_short): Likewise.
2495 (arc_attr_type): Likewise.
2496 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
2497 (walk_stores): Likewise.
2498 (arc_address_cost): Make it static.
2499 (arc_verify_short): Likewise.
2500 (branch_dest): Likewise.
2501 (arc_attr_type): Likewise.
2502 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
2503 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
2504 (arc_final_prescan_insn): Remove inserting the nops due to
2505 hardware hazards. It is done in reorg step.
2506 (insn_length_variant_t): Remove.
2507 (insn_length_parameters_t): Likewise.
2508 (arc_insn_length_parameters): Likewise.
2509 (arc_get_insn_variants): Likewise.
2510 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
2512 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2515 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
2518 PR tree-optimization/84687
2519 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
2521 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
2523 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2525 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
2526 Rename to ppc_speculation_barrier.
2527 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
2528 __builtin_ppc_speculation_barrier.
2530 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2533 * combine.c (combine_simplify_rtx): Don't try to simplify if
2534 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
2537 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
2539 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
2540 to 32 bytes when compiling for POWER9.
2542 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2545 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
2546 regparm >= 3 with no arg reg available also for calls with
2547 flag_force_indirect_call. Pass decl to ix86_function_regparm.
2550 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
2552 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
2554 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
2557 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
2559 2018-03-05 Richard Biener <rguenther@suse.de>
2561 PR tree-optimization/84486
2562 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
2563 When inserting a __builtin_assume_aligned call set the LHS
2564 SSA name alignment info accordingly.
2566 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
2568 PR tree-optimization/84114
2569 * config/aarch64/aarch64.c (aarch64_reassociation_width)
2570 Avoid reassociation of FLOAT_MODE addition.
2572 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
2574 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
2575 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
2576 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
2577 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
2578 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
2579 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
2580 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
2582 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
2583 __builtin_ia32_wbinvd): New builtins.
2584 (SPECIAL_ARGS2): New.
2585 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
2586 (SPECIAL_ARGS2): New.
2587 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
2588 (ix86_valid_target_attribute_inner_p): Ditto.
2589 (ix86_init_mmx_sse_builtins): Add special_args2.
2590 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
2591 TARGET_WBNOINVD_P): New.
2592 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
2593 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
2594 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
2595 * config/i386/immintrin.h (_wbinvd): New intrinsic.
2596 * config/i386/pconfigintrin.h: New file.
2597 * config/i386/wbnoinvdintrin.h: Ditto.
2598 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
2599 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
2601 2018-03-05 Richard Biener <rguenther@suse.de>
2603 PR tree-optimization/84670
2604 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2606 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2607 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2608 (compute_antic_aux): Only assert the number of values in ANTIC_IN
2609 doesn't grow if all successors (recursively) were visited at least
2612 2018-03-05 Richard Biener <rguenther@suse.de>
2614 PR tree-optimization/84650
2615 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
2616 if executed in the loop pipeline.
2618 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
2620 * doc/configfiles.texi (Configuration Files): Move info about
2621 conditionalizing $target-protos.h to...
2622 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
2623 differs from $target-protos.h.
2625 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
2626 Chung-Ju Wu <jasonwucj@gmail.com>
2628 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
2629 * config/nds32/nds32-multiple.md (setmemsi): Define.
2630 * config/nds32/nds32-memory-manipulation.c
2631 (nds32_gen_dup_4_byte_to_word_value): New.
2632 (emit_setmem_word_loop): New.
2633 (emit_setmem_byte_loop): New.
2634 (nds32_expand_setmem_loop): New.
2635 (nds32_expand_setmem_loop_v3m): New.
2636 (nds32_expand_setmem_unroll): New.
2637 (nds32_expand_setmem): New.
2639 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2640 Chung-Ju Wu <jasonwucj@gmail.com>
2642 * config/nds32/nds32-memory-manipulation.c
2643 (nds32_emit_load_store): New.
2644 (nds32_emit_post_inc_load_store): New.
2645 (nds32_emit_mem_move): New.
2646 (nds32_emit_mem_move_block): New.
2647 (nds32_expand_movmemsi_loop_unknown_size): New.
2648 (nds32_expand_movmemsi_loop_known_size): New.
2649 (nds32_expand_movmemsi_loop): New.
2650 (nds32_expand_movmemsi_unroll): New.
2651 (nds32_expand_movmemqi): Rename ...
2652 (nds32_expand_movmemsi): ... to this.
2653 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
2654 (movmemsi): ... to this.
2655 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
2656 (nds32_expand_movmemsi): ... to this.
2658 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2659 Monk Chiang <sh.chiang04@gmail.com>
2660 Chung-Ju Wu <jasonwucj@gmail.com>
2662 * config/nds32/nds32-protos.h
2663 (nds32_expand_load_multiple): New arguments.
2664 (nds32_expand_store_multiple): Ditto.
2665 (nds32_valid_multiple_load_store): Rename ...
2666 (nds32_valid_multiple_load_store_p): ... to this.
2667 * config/nds32/nds32-memory-manipulation.c
2668 (nds32_expand_load_multiple): Refine implementation.
2669 (nds32_expand_store_multiple): Ditto.
2670 * config/nds32/nds32-multiple.md
2671 (load_multiple): Update nds32_expand_load_multiple interface.
2672 (store_multiple): Update nds32_expand_store_multiple interface.
2673 * config/nds32/nds32-predicates.c
2674 (nds32_valid_multiple_load_store): Rename ...
2675 (nds32_valid_multiple_load_store_p): ... to this and refine
2677 * config/nds32/predicates.md
2678 (nds32_load_multiple_and_update_address_operation): New predicate.
2679 (nds32_store_multiple_and_update_address_operation): New predicate.
2681 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2682 Chung-Ju Wu <jasonwucj@gmail.com>
2684 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
2685 (combo): New attribute.
2686 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
2688 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
2690 * config/nds32/nds32.opt: Change -mcmodel= default value.
2692 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
2693 Monk Chiang <sh.chiang04@gmail.com>
2694 Chung-Ju Wu <jasonwucj@gmail.com>
2696 * config/nds32/constants.md (unspec_element): New enum.
2697 * config/nds32/constraints.md (Umw): New constraint.
2698 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
2699 * config/nds32/nds32-intrinsic.md: Likewise.
2700 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
2701 (nds32_valid_smw_lwm_base_p): New.
2702 (nds32_output_smw_single_word): New.
2703 (nds32_output_lmw_single_word): New.
2704 (nds32_expand_unaligned_load): New.
2705 (nds32_expand_unaligned_store): New.
2706 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
2707 (nds32_output_smw_single_word): Declare.
2708 (nds32_output_lmw_single_word): Declare.
2709 (nds32_expand_unaligned_load): Declare.
2710 (nds32_expand_unaligned_store): Declare.
2711 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
2712 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
2713 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
2714 NDS32_BUILTIN_UASTORE_DW.
2715 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
2718 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
2719 Kito Cheng <kito.cheng@gmail.com>
2720 Chung-Ju Wu <jasonwucj@gmail.com>
2722 * config/nds32/nds32-intrinsic.c
2723 (nds32_expand_builtin_null_ftype_reg): Delete.
2724 (nds32_expand_builtin_reg_ftype_imm): Ditto.
2725 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
2726 (nds32_read_argument): New.
2727 (nds32_legitimize_target): Ditto.
2728 (nds32_legitimize_argument): Ditto.
2729 (nds32_check_constant_argument): Ditto.
2730 (nds32_expand_unop_builtin): Ditto.
2731 (nds32_expand_unopimm_builtin): Ditto.
2732 (nds32_expand_binop_builtin): Ditto.
2733 (nds32_builtin_decl_impl): Ditto.
2734 (builtin_description): Ditto.
2735 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
2736 (nds32_init_builtins_impl): Ditto.
2737 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
2738 (nds32_builtin_decl): New.
2739 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
2740 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
2742 2018-03-02 Jeff Law <law@redhat.com>
2744 * reorg.c (stop_search_p): Handle DEBUG_INSN.
2745 (redundant_insn, fill_simple_delay_slots): Likewise.
2746 (fill_slots_from_thread): Likewise.
2747 * resource.c (mark_referenced_resources): Likewise.
2748 (mark_set_resources, find_dead_or_set_registers): Likewise.
2750 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2752 * substring-locations.h (format_warning_va): Formatting fix for
2754 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
2756 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
2757 * substring-locations.c: Include intl.h.
2758 (format_warning_va): Turned into small wrapper around
2759 format_warning_n_va, renamed to ...
2760 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
2761 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
2763 (format_warning_at_substring_n): New function.
2764 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
2765 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
2766 format_warning_at_substring with just a shorter name instead of
2767 const function pointer.
2768 (fmtwarn_n): New function.
2769 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
2770 appropriate, get rid of all the fmtstr temporaries, move conditionals
2771 with G_() wrapped string literals directly into fmtwarn arguments,
2772 cast dir.len to (int), formatting fixes.
2774 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
2776 * doc/invoke.texi: Remove "Cilk Plus" references.
2778 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2779 Richard Biener <rguenther@suse.de>
2782 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
2783 for error or warning attributes if CALL_FROM_THUNK_P is set.
2786 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2789 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
2790 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
2793 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
2794 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
2796 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
2797 instead of -1U in last predictors element's probability member.
2799 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2802 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2803 arguments if they are comparable.
2805 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
2807 PR tree-optimization/84634
2808 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
2809 masks and masked_loop_p with a single loop_masks, making sure it's
2810 null for bb vectorization.
2812 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
2814 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
2815 (vect_analyze_data_ref_access): Use loop->safe_len rather than
2816 loop->force_vectorize to check whether there is no alias.
2818 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2821 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
2823 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
2825 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
2826 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
2827 instead of a loop around prev_real_insn.
2828 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
2832 * config/i386/i386.c (ix86_print_operand): Use conditional
2833 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
2836 2018-03-02 Richard Biener <rguenther@suse.de>
2838 PR tree-optimization/84427
2839 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
2840 (bitmap_set_subtract_values): Rewrite to handle multiple
2843 (prune_clobbered_mems): Likewise.
2844 (phi_translate): Take edge instead of pred/phiblock.
2845 (phi_translate_1): Likewise.
2846 (phi_translate_set): Likewise. Insert all translated
2847 exprs for a value into the set, keeping possibly multiple
2848 expressions per value.
2849 (compute_antic_aux): Adjust for phi_translate changes.
2850 When intersecting union the expressions and prune those
2851 not in the final value set, keeping possibly multiple
2852 expressions per value. Do not use value-insertion
2853 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
2854 all expressions. Add verification that the value-sets
2855 only shrink during iteration.
2856 (compute_partial_antic_aux): Adjust for the phi_translate changes.
2857 (do_pre_regular_insertion): Likewise.
2858 (do_pre_partial_partial_insertion): Likewise.
2860 2018-03-02 Richard Biener <rguenther@suse.de>
2863 * config/darwin.c (saved_debug_info_level): New static global.
2864 (darwin_asm_lto_start): Disable debug info generation for LTO out.
2865 (darwin_asm_lto_end): Restore debug info generation settings.
2867 2018-03-01 Martin Liska <mliska@suse.cz>
2870 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
2873 2018-03-01 Richard Biener <rguenther@suse.de>
2876 * dwarf2out.c (gen_variable_die): Properly handle late VLA
2877 type annotation with LTO when debug was disabled at compile-time.
2879 2018-03-01 Matthew Fortune <mfortune@gmail.com>
2881 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
2883 (mips_final_postscan_insn): Likewise.
2885 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
2887 PR rtl-optimization/84528
2888 * alias.c (init_alias_target): Add commentary.
2889 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
2890 a unique base value if the frame pointer is not eliminated
2891 to the stack pointer.
2893 2018-03-01 Tom de Vries <tom@codesourcery.com>
2895 PR rtl-optimization/83327
2896 * lra-int.h (hard_regs_spilled_into): Declare.
2897 * lra.c (hard_regs_spilled_into): Define.
2898 (init_reg_info): Init hard_regs_spilled_into.
2899 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
2900 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
2901 (process_bb_lives): Handle hard_regs_spilled_into.
2902 (lra_create_live_ranges_1): Before doing liveness propagation, clear
2903 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
2905 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
2907 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
2908 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
2909 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
2910 * config/rs6000/aix72.h: New file.
2912 2018-02-28 Jakub Jelinek <jakub@redhat.com>
2914 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
2915 instead of warning_at with conditional singular and plural messages
2919 * stor-layout.c (update_alignment_for_field): For
2920 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
2921 && !DECL_PACKED (field), do the alignment update, just use
2922 only desired_align instead of MAX (type_align, desired_align)
2924 (place_field): Don't do known_align < desired_align handling
2925 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
2926 is non-NULL, instead do it after rli->prev_field handling and
2927 only if not within a bitfield word. For DECL_PACKED (field)
2928 use type_align of BITS_PER_UNIT.
2930 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
2932 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
2933 superfluous parentheses and trailing spaces.
2935 2018-02-28 Richard Biener <rguenther@suse.de>
2937 PR tree-optimization/84584
2938 * graphite-scop-detection.c (scop_detection::add_scop): Discard
2939 SCoPs with fake exit edge.
2941 2018-02-28 Martin Liska <mliska@suse.cz>
2944 * timevar.c (timer::print): Fix format to properly print 100%
2947 2018-02-28 Richard Biener <rguenther@suse.de>
2950 * genmatch.c (capture_info::walk_match): Do not mark
2951 captured expressions without operands as expr_p given
2952 they act more like predicates and should be subject to
2953 "lost tail" side-effect preserving.
2955 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
2957 PR rtl-optimization/81611
2958 * auto-inc-dec.c (attempt_change): Move dead note from
2959 mem_insn if it's the next use of regno
2960 (find_address): Take address use of reg holding
2961 non-incremented value. Add parm to limit search to the named
2963 (merge_in_block): Attempt to use a mem insn that is the next
2964 use of the original regno.
2966 2018-02-27 Martin Sebor <msebor@redhat.com>
2969 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
2970 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
2972 2018-02-27 Martin Sebor <msebor@redhat.com>
2974 PR translation/84207
2975 * diagnostic-core.h (warning_n, error_n, inform_n): Change
2976 n argument to unsigned HOST_WIDE_INT.
2977 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
2978 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
2979 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
2980 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
2982 2018-02-27 Richard Biener <rguenther@suse.de>
2984 PR tree-optimization/84512
2985 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
2986 Do not use the estimate returned from record_stmt_cost for
2987 the scalar iteration cost but sum properly using add_stmt_cost.
2989 2018-02-27 Richard Biener <rguenther@suse.de>
2991 PR tree-optimization/84466
2992 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2993 Adjust last change to less strictly validate use operands.
2995 2018-02-27 Martin Liska <mliska@suse.cz>
2997 PR gcov-profile/84548
2998 * gcov.c (process_file): Allow partial overlap and consider it
2999 also as group functions.
3000 (output_lines): Properly calculate range of lines for a group.
3002 2018-02-27 Martin Liska <mliska@suse.cz>
3004 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
3005 'ggc' suffixes. Change first column width.
3006 (timer::print): Fix formatting of the column.
3008 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
3010 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
3011 preserve inline entry blocks for the sake of debug inline
3012 entry point markers alone.
3013 (remove_unused_locals): Suggest in comments a better place to
3014 force the preservation of inline entry blocks that are
3015 otherwise unused, but do not preserve them.
3017 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3019 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
3021 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3024 * config/i386/constraints.md (Bs): Replace
3025 ix86_indirect_branch_register with
3026 TARGET_INDIRECT_BRANCH_REGISTER.
3028 * config/i386/i386.md (indirect_jump): Likewise.
3029 (tablejump): Likewise.
3030 (*sibcall_memory): Likewise.
3031 (*sibcall_value_memory): Likewise.
3032 Peepholes of indirect call and jump via memory: Likewise.
3033 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
3034 (*sibcall_value_GOT_32): Likewise.
3035 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3036 (GOT_memory_operand): Likewise.
3037 (call_insn_operand): Likewise.
3038 (sibcall_insn_operand): Likewise.
3039 (GOT32_symbol_operand): Likewise.
3040 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
3042 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3044 PR rtl-optimization/83496
3045 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
3046 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
3047 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
3048 redundant insn, if any.
3049 (relax_delay_slots): Likewise.
3050 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
3052 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
3054 PR tree-optimization/83965
3055 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
3056 that grouped statements are part of a reduction chain. Return
3057 true if the statement is not marked as a reduction itself but
3059 (vect_recog_dot_prod_pattern): Don't check whether the statement
3060 is part of a group here.
3061 (vect_recog_sad_pattern): Likewise.
3062 (vect_recog_widen_sum_pattern): Likewise.
3064 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3067 * final.c (rest_of_clean_state): Also look for calls inside sequences.
3069 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3072 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
3074 (ix86_output_indirect_function_return): New prototype.
3075 (ix86_split_simple_return_pop_internal): Likewise.
3076 * config/i386/i386.c (indirect_return_via_cx): New.
3077 (indirect_return_via_cx_bnd): Likewise.
3078 (indirect_thunk_name): Handle return va CX_REG.
3079 (output_indirect_thunk_function): Create alias for
3080 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
3081 (ix86_output_indirect_jmp): Remove the bool argument.
3082 (ix86_output_indirect_function_return): New function.
3083 (ix86_split_simple_return_pop_internal): Likewise.
3084 * config/i386/i386.md (*indirect_jump): Don't pass false
3085 to ix86_output_indirect_jmp.
3086 (*tablejump_1): Likewise.
3087 (simple_return_pop_internal): Change it to define_insn_and_split.
3088 Call ix86_split_simple_return_pop_internal to split it for
3090 (simple_return_indirect_internal): Call
3091 ix86_output_indirect_function_return instead of
3092 ix86_output_indirect_jmp.
3094 2018-02-26 Jakub Jelinek <jakub@redhat.com>
3097 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
3098 memset and value initialization afterwards.
3100 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
3102 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
3104 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3107 * common/config/aarch64/aarch64-common.c
3108 (aarch_option_optimization_table[]): Switch
3109 off fomit-frame-pointer
3111 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3112 Chung-Ju Wu <jasonwucj@gmail.com>
3114 * config/nds32/nds32-multiple.md (load_multiple): Disallow
3116 (store_multiple): Ditto.
3118 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3120 * config.gcc: Add --with-cpu support for nds32 target.
3121 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
3122 * config/nds32/nds32.opt: Add -mcpu= option.
3124 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
3126 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
3127 isel=yes): Warn for these deprecated options.
3129 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
3131 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
3132 ISA_2_5_MASKS_EMBEDDED.
3134 2018-02-23 Jakub Jelinek <jakub@redhat.com>
3136 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
3137 p->max as pointers rather than using iterative_hash_expr.
3139 2018-02-23 Carl Love <cel@us.ibm.com>
3141 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
3142 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
3144 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
3145 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
3146 P8V_BUILTIN_VEC_VUNSIGNED2.
3148 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
3151 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
3152 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
3154 * lra-constraints.c (curr_insn_transform): Set up
3155 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
3157 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
3158 * lra-eliminations.c (spill_pseudos): Ditto.
3159 (process_insn_for_elimination): Ditto.
3160 * lra-lives.c (reg_early_clobber_p): Use the new macros.
3161 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
3162 LRA_NON_CLOBBERED_ALT.
3164 2018-02-22 Martin Sebor <msebor@redhat.com>
3166 PR tree-optimization/84480
3167 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
3168 to maybe_diag_stxncpy_trunc. Call it.
3169 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
3170 from gimple_fold_builtin_strcpy. Print inlining stack.
3171 (handle_builtin_stxncpy): Print inlining stack.
3172 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
3174 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
3177 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
3178 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
3179 and -fcheck-pointer-bounds are used together.
3180 (indirect_thunk_prefix): New enum.
3181 (indirect_thunk_need_prefix): New function.
3182 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
3183 "_nt" instead of "_bnd" for NOTRACK prefix.
3184 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
3185 (output_indirect_thunk_function): Likewise.
3187 (ix86_code_end): Update output_indirect_thunk_function calls.
3188 (ix86_output_indirect_branch_via_reg): Replace
3189 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
3190 (ix86_output_indirect_branch_via_push): Likewise.
3191 (ix86_output_function_return): Likewise.
3192 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
3193 incompatible with -fcf-protection=branch and
3194 -fcheck-pointer-bounds.
3196 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3199 * config/aarch64/aarch64.c (aarch64_print_address_internal):
3200 Change gcc_assert call to output_operand_lossage.
3202 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3204 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
3206 2018-02-22 DJ Delorie <dj@redhat.com>
3207 Sebastian Perta <sebastian.perta@renesas.com>
3208 Oleg Endo <olegendo@gcc.gnu.org>
3210 * config/rx/rx.c (rx_rtx_costs): New function.
3211 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
3213 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
3215 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
3217 2018-02-22 Martin Liska <mliska@suse.cz>
3220 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
3221 Add "native" as a possible value.
3223 2018-02-22 Martin Liska <mliska@suse.cz>
3226 * config/i386/i386.c (ix86_option_override_internal):
3227 Add "native" as a possible value for -march and -mtune.
3229 2018-02-22 Jakub Jelinek <jakub@redhat.com>
3232 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
3233 to all type variants.
3235 PR tree-optimization/84503
3236 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
3237 width as info->bitpos + info->bitsize - start.
3238 (merged_store_group::merge_overlapping): Simplify width computation.
3239 (check_no_overlap): New function.
3240 (imm_store_chain_info::try_coalesce_bswap): Compute expected
3241 start + width and last_order of the group, fail if check_no_overlap
3243 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
3244 to group if check_no_overlap fails.
3246 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3248 * config/rs6000/altivec.md: Delete contraint arguments to
3249 define_expand, define_split, and define_peephole2, and in
3250 define_insn_and_split if always unused.
3251 * config/rs6000/darwin.md: Ditto.
3252 * config/rs6000/dfp.md: Ditto.
3253 * config/rs6000/rs6000.md: Ditto.
3254 * config/rs6000/sync.md: Ditto.
3255 * config/rs6000/vector.md: Ditto.
3256 * config/rs6000/vsx.md: Ditto.
3258 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3260 * config/rs6000/altivec.md: Write output control strings as braced
3261 blocks instead of double-quoted strings.
3262 * config/rs6000/darwin.md: Ditto.
3263 * config/rs6000/rs6000.md: Ditto.
3264 * config/rs6000/vector.md: Ditto.
3265 * config/rs6000/vsx.md: Ditto.
3267 2018-02-21 Jason Merrill <jason@redhat.com>
3269 PR c++/84314 - ICE with templates and fastcall attribute.
3270 * attribs.c (build_type_attribute_qual_variant): Remove assert.
3272 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3274 * ipa-cp.c (determine_versionability): Fix comment typos.
3276 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3279 * ipa-cp.c (determine_versionability): Do not version functions caling
3282 2018-02-21 Martin Liska <mliska@suse.cz>
3285 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
3286 Add "native" as a possible value.
3287 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
3288 the macro when native cpu detection is available.
3290 2018-02-21 Martin Liska <mliska@suse.cz>
3293 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
3294 Add "native" as a possible value.
3295 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
3296 when native cpu detection is available.
3298 2018-02-21 Jakub Jelinek <jakub@redhat.com>
3299 Martin Sebor <msebor@redhat.com>
3301 PR tree-optimization/84478
3302 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
3304 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
3305 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
3306 support which is conservatively correct, for 2 only stay conservative
3307 for maxlen. Formatting and comment capitalization fixes. Add STRICT
3308 argument to the 2 argument get_range_strlen, adjust 6 arg
3309 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
3311 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
3312 (gimple_fold_builtin_strlen): Pass true as last argument to
3315 2018-02-20 Martin Sebor <msebor@redhat.com>
3318 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
3319 (builtin_memref::set_base_and_offset): Same. Handle inner references.
3320 (builtin_memref::builtin_memref): Factor out parts into
3321 set_base_and_offset and call it.
3323 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3326 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
3327 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
3328 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
3329 search at the associated MODE_INT.
3331 2018-02-20 Jeff Law <law@redhat.com>
3334 PR tree-optimization/81592
3336 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
3337 for range data rather than using global data.
3338 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
3339 range data rather than using global data.
3340 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
3341 pass it to children as needed.
3342 (struct directive::fmtresult): Similarly.
3343 (struct directive::set_width): Similarly.
3344 (struct directive::set_precision): Similarly.
3345 (format_integer, format_directive, parse_directive): Similarly.
3346 (format_none): Accept unnamed vr_values parameter.
3347 (format_percent, format_floating, format_character): Similarly.
3348 (format_string, format_plain): Similarly.
3349 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
3350 the EVRP range analyzer for range data rather than using global data.
3351 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
3352 gimple-ssa-evrp-analyze.h
3353 (class sprintf_dom_walker): Add after_dom_children member function.
3354 Add evrp_range_analyzer member.
3355 (sprintf_dom_walker::before_dom_children): Call into the EVRP
3356 range analyzer as needed.
3357 (sprintf_dom_walker::after_dom_children): New member function.
3358 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
3360 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
3361 (evrp_range_analyzer::pop_to_marker): Likewise.
3363 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3365 PR tree-optimization/84419
3366 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
3367 with the required type if its current type is compatible but
3370 2018-02-20 Jakub Jelinek <jakub@redhat.com>
3373 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
3374 after vectorization.
3376 2018-02-20 Martin Liska <mliska@suse.cz>
3379 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
3380 possible values if we don't have a hint.
3382 2018-02-20 Martin Liska <mliska@suse.cz>
3386 * final.c (shorten_branches): Build align_tab array with one
3388 * opts.c (finish_options): Add alignment option limit check.
3389 (MAX_CODE_ALIGN): Likewise.
3390 (MAX_CODE_ALIGN_VALUE): Likewise.
3391 * doc/invoke.texi: Document maximum allowed option value for
3392 all -falign-* options.
3394 2018-02-19 Jakub Jelinek <jakub@redhat.com>
3397 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
3398 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
3399 * var-tracking.c (emit_note_insn_var_location): Remove all references
3400 to NOTE_INSN_CALL_ARG_LOCATION.
3401 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
3402 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
3403 Use copy_rtx_if_shared.
3404 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
3405 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
3406 (dwarf2out_var_location): Remove handling of
3407 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
3409 * final.c (final_scan_insn): Remove all references to
3410 NOTE_INSN_CALL_ARG_LOCATION.
3411 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
3412 before dumping final insns.
3413 * except.c (emit_note_eh_region_end): Remove all references to
3414 NOTE_INSN_CALL_ARG_LOCATION.
3415 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
3416 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
3417 * config/arc/arc.c (hwloop_optimize): Likewise.
3418 * config/arm/arm.c (create_fix_barrier): Likewise.
3419 * config/s390/s390.c (s390_chunkify_start): Likewise.
3420 * config/sh/sh.c (find_barrier): Likewise.
3421 * config/i386/i386.c (rest_of_insert_endbranch,
3422 ix86_seh_fixup_eh_fallthru): Likewise.
3423 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
3424 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
3425 * config/frv/frv.c (frv_function_prologue): Likewise.
3426 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
3428 (note_outside_basic_block_p): Remove all references to
3429 NOTE_INSN_CALL_ARG_LOCATION.
3430 * gengtype.c (adjust_field_rtx_def): Likewise.
3431 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
3433 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
3434 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
3437 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
3440 PR tree-optimization/84452
3441 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
3442 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
3445 2018-02-19 Martin Liska <mliska@suse.cz>
3448 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
3450 2018-02-19 Martin Liska <mliska@suse.cz>
3451 Richard Sandiford <richard.sandiford@linaro.org>
3453 PR tree-optimization/82491
3454 * gimple-fold.c (get_base_constructor): Make earlier bail out
3457 2018-02-19 Carl Love <cel@us.ibm.com>
3459 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
3460 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
3462 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
3463 P8V_BUILTIN_VEC_NEG.
3465 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
3467 * config/rl78/rl78.md (movdf): New define expand.
3469 2018-02-19 Martin Liska <mliska@suse.cz>
3472 * doc/invoke.texi: Fix typo.
3473 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
3475 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
3477 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
3478 handle rs6000_single_float and rs6000_double_float specially for
3481 2018-02-16 Jeff Law <law@redhat.com>
3483 * config/rx/rx.c (add_pop_cfi_notes): New function.;
3486 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3489 * ipa-inline.c (inline_small_functions): Fix a typo.
3491 2018-02-16 Nathan Sidwell <nathan@acm.org>
3493 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
3495 2018-02-16 Carl Love <cel@us.ibm.com>
3497 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
3498 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
3499 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
3500 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
3501 expansion to P8V_BUILTIN_VEC_FLOAT2.
3503 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
3505 PR rtl-optimization/70023
3506 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
3507 src_regno into account.
3509 2018-02-16 Carl Love <cel@us.ibm.com>
3511 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
3512 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
3513 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
3514 * config/rs6000/rs6000.c: Remove case statements for
3515 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
3516 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
3517 and P9V_BUILTIN_VEC_VINSERT4B.
3518 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
3519 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
3520 * config/rs6000/vsx.md:
3521 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
3524 2018-02-16 Carl Love <cel@us.ibm.com>
3526 * config/rs6000/altivec.h: Add builtin names vec_extract4b
3528 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
3530 * config/rs6000/rs6000-c.c: Add the definitions for
3531 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
3532 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
3533 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
3534 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
3535 definition for insert4b and define insn *insert3b_internal.
3536 * doc/extend.texi: Add documentation for vec_extract4b.
3538 2018-02-16 Nathan Sidwell <nathan@acm.org>
3540 * doc/extend.texi (Backwards Compatibility): Mention friend
3541 injection. Note for-scope is deprecated.
3542 * doc/invoke.texi (-ffriend-injection): Deprecate.
3544 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
3546 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
3547 that moved to I2, also allow destinations that are a paradoxical
3548 subreg (instead of a normal reg).
3550 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
3553 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
3556 2018-02-16 Richard Biener <rguenther@suse.de>
3558 PR tree-optimization/84037
3559 PR tree-optimization/84016
3561 * config/i386/i386.c (ix86_builtin_vectorization_cost):
3562 Adjust vec_construct for the fact we need additional higher latency
3563 128bit inserts for AVX256 and AVX512 vector builds.
3564 (ix86_add_stmt_cost): Scale vector construction cost for
3567 2018-02-16 Richard Biener <rguenther@suse.de>
3569 PR tree-optimization/84417
3570 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
3571 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
3572 (non_rewritable_lvalue_p): Likewise, use poly-ints.
3574 2018-02-16 Martin Liska <mliska@suse.cz>
3577 * internal-fn.def (ASAN_CHECK): Set proper flags.
3578 (ASAN_MARK): Likewise.
3580 2018-02-16 Julia Koval <julia.koval@intel.com>
3582 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
3583 from PTA_CANNONLAKE.
3585 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3588 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
3589 Use ++iter rather than iter++ for std::list iterators.
3590 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
3591 defer deleting them until all nodes in the forest are processed. Do
3592 free even leaf nodes. Change to_process into auto_vec.
3595 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
3596 * vec.h (vec_default_construct): Use memset instead of placement new
3597 if BROKEN_VALUE_INITIALIZATION is defined.
3598 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
3599 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
3602 PR rtl-optimization/83723
3603 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
3604 * lra.c (lra_substitute_pseudo): Likewise. If true, use
3605 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
3607 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
3609 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
3611 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
3613 PR rtl-optimization/81443
3614 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
3615 from inner REGs to paradoxical SUBREGs.
3617 2018-02-16 Richard Biener <rguenther@suse.de>
3619 PR tree-optimization/84399
3620 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3621 For operands we can analyze at their definition make sure we can
3622 analyze them at each use as well.
3624 2018-02-16 Richard Biener <rguenther@suse.de>
3626 PR tree-optimization/84190
3627 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
3628 volatile accesses if the decl isn't volatile.
3630 2018-02-15 Jason Merrill <jason@redhat.com>
3632 PR c++/84314 - ICE with templates and fastcall attribute.
3633 * attribs.c (build_type_attribute_qual_variant): Don't clobber
3634 TYPE_CANONICAL on an existing type.
3636 2018-02-15 Jakub Jelinek <jakub@redhat.com>
3638 PR tree-optimization/84383
3639 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
3640 dstoff nor call operand_equal_p if dstbase is NULL.
3642 PR tree-optimization/84334
3643 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
3644 also a CONSTANT_CLASS_P, punt.
3646 2018-02-14 Jim Wilson <jimw@sifive.com>
3648 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
3649 first SMALL_OPERAND check. New local min_second_step. Move assert
3650 to where locals are set. Add TARGET_RVC support.
3651 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
3653 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
3655 * doc/invoke.texi: Correct -Wformat-overflow code sample.
3657 2018-02-14 Martin Sebor <msebor@redhat.com>
3659 PR tree-optimization/83698
3660 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
3661 arrays constrain the offset range to their bounds.
3662 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
3663 (builtin_access::overlap): Avoid setting the size of overlap if it's
3665 (maybe_diag_overlap): Also consider arrays when deciding what values
3666 of offsets to include in diagnostics.
3668 2018-02-14 Martin Sebor <msebor@redhat.com>
3671 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
3672 that correspond to the kind of a declaration.
3674 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
3677 * config/pa/pa.md: Load address of PIC label using the linkage table
3678 if the label is nonlocal.
3680 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
3682 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
3683 warning message if user requests -maltivec=be.
3684 * doc/invoke.texi: Document deprecation of -maltivec=be.
3686 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
3689 * config/rs6000/rs6000-c.c: Update definitions for
3690 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
3691 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
3693 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3696 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
3697 add _get_ssp intrinsics. Remove argument from
3698 __builtin_ia32_rdssp[d|q].
3699 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
3700 * config/i386/i386-builtin.def: Remove argument from
3701 __builtin_ia32_rdssp[d|q].
3702 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
3703 ix86_expand_special_args_builtin for _rdssp[d|q].
3704 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
3705 Clear register before usage.
3706 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
3707 Add documentation for new _get_ssp and _inc_ssp intrinsics.
3709 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
3711 PR tree-optimization/84357
3712 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
3713 operand 1 of an ARRAY_REF too.
3715 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
3718 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
3719 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3721 (set_of_reg): New struct.
3722 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
3723 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
3724 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3726 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
3727 Split into bitclr, bitset, bitinvert patterns if appropriate.
3728 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
3729 use rx_fuse_in_memory_bitop.
3730 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
3731 to named insn, correct maximum insn length.
3733 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
3736 * machmode.def: Define a complex mode for PARTIAL_INT.
3737 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
3739 * doc/rtl.texi: Document CSPImode.
3740 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
3742 (msp430_hard_regno_nregs_with_padding): Likewise.
3744 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
3747 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
3749 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
3751 PR rtl-optimization/84169
3752 * combine.c (try_combine): New variable split_i2i3. Set it to true if
3753 we generated a parallel as new i3 and we split that to new i2 and i3
3754 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
3755 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
3756 those to i2, not i1. Partially rewrite this scan code.
3758 2018-02-13 Jakub Jelinek <jakub@redhat.com>
3761 * stor-layout.c (place_field): For variable length fields, adjust
3762 offset_align afterwards not just based on the field's alignment,
3763 but also on the size.
3766 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
3767 of exps and logs in the use_exp2 case.
3769 2018-02-13 Jeff Law <law@redhat.com>
3771 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
3774 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
3777 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
3781 * common.opt (gas-loc-support, gas-locview-support): New.
3782 (ginline-points, ginternal-reset-location-views): New.
3783 * doc/invoke.texi: Document them. Use @itemx where intended.
3784 (gvariable-location-views): Adjust.
3785 * target.def (reset_location_view): New.
3786 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
3787 (TARGET_RESET_LOCATION_VIEW): New.
3788 * doc/tm.texi: Rebuilt.
3789 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
3790 (dwarf2out_default_as_locview_support): New.
3791 (output_asm_line_debug_info): Use option variables.
3792 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
3793 (output_loc_list): Likewise.
3794 (add_high_low_attributes): Check option variables.
3795 Don't output entry view attribute in strict mode.
3796 (gen_inlined_subroutine_die): Check option variables.
3797 (dwarf2out_inline_entry): Likewise.
3798 (init_sections_and_labels): Likewise.
3799 (dwarf2out_early_finish): Likewise.
3800 (maybe_reset_location_view): New, from...
3801 (dwarf2out_var_location): ... here. Call it.
3802 * debug.h (dwarf2out_default_as_loc_support): Declare.
3803 (dwarf2out_default_as_locview_support): Declare.
3804 * hooks.c (hook_int_rtx_insn_0): New.
3805 * hooks.h (hook_int_rtx_insn_0): Declare.
3806 * toplev.c (process_options): Take -gas-loc-support and
3807 -gas-locview-support from dwarf2out. Enable
3808 -gvariable-location-views by default only with locview
3809 assembler support. Enable -ginternal-reset-location-views by
3810 default only if the target defines the corresponding hook.
3811 Enable -ginline-points by default if location views are
3812 enabled; force it disabled if statement frontiers are
3814 * tree-inline.c (expand_call_inline): Check option variables.
3815 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
3817 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
3819 PR tree-optimization/84321
3820 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
3821 handling. Also check whether the anti-range contains any values
3822 that satisfy the mask; switch to a VR_RANGE if not.
3824 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
3827 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
3829 2018-02-13 Martin Jambor <mjambor@suse.cz>
3832 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
3833 of call statements, also set location of a load to a temporary.
3835 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
3837 * config/rl78/rl78.c (add_vector_labels): New function.
3838 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
3839 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
3840 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
3841 which checks that no arguments are passed.
3842 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
3843 * doc/extend.texi: Documentation for the new attribute.
3845 2018-02-13 Andreas Schwab <schwab@suse.de>
3847 * config/riscv/linux.h (CPP_SPEC): Define.
3849 2018-02-13 Jakub Jelinek <jakub@redhat.com>
3852 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
3853 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
3854 OPTION_MASK_ISA_AES as first argument to def_builtin_const
3855 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
3856 instead of OPTION_MASK_ISA_PCLMUL as first argument to
3857 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
3858 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
3859 temporarily for AES and PCLMUL builtins.
3861 PR tree-optimization/84339
3862 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
3863 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
3867 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
3868 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
3869 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
3871 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
3873 * omp-simd-clone.h: New file.
3874 * omp-simd-clone.c: Include omp-simd-clone.h.
3875 (expand_simd_clones): No longer static.
3876 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
3877 cgraph.h and omp-simd-clone.h.
3878 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
3879 (vect_recog_widen_shift_pattern): Formatting fix.
3880 (vect_pattern_recog_1): Don't check optab for calls.
3883 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
3884 operands[2] into a REG before using gen_lowpart on it.
3886 2018-02-12 Jeff Law <law@redhat.com>
3889 * config/sh/sh.c (find_barrier): Consider a sibling call
3892 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
3893 successfully back substituting a reg.
3895 2018-02-12 Richard Biener <rguenther@suse.de>
3897 PR tree-optimization/84037
3898 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
3899 parameter, move visited init to caller.
3900 (vect_slp_analyze_operations): Separate cost from validity
3901 check, initialize visited once for all instances.
3902 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
3904 * tree-vect-stmts.c (vect_model_simple_cost): Make early
3906 (vect_model_promotion_demotion_cost): Likewise.
3907 (vectorizable_bswap): Guard cost modeling with !slp_node
3908 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
3910 (vectorizable_call): Likewise.
3911 (vectorizable_conversion): Likewise.
3912 (vectorizable_assignment): Likewise.
3913 (vectorizable_shift): Likewise.
3914 (vectorizable_operation): Likewise.
3915 (vectorizable_store): Likewise.
3916 (vectorizable_load): Likewise.
3917 (vectorizable_condition): Likewise.
3918 (vectorizable_comparison): Likewise.
3920 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
3923 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
3924 (ASAN_MARK): Fix fnspec to account for return value, change pointer
3925 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
3927 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3930 * params.def (inline-min-speedup): Increase from 8 to 15.
3931 (max-inline-insns-auto): Decrease from 40 to 30.
3932 * ipa-split.c (consider_split): Add some buffer for function to
3933 be considered inlining candidate.
3934 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
3937 2018-02-12 Richard Biener <rguenther@suse.de>
3939 PR tree-optimization/84037
3940 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
3941 matched stmts if we cannot swap the non-matched ones.
3943 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3945 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
3946 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
3947 _mm_maskz_scalef_round_ss): New intrinsics.
3948 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
3949 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
3950 __builtin_ia32_scalefss_round): Remove.
3951 (__builtin_ia32_scalefsd_mask_round,
3952 __builtin_ia32_scalefss_mask_round): New intrinsics.
3953 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
3954 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
3955 ((match_operand:VF_128 2 "<round_nimm_predicate>"
3956 "<round_constraint>")): Changed to ...
3957 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
3958 "<round_scalar_constraint>")): ... this.
3959 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
3960 %0, %1, %2<round_op3>}"): Changed to ...
3961 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
3962 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
3963 %2<round_scalar_mask_op3>}"): ... this.
3964 * config/i386/subst.md (round_scalar_nimm_predicate): New.
3966 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3968 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
3969 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
3970 (_mm_maskz_sqrt_round_ss): New intrinsics.
3971 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
3972 (__builtin_ia32_sqrtsd_mask_round)
3973 (__builtin_ia32_sqrtss_mask_round): New builtins.
3974 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
3975 (__builtin_ia32_sqrtss_round): Remove.
3976 (__builtin_ia32_sqrtsd_mask_round)
3977 (__builtin_ia32_sqrtss_mask_round): New builtins.
3978 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
3979 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
3980 ((match_operand:VF_128 1 "vector_operand"
3981 "xBm,<round_constraint>")): Changed to ...
3982 ((match_operand:VF_128 1 "vector_operand"
3983 "xBm,<round_scalar_constraint>")): ... this.
3984 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
3985 %0, %2, %<iptr>1<round_op3>}): Changed to ...
3986 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
3987 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
3988 %<iptr>1<round_scalar_mask_op3>}): ... this.
3989 ((set_attr "prefix" "<round_prefix>")): Changed to ...
3990 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
3992 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
3995 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
3996 Cast vec_cmpeq result to correct type.
3997 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
3998 Cast vec_cmpgt result to correct type.
4000 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
4002 * final.c (final_scan_insn_1): Renamed from...
4003 (final_scan_insn): ... this. New wrapper, to recover
4004 seen from the outermost call in recursive ones.
4005 * config/sparc/sparc.c (output_return): Drop seen from call.
4006 (output_sibcall): Likewise.
4007 * config/visium/visium.c (output_branch): Likewise.
4009 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
4011 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
4014 2018-02-10 Alan Modra <amodra@gmail.com>
4017 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
4018 Specify LR as an input.
4020 2018-02-10 Jakub Jelinek <jakub@redhat.com>
4023 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
4024 remove_member_access_dummy_vars): New functions.
4025 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
4026 lower_omp_1, execute_lower_omp): Use them.
4028 PR rtl-optimization/84308
4029 * shrink-wrap.c (spread_components): Release todo vector.
4031 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
4033 PR rtl-optimization/57193
4034 * ira-color.c (struct allocno_color_data): Add member
4035 conflict_allocno_hard_prefs.
4036 (update_conflict_allocno_hard_prefs): New.
4037 (bucket_allocno_compare_func): Add a preference based on
4038 conflict_allocno_hard_prefs.
4039 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
4040 (color_allocnos): Remove a dead code. Initiate
4041 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
4043 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4046 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
4047 constraint from =wa to wa. Avoid a subreg on the output operand,
4048 instead use a pseudo and subreg it in a move.
4049 (p9_xxbrd_<mode>): Changed to ...
4050 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
4051 (p9_xxbrd_v2df): New expander.
4052 (p9_xxbrw_<mode>): Changed to ...
4053 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
4054 (p9_xxbrw_v4sf): New expander.
4056 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4058 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
4060 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
4063 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
4064 multiply in 32-bit mode.
4065 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
4066 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
4069 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4071 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
4072 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
4073 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
4074 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
4076 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
4079 * dwarf2out.c (is_trivial_indirect_ref): New function.
4080 (dwarf2out_late_global_decl): Do not generate a location
4081 attribute for variables that have a non-trivial DECL_VALUE_EXPR
4082 and that are not defined in the current unit.
4084 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4086 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
4087 instead of a libcall for UNORDERED.
4089 2018-02-09 Tamar Christina <tamar.christina@arm.com>
4092 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
4093 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
4095 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4098 * config/s390/s390.c (s390_set_current_function): Invoke
4099 s390_indirect_branch_settings also if fndecl didn't change.
4101 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4103 * config/rs6000/rs6000.md (blockage): Set length to zero.
4105 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4107 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
4109 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4112 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
4113 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
4117 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
4118 PARALLEL incoming that failed vt_get_decl_and_offset check.
4121 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
4122 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
4124 (get_variable_section): For decls in named .bss* sections pass true as
4125 second argument to bss_initializer_p.
4127 2018-02-09 Marek Polacek <polacek@redhat.com>
4128 Jakub Jelinek <jakub@redhat.com>
4131 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
4132 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
4133 Sync some changes from cxx_fold_indirect_ref.
4135 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4137 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
4139 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
4140 (BLOCK_INLINE_ENTRY_LABEL): New.
4141 (dwarf2out_var_location): Disregard inline entry markers.
4142 (inline_entry_data): New struct.
4143 (inline_entry_data_hasher): New hashtable type.
4144 (inline_entry_data_hasher::hash): New.
4145 (inline_entry_data_hasher::equal): New.
4146 (inline_entry_data_table): New variable.
4147 (add_high_low_attributes): Add DW_AT_entry_pc and
4148 DW_AT_GNU_entry_view attributes if a pending entry is found
4149 in inline_entry_data_table. Add old entry_pc attribute only
4150 if debug nonbinding markers are disabled.
4151 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
4152 markers are enabled.
4153 (block_within_block_p, dwarf2out_inline_entry): New.
4154 (dwarf2out_finish): Check that no entries remained in
4155 inline_entry_data_table.
4156 * final.c (reemit_insn_block_notes): Handle inline entry notes.
4157 (final_scan_insn, notice_source_line): Likewise.
4158 (rest_of_clean_state): Skip inline entry markers.
4159 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
4161 * gimple.c (gimple_build_debug_inline_entry): New.
4162 * gimple.h (enum gimple_debug_subcode): Add
4163 GIMPLE_DEBUG_INLINE_ENTRY.
4164 (gimple_build_debug_inline_entry): Declare.
4165 (gimple_debug_inline_entry_p): New.
4166 (gimple_debug_nonbind_marker_p): Adjust.
4167 * insn-notes.def (INLINE_ENTRY): New.
4168 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
4169 inline entry marker notes.
4170 (print_insn): Likewise.
4171 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
4172 (INSN_DEBUG_MARKER_KIND): Likewise.
4173 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
4174 * tree-inline.c (expand_call_inline): Build and insert
4175 debug_inline_entry stmt.
4176 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
4177 inline entry blocks early, if nonbind markers are enabled.
4178 (dump_scope_block): Dump fragment info.
4179 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
4180 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
4181 (gimple_build_debug_inline_entry): New.
4182 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
4183 Enable/disable inline entry points too.
4184 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
4185 (DEBUG_INSN): Describe inline entry markers.
4187 * common.opt (gvariable-location-views): New.
4188 (gvariable-location-views=incompat5): New.
4189 * config.in: Rebuilt.
4190 * configure: Rebuilt.
4191 * configure.ac: Test assembler for view support.
4192 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
4193 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
4194 * dwarf2out.c (var_loc_view): New typedef.
4195 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
4196 (dwarf2out_locviews_in_attribute): New.
4197 (dwarf2out_locviews_in_loclist): New.
4198 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
4199 (enum dw_line_info_opcode): Add LI_adv_address.
4200 (struct dw_line_info_table): Add view.
4201 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
4202 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
4203 (zero_view_p): New variable.
4204 (ZERO_VIEW_P): New macro.
4205 (output_asm_line_debug_info): New.
4206 (struct var_loc_node): Add view.
4207 (add_AT_view_list, AT_loc_list): New.
4208 (add_var_loc_to_decl): Add view param. Test it against last.
4209 (new_loc_list): Add view params. Record them.
4210 (AT_loc_list_ptr): Handle loc and view lists.
4211 (view_list_to_loc_list_val_node): New.
4212 (print_dw_val): Handle dw_val_class_view_list.
4213 (size_of_die): Likewise.
4214 (value_format): Likewise.
4215 (loc_list_has_views): New.
4216 (gen_llsym): Set vl_symbol too.
4217 (maybe_gen_llsym, skip_loc_list_entry): New.
4218 (dwarf2out_maybe_output_loclist_view_pair): New.
4219 (output_loc_list): Output view list or entries too.
4220 (output_view_list_offset): New.
4221 (output_die): Handle dw_val_class_view_list.
4222 (output_dwarf_version): New.
4223 (output_compilation_unit_header): Use it.
4224 (output_skeleton_debug_sections): Likewise.
4225 (output_rnglists, output_line_info): Likewise.
4226 (output_pubnames, output_aranges): Update version comments.
4227 (output_one_line_info_table): Output view numbers in asm comments.
4228 (dw_loc_list): Determine current endview, pass it to new_loc_list.
4229 Call maybe_gen_llsym.
4230 (loc_list_from_tree_1): Adjust.
4231 (add_AT_location_description): Create view list attribute if
4232 needed, check it's absent otherwise.
4233 (convert_cfa_to_fb_loc_list): Adjust.
4234 (maybe_emit_file): Call output_asm_line_debug_info for test.
4235 (dwarf2out_var_location): Reset views as needed. Precompute
4236 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
4237 attribute. Set view.
4238 (new_line_info_table): Reset next view.
4239 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
4240 (dwarf2out_source_line): Likewise. Output view resets and labels to
4241 the assembler, or select appropriate line info opcodes.
4242 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
4243 (optimize_string_length): Catch it. Adjust.
4244 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
4245 dw_val_class_view_list, and remove it if no longer needed.
4246 (hash_loc_list): Hash view numbers.
4247 (loc_list_hasher::equal): Compare them.
4248 (optimize_location_lists): Check whether a view list symbol is
4249 needed, and whether the locview attribute is present, and
4250 whether they match. Remove the locview attribute if no longer
4252 (index_location_lists): Call skip_loc_list_entry for test.
4253 (dwarf2out_finish): Call output_asm_line_debug_info for test.
4254 Use output_dwarf_version.
4255 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
4256 (struct dw_val_node): Add val_view_list.
4257 * final.c (SEEN_NEXT_VIEW): New.
4258 (set_next_view_needed): New.
4259 (clear_next_view_needed): New.
4260 (maybe_output_next_view): New.
4261 (final_start_function): Rename to...
4262 (final_start_function_1): ... this. Take pointer to FIRST,
4263 add SEEN parameter. Emit param bindings in the initial view.
4264 (final_start_function): Reintroduce SEEN-less interface.
4265 (final): Rename to...
4266 (final_1): ... this. Take SEEN parameter. Output final pending
4267 next view at the end.
4268 (final): Reintroduce seen-less interface.
4269 (final_scan_insn): Output pending next view before switching
4270 sections or ending a block. Mark the next view as needed when
4271 outputting variable locations. Notify debug backend of section
4272 changes, and of location view changes.
4273 (rest_of_handle_final): Adjust.
4274 * toplev.c (process_options): Autodetect value for debug variable
4275 location views option. Warn on incompat5 without -gdwarf-5.
4276 * doc/invoke.texi (gvariable-location-views): New.
4277 (gvariable-location-views=incompat5): New.
4278 (gno-variable-location-views): New.
4280 2018-02-08 David Malcolm <dmalcolm@redhat.com>
4282 PR tree-optimization/84136
4283 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
4284 that the result of find_edge is non-NULL.
4286 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4289 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
4290 storing integer register in SImode. Fix cost of 256 and 512
4291 byte aligned SSE register store.
4293 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4295 * config/i386/i386.c (ix86_multiplication_cost): Fix
4296 multiplication cost for TARGET_AVX512DQ.
4298 2018-02-08 Marek Polacek <polacek@redhat.com>
4300 PR tree-optimization/84238
4301 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
4304 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4306 PR tree-optimization/84265
4307 * tree-vect-stmts.c (vectorizable_store): Don't treat
4308 VMAT_CONTIGUOUS accesses as grouped.
4309 (vectorizable_load): Likewise.
4311 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4313 PR tree-optimization/81635
4314 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
4315 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
4316 (test_round_for_mask): New functions.
4317 (wide_int_cc_tests): Call test_round_for_mask.
4318 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
4319 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
4320 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
4321 range returned by get_range_info.
4323 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4326 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
4327 * symtab.c: Include builtins.h
4328 (symtab_node::output_to_lto_symbol_table_p): Move here
4329 from lto-streamer-out.c:output_symbol_p.
4330 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
4331 (output_symbol_p): Move all logic to symtab.c
4332 (produce_symtab): Update.
4334 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4336 * config/s390/s390-opts.h (enum indirect_branch): Define.
4337 * config/s390/s390-protos.h (s390_return_addr_from_memory)
4338 (s390_indirect_branch_via_thunk)
4339 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
4340 (enum s390_indirect_branch_type): Define.
4341 * config/s390/s390.c (struct s390_frame_layout, struct
4342 machine_function): Remove.
4343 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
4344 (indirect_branch_table_label_no, indirect_branch_table_name):
4346 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
4347 (enum s390_indirect_branch_option): Define.
4348 (s390_return_addr_from_memory): New function.
4349 (s390_handle_string_attribute): New function.
4350 (s390_attribute_table): Add new attribute handler.
4351 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
4352 (s390_indirect_branch_via_thunk): New function.
4353 (s390_indirect_branch_via_inline_thunk): New function.
4354 (s390_function_ok_for_sibcall): When jumping via thunk disallow
4355 sibling call optimization for non z10 compiles.
4356 (s390_emit_call): Force indirect branch target to be a single
4357 register. Add r1 clobber for non-z10 compiles.
4358 (s390_emit_epilogue): Emit return jump via return_use expander.
4359 (s390_reorg): Handle JUMP_INSNs as execute targets.
4360 (s390_option_override_internal): Perform validity checks for the
4361 new command line options.
4362 (s390_indirect_branch_attrvalue): New function.
4363 (s390_indirect_branch_settings): New function.
4364 (s390_set_current_function): Invoke s390_indirect_branch_settings.
4365 (s390_output_indirect_thunk_function): New function.
4366 (s390_code_end): Implement target hook.
4367 (s390_case_values_threshold): Implement target hook.
4368 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
4370 * config/s390/s390.h (struct s390_frame_layout)
4371 (struct machine_function): Move here from s390.c.
4372 (TARGET_INDIRECT_BRANCH_NOBP_RET)
4373 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
4374 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
4375 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
4376 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
4377 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
4378 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
4379 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
4380 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
4381 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
4382 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
4383 (mnemonic attribute): Add values which aren't recognized
4385 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
4386 pattern for branch conversion. Fix mnemonic attribute.
4387 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
4388 indirect branch via thunk if requested.
4389 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
4390 ("*indirect_jump"): Disable for branch conversion using out of
4392 ("indirect_jump_via_thunk<mode>_z10")
4393 ("indirect_jump_via_thunk<mode>")
4394 ("indirect_jump_via_inlinethunk<mode>_z10")
4395 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
4396 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
4397 ("casesi_jump_via_inlinethunk<mode>_z10")
4398 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
4399 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
4400 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
4401 ("*indirect2_jump"): Disable for branch conversion.
4402 ("casesi_jump"): Turn into expander and expand patterns for branch
4404 ("return_use"): New expander.
4405 ("*return"): Emit return via thunk and rename it to ...
4406 ("*return<mode>"): ... this one.
4407 * config/s390/s390.opt: Add new options and and enum for the
4410 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4412 * lra-constraints.c (match_reload): Unconditionally use
4413 gen_lowpart_SUBREG, rather than selecting between that
4414 and equivalent gen_rtx_SUBREG code.
4416 2018-02-08 Richard Biener <rguenther@suse.de>
4418 PR tree-optimization/84233
4419 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
4420 changed flag instead of boguously re-using phi_inserted.
4422 2018-02-08 Martin Jambor <mjambor@suse.cz>
4424 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
4425 static local variables.
4427 2018-02-08 Richard Biener <rguenther@suse.de>
4429 PR tree-optimization/84278
4430 * tree-vect-stmts.c (vectorizable_store): When looking for
4431 smaller vector types to perform grouped strided loads/stores
4432 make sure the mode is supported by the target.
4433 (vectorizable_load): Likewise.
4435 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4437 * config/aarch64/aarch64.c (aarch64_components_for_bb):
4438 Increase LDP/STP opportunities by adding adjacent callee-saves.
4440 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4442 PR rtl-optimization/84068
4443 PR rtl-optimization/83459
4444 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
4446 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
4448 PR tree-optimization/84224
4449 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
4450 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
4453 2018-02-07 Iain Sandoe <iain@codesourcery.com>
4456 * config/rs6000/altivec.md (*restore_world): Remove LR use.
4457 * config/rs6000/predicates.md (restore_world_operation): Adjust op
4458 count, remove one USE.
4460 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
4462 * doc/install.texi (Configuration): Document the
4463 --with-long-double-format={ibm,ieee} PowerPC configuration
4467 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
4468 Convert from define_expand to be define_insn_and_split. Rework
4469 float/double/_Float128 conversions to QI/HI/SImode to work with
4470 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
4471 conversions to QI/HImode types did a store and then a load to
4472 truncate the value. For conversions to VSX registers, don't split
4473 the insn, instead emit the code directly. Use the code iterator
4474 any_fix to combine signed and unsigned conversions.
4475 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
4476 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
4477 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4478 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
4479 (fix_<mode>di2_hw): Likewise.
4480 (fixuns_<mode>di2_hw): Likewise.
4481 (fix_<mode>si2_hw): Likewise.
4482 (fixuns_<mode>si2_hw): Likewise.
4483 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
4484 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4485 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
4486 fix<uns>_trunc<SFDF:mode>si2_p8.
4487 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
4489 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
4490 (fix<uns>_<mode>_mem): Likewise.
4491 (fctiw<u>z_<mode>_mem): Likewise.
4492 (fix<uns>_<mode>_mem): Likewise.
4493 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
4494 the register allocator from doing a direct move to the GPRs to do
4495 a store, and instead use the ISA 3.0 store byte/half-word from
4496 vector register instruction. For IEEE 128-bit floating point,
4497 also optimize stores of 32-bit ints.
4498 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
4500 2018-02-07 Alan Hayward <alan.hayward@arm.com>
4502 * genextract.c (push_pathstr_operand): New function to support
4504 (walk_rtx): Call push_pathstr_operand.
4505 (print_path): Support [a-zA-Z].
4507 2018-02-07 Richard Biener <rguenther@suse.de>
4509 PR tree-optimization/84037
4510 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
4511 (cse_and_gimplify_to_preheader): Declare.
4512 (vect_get_place_in_interleaving_chain): Likewise.
4513 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4515 (_loop_vec_info::~_loop_vec_info): Delete it.
4516 (cse_and_gimplify_to_preheader): New function.
4517 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
4518 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
4519 (vectorizable_load): Likewise. For grouped stores always base
4520 the IV on the first element.
4521 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
4522 condition before gimplifying.
4524 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4526 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
4527 *DIV_EXPR and *MOD_EXPR.
4529 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
4532 * config/i386/i386.c (ix86_option_override_internal): Mask out
4533 the CF_SET bit when checking -fcf-protection.
4535 2018-02-07 Tom de Vries <tom@codesourcery.com>
4538 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
4541 2018-02-07 Richard Biener <rguenther@suse.de>
4543 PR tree-optimization/84204
4544 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
4547 PR tree-optimization/84205
4548 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
4549 special-case isl_ast_op_zdiv_r.
4551 PR tree-optimization/84223
4552 * graphite-scop-detection.c (gather_bbs::before_dom_children):
4553 Only add conditions from within the region.
4554 (gather_bbs::after_dom_children): Adjust.
4556 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
4559 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
4560 * config/avr/avr.md: Only post-reload split REG-REG moves if
4561 either register is GENERAL_REG_P.
4563 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4565 PR tree-optimization/84235
4566 * tree-ssa-scopedtables.c
4567 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
4568 if the subtraction is performed in floating point type where NaNs are
4569 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
4570 build 1. Formatting fix.
4572 2018-02-06 Jakub Jelinek <jakub@redhat.com>
4575 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
4576 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
4577 and skip it regardless of bb boundaries. Use CALL_P macro,
4578 don't test INSN_P (insn) together with CALL_P or JUMP_P check
4579 unnecessarily, formatting fix.
4581 2018-02-06 Michael Collison <michael.collison@arm.com>
4583 * config/arm/thumb2.md:
4584 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
4585 (*thumb_mov_notscc): Ditto.
4587 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
4590 * config/rs6000/rs6000.md (su code attribute): Use "u" for
4591 unsigned_fix, not "s".
4593 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4595 * configure.ac (gcc_fn_eh_frame_ro): New function.
4596 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
4597 correct .eh_frame permissions.
4598 * configure: Regenerate.
4600 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
4602 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
4605 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4607 * config/rs6000/rs6000.c (rs6000_option_override_internal):
4608 Display warning message for -mno-speculate-indirect-jumps.
4610 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
4612 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
4614 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
4616 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
4618 PR tree-optimization/84225
4619 * tree-eh.c (find_trapping_overflow): Only call
4620 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
4622 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4625 * config/i386/i386.c: Reimplement the check of possible options
4626 -mibt/-mshstk conbination. Change error messages.
4627 * doc/invoke.texi: Fix a typo: remove extra '='.
4629 2018-02-06 Marek Polacek <polacek@redhat.com>
4631 PR tree-optimization/84228
4632 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
4634 2018-02-06 Tamar Christina <tamar.christina@arm.com>
4637 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
4638 emitted arch directives.
4639 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
4640 __ARM_FEATURE_COPROC before changing architectures.
4642 2018-02-06 Richard Biener <rguenther@suse.de>
4644 * config/i386/i386.c (print_reg): Fix typo.
4645 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
4647 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
4649 * configure: Regenerate.
4651 2018-02-05 Martin Sebor <msebor@redhat.com>
4653 PR tree-optimization/83369
4654 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
4657 2018-02-05 Martin Liska <mliska@suse.cz>
4659 * doc/invoke.texi: Cherry-pick upstream r323995.
4661 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
4663 * ira.c (ira_init_register_move_cost): Adjust comment.
4665 2018-02-05 Martin Liska <mliska@suse.cz>
4667 PR gcov-profile/84137
4668 * doc/gcov.texi: Fix typo in documentation.
4670 2018-02-05 Martin Liska <mliska@suse.cz>
4672 PR gcov-profile/83879
4673 * doc/gcov.texi: Document necessity of --dynamic-list-data when
4674 using dlopen functionality.
4676 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
4678 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
4679 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
4680 _mm_maskz_range_ss, _mm_mask_range_round_ss,
4681 _mm_maskz_range_round_ss): New intrinsics.
4682 (__builtin_ia32_rangesd128_round)
4683 (__builtin_ia32_rangess128_round): Remove.
4684 (__builtin_ia32_rangesd128_mask_round,
4685 __builtin_ia32_rangess128_mask_round): New builtins.
4686 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
4687 __builtin_ia32_rangess128_round): Remove.
4688 (__builtin_ia32_rangesd128_mask_round,
4689 __builtin_ia32_rangess128_mask_round): New builtins.
4690 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
4691 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
4692 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
4693 "<round_saeonly_constraint>")): Changed to ...
4694 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
4695 "<round_saeonly_scalar_constraint>")): ... this.
4696 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
4697 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
4698 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
4699 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
4700 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
4702 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
4704 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
4706 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
4707 Remove all values except native, 8540 and 8548.
4709 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
4711 * config/i386/i386.c (ix86_output_function_return): Pass
4712 INVALID_REGNUM, instead of -1, as invalid register number to
4713 indirect_thunk_name and output_indirect_thunk.
4715 2018-02-02 Julia Koval <julia.koval@intel.com>
4717 * config.gcc: Add -march=icelake.
4718 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
4719 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
4720 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
4721 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
4722 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
4723 (processor_target_table): Add icelake.
4724 (ix86_option_override_internal): Handle new PTAs.
4725 (get_builtin_code_for_version): Handle icelake.
4726 (M_INTEL_COREI7_ICELAKE): New.
4727 (fold_builtin_cpu): Handle icelake.
4728 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
4729 * doc/invoke.texi: Add -march=icelake.
4731 2018-02-02 Julia Koval <julia.koval@intel.com>
4733 * config/i386/i386.c (ix86_option_override_internal): Change flags type
4734 to wide_int_bitmask.
4735 * wide-int-bitmask.h: New.
4737 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4740 * config/i386/i386.md: Replace Pmode with word_mode in
4741 builtin_setjmp_setup and builtin_longjmp to support x32.
4743 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
4747 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
4749 (rs6000_supported_cpu_names): New static variable.
4750 (linux_cpu_translation_table): Likewise.
4751 (elf_platform) <cpu>: Define new static variable and use it.
4752 Translate kernel AT_PLATFORM name to canonical name if needed.
4753 Error if platform name is unknown.
4755 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
4758 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
4760 2018-02-01 Jeff Law <law@redhat.com>
4763 * config/i386/i386.c (release_scratch_register_on_entry): Add new
4764 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
4765 the scratch if RELEASE_VIA_POP is false.
4766 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
4767 If we have to save a temporary register, decrement SIZE appropriately.
4768 Pass new arguments to release_scratch_register_on_entry.
4769 (ix86_adjust_stack_and_probe): Likewise.
4770 (ix86_emit_probe_stack_range): Pass new arguments to
4771 release_scratch_register_on_entry.
4773 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
4775 PR rtl-optimization/84157
4776 * combine.c (change_zero_ext): Use REG_P predicate in
4777 front of HARD_REGISTER_P predicate.
4779 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
4781 * config/avr/avr.c (avr_option_override): Move disabling of
4782 -fdelete-null-pointer-checks to...
4783 * common/config/avr/avr-common.c (avr_option_optimization_table):
4786 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4788 PR tree-optimization/81635
4789 * tree-data-ref.c (split_constant_offset_1): For types that
4790 wrap on overflow, try to use range info to prove that wrapping
4793 2018-02-01 Renlin Li <renlin.li@arm.com>
4796 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
4798 (aarch64_register_move_cost): Likewise.
4799 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
4801 (REG_CLASS_NAMES): Likewise.
4802 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
4803 TAILCALL_ADDR_REGS. Remove IP registers.
4804 * config/aarch64/aarch64.md (Ucs): Update register constraint.
4806 2018-02-01 Richard Biener <rguenther@suse.de>
4808 * domwalk.h (dom_walker::dom_walker): Add additional constructor
4809 for specifying RPO order and allow NULL for that.
4810 * domwalk.c (dom_walker::dom_walker): Likewise.
4811 (dom_walker::walk): Handle NULL RPO order.
4812 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
4814 (rewrite_update_dom_walker): Likewise.
4815 (mark_def_dom_walker): Likewise.
4817 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4819 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
4820 (aarch64_maybe_expand_sve_subreg_move): Declare.
4821 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
4822 * config/aarch64/predicates.md (aarch64_any_register_operand): New
4824 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
4825 that are semantically a reverse operation.
4826 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
4827 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
4828 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
4830 (aarch64_can_change_mode_class): For big-endian, forbid changes
4831 between two SVE modes if they have different element sizes.
4833 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4835 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
4836 the TImode handling for big-endian targets.
4838 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4840 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
4841 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
4843 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
4844 Remove BSWAP handing for big-endian targets and use the form of
4845 LD1RQ appropariate for the mode.
4847 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4849 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
4850 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
4853 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4856 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
4857 check for operands that need to go through aarch64_sve_reload_be.
4859 2018-02-01 Jakub Jelinek <jakub@redhat.com>
4861 PR tree-optimization/81661
4862 PR tree-optimization/84117
4863 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
4864 * tree-eh.c: Include gimplify.h.
4865 (find_trapping_overflow, replace_trapping_overflow,
4866 rewrite_to_non_trapping_overflow): New functions.
4867 * tree-vect-loop.c: Include tree-eh.h.
4868 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
4869 * tree-data-ref.c: Include tree-eh.h.
4870 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
4872 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
4874 PR rtl-optimization/84123
4875 * combine.c (change_zero_ext): Check if hard register satisfies
4876 can_change_dest_mode before calling gen_lowpart_SUBREG.
4878 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
4881 * ira.c (ira_init_register_move_cost): Remove assert.
4883 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4885 PR rtl-optimization/84071
4886 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
4887 * doc/tm.texi: Regenerate.
4889 2018-01-31 Richard Biener <rguenther@suse.de>
4891 PR tree-optimization/84132
4892 * tree-data-ref.c (analyze_miv_subscript): Properly
4893 check whether evolution_function_is_affine_multivariate_p
4894 before calling gcd_of_steps_may_divide_p.
4896 2018-01-31 Julia Koval <julia.koval@intel.com>
4899 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
4900 * config/i386/i386.md (rdpid_rex64) New.
4901 (rdpid): Make 32bit only.
4903 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
4906 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
4907 an IDENTIFIER_NODE for FUNCTION_TYPE's.
4909 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4912 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4914 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
4916 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4918 PR rtl-optimization/84071
4919 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
4920 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
4922 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4924 * config/arc/arc.c (arc_handle_aux_attribute): New function.
4925 (arc_attribute_table): Add 'aux' attribute.
4926 (arc_in_small_data_p): Consider aux like variables.
4927 (arc_is_aux_reg_p): New function.
4928 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
4929 (arc_get_aux_arg): New function.
4930 (prepare_move_operands): Handle aux-register access.
4931 (arc_handle_aux_attribute): New function.
4932 * doc/extend.texi (ARC Variable attributes): Add subsection.
4934 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4936 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
4937 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
4938 (arc_attribute_table): Add 'uncached' attribute.
4939 (arc_print_operand): Print '.di' flag for uncached memory
4941 (arc_in_small_data_p): Do not consider for small data the uncached
4943 (arc_is_uncached_mem_p): New function.
4944 * config/arc/predicates.md (compact_store_memory_operand): Check
4945 for uncached memory accesses.
4946 (nonvol_nonimm_operand): Likewise.
4947 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
4949 2018-01-31 Jakub Jelinek <jakub@redhat.com>
4952 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
4953 falign-loops=): Add Optimization flag.
4955 2018-01-30 Jeff Law <law@redhat.com>
4958 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
4959 INT_REGISTERS_SAVED. Check it prior to calling
4960 get_scratch_register_on_entry.
4961 (ix86_adjust_stack_and_probe): Similarly.
4962 (ix86_emit_probe_stack_range): Similarly.
4963 (ix86_expand_prologue): Corresponding changes.
4965 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4968 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
4969 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
4971 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
4974 * lra-constraints.c (curr_insn_transform): Process AND in the
4977 2018-01-30 Jakub Jelinek <jakub@redhat.com>
4979 PR rtl-optimization/83986
4980 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
4981 dependence against last_pending_memory_flush in addition to
4984 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
4986 PR tree-optimization/81611
4987 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
4990 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4993 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
4996 2018-01-30 Richard Biener <rguenther@suse.de>
4997 Jakub Jelinek <jakub@redhat.com>
4999 PR tree-optimization/84111
5000 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
5001 inner loops added during recursion, as they don't have up-to-date
5004 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5007 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
5008 (can_inline_edge_by_limits_p): ... here.
5009 (can_early_inline_edge_p, check_callers,
5010 update_caller_keys, update_callee_keys, recursive_inlining,
5011 add_new_edges_to_heap, speculation_useful_p,
5012 inline_small_functions,
5013 inline_small_functions, flatten_function,
5014 inline_to_all_callers_1): Update.
5016 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5018 * profile-count.c (profile_count::combine_with_ipa_count): Handle
5021 2018-01-30 Richard Biener <rguenther@suse.de>
5023 PR tree-optimization/83008
5024 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
5025 invariant and constant vector uses in stmts when they need
5028 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5031 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
5032 * configure: Regenerate.
5034 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5036 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
5038 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
5039 Use gen_rtx_REG rather than gen_lowpart.
5041 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5043 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
5044 rather than 0 when creating partial subregs.
5046 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5048 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
5051 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
5054 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
5055 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
5056 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
5057 flags. This restores the settings used before the 2017-07-24.
5058 Turning off pre increment/decrement/modify allows IVOPTS to
5059 optimize DF/SF loops where the index is an int.
5061 2018-01-29 Richard Biener <rguenther@suse.de>
5062 Kelvin Nilsen <kelvin@gcc.gnu.org>
5065 * tree-vect-stmts.c (vectorizable_call): Don't call
5066 targetm.vectorize_builtin_md_vectorized_function if callee is
5069 2018-01-22 Carl Love <cel@us.ibm.com>
5071 * doc/extend.tex: Fix typo in second arg in
5072 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
5074 2018-01-29 Richard Biener <rguenther@suse.de>
5076 PR tree-optimization/84086
5077 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
5078 (flush_ssaname_freelist): When SSA names were released reset
5079 the SCEV hash table.
5081 2018-01-29 Richard Biener <rguenther@suse.de>
5083 PR tree-optimization/84057
5084 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
5085 removed paths when removing edges.
5087 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
5089 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
5090 -mfunction-return=@var{choice}.
5092 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
5095 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
5097 (layout::print_source_line): Likewise.
5098 (test_get_line_width_without_trailing_whitespace): Add test cases.
5100 2018-01-27 Jakub Jelinek <jakub@redhat.com>
5103 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
5106 2018-01-26 Jim Wilson <jimw@sifive.com>
5108 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
5110 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
5113 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5115 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
5116 and CMP + SUB-immediate -> SUBS.
5118 2018-01-26 Martin Sebor <msebor@redhat.com>
5120 PR tree-optimization/83896
5121 * tree-ssa-strlen.c (get_string_len): Rename...
5122 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
5123 Avoid assuming length is constant.
5124 (handle_char_store): Use HOST_WIDE_INT for string length.
5126 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
5129 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
5130 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
5132 2018-01-26 Richard Biener <rguenther@suse.de>
5134 PR rtl-optimization/84003
5135 * dse.c (record_store): Only record redundant stores when
5136 the earlier store aliases at least all accesses the later one does.
5138 2018-01-26 Jakub Jelinek <jakub@redhat.com>
5140 PR rtl-optimization/83985
5141 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
5142 REG_CFA_RESTORE insns.
5143 (delete_unmarked_insns): Don't ignore separate shrink wrapping
5144 REG_CFA_RESTORE insns here.
5147 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
5148 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
5150 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5152 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
5153 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
5154 (arc_init): Likewise.
5155 (arc_override_options): Likewise.
5156 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
5158 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
5160 * config/arc/arc.h (TARGET_DBNZ): Define.
5161 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
5162 properly set the tune attribute.
5163 (dbnz): Use TARGET_DBNZ guard.
5164 * config/arc/arc.opt (mtune): Add core3 option.
5166 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5168 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
5169 recognize new pic like addresses.
5170 (arc_delegitimize_address): Clean up.
5172 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5174 * config/arc/arc-arches.def: Option mrf16 valid for all
5176 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
5177 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
5178 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
5179 * config/arc/arc-tables.opt: Regenerate.
5180 * config/arc/arc.c (arc_conditional_register_usage): Handle
5181 reduced register file case.
5182 (arc_file_start): Set must have build attributes.
5183 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
5185 * config/arc/arc.opt (mrf16): Add new option.
5186 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
5187 * config/arc/genmultilib.awk: Handle new mrf16 option.
5188 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
5189 * config/arc/t-multilib: Regenerate.
5190 * doc/invoke.texi (ARC Options): Document mrf16 option.
5192 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5194 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
5195 * config/arc/arc.c (arc_handle_secure_attribute): New function.
5196 (arc_attribute_table): Add 'secure_call' attribute.
5197 (arc_print_operand): Print secure call operand.
5198 (arc_function_ok_for_sibcall): Don't optimize tail calls when
5200 (arc_is_secure_call_p): New function. * config/arc/arc.md
5201 (call_i): Add support for sjli instruction.
5202 (call_value_i): Likewise.
5203 * config/arc/constraints.md (Csc): New constraint.
5205 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5206 John Eric Martin <John.Martin@emmicro-us.com>
5208 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
5209 * config/arc/arc.c (_arc_jli_section): New struct.
5210 (arc_jli_section): New type.
5211 (rc_jli_sections): New static variable.
5212 (arc_handle_jli_attribute): New function.
5213 (arc_attribute_table): Add jli_always and jli_fixed attribute.
5214 (arc_file_end): New function.
5215 (TARGET_ASM_FILE_END): Define.
5216 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
5217 (arc_add_jli_section): New function.
5218 (jli_call_scan): Likewise.
5219 (arc_reorg): Call jli_call_scan.
5220 (arc_output_addsi): Remove 'S' from printing asm operand.
5221 (arc_is_jli_call_p): New function.
5222 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
5224 (movhi_insn): Likewise.
5225 (movsi_insn): Likewise.
5226 (movsi_set_cc_insn): Likewise.
5227 (loadqi_update): Likewise.
5228 (load_zeroextendqisi_update): Likewise.
5229 (load_signextendqisi_update): Likewise.
5230 (loadhi_update): Likewise.
5231 (load_zeroextendhisi_update): Likewise.
5232 (load_signextendhisi_update): Likewise.
5233 (loadsi_update): Likewise.
5234 (loadsf_update): Likewise.
5235 (movsicc_insn): Likewise.
5236 (bset_insn): Likewise.
5237 (bxor_insn): Likewise.
5238 (bclr_insn): Likewise.
5239 (bmsk_insn): Likewise.
5240 (bicsi3_insn): Likewise.
5241 (cmpsi_cc_c_insn): Likewise.
5242 (movsi_ne): Likewise.
5243 (movsi_cond_exec): Likewise.
5244 (clrsbsi2): Likewise.
5255 (call_i): Remove 'S' asm letter, add jli instruction.
5256 (call_value_i): Likewise.
5257 * config/arc/arc.op (mjli-always): New option.
5258 * config/arc/constraints.md (Cji): New constraint.
5259 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
5261 (subsf3_fpx): Likewise.
5262 (mulsf3_fpx): Likewise.
5263 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
5265 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
5267 * doc/invoke.texi (ARC): Document mjli-always option.
5269 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
5271 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
5272 avoid addition with 0 and use incw and decw where possible.
5274 2018-01-26 Richard Biener <rguenther@suse.de>
5276 PR tree-optimization/81082
5277 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
5278 association if it requires casting to unsigned.
5279 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
5280 from fold_plusminus_mult_expr to catch important cases late when
5281 range info is available.
5283 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5285 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
5286 * configure.ac (hidden_linkonce): New test.
5287 * configure: Regenerate.
5288 * config.in: Regenerate.
5290 2018-01-26 Julia Koval <julia.koval@intel.com>
5292 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
5293 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
5294 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
5295 _mm_mask_bitshuffle_epi64_mask): Fix type.
5296 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
5297 USI_FTYPE_V4DI_V4DI_USI): Remove.
5298 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
5299 __builtin_ia32_vpshufbitqmb256_mask,
5300 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
5301 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
5302 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
5304 2018-01-26 Alan Modra <amodra@gmail.com>
5307 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
5308 UNSPEC_VBPERMQ. Sort other unspecs.
5310 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
5312 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
5314 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
5317 * predict.c (drop_profile): Do not push/pop cfun; update also
5319 (handle_missing_profiles): Fix logic looking for zero profiles.
5321 2018-01-25 Jakub Jelinek <jakub@redhat.com>
5324 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
5325 on functions with #pragma omp declare simd or functions with simd
5327 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
5328 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
5329 Remove trailing \n from warning_at calls.
5331 2018-01-25 Tom de Vries <tom@codesourcery.com>
5334 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5335 for neutered workers.
5337 2018-01-24 Joseph Myers <joseph@codesourcery.com>
5340 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
5341 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
5343 2018-01-24 Jeff Law <law@redhat.com>
5346 * i386.c (get_probe_interval): Move to earlier point.
5347 (ix86_compute_frame_layout): If -fstack-clash-protection and
5348 the frame is larger than the probe interval, then use pushes
5349 to save registers rather than reg->mem moves.
5350 (ix86_expand_prologue): Remove conditional for int_registers_saved
5353 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
5356 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
5357 min/max for never referenced object.
5359 2018-01-24 Jakub Jelinek <jakub@redhat.com>
5362 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
5364 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
5365 attributes from DECL_ATTRIBUTES (decl) without affecting
5366 DECL_ATTRIBUTES (current_function_decl).
5367 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
5368 functions with non-NULL DECL_ABSTRACT_ORIGIN.
5370 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
5372 PR tree-optimization/83979
5373 * fold-const.c (fold_comparison): Use constant_boolean_node
5374 instead of boolean_{true,false}_node.
5376 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
5378 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
5381 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5383 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5384 Simplify the clause that sets the length attribute.
5385 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5386 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
5387 clause that sets the length attribute.
5388 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5390 2018-01-24 Tom de Vries <tom@codesourcery.com>
5393 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
5394 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
5395 Add strict parameter.
5396 (prevent_branch_around_nothing): Insert dummy insn between branch to
5397 label and label with no ptx insn inbetween.
5398 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
5400 2018-01-24 Tom de Vries <tom@codesourcery.com>
5403 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5404 for neutered threads in warp.
5405 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
5407 2018-01-24 Richard Biener <rguenther@suse.de>
5409 PR tree-optimization/83176
5410 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
5413 2018-01-24 Richard Biener <rguenther@suse.de>
5415 PR tree-optimization/82819
5416 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
5417 code generating pluses that are no-ops in the target precision.
5419 2018-01-24 Richard Biener <rguenther@suse.de>
5422 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
5424 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5426 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
5427 to merge probabilities.
5428 * predict.c (probably_never_executed): Also mark as cold functions
5429 with global 0 profile and guessed local profile.
5430 * profile-count.c (profile_probability::combine_with_count): New
5432 * profile-count.h (profile_probability::operator*,
5433 profile_probability::operator*=, profile_probability::operator/,
5434 profile_probability::operator/=): Reduce precision to adjusted
5435 and set value to guessed on contradictory divisions.
5436 (profile_probability::combine_with_freq): Remove.
5437 (profile_probability::combine_wiht_count): Declare.
5438 (profile_count::force_nonzero):: Set to adjusted.
5439 (profile_count::probability_in):: Set quality to adjusted.
5440 * tree-ssa-tail-merge.c (replace_block_by): Use
5443 2018-01-23 Andrew Waterman <andrew@sifive.com>
5444 Jim Wilson <jimw@sifive.com>
5446 * config/riscv/riscv.c (riscv_stack_boundary): New.
5447 (riscv_option_override): Set riscv_stack_boundary. Handle
5448 riscv_preferred_stack_boundary_arg.
5449 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
5450 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
5451 (STACK_BOUNDARY): Set to riscv_stack_boundary.
5452 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
5453 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
5454 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
5456 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
5459 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
5460 of struct ix86_frame.
5461 (ix86_expand_epilogue): Likewise. Add a local variable for
5462 the reg_save_offset field in struct ix86_frame.
5464 2018-01-23 Bin Cheng <bin.cheng@arm.com>
5466 PR tree-optimization/82604
5467 * tree-loop-distribution.c (enum partition_kind): New enum item
5468 PKIND_PARTIAL_MEMSET.
5469 (partition_builtin_p): Support above new enum item.
5470 (generate_code_for_partition): Ditto.
5471 (compute_access_range): Differentiate cases that equality can be
5472 proven at all loops, the innermost loops or no loops.
5473 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
5474 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
5475 (finalize_partitions, distribute_loop): Don't fuse partition of
5476 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
5477 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
5480 2018-01-23 Martin Liska <mliska@suse.cz>
5482 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
5483 order to ignore the predictor.
5484 (PRED_POLYMORPHIC_CALL): Likewise.
5485 (PRED_RECURSIVE_CALL): Likewise.
5487 2018-01-23 Martin Liska <mliska@suse.cz>
5489 * tree-profile.c (tree_profiling): Print function header to
5490 aware reader which function we are working on.
5491 * value-prof.c (gimple_find_values_to_profile): Do not print
5492 not interesting value histograms.
5494 2018-01-23 Martin Liska <mliska@suse.cz>
5496 * profile-count.h (enum profile_quality): Add
5497 profile_uninitialized as the first value. Do not number values
5498 as they are zero based.
5499 (profile_count::verify): Update sanity check.
5500 (profile_probability::verify): Likewise.
5502 2018-01-23 Nathan Sidwell <nathan@acm.org>
5504 * doc/invoke.texi (ffor-scope): Deprecate.
5506 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5508 PR tree-optimization/83510
5509 * domwalk.c (set_all_edges_as_executable): New function.
5510 (dom_walker::dom_walker): Convert bool param
5511 "skip_unreachable_blocks" to enum reachability. Move setup of
5512 edge flags to set_all_edges_as_executable and only do it when
5513 reachability is REACHABLE_BLOCKS.
5514 * domwalk.h (enum dom_walker::reachability): New enum.
5515 (dom_walker::dom_walker): Convert bool param
5516 "skip_unreachable_blocks" to enum reachability.
5517 (set_all_edges_as_executable): New decl.
5518 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
5519 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
5521 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
5522 but converting true to REACHABLE_BLOCKS.
5523 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
5525 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
5526 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
5527 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
5529 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
5530 if check_all_array_refs will be called.
5532 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5534 * tree.c (selftest::test_location_wrappers): Add more test
5537 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5539 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
5540 (selftest::test_bit_in_range): Likewise.
5542 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5545 * doc/sourcebuild.texi (vect_float): Say that the selector
5546 only describes the situation when -funsafe-math-optimizations is on.
5547 (vect_float_strict): Document.
5549 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5551 PR tree-optimization/83965
5552 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
5553 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
5554 instead of checking only for a reduction.
5555 (vect_recog_widen_sum_pattern): Likewise.
5557 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5559 * predict.c (probably_never_executed): Only use precise profile info.
5560 (compute_function_frequency): Skip after inlining hack since we now
5561 have quality checking.
5563 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5565 * profile-count.h (profile_probability::very_unlikely,
5566 profile_probability::unlikely, profile_probability::even): Set
5567 precision to guessed.
5569 2018-01-23 Richard Biener <rguenther@suse.de>
5571 PR tree-optimization/83963
5572 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
5573 Properly terminate dominator walk when crossing the exit edge not
5574 when visiting its source block.
5576 2018-01-23 Jakub Jelinek <jakub@redhat.com>
5579 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
5580 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
5582 2018-01-22 Jakub Jelinek <jakub@redhat.com>
5584 PR tree-optimization/83957
5585 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
5586 semicolon after for body surrounded by braces.
5588 PR tree-optimization/83081
5589 * profile-count.h (profile_probability::split): New method.
5590 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
5591 Use profile_probability::split.
5592 (do_compare_rtx_and_jump): Fix adjustment of probabilities
5593 when splitting a single conditional jump into 2.
5595 2018-01-22 David Malcolm <dmalcolm@redhat.com>
5597 PR tree-optimization/69452
5598 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
5601 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5603 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
5604 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
5605 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
5607 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5609 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
5610 * config/rl78/rl78.md (movdi): New define_expand.
5611 * config/rl78/rl78.c (rl78_split_movdi): New function.
5613 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
5616 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
5618 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
5619 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
5620 128-bit to produce an UNSPEC move to get the double word with the
5621 signbit and then a shift directly to do signbit.
5622 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
5623 implementation with a new version that just does either a direct
5624 move or a regular move. Move memory interface to separate insns.
5625 Move insns so they are next to the expander.
5626 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
5627 with signbit move. Split big and little endian case.
5628 (signbit<mode>2_dm_mem_le): Likewise.
5629 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
5630 (signbit<mode>2_dm2): Likewise.
5632 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5634 * config/rl78/rl78.md (anddi3): New define_expand.
5636 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5638 * config/rl78/rl78.md (umindi3): New define_expand.
5640 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5642 * config/rl78/rl78.md (smindi3): New define_expand.
5644 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5646 * config/rl78/rl78.md (smaxdi3): New define_expand.
5648 2018-01-22 Carl Love <cel@us.ibm.com>
5650 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
5651 LVX_V1TI): Add macro expansion.
5652 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
5653 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
5654 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
5655 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
5656 Change check to determine if the instruction is a byte reversing
5657 entry. Fix typo in comment.
5658 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
5659 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
5660 Add def_builtin calls for new builtins.
5661 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
5662 Add define_insn expansion.
5664 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5666 * config/rl78/rl78.md (umaxdi3): New define_expand.
5668 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5670 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
5671 for non-QImode registers.
5673 2018-01-22 Richard Biener <rguenther@suse.de>
5675 PR tree-optimization/83963
5676 * graphite-scop-detection.c (scop_detection::get_sese): Delay
5677 including the loop exit block.
5678 (scop_detection::merge_sese): Likewise.
5679 (scop_detection::add_scop): Do it here instead.
5681 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5683 * doc/sourcebuild.texi (arm_softfloat): Document.
5685 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
5688 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
5689 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
5690 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
5692 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5693 David Edelsohn <dje.gcc@gmail.com>
5696 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5697 Change "crset eq" to "crset 2".
5698 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5699 (*call_indirect_aix<mode>_nospec): Likewise.
5700 (*call_value_indirect_aix<mode>_nospec): Likewise.
5701 (*call_indirect_elfv2<mode>_nospec): Likewise.
5702 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
5703 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
5704 change assembly output from . to $.
5705 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5706 (indirect_jump<mode>_nospec): Change assembly output from . to $.
5707 (*tablejump<mode>_internal1_nospec): Likewise.
5709 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
5712 * config/sh/sh_optimize_sett_clrt.cc:
5713 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
5715 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
5717 PR tree-optimization/83940
5718 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
5719 offset_dt to vect_constant_def rather than vect_unknown_def_type.
5720 (vect_check_load_store_mask): Add a mask_dt_out parameter and
5721 use it to pass back the definition type.
5722 (vect_check_store_rhs): Likewise rhs_dt_out.
5723 (vect_build_gather_load_calls): Add a mask_dt argument and use
5724 it instead of a call to vect_is_simple_use.
5725 (vectorizable_store): Update calls to vect_check_load_store_mask
5726 and vect_check_store_rhs. Use the dt returned by the latter instead
5727 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
5728 instead of calls to vect_is_simple_use. Pass the scalar rather
5729 than the vector operand to vect_is_simple_use when handling
5730 second and subsequent copies of an rhs value.
5731 (vectorizable_load): Update calls to vect_check_load_store_mask
5732 and vect_build_gather_load_calls. Use the cached mask_dt and
5733 gs_info.offset_dt instead of calls to vect_is_simple_use.
5735 2018-01-20 Jakub Jelinek <jakub@redhat.com>
5738 * tree-emutls.c: Include gimplify.h.
5739 (lower_emutls_2): New function.
5740 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
5741 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
5742 it before further processing.
5745 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
5746 UINTVAL (trueop1) instead of INTVAL (op1).
5748 2018-01-19 Jakub Jelinek <jakub@redhat.com>
5752 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
5753 INCOMING_FRAME_SP_OFFSET if not defined.
5754 (scan_trace): Add ENTRY argument. If true and
5755 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
5756 emit a note to adjust the CFA offset.
5757 (create_cfi_notes): Adjust scan_trace callers.
5758 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
5759 INCOMING_FRAME_SP_OFFSET in the CIE.
5760 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
5761 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
5763 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
5764 * doc/tm.texi: Regenerated.
5766 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5768 PR rtl-optimization/83147
5769 * lra-constraints.c (remove_inheritance_pseudos): Use
5770 lra_substitute_pseudo_within_insn.
5772 2018-01-19 Tom de Vries <tom@codesourcery.com>
5773 Cesar Philippidis <cesar@codesourcery.com>
5776 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
5778 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
5781 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
5782 spaces for function labels.
5784 2018-01-19 Martin Liska <mliska@suse.cz>
5786 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
5787 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
5788 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
5789 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
5790 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
5791 (PRED_CONST_RETURN): Change from 69 to 65.
5792 (PRED_NULL_RETURN): Change from 91 to 71.
5793 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
5794 (PRED_LOOP_GUARD): Change from 66 to 73.
5796 2018-01-19 Martin Liska <mliska@suse.cz>
5798 * predict.c (predict_insn_def): Add new assert.
5799 (struct branch_predictor): Change type to signed integer.
5800 (test_prediction_value_range): Amend test to cover
5802 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
5803 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
5804 (PRED_LOOP_ITERATIONS_MAX): Likewise.
5805 (PRED_LOOP_IV_COMPARE): Likewise.
5806 * predict.h (PROB_UNINITIALIZED): Define new constant.
5808 2018-01-19 Martin Liska <mliska@suse.cz>
5810 * predict.c (dump_prediction): Add new format for
5811 analyze_brprob.py script which is enabled with -details
5813 * profile-count.h (precise_p): New function.
5815 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
5817 PR tree-optimization/83922
5818 * tree-vect-loop.c (vect_verify_full_masking): Return false if
5819 there are no statements that need masking.
5820 (vect_active_double_reduction_p): New function.
5821 (vect_analyze_loop_operations): Use it when handling phis that
5822 are not in the loop header.
5824 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
5826 PR tree-optimization/83914
5827 * tree-vect-loop.c (vectorizable_induction): Don't convert
5828 init_expr or apply the peeling adjustment for inductions
5829 that are nested within the vectorized loop.
5831 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5833 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
5836 2018-01-18 Jakub Jelinek <jakub@redhat.com>
5840 * function.h (gimplify_parameters): Add gimple_seq * argument.
5841 * function.c: Include gimple.h and options.h.
5842 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
5843 for the added local temporaries if needed.
5844 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
5845 if there are any parameter cleanups, wrap whole body into a
5846 try/finally with the cleanups.
5848 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
5851 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
5852 Use GET_MODE_CLASS for scalar floating point.
5854 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
5858 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
5859 Fix call of call_cgraph_insertion_hooks.
5861 2018-01-18 Martin Sebor <msebor@redhat.com>
5863 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
5865 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
5868 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
5871 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
5874 * common.opt: (-ffile-prefix-map): New option.
5875 * opts.c (common_handle_option): Defer it.
5876 * opts-global.c (handle_common_deferred_options): Handle it.
5877 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
5878 * file-prefix-map.h: New file.
5879 (remap_debug_filename, add_debug_prefix_map): ...here.
5880 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
5881 * final.c (debug_prefix_map, add_debug_prefix_map
5882 remap_debug_filename): Move to...
5883 * file-prefix-map.c: New file.
5884 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
5885 generalize, get rid of alloca(), use strrchr() instead of strchr().
5886 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
5887 Implement in terms of add_prefix_map().
5888 (remap_macro_filename, remap_debug_filename): Implement in term of
5890 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
5891 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
5892 * dbxout.c: Include file-prefix-map.h.
5893 * varasm.c: Likewise.
5894 * vmsdbgout.c: Likewise.
5895 * xcoffout.c: Likewise.
5896 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
5897 * doc/cppopts.texi (-fmacro-prefix-map): Document.
5898 * doc/invoke.texi (-ffile-prefix-map): Document.
5899 (-fdebug-prefix-map): Update description.
5901 2018-01-18 Martin Liska <mliska@suse.cz>
5903 * config/i386/i386.c (indirect_thunk_name): Document that also
5905 (output_indirect_thunk): Document why both instructions
5906 (pause and lfence) are generated.
5908 2018-01-18 Richard Biener <rguenther@suse.de>
5910 PR tree-optimization/83887
5911 * graphite-scop-detection.c
5912 (scop_detection::get_nearest_dom_with_single_entry): Remove.
5913 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
5914 (scop_detection::merge_sese): Re-implement with a flood-fill
5915 algorithm that properly finds a SESE region if it exists.
5917 2018-01-18 Jakub Jelinek <jakub@redhat.com>
5920 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
5921 pointer_diff optimizations use view_convert instead of convert.
5923 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5925 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5926 Generate different code for -mno-speculate-indirect-jumps.
5927 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5928 (*call_indirect_aix<mode>): Disable for
5929 -mno-speculate-indirect-jumps.
5930 (*call_indirect_aix<mode>_nospec): New define_insn.
5931 (*call_value_indirect_aix<mode>): Disable for
5932 -mno-speculate-indirect-jumps.
5933 (*call_value_indirect_aix<mode>_nospec): New define_insn.
5934 (*sibcall_nonlocal_sysv<mode>): Generate different code for
5935 -mno-speculate-indirect-jumps.
5936 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5938 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
5940 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
5941 long double type, set the flags for noting the default long double
5942 type, even if we don't pass or return a long double type.
5944 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
5947 * ipa-inline.c (flatten_function): Do not overwrite final inlining
5950 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
5952 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
5953 support for merge[hl].
5954 (fold_mergehl_helper): New helper function.
5955 (tree-vector-builder.h): New #include for tree_vector_builder usage.
5956 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
5957 (altivec_vmrglw_direct): Add xxmrglw insn.
5959 2018-01-17 Andrew Waterman <andrew@sifive.com>
5961 * config/riscv/riscv.c (riscv_conditional_register_usage): If
5962 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
5964 2018-01-17 David Malcolm <dmalcolm@redhat.com>
5967 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
5968 call the lto_location_cache before reading the
5969 DECL_SOURCE_LOCATION of the types.
5971 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
5972 Richard Sandiford <richard.sandiford@linaro.org>
5974 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
5975 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
5976 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
5977 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
5978 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
5980 * config/aarch64/constraints.md (aarch64_movti_operand):
5982 * config/aarch64/predicates.md (Uti): Add new constraint.
5984 2018-01-17 Carl Love <cel@us.ibm.com>
5986 * config/rs6000/vsx.md (define_expand xl_len_r,
5987 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
5988 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
5990 (define_expand, define_insn): Move the shift left from the
5991 define_insn to the define_expand for lxvl and stxvl instructions.
5992 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
5993 and XL_LEN_R definitions to PURE.
5995 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
5997 * config/i386/i386.c (indirect_thunk_name): Declare regno
5998 as unsigned int. Compare regno with INVALID_REGNUM.
5999 (output_indirect_thunk): Ditto.
6000 (output_indirect_thunk_function): Ditto.
6001 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
6002 in the call to output_indirect_thunk_function.
6004 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
6007 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
6008 rather than the size of inner_type to determine the stack slot size
6009 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
6011 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
6014 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
6017 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
6019 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
6020 endian Linux systems to optionally enable multilibs for selecting
6021 the long double type if the user configured an explicit type.
6022 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
6023 have no long double multilibs if not defined.
6024 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
6025 warn if the user used -mabi={ieee,ibm}longdouble and we built
6026 multilibs for long double.
6027 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
6028 appropriate multilib option.
6029 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
6031 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
6032 for building long double multilibs.
6033 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
6035 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
6037 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
6040 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
6042 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
6045 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
6048 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
6051 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6053 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
6054 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
6056 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
6058 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
6059 different rtl trees depending on TARGET_64BIT.
6060 (rs6000_gen_lvx): Likewise.
6062 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6064 * config/visium/visium.md (nop): Tweak comment.
6065 (hazard_nop): Likewise.
6067 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6069 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
6070 -mspeculate-indirect-jumps.
6071 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
6072 for -mno-speculate-indirect-jumps.
6073 (*call_indirect_elfv2<mode>_nospec): New define_insn.
6074 (*call_value_indirect_elfv2<mode>): Disable for
6075 -mno-speculate-indirect-jumps.
6076 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
6077 (indirect_jump): Emit different RTL for
6078 -mno-speculate-indirect-jumps.
6079 (*indirect_jump<mode>): Disable for
6080 -mno-speculate-indirect-jumps.
6081 (*indirect_jump<mode>_nospec): New define_insn.
6082 (tablejump): Emit different RTL for
6083 -mno-speculate-indirect-jumps.
6084 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
6085 (tablejumpsi_nospec): New define_expand.
6086 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
6087 (tablejumpdi_nospec): New define_expand.
6088 (*tablejump<mode>_internal1): Disable for
6089 -mno-speculate-indirect-jumps.
6090 (*tablejump<mode>_internal1_nospec): New define_insn.
6091 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
6094 2018-01-16 Artyom Skrobov tyomitch@gmail.com
6096 * caller-save.c (insert_save): Drop unnecessary parameter. All
6099 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6100 Richard Biener <rguenth@suse.de>
6103 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
6104 return early, inline manually is_gimple_sizepos. Make sure if we
6105 call gimplify_expr we don't end up with a gimple constant.
6106 * tree.c (variably_modified_type_p): Don't return true for
6107 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
6108 * gimplify.h (is_gimple_sizepos): Remove.
6110 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6112 PR tree-optimization/83857
6113 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
6114 vectorizable_live_operation for pure SLP statements.
6115 (vectorizable_live_operation): Handle PHIs.
6117 2018-01-16 Richard Biener <rguenther@suse.de>
6119 PR tree-optimization/83867
6120 * tree-vect-stmts.c (vect_transform_stmt): Precompute
6121 nested_in_vect_loop_p since the scalar stmt may get invalidated.
6123 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6126 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
6127 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
6128 If off is not INTEGER_CST, issue a may not be aligned warning
6129 rather than isn't aligned. Use isn%'t rather than isn't.
6130 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
6132 <case MULT_EXPR>: Improve the case when bottom and one of the
6133 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
6134 operand, in that case check if the other operand is multiple of
6135 bottom divided by the INTEGER_CST operand.
6137 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6140 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
6141 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
6142 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
6143 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
6144 * config/pa/pa.c (pa_function_arg_advance): Likewise.
6145 (pa_function_arg, pa_arg_partial_bytes): Likewise.
6146 (pa_function_arg_size): New function.
6148 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6150 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
6151 in a separate statement.
6153 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6155 PR tree-optimization/83847
6156 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
6157 group gathers and scatters.
6159 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6161 PR rtl-optimization/86620
6162 * params.def (max-sched-ready-insns): Bump minimum value to 1.
6164 PR rtl-optimization/83213
6165 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
6166 to last if both are JUMP_INSNs.
6168 PR tree-optimization/83843
6169 * gimple-ssa-store-merging.c
6170 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
6171 store_immediate_info for bswap/nop orig_stores.
6173 2018-01-15 Andrew Waterman <andrew@sifive.com>
6175 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
6177 <UDIV>: Increase cost if !TARGET_DIV.
6179 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
6181 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
6182 (define_attr "cr_logical_3op"): New.
6183 (cceq_ior_compare): Adjust.
6184 (cceq_ior_compare_complement): Adjust.
6185 (*cceq_rev_compare): Adjust.
6186 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
6187 (is_cracked_insn): Adjust.
6188 (insn_must_be_first_in_group): Adjust.
6189 * config/rs6000/40x.md: Adjust.
6190 * config/rs6000/440.md: Adjust.
6191 * config/rs6000/476.md: Adjust.
6192 * config/rs6000/601.md: Adjust.
6193 * config/rs6000/603.md: Adjust.
6194 * config/rs6000/6xx.md: Adjust.
6195 * config/rs6000/7450.md: Adjust.
6196 * config/rs6000/7xx.md: Adjust.
6197 * config/rs6000/8540.md: Adjust.
6198 * config/rs6000/cell.md: Adjust.
6199 * config/rs6000/e300c2c3.md: Adjust.
6200 * config/rs6000/e500mc.md: Adjust.
6201 * config/rs6000/e500mc64.md: Adjust.
6202 * config/rs6000/e5500.md: Adjust.
6203 * config/rs6000/e6500.md: Adjust.
6204 * config/rs6000/mpc.md: Adjust.
6205 * config/rs6000/power4.md: Adjust.
6206 * config/rs6000/power5.md: Adjust.
6207 * config/rs6000/power6.md: Adjust.
6208 * config/rs6000/power7.md: Adjust.
6209 * config/rs6000/power8.md: Adjust.
6210 * config/rs6000/power9.md: Adjust.
6211 * config/rs6000/rs64.md: Adjust.
6212 * config/rs6000/titan.md: Adjust.
6214 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6216 * config/i386/predicates.md (indirect_branch_operand): Rewrite
6217 ix86_indirect_branch_register logic.
6219 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6221 * config/i386/constraints.md (Bs): Update
6222 ix86_indirect_branch_register check. Don't check
6223 ix86_indirect_branch_register with GOT_memory_operand.
6225 * config/i386/predicates.md (GOT_memory_operand): Don't check
6226 ix86_indirect_branch_register here.
6227 (GOT32_symbol_operand): Likewise.
6229 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6231 * config/i386/predicates.md (constant_call_address_operand):
6232 Rewrite ix86_indirect_branch_register logic.
6233 (sibcall_insn_operand): Likewise.
6235 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6237 * config/i386/constraints.md (Bs): Replace
6238 ix86_indirect_branch_thunk_register with
6239 ix86_indirect_branch_register.
6241 * config/i386/i386.md (indirect_jump): Likewise.
6242 (tablejump): Likewise.
6243 (*sibcall_memory): Likewise.
6244 (*sibcall_value_memory): Likewise.
6245 Peepholes of indirect call and jump via memory: Likewise.
6246 * config/i386/i386.opt: Likewise.
6247 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6248 (GOT_memory_operand): Likewise.
6249 (call_insn_operand): Likewise.
6250 (sibcall_insn_operand): Likewise.
6251 (GOT32_symbol_operand): Likewise.
6253 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6256 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
6257 type rather than type addr's type points to.
6258 (expand_omp_atomic_mutex): Likewise.
6259 (expand_omp_atomic): Likewise.
6261 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6264 * config/i386/i386.c (output_indirect_thunk_function): Use
6265 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
6266 for __x86_return_thunk.
6268 2018-01-15 Richard Biener <rguenther@suse.de>
6271 * expmed.c (extract_bit_field_1): Fix typo.
6273 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6276 * config/arm/iterators.md (VF): New mode iterator.
6277 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
6278 Remove integer-related logic from pattern.
6279 (neon_vabd<mode>_3): Likewise.
6281 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6284 * common.opt (fstrict-overflow): No longer an alias.
6285 (fwrapv-pointer): New option.
6286 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
6287 also for pointer types based on flag_wrapv_pointer.
6288 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
6289 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
6290 opts->x_flag_wrapv got set.
6291 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
6292 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
6293 POINTER_TYPE_OVERFLOW_UNDEFINED.
6294 * match.pd: Likewise in address comparison pattern.
6295 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
6297 2018-01-15 Richard Biener <rguenther@suse.de>
6300 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
6301 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
6302 Reset type names to their identifier if their TYPE_DECL doesn't
6303 have linkage (and thus is used for ODR and devirt).
6304 (save_debug_info_for_decl): Remove.
6305 (save_debug_info_for_type): Likewise.
6306 (add_tree_to_fld_list): Adjust.
6307 * tree-pretty-print.c (dump_generic_node): Make dumping of
6308 type names more robust.
6310 2018-01-15 Richard Biener <rguenther@suse.de>
6312 * BASE-VER: Bump to 8.0.1.
6314 2018-01-14 Martin Sebor <msebor@redhat.com>
6317 * builtins.c (check_access): Avoid warning when the no-warning bit
6320 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
6322 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
6323 * ira-color (allocno_hard_regs_compare): Likewise.
6325 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
6328 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
6329 Use .pushsection/.popsection.
6331 2018-01-14 Martin Sebor <msebor@redhat.com>
6334 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
6336 2018-01-14 Jakub Jelinek <jakub@redhat.com>
6338 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
6339 entry from extra_headers.
6340 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
6341 extra_headers, make the list bitwise identical to the i?86-*-* one.
6343 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6345 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
6346 -mcmodel=large with -mindirect-branch=thunk,
6347 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
6348 -mfunction-return=thunk-extern.
6349 * doc/invoke.texi: Document -mcmodel=large is incompatible with
6350 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
6351 -mfunction-return=thunk and -mfunction-return=thunk-extern.
6353 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6355 * config/i386/i386.c (print_reg): Print the name of the full
6356 integer register without '%'.
6357 (ix86_print_operand): Handle 'V'.
6358 * doc/extend.texi: Document 'V' modifier.
6360 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6362 * config/i386/constraints.md (Bs): Disallow memory operand for
6363 -mindirect-branch-register.
6365 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6366 (GOT_memory_operand): Likewise.
6367 (call_insn_operand): Likewise.
6368 (sibcall_insn_operand): Likewise.
6369 (GOT32_symbol_operand): Likewise.
6370 * config/i386/i386.md (indirect_jump): Call convert_memory_address
6371 for -mindirect-branch-register.
6372 (tablejump): Likewise.
6373 (*sibcall_memory): Likewise.
6374 (*sibcall_value_memory): Likewise.
6375 Disallow peepholes of indirect call and jump via memory for
6376 -mindirect-branch-register.
6377 (*call_pop): Replace m with Bw.
6378 (*call_value_pop): Likewise.
6379 (*sibcall_pop_memory): Replace m with Bs.
6380 * config/i386/i386.opt (mindirect-branch-register): New option.
6381 * doc/invoke.texi: Document -mindirect-branch-register option.
6383 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6385 * config/i386/i386-protos.h (ix86_output_function_return): New.
6386 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
6387 set function_return_type.
6388 (indirect_thunk_name): Add ret_p to indicate thunk for function
6390 (output_indirect_thunk_function): Pass false to
6391 indirect_thunk_name.
6392 (ix86_output_indirect_branch_via_reg): Likewise.
6393 (ix86_output_indirect_branch_via_push): Likewise.
6394 (output_indirect_thunk_function): Create alias for function
6395 return thunk if regno < 0.
6396 (ix86_output_function_return): New function.
6397 (ix86_handle_fndecl_attribute): Handle function_return.
6398 (ix86_attribute_table): Add function_return.
6399 * config/i386/i386.h (machine_function): Add
6400 function_return_type.
6401 * config/i386/i386.md (simple_return_internal): Use
6402 ix86_output_function_return.
6403 (simple_return_internal_long): Likewise.
6404 * config/i386/i386.opt (mfunction-return=): New option.
6405 (indirect_branch): Mention -mfunction-return=.
6406 * doc/extend.texi: Document function_return function attribute.
6407 * doc/invoke.texi: Document -mfunction-return= option.
6409 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6411 * config/i386/i386-opts.h (indirect_branch): New.
6412 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
6413 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
6414 with local indirect jump when converting indirect call and jump.
6415 (ix86_set_indirect_branch_type): New.
6416 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
6417 (indirectlabelno): New.
6418 (indirect_thunk_needed): Likewise.
6419 (indirect_thunk_bnd_needed): Likewise.
6420 (indirect_thunks_used): Likewise.
6421 (indirect_thunks_bnd_used): Likewise.
6422 (INDIRECT_LABEL): Likewise.
6423 (indirect_thunk_name): Likewise.
6424 (output_indirect_thunk): Likewise.
6425 (output_indirect_thunk_function): Likewise.
6426 (ix86_output_indirect_branch_via_reg): Likewise.
6427 (ix86_output_indirect_branch_via_push): Likewise.
6428 (ix86_output_indirect_branch): Likewise.
6429 (ix86_output_indirect_jmp): Likewise.
6430 (ix86_code_end): Call output_indirect_thunk_function if needed.
6431 (ix86_output_call_insn): Call ix86_output_indirect_branch if
6433 (ix86_handle_fndecl_attribute): Handle indirect_branch.
6434 (ix86_attribute_table): Add indirect_branch.
6435 * config/i386/i386.h (machine_function): Add indirect_branch_type
6436 and has_local_indirect_jump.
6437 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
6439 (tablejump): Likewise.
6440 (*indirect_jump): Use ix86_output_indirect_jmp.
6441 (*tablejump_1): Likewise.
6442 (simple_return_indirect_internal): Likewise.
6443 * config/i386/i386.opt (mindirect-branch=): New option.
6444 (indirect_branch): New.
6447 (thunk-inline): Likewise.
6448 (thunk-extern): Likewise.
6449 * doc/extend.texi: Document indirect_branch function attribute.
6450 * doc/invoke.texi: Document -mindirect-branch= option.
6452 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
6455 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
6457 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
6459 * ipa-inline.c (want_inline_small_function_p): Return false if
6460 inlining has already failed with CIF_FINAL_ERROR.
6461 (update_caller_keys): Call want_inline_small_function_p before
6463 (update_callee_keys): Likewise.
6465 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6467 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
6469 (rs6000_quadword_masked_address_p): Likewise.
6470 (quad_aligned_load_p): Likewise.
6471 (quad_aligned_store_p): Likewise.
6472 (const_load_sequence_p): Add comment to describe the outer-most loop.
6473 (mimic_memory_attributes_and_flags): New function.
6474 (rs6000_gen_stvx): Likewise.
6475 (replace_swapped_aligned_store): Likewise.
6476 (rs6000_gen_lvx): Likewise.
6477 (replace_swapped_aligned_load): Likewise.
6478 (replace_swapped_load_constant): Capitalize argument name in
6479 comment describing this function.
6480 (rs6000_analyze_swaps): Add a third pass to search for vector loads
6481 and stores that access quad-word aligned addresses and replace
6482 with stvx or lvx instructions when appropriate.
6483 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
6484 New function prototype.
6485 (rs6000_quadword_masked_address_p): Likewise.
6486 (rs6000_gen_lvx): Likewise.
6487 (rs6000_gen_stvx): Likewise.
6488 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
6489 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
6490 when memory address is aligned.
6491 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
6492 this split to select lvx instruction when memory address is aligned.
6493 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
6494 instruction when memory address is aligned.
6495 (*vsx_le_perm_load_v16qi): Likewise.
6496 (four unnamed splitters): Modify to select the stvx instruction
6497 when memory is aligned.
6499 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
6501 * predict.c (determine_unlikely_bbs): Handle correctly BBs
6502 which appears in the queue multiple times.
6504 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6505 Alan Hayward <alan.hayward@arm.com>
6506 David Sherwood <david.sherwood@arm.com>
6508 * tree-vectorizer.h (vec_lower_bound): New structure.
6509 (_loop_vec_info): Add check_nonzero and lower_bounds.
6510 (LOOP_VINFO_CHECK_NONZERO): New macro.
6511 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
6512 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
6513 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
6514 fields. Make seg_len the distance travelled, not including the
6516 (dr_direction_indicator): Declare.
6517 (dr_zero_step_indicator): Likewise.
6518 (dr_known_forward_stride_p): Likewise.
6519 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
6521 (runtime_alias_check_p): Allow runtime alias checks with
6523 (operator ==): Compare access_size and align.
6524 (prune_runtime_alias_test_list): Rework for new distinction between
6525 the access_size and seg_len.
6526 (create_intersect_range_checks_index): Likewise. Cope with polynomial
6528 (get_segment_min_max): New function.
6529 (create_intersect_range_checks): Use it.
6530 (dr_step_indicator): New function.
6531 (dr_direction_indicator): Likewise.
6532 (dr_zero_step_indicator): Likewise.
6533 (dr_known_forward_stride_p): Likewise.
6534 * tree-loop-distribution.c (data_ref_segment_size): Return
6535 DR_STEP * (niters - 1).
6536 (compute_alias_check_pairs): Update call to the dr_with_seg_len
6538 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
6539 (vect_preserves_scalar_order_p): New function, split out from...
6540 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
6541 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
6542 (vect_vfa_access_size): New function.
6543 (vect_vfa_align): Likewise.
6544 (vect_compile_time_alias): Take access_size_a and access_b arguments.
6545 (dump_lower_bound): New function.
6546 (vect_check_lower_bound): Likewise.
6547 (vect_small_gap_p): Likewise.
6548 (vectorizable_with_step_bound_p): Likewise.
6549 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
6550 depencies if the vectorization factor is 1. Convert the checks
6551 for nonzero steps into checks on the bounds of DR_STEP. Try using
6552 a bunds check for variable steps if the minimum required step is
6553 relatively small. Update calls to the dr_with_seg_len
6554 constructor and to vect_compile_time_alias.
6555 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
6557 (vect_loop_versioning): Call it.
6558 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
6560 (vect_estimate_min_profitable_iters): Account for any bounds checks.
6562 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6563 Alan Hayward <alan.hayward@arm.com>
6564 David Sherwood <david.sherwood@arm.com>
6566 * doc/sourcebuild.texi (vect_scatter_store): Document.
6567 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
6569 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
6571 * genopinit.c (main): Add supports_vec_scatter_store and
6572 supports_vec_scatter_store_cached to target_optabs.
6573 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
6574 IFN_MASK_SCATTER_STORE.
6575 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
6577 * internal-fn.h (internal_store_fn_p): Declare.
6578 (internal_fn_stored_value_index): Likewise.
6579 * internal-fn.c (scatter_store_direct): New macro.
6580 (expand_scatter_store_optab_fn): New function.
6581 (direct_scatter_store_optab_supported_p): New macro.
6582 (internal_store_fn_p): New function.
6583 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
6584 IFN_MASK_SCATTER_STORE.
6585 (internal_fn_mask_index): Likewise.
6586 (internal_fn_stored_value_index): New function.
6587 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
6589 * optabs-query.h (supports_vec_scatter_store_p): Declare.
6590 * optabs-query.c (supports_vec_scatter_store_p): New function.
6591 * tree-vectorizer.h (vect_get_store_rhs): Declare.
6592 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
6593 true for scatter stores.
6594 (vect_gather_scatter_fn_p): Handle scatter stores too.
6595 (vect_check_gather_scatter): Consider using scatter stores if
6596 supports_vec_scatter_store_p.
6597 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
6599 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6600 internal_fn_stored_value_index.
6601 (check_load_store_masking): Handle scatter stores too.
6602 (vect_get_store_rhs): Make public.
6603 (vectorizable_call): Use internal_store_fn_p.
6604 (vectorizable_store): Handle scatter store internal functions.
6605 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
6606 when deciding whether the end of the group has been reached.
6607 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
6608 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
6609 (mask_scatter_store<mode>): New insns.
6611 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6612 Alan Hayward <alan.hayward@arm.com>
6613 David Sherwood <david.sherwood@arm.com>
6615 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
6616 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
6617 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
6619 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
6620 Use vect_truncate_gather_scatter_offset if we can't treat the
6621 operation as a normal gather load or scatter store.
6622 (get_group_load_store_type): Take the gather_scatter_info
6623 as argument. Try using a gather load or scatter store for
6624 single-element groups.
6625 (get_load_store_type): Update calls to get_group_load_store_type
6626 and vect_use_strided_gather_scatters_p.
6628 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6629 Alan Hayward <alan.hayward@arm.com>
6630 David Sherwood <david.sherwood@arm.com>
6632 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
6633 optional tree argument.
6634 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
6636 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
6637 but continue to use the current value as a fallback.
6638 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
6639 to compare the updates.
6640 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
6641 (get_load_store_type): Use it when handling a strided access.
6642 (vect_get_strided_load_store_ops): New function.
6643 (vect_get_data_ptr_increment): Likewise.
6644 (vectorizable_load): Handle strided gather loads. Always pass
6645 a step to vect_create_data_ref_ptr and bump_vector_ptr.
6647 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6648 Alan Hayward <alan.hayward@arm.com>
6649 David Sherwood <david.sherwood@arm.com>
6651 * doc/md.texi (gather_load@var{m}): Document.
6652 (mask_gather_load@var{m}): Likewise.
6653 * genopinit.c (main): Add supports_vec_gather_load and
6654 supports_vec_gather_load_cached to target_optabs.
6655 * optabs-tree.c (init_tree_optimization_optabs): Use
6656 ggc_cleared_alloc to allocate target_optabs.
6657 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
6658 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
6660 * internal-fn.h (internal_load_fn_p): Declare.
6661 (internal_gather_scatter_fn_p): Likewise.
6662 (internal_fn_mask_index): Likewise.
6663 (internal_gather_scatter_fn_supported_p): Likewise.
6664 * internal-fn.c (gather_load_direct): New macro.
6665 (expand_gather_load_optab_fn): New function.
6666 (direct_gather_load_optab_supported_p): New macro.
6667 (direct_internal_fn_optab): New function.
6668 (internal_load_fn_p): Likewise.
6669 (internal_gather_scatter_fn_p): Likewise.
6670 (internal_fn_mask_index): Likewise.
6671 (internal_gather_scatter_fn_supported_p): Likewise.
6672 * optabs-query.c (supports_at_least_one_mode_p): New function.
6673 (supports_vec_gather_load_p): Likewise.
6674 * optabs-query.h (supports_vec_gather_load_p): Declare.
6675 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
6676 and memory_type field.
6677 (NUM_PATTERNS): Bump to 15.
6678 * tree-vect-data-refs.c: Include internal-fn.h.
6679 (vect_gather_scatter_fn_p): New function.
6680 (vect_describe_gather_scatter_call): Likewise.
6681 (vect_check_gather_scatter): Try using internal functions for
6682 gather loads. Recognize existing calls to a gather load function.
6683 (vect_analyze_data_refs): Consider using gather loads if
6684 supports_vec_gather_load_p.
6685 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
6686 (vect_get_gather_scatter_offset_type): Likewise.
6687 (vect_convert_mask_for_vectype): Likewise.
6688 (vect_add_conversion_to_patterm): Likewise.
6689 (vect_try_gather_scatter_pattern): Likewise.
6690 (vect_recog_gather_scatter_pattern): New pattern recognizer.
6691 (vect_vect_recog_func_ptrs): Add it.
6692 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6693 internal_fn_mask_index and internal_gather_scatter_fn_p.
6694 (check_load_store_masking): Take the gather_scatter_info as an
6695 argument and handle gather loads.
6696 (vect_get_gather_scatter_ops): New function.
6697 (vectorizable_call): Check internal_load_fn_p.
6698 (vectorizable_load): Likewise. Handle gather load internal
6700 (vectorizable_store): Update call to check_load_store_masking.
6701 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
6702 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
6703 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
6704 (aarch64_gather_scale_operand_d): New predicates.
6705 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
6706 (mask_gather_load<mode>): New insns.
6708 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6709 Alan Hayward <alan.hayward@arm.com>
6710 David Sherwood <david.sherwood@arm.com>
6712 * optabs.def (fold_left_plus_optab): New optab.
6713 * doc/md.texi (fold_left_plus_@var{m}): Document.
6714 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
6715 * internal-fn.c (fold_left_direct): Define.
6716 (expand_fold_left_optab_fn): Likewise.
6717 (direct_fold_left_optab_supported_p): Likewise.
6718 * fold-const-call.c (fold_const_fold_left): New function.
6719 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
6720 * tree-parloops.c (valid_reduction_p): New function.
6721 (gather_scalar_reductions): Use it.
6722 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
6723 (vect_finish_replace_stmt): Declare.
6724 * tree-vect-loop.c (fold_left_reduction_fn): New function.
6725 (needs_fold_left_reduction_p): New function, split out from...
6726 (vect_is_simple_reduction): ...here. Accept reductions that
6727 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
6728 (vect_force_simple_reduction): Also store the reduction type in
6729 the assignment's STMT_VINFO_REDUC_TYPE.
6730 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
6731 (merge_with_identity): New function.
6732 (vect_expand_fold_left): Likewise.
6733 (vectorize_fold_left_reduction): Likewise.
6734 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
6735 scalar phi in place for it. Check for target support and reject
6736 cases that would reassociate the operation. Defer the transform
6737 phase to vectorize_fold_left_reduction.
6738 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
6739 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
6740 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
6742 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6744 * tree-if-conv.c (predicate_mem_writes): Remove redundant
6745 call to ifc_temp_var.
6747 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6748 Alan Hayward <alan.hayward@arm.com>
6749 David Sherwood <david.sherwood@arm.com>
6751 * target.def (legitimize_address_displacement): Take the original
6752 offset as a poly_int.
6753 * targhooks.h (default_legitimize_address_displacement): Update
6755 * targhooks.c (default_legitimize_address_displacement): Likewise.
6756 * doc/tm.texi: Regenerate.
6757 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
6758 as an argument, moving assert of ad->disp == ad->disp_term to...
6759 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
6760 Try calling targetm.legitimize_address_displacement before expanding
6761 the address rather than afterwards, and adjust for the new interface.
6762 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
6763 Match the new hook interface. Handle SVE addresses.
6764 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
6767 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6769 * Makefile.in (OBJS): Add early-remat.o.
6770 * target.def (select_early_remat_modes): New hook.
6771 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
6772 * doc/tm.texi: Regenerate.
6773 * targhooks.h (default_select_early_remat_modes): Declare.
6774 * targhooks.c (default_select_early_remat_modes): New function.
6775 * timevar.def (TV_EARLY_REMAT): New timevar.
6776 * passes.def (pass_early_remat): New pass.
6777 * tree-pass.h (make_pass_early_remat): Declare.
6778 * early-remat.c: New file.
6779 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
6781 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
6783 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6784 Alan Hayward <alan.hayward@arm.com>
6785 David Sherwood <david.sherwood@arm.com>
6787 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
6788 vfm1 with a bound_epilog parameter.
6789 (vect_do_peeling): Update calls accordingly, and move the prologue
6790 call earlier in the function. Treat the base bound_epilog as 0 for
6791 fully-masked loops and retain vf - 1 for other loops. Add 1 to
6792 this base when peeling for gaps.
6793 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
6794 with fully-masked loops.
6795 (vect_estimate_min_profitable_iters): Handle the single peeled
6796 iteration in that case.
6798 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6799 Alan Hayward <alan.hayward@arm.com>
6800 David Sherwood <david.sherwood@arm.com>
6802 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
6803 single-element interleaving even if the size is not a power of 2.
6804 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
6805 accesses for single-element interleaving if the group size is
6808 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6809 Alan Hayward <alan.hayward@arm.com>
6810 David Sherwood <david.sherwood@arm.com>
6812 * doc/md.texi (fold_extract_last_@var{m}): Document.
6813 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
6814 * optabs.def (fold_extract_last_optab): New optab.
6815 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
6816 * internal-fn.c (fold_extract_direct): New macro.
6817 (expand_fold_extract_optab_fn): Likewise.
6818 (direct_fold_extract_optab_supported_p): Likewise.
6819 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
6820 * tree-vect-loop.c (vect_model_reduction_cost): Handle
6821 EXTRACT_LAST_REDUCTION.
6822 (get_initial_def_for_reduction): Do not create an initial vector
6823 for EXTRACT_LAST_REDUCTION reductions.
6824 (vectorizable_reduction): Leave the scalar phi in place for
6825 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
6826 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
6827 epilogue code for EXTRACT_LAST_REDUCTION and defer the
6828 transform phase to vectorizable_condition.
6829 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
6831 (vect_finish_stmt_generation): ...here.
6832 (vect_finish_replace_stmt): New function.
6833 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
6834 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
6836 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
6838 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6839 Alan Hayward <alan.hayward@arm.com>
6840 David Sherwood <david.sherwood@arm.com>
6842 * doc/md.texi (extract_last_@var{m}): Document.
6843 * optabs.def (extract_last_optab): New optab.
6844 * internal-fn.def (EXTRACT_LAST): New internal function.
6845 * internal-fn.c (cond_unary_direct): New macro.
6846 (expand_cond_unary_optab_fn): Likewise.
6847 (direct_cond_unary_optab_supported_p): Likewise.
6848 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
6849 loops using EXTRACT_LAST.
6850 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
6851 (extract_last_<mode>): ...this optab.
6852 (vec_extract<mode><Vel>): Update accordingly.
6854 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6855 Alan Hayward <alan.hayward@arm.com>
6856 David Sherwood <david.sherwood@arm.com>
6858 * target.def (empty_mask_is_expensive): New hook.
6859 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
6860 * doc/tm.texi: Regenerate.
6861 * targhooks.h (default_empty_mask_is_expensive): Declare.
6862 * targhooks.c (default_empty_mask_is_expensive): New function.
6863 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
6864 if the target says that empty masks are expensive.
6865 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
6867 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
6869 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6870 Alan Hayward <alan.hayward@arm.com>
6871 David Sherwood <david.sherwood@arm.com>
6873 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
6874 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
6875 (vect_use_loop_mask_for_alignment_p): New function.
6876 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
6877 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
6878 niters_skip argument. Make sure that the first niters_skip elements
6879 of the first iteration are inactive.
6880 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
6881 Update call to vect_set_loop_masks_directly.
6882 (get_misalign_in_elems): New function, split out from...
6883 (vect_gen_prolog_loop_niters): ...here.
6884 (vect_update_init_of_dr): Take a code argument that specifies whether
6885 the adjustment should be added or subtracted.
6886 (vect_update_init_of_drs): Likewise.
6887 (vect_prepare_for_masked_peels): New function.
6888 (vect_do_peeling): Skip prologue peeling if we're using a mask
6889 instead. Update call to vect_update_inits_of_drs.
6890 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6892 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
6893 alignment. Do not include the number of peeled iterations in
6894 the minimum threshold in that case.
6895 (vectorizable_induction): Adjust the start value down by
6896 LOOP_VINFO_MASK_SKIP_NITERS iterations.
6897 (vect_transform_loop): Call vect_prepare_for_masked_peels.
6898 Take the number of skipped iterations into account when calculating
6900 * tree-vect-stmts.c (vect_gen_while_not): New function.
6902 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6903 Alan Hayward <alan.hayward@arm.com>
6904 David Sherwood <david.sherwood@arm.com>
6906 * doc/sourcebuild.texi (vect_fully_masked): Document.
6907 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
6909 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
6911 (vect_analyze_loop_2): ...here. Don't check the vectorization
6912 factor against the number of loop iterations if the loop is
6915 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6916 Alan Hayward <alan.hayward@arm.com>
6917 David Sherwood <david.sherwood@arm.com>
6919 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
6920 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
6921 (dump_groups): Update accordingly.
6922 (iv_use::mem_type): New member variable.
6923 (address_p): New function.
6924 (record_use): Add a mem_type argument and initialize the new
6926 (record_group_use): Add a mem_type argument. Use address_p.
6927 Remove obsolete null checks of base_object. Update call to record_use.
6928 (find_interesting_uses_op): Update call to record_group_use.
6929 (find_interesting_uses_cond): Likewise.
6930 (find_interesting_uses_address): Likewise.
6931 (get_mem_type_for_internal_fn): New function.
6932 (find_address_like_use): Likewise.
6933 (find_interesting_uses_stmt): Try find_address_like_use before
6934 calling find_interesting_uses_op.
6935 (addr_offset_valid_p): Use the iv mem_type field as the type
6936 of the addressed memory.
6937 (add_autoinc_candidates): Likewise.
6938 (get_address_cost): Likewise.
6939 (split_small_address_groups_p): Use address_p.
6940 (split_address_groups): Likewise.
6941 (add_iv_candidate_for_use): Likewise.
6942 (autoinc_possible_for_pair): Likewise.
6943 (rewrite_groups): Likewise.
6944 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
6945 (determine_group_iv_cost): Update after split of USE_ADDRESS.
6946 (get_alias_ptr_type_for_ptr_address): New function.
6947 (rewrite_use_address): Rewrite address uses in calls that were
6948 identified by find_address_like_use.
6950 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6951 Alan Hayward <alan.hayward@arm.com>
6952 David Sherwood <david.sherwood@arm.com>
6954 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
6956 * gimple-expr.h (is_gimple_addressable: Likewise.
6957 * gimple-expr.c (is_gimple_address): Likewise.
6958 * internal-fn.c (expand_call_mem_ref): New function.
6959 (expand_mask_load_optab_fn): Use it.
6960 (expand_mask_store_optab_fn): Likewise.
6962 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6963 Alan Hayward <alan.hayward@arm.com>
6964 David Sherwood <david.sherwood@arm.com>
6966 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
6967 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
6968 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
6969 (cond_umax@var{mode}): Document.
6970 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
6971 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
6972 (cond_umin_optab, cond_umax_optab): New optabs.
6973 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
6974 (COND_IOR, COND_XOR): New internal functions.
6975 * internal-fn.h (get_conditional_internal_fn): Declare.
6976 * internal-fn.c (cond_binary_direct): New macro.
6977 (expand_cond_binary_optab_fn): Likewise.
6978 (direct_cond_binary_optab_supported_p): Likewise.
6979 (get_conditional_internal_fn): New function.
6980 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
6981 Cope with reduction statements that are vectorized as calls rather
6983 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
6984 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
6985 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
6986 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
6987 (UNSPEC_COND_EOR): New unspecs.
6988 (optab): Add mappings for them.
6989 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
6990 (sve_int_op, sve_fp_op): New int attributes.
6992 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6993 Alan Hayward <alan.hayward@arm.com>
6994 David Sherwood <david.sherwood@arm.com>
6996 * optabs.def (while_ult_optab): New optab.
6997 * doc/md.texi (while_ult@var{m}@var{n}): Document.
6998 * internal-fn.def (WHILE_ULT): New internal function.
6999 * internal-fn.h (direct_internal_fn_supported_p): New override
7000 that takes two types as argument.
7001 * internal-fn.c (while_direct): New macro.
7002 (expand_while_optab_fn): New function.
7003 (convert_optab_supported_p): Likewise.
7004 (direct_while_optab_supported_p): New macro.
7005 * wide-int.h (wi::udiv_ceil): New function.
7006 * tree-vectorizer.h (rgroup_masks): New structure.
7007 (vec_loop_masks): New typedef.
7008 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
7010 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
7011 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
7012 (vect_max_vf): New function.
7013 (slpeel_make_loop_iterate_ntimes): Delete.
7014 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
7015 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
7016 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
7017 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
7018 internal-fn.h, stor-layout.h and optabs-query.h.
7019 (vect_set_loop_mask): New function.
7020 (add_preheader_seq): Likewise.
7021 (add_header_seq): Likewise.
7022 (interleave_supported_p): Likewise.
7023 (vect_maybe_permute_loop_masks): Likewise.
7024 (vect_set_loop_masks_directly): Likewise.
7025 (vect_set_loop_condition_masked): Likewise.
7026 (vect_set_loop_condition_unmasked): New function, split out from
7027 slpeel_make_loop_iterate_ntimes.
7028 (slpeel_make_loop_iterate_ntimes): Rename to..
7029 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
7030 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
7031 (vect_do_peeling): Update call accordingly.
7032 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
7034 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7035 mask_compare_type, can_fully_mask_p and fully_masked_p.
7036 (release_vec_loop_masks): New function.
7037 (_loop_vec_info): Use it to free the loop masks.
7038 (can_produce_all_loop_masks_p): New function.
7039 (vect_get_max_nscalars_per_iter): Likewise.
7040 (vect_verify_full_masking): Likewise.
7041 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
7042 retries, and free the mask rgroups before retrying. Check loop-wide
7043 reasons for disallowing fully-masked loops. Make the final decision
7044 about whether use a fully-masked loop or not.
7045 (vect_estimate_min_profitable_iters): Do not assume that peeling
7046 for the number of iterations will be needed for fully-masked loops.
7047 (vectorizable_reduction): Disable fully-masked loops.
7048 (vectorizable_live_operation): Likewise.
7049 (vect_halve_mask_nunits): New function.
7050 (vect_double_mask_nunits): Likewise.
7051 (vect_record_loop_mask): Likewise.
7052 (vect_get_loop_mask): Likewise.
7053 (vect_transform_loop): Handle the case in which the final loop
7054 iteration might handle a partial vector. Call vect_set_loop_condition
7055 instead of slpeel_make_loop_iterate_ntimes.
7056 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
7057 (check_load_store_masking): New function.
7058 (prepare_load_store_mask): Likewise.
7059 (vectorizable_store): Handle fully-masked loops.
7060 (vectorizable_load): Likewise.
7061 (supportable_widening_operation): Use vect_halve_mask_nunits for
7063 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
7064 (vect_gen_while): New function.
7065 * config/aarch64/aarch64.md (umax<mode>3): New expander.
7066 (aarch64_uqdec<mode>): New insn.
7068 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7069 Alan Hayward <alan.hayward@arm.com>
7070 David Sherwood <david.sherwood@arm.com>
7072 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
7073 (reduc_xor_scal_optab): New optabs.
7074 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
7075 (reduc_xor_scal_@var{m}): Document.
7076 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
7077 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
7079 * fold-const-call.c (fold_const_call): Handle them.
7080 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
7081 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
7082 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
7083 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
7084 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
7085 (UNSPEC_XORV): New unspecs.
7086 (optab): Add entries for them.
7087 (BITWISEV): New int iterator.
7088 (bit_reduc_op): New int attributes.
7090 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7091 Alan Hayward <alan.hayward@arm.com>
7092 David Sherwood <david.sherwood@arm.com>
7094 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
7095 * internal-fn.def (VEC_SHL_INSERT): New internal function.
7096 * optabs.def (vec_shl_insert_optab): New optab.
7097 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
7098 (duplicate_and_interleave): Likewise.
7099 * tree-vect-loop.c: Include internal-fn.h.
7100 (neutral_op_for_slp_reduction): New function, split out from
7101 get_initial_defs_for_reduction.
7102 (get_initial_def_for_reduction): Handle option 2 for variable-length
7103 vectors by loading the neutral value into a vector and then shifting
7104 the initial value into element 0.
7105 (get_initial_defs_for_reduction): Replace the code argument with
7106 the neutral value calculated by neutral_op_for_slp_reduction.
7107 Use gimple_build_vector for constant-length vectors.
7108 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
7109 but the first group_size elements have a neutral value.
7110 Use duplicate_and_interleave otherwise.
7111 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
7112 Update call to get_initial_defs_for_reduction. Handle SLP
7113 reductions for variable-length vectors by creating one vector
7114 result for each scalar result, with the elements associated
7115 with other scalar results stubbed out with the neutral value.
7116 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
7117 Require IFN_VEC_SHL_INSERT for double reductions on
7118 variable-length vectors, or SLP reductions that have
7119 a neutral value. Require can_duplicate_and_interleave_p
7120 support for variable-length unchained SLP reductions if there
7121 is no neutral value, such as for MIN/MAX reductions. Also require
7122 the number of vector elements to be a multiple of the number of
7123 SLP statements when doing variable-length unchained SLP reductions.
7124 Update call to vect_create_epilog_for_reduction.
7125 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
7126 and remove initial values.
7127 (duplicate_and_interleave): Make public.
7128 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
7129 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
7131 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7132 Alan Hayward <alan.hayward@arm.com>
7133 David Sherwood <david.sherwood@arm.com>
7135 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
7136 (can_duplicate_and_interleave_p): New function.
7137 (vect_get_and_check_slp_defs): Take the vector of statements
7138 rather than just the current one. Remove excess parentheses.
7139 Restriction rejectinon of vect_constant_def and vect_external_def
7140 for variable-length vectors to boolean types, or types for which
7141 can_duplicate_and_interleave_p is false.
7142 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
7143 (duplicate_and_interleave): New function.
7144 (vect_get_constant_vectors): Use gimple_build_vector for
7145 constant-length vectors and suitable variable-length constant
7146 vectors. Use duplicate_and_interleave for other variable-length
7147 vectors. Don't defer the update when inserting new statements.
7149 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7150 Alan Hayward <alan.hayward@arm.com>
7151 David Sherwood <david.sherwood@arm.com>
7153 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
7154 min_profitable_iters doesn't go negative.
7156 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7157 Alan Hayward <alan.hayward@arm.com>
7158 David Sherwood <david.sherwood@arm.com>
7160 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
7161 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
7162 * optabs.def (vec_mask_load_lanes_optab): New optab.
7163 (vec_mask_store_lanes_optab): Likewise.
7164 * internal-fn.def (MASK_LOAD_LANES): New internal function.
7165 (MASK_STORE_LANES): Likewise.
7166 * internal-fn.c (mask_load_lanes_direct): New macro.
7167 (mask_store_lanes_direct): Likewise.
7168 (expand_mask_load_optab_fn): Handle masked operations.
7169 (expand_mask_load_lanes_optab_fn): New macro.
7170 (expand_mask_store_optab_fn): Handle masked operations.
7171 (expand_mask_store_lanes_optab_fn): New macro.
7172 (direct_mask_load_lanes_optab_supported_p): Likewise.
7173 (direct_mask_store_lanes_optab_supported_p): Likewise.
7174 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
7176 (vect_load_lanes_supported): Likewise.
7177 * tree-vect-data-refs.c (strip_conversion): New function.
7178 (can_group_stmts_p): Likewise.
7179 (vect_analyze_data_ref_accesses): Use it instead of checking
7180 for a pair of assignments.
7181 (vect_store_lanes_supported): Take a masked_p parameter.
7182 (vect_load_lanes_supported): Likewise.
7183 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
7184 vect_store_lanes_supported and vect_load_lanes_supported.
7185 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7186 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
7187 parameter. Don't allow gaps for masked accesses.
7188 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
7189 and vect_load_lanes_supported.
7190 (get_load_store_type): Take a masked_p parameter and update
7191 call to get_group_load_store_type.
7192 (vectorizable_store): Update call to get_load_store_type.
7193 Handle IFN_MASK_STORE_LANES.
7194 (vectorizable_load): Update call to get_load_store_type.
7195 Handle IFN_MASK_LOAD_LANES.
7197 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7198 Alan Hayward <alan.hayward@arm.com>
7199 David Sherwood <david.sherwood@arm.com>
7201 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
7203 * config/aarch64/aarch64-protos.h
7204 (aarch64_sve_struct_memory_operand_p): Declare.
7205 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
7206 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
7207 (VPRED, vpred): Handle SVE structure modes.
7208 * config/aarch64/constraints.md (Utx): New constraint.
7209 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
7210 (aarch64_sve_struct_nonimmediate_operand): New predicates.
7211 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
7212 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
7213 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
7214 structure modes. Split into pieces after RA.
7215 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
7216 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
7218 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
7219 SVE structure modes.
7220 (aarch64_classify_address): Likewise.
7221 (sizetochar): Move earlier in file.
7222 (aarch64_print_operand): Handle SVE register lists.
7223 (aarch64_array_mode): New function.
7224 (aarch64_sve_struct_memory_operand_p): Likewise.
7225 (TARGET_ARRAY_MODE): Redefine.
7227 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7228 Alan Hayward <alan.hayward@arm.com>
7229 David Sherwood <david.sherwood@arm.com>
7231 * target.def (array_mode): New target hook.
7232 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
7233 * doc/tm.texi: Regenerate.
7234 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
7235 * hooks.c (hook_optmode_mode_uhwi_none): New function.
7236 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
7238 * stor-layout.c (mode_for_array): Likewise. Support polynomial
7241 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7242 Alan Hayward <alan.hayward@arm.com>
7243 David Sherwood <david.sherwood@arm.com>
7245 * fold-const.c (fold_binary_loc): Check the argument types
7246 rather than the result type when testing for a vector operation.
7248 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7250 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
7251 * doc/tm.texi: Regenerate.
7253 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7254 Alan Hayward <alan.hayward@arm.com>
7255 David Sherwood <david.sherwood@arm.com>
7257 * doc/invoke.texi (-msve-vector-bits=): Document new option.
7258 (sve): Document new AArch64 extension.
7259 * doc/md.texi (w): Extend the description of the AArch64
7260 constraint to include SVE vectors.
7261 (Upl, Upa): Document new AArch64 predicate constraints.
7262 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
7264 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
7265 (msve-vector-bits=): New option.
7266 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
7267 SVE when these are disabled.
7268 (sve): New extension.
7269 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
7270 modes. Adjust their number of units based on aarch64_sve_vg.
7271 (MAX_BITSIZE_MODE_ANY_MODE): Define.
7272 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
7273 aarch64_addr_query_type.
7274 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
7275 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
7276 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
7277 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
7278 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
7279 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
7280 (aarch64_simd_imm_zero_p): Delete.
7281 (aarch64_check_zero_based_sve_index_immediate): Declare.
7282 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7283 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7284 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7285 (aarch64_sve_float_mul_immediate_p): Likewise.
7286 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7288 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
7289 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
7290 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
7291 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
7292 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
7293 (aarch64_regmode_natural_size): Likewise.
7294 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
7295 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
7297 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
7298 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
7299 for VG and the SVE predicate registers.
7300 (V_ALIASES): Add a "z"-prefixed alias.
7301 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
7302 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
7303 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
7304 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
7305 (REG_CLASS_NAMES): Add entries for them.
7306 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
7307 and the predicate registers.
7308 (aarch64_sve_vg): Declare.
7309 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
7310 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
7311 (REGMODE_NATURAL_SIZE): Define.
7312 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
7314 * config/aarch64/aarch64.c: Include cfgrtl.h.
7315 (simd_immediate_info): Add a constructor for series vectors,
7316 and an associated step field.
7317 (aarch64_sve_vg): New variable.
7318 (aarch64_dbx_register_number): Handle VG and the predicate registers.
7319 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
7320 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
7321 (VEC_ANY_DATA, VEC_STRUCT): New constants.
7322 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
7323 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
7324 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
7325 (aarch64_get_mask_mode): New functions.
7326 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
7327 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
7328 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
7329 predicate modes and predicate registers. Explicitly restrict
7330 GPRs to modes of 16 bytes or smaller. Only allow FP registers
7331 to store a vector mode if it is recognized by
7332 aarch64_classify_vector_mode.
7333 (aarch64_regmode_natural_size): New function.
7334 (aarch64_hard_regno_caller_save_mode): Return the original mode
7336 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
7337 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
7338 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
7339 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
7341 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
7342 does not overlap dest if the function is frame-related. Handle
7344 (aarch64_split_add_offset): New function.
7345 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
7346 them aarch64_add_offset.
7347 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
7348 and update call to aarch64_sub_sp.
7349 (aarch64_add_cfa_expression): New function.
7350 (aarch64_expand_prologue): Pass extra temporary registers to the
7351 functions above. Handle the case in which we need to emit new
7352 DW_CFA_expressions for registers that were originally saved
7353 relative to the stack pointer, but now have to be expressed
7354 relative to the frame pointer.
7355 (aarch64_output_mi_thunk): Pass extra temporary registers to the
7357 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
7358 IP0 and IP1 values for SVE frames.
7359 (aarch64_expand_vec_series): New function.
7360 (aarch64_expand_sve_widened_duplicate): Likewise.
7361 (aarch64_expand_sve_const_vector): Likewise.
7362 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
7363 Handle SVE constants. Use emit_move_insn to move a force_const_mem
7364 into the register, rather than emitting a SET directly.
7365 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
7366 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
7367 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
7368 (offset_9bit_signed_scaled_p): New functions.
7369 (aarch64_replicate_bitmask_imm): New function.
7370 (aarch64_bitmask_imm): Use it.
7371 (aarch64_cannot_force_const_mem): Reject expressions involving
7372 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
7373 (aarch64_classify_index): Handle SVE indices, by requiring
7374 a plain register index with a scale that matches the element size.
7375 (aarch64_classify_address): Handle SVE addresses. Assert that
7376 the mode of the address is VOIDmode or an integer mode.
7377 Update call to aarch64_classify_symbol.
7378 (aarch64_classify_symbolic_expression): Update call to
7379 aarch64_classify_symbol.
7380 (aarch64_const_vec_all_in_range_p): New function.
7381 (aarch64_print_vector_float_operand): Likewise.
7382 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
7383 "vN" for FP registers with SVE modes. Handle (const ...) vectors
7384 and the FP immediates 1.0 and 0.5.
7385 (aarch64_print_address_internal): Handle SVE addresses.
7386 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
7387 (aarch64_regno_regclass): Handle predicate registers.
7388 (aarch64_secondary_reload): Handle big-endian reloads of SVE
7390 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
7391 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
7392 (aarch64_convert_sve_vector_bits): New function.
7393 (aarch64_override_options): Use it to handle -msve-vector-bits=.
7394 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7396 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
7397 Handle SVE vector and predicate modes. Accept VL-based constants
7398 that need only one temporary register, and VL offsets that require
7399 no temporary registers.
7400 (aarch64_conditional_register_usage): Mark the predicate registers
7401 as fixed if SVE isn't available.
7402 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
7403 Return true for SVE vector and predicate modes.
7404 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
7405 rather than an unsigned int. Handle SVE modes.
7406 (aarch64_preferred_simd_mode): Update call accordingly. Handle
7408 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
7410 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7411 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7412 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7413 (aarch64_sve_float_mul_immediate_p): New functions.
7414 (aarch64_sve_valid_immediate): New function.
7415 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
7416 Explicitly reject structure modes. Check for INDEX constants.
7417 Handle PTRUE and PFALSE constants.
7418 (aarch64_check_zero_based_sve_index_immediate): New function.
7419 (aarch64_simd_imm_zero_p): Delete.
7420 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
7421 vector modes. Accept constants in the range of CNT[BHWD].
7422 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
7423 ask for an Advanced SIMD mode.
7424 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
7425 (aarch64_simd_vector_alignment): Handle SVE predicates.
7426 (aarch64_vectorize_preferred_vector_alignment): New function.
7427 (aarch64_simd_vector_alignment_reachable): Use it instead of
7429 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
7430 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
7432 (MAX_VECT_LEN): Delete.
7433 (expand_vec_perm_d): Add a vec_flags field.
7434 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
7435 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
7436 (aarch64_evpc_ext): Don't apply a big-endian lane correction
7438 (aarch64_evpc_rev): Rename to...
7439 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
7440 (aarch64_evpc_rev_global): New function.
7441 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
7442 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
7444 (aarch64_evpc_sve_tbl): New function.
7445 (aarch64_expand_vec_perm_const_1): Update after rename of
7446 aarch64_evpc_rev. Handle SVE permutes too, trying
7447 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
7448 than aarch64_evpc_tbl.
7449 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
7450 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
7451 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
7452 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
7453 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
7454 (aarch64_expand_sve_vcond): New functions.
7455 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
7456 of aarch64_vector_mode_p.
7457 (aarch64_dwarf_poly_indeterminate_value): New function.
7458 (aarch64_compute_pressure_classes): Likewise.
7459 (aarch64_can_change_mode_class): Likewise.
7460 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
7461 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
7462 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
7463 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
7464 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
7465 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
7466 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
7467 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
7469 (Dn, Dl, Dr): Accept const as well as const_vector.
7470 (Dz): Likewise. Compare against CONST0_RTX.
7471 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
7472 of "vector" where appropriate.
7473 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
7474 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
7475 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
7476 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
7477 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
7478 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
7479 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
7480 (v_int_equiv): Extend to SVE modes.
7481 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
7483 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
7484 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
7485 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
7486 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
7487 (SVE_COND_FP_CMP): New int iterators.
7488 (perm_hilo): Handle the new unpack unspecs.
7489 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
7491 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
7492 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
7493 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
7494 (aarch64_equality_operator, aarch64_constant_vector_operand)
7495 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
7496 (aarch64_sve_nonimmediate_operand): Likewise.
7497 (aarch64_sve_general_operand): Likewise.
7498 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
7499 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
7500 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
7501 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
7502 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
7503 (aarch64_sve_float_arith_immediate): Likewise.
7504 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
7505 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
7506 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
7507 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
7508 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
7509 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
7510 (aarch64_sve_float_arith_operand): Likewise.
7511 (aarch64_sve_float_arith_with_sub_operand): Likewise.
7512 (aarch64_sve_float_mul_operand): Likewise.
7513 (aarch64_sve_vec_perm_operand): Likewise.
7514 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
7515 (aarch64_mov_operand): Accept const_poly_int and const_vector.
7516 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
7517 as well as const_vector.
7518 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
7519 in file. Use CONST0_RTX and CONSTM1_RTX.
7520 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
7521 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
7522 Use aarch64_simd_imm_zero.
7523 * config/aarch64/aarch64-sve.md: New file.
7524 * config/aarch64/aarch64.md: Include it.
7525 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
7526 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
7527 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
7528 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
7529 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
7530 (sve): New attribute.
7531 (enabled): Disable instructions with the sve attribute unless
7533 (movqi, movhi): Pass CONST_POLY_INT operaneds through
7534 aarch64_expand_mov_immediate.
7535 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
7536 CNT[BHSD] immediates.
7537 (movti): Split CONST_POLY_INT moves into two halves.
7538 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
7539 Split additions that need a temporary here if the destination
7540 is the stack pointer.
7541 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
7542 (*add<mode>3_poly_1): New instruction.
7543 (set_clobber_cc): New expander.
7545 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7547 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
7548 parameter and use it instead of GET_MODE_SIZE (innermode). Use
7549 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
7550 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
7551 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
7552 Change innermode from fixed_mode_size to machine_mode.
7553 (simplify_subreg): Update call accordingly. Handle a constant-sized
7554 subreg of a variable-length CONST_VECTOR.
7556 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7557 Alan Hayward <alan.hayward@arm.com>
7558 David Sherwood <david.sherwood@arm.com>
7560 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
7561 (add_offset_to_base): New function, split out from...
7562 (create_mem_ref): ...here. When handling a scale other than 1,
7563 check first whether the address is valid without the offset.
7564 Add it into the base if so, leaving the index and scale as-is.
7566 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7569 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
7570 fold_for_warn before checking if arg2 is INTEGER_CST.
7572 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
7574 * config/rs6000/predicates.md (load_multiple_operation): Delete.
7575 (store_multiple_operation): Delete.
7576 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
7577 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
7578 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
7579 guarded by TARGET_STRING.
7580 (rs6000_output_load_multiple): Delete.
7581 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
7582 OPTION_MASK_STRING / TARGET_STRING handling.
7583 (print_operand) <'N', 'O'>: Add comment that these are unused now.
7584 (const rs6000_opt_masks) <"string">: Change mask to 0.
7585 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
7586 (MASK_STRING): Delete.
7587 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
7589 (load_multiple): Delete.
7596 (store_multiple): Delete.
7603 (movmemsi_8reg): Delete.
7604 (corresponding unnamed define_insn): Delete.
7605 (movmemsi_6reg): Delete.
7606 (corresponding unnamed define_insn): Delete.
7607 (movmemsi_4reg): Delete.
7608 (corresponding unnamed define_insn): Delete.
7609 (movmemsi_2reg): Delete.
7610 (corresponding unnamed define_insn): Delete.
7611 (movmemsi_1reg): Delete.
7612 (corresponding unnamed define_insn): Delete.
7613 * config/rs6000/rs6000.opt (mno-string): New.
7614 (mstring): Replace by deprecation warning stub.
7615 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
7617 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7619 * regrename.c (regrename_do_replace): If replacing the same
7620 reg multiple times, try to reuse last created gen_raw_REG.
7623 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
7624 main to workaround a bug in GDB.
7626 2018-01-12 Tom de Vries <tom@codesourcery.com>
7629 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
7631 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
7633 PR rtl-optimization/80481
7634 * ira-color.c (get_cap_member): New function.
7635 (allocnos_conflict_by_live_ranges_p): Use it.
7636 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
7637 (setup_slot_coalesced_allocno_live_ranges): Ditto.
7639 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
7642 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
7643 (*saddl_se_1): Ditto.
7645 (*ssubl_se_1): Ditto.
7647 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7649 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
7650 rather than wi::to_widest for DR_INITs.
7651 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
7652 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
7653 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
7655 (vect_analyze_group_access_1): Note that here.
7657 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7659 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
7660 polynomial type sizes.
7662 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7664 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
7665 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
7666 (gimple_add_tmp_var): Likewise.
7668 2018-01-12 Martin Liska <mliska@suse.cz>
7670 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
7671 (gimple_alloc_sizes): Likewise.
7672 (dump_gimple_statistics): Use PRIu64 in printf format.
7673 * gimple.h: Change uint64_t to int.
7675 2018-01-12 Martin Liska <mliska@suse.cz>
7677 * tree-core.h: Use uint64_t instead of int.
7678 * tree.c (tree_node_counts): Likewise.
7679 (tree_node_sizes): Likewise.
7680 (dump_tree_statistics): Use PRIu64 in printf format.
7682 2018-01-12 Martin Liska <mliska@suse.cz>
7684 * Makefile.in: As qsort_chk is implemented in vec.c, add
7685 vec.o to linkage of gencfn-macros.
7686 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
7687 passing the info to record_node_allocation_statistics.
7688 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
7690 * ggc-common.c (struct ggc_usage): Add operator== and use
7691 it in operator< and compare function.
7692 * mem-stats.h (struct mem_usage): Likewise.
7693 * vec.c (struct vec_usage): Remove operator< and compare
7694 function. Can be simply inherited.
7696 2018-01-12 Martin Jambor <mjambor@suse.cz>
7699 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
7700 * tree-ssa-math-opts.c: Include domwalk.h.
7701 (convert_mult_to_fma_1): New function.
7702 (fma_transformation_info): New type.
7703 (fma_deferring_state): Likewise.
7704 (cancel_fma_deferring): New function.
7705 (result_of_phi): Likewise.
7706 (last_fma_candidate_feeds_initial_phi): Likewise.
7707 (convert_mult_to_fma): Added deferring logic, split actual
7708 transformation to convert_mult_to_fma_1.
7709 (math_opts_dom_walker): New type.
7710 (math_opts_dom_walker::after_dom_children): New method, body moved
7711 here from pass_optimize_widening_mul::execute, added deferring logic
7713 (pass_optimize_widening_mul::execute): Moved most of code to
7714 math_opts_dom_walker::after_dom_children.
7715 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
7716 * config/i386/i386.c (ix86_option_override_internal): Added
7717 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
7719 2018-01-12 Richard Biener <rguenther@suse.de>
7722 * dwarf2out.c (gen_variable_die): Do not reset old_die for
7723 inline instance vars.
7725 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
7728 * config/rx/rx.c (rx_is_restricted_memory_address):
7731 2018-01-12 Richard Biener <rguenther@suse.de>
7733 PR tree-optimization/80846
7734 * target.def (split_reduction): New target hook.
7735 * targhooks.c (default_split_reduction): New function.
7736 * targhooks.h (default_split_reduction): Declare.
7737 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
7738 target requests first reduce vectors by combining low and high
7740 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
7741 (get_vectype_for_scalar_type_and_size): Export.
7742 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
7743 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
7744 * doc/tm.texi: Regenerate.
7745 * config/i386/i386.c (ix86_split_reduction): Implement
7746 TARGET_VECTORIZE_SPLIT_REDUCTION.
7748 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7751 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
7752 in PIC mode except for TARGET_VXWORKS_RTP.
7753 * config/sparc/sparc.c: Include cfgrtl.h.
7754 (TARGET_INIT_PIC_REG): Define.
7755 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
7756 (sparc_pic_register_p): New predicate.
7757 (sparc_legitimate_address_p): Use it.
7758 (sparc_legitimize_pic_address): Likewise.
7759 (sparc_delegitimize_address): Likewise.
7760 (sparc_mode_dependent_address_p): Likewise.
7761 (gen_load_pcrel_sym): Remove 4th parameter.
7762 (load_got_register): Adjust call to above. Remove obsolete stuff.
7763 (sparc_expand_prologue): Do not call load_got_register here.
7764 (sparc_flat_expand_prologue): Likewise.
7765 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
7766 (sparc_use_pseudo_pic_reg): New function.
7767 (sparc_init_pic_reg): Likewise.
7768 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7769 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
7771 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
7773 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
7774 Add item for branch_cost.
7776 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7778 PR rtl-optimization/83565
7779 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
7780 not extend the result to a larger mode for rotate operations.
7781 (num_sign_bit_copies1): Likewise.
7783 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7786 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
7788 Use values-Xc.o for -pedantic.
7789 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
7791 2018-01-12 Martin Liska <mliska@suse.cz>
7794 * ipa-devirt.c (final_warning_record::grow_type_warnings):
7796 (possible_polymorphic_call_targets): Use it.
7797 (ipa_devirt): Likewise.
7799 2018-01-12 Martin Liska <mliska@suse.cz>
7801 * profile-count.h (enum profile_quality): Use 0 as invalid
7802 enum value of profile_quality.
7804 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
7806 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
7807 -mext-string options.
7809 2018-01-12 Richard Biener <rguenther@suse.de>
7811 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
7812 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
7813 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
7815 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
7817 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
7819 * configure.ac (--with-long-double-format): Add support for the
7820 configuration option to change the default long double format on
7822 * config.gcc (powerpc*-linux*-*): Likewise.
7823 * configure: Regenerate.
7824 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
7825 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
7826 used without modification.
7828 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7830 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
7831 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
7832 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
7833 MISC_BUILTIN_SPEC_BARRIER.
7834 (rs6000_init_builtins): Likewise.
7835 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
7837 (speculation_barrier): New define_insn.
7838 * doc/extend.texi: Document __builtin_speculation_barrier.
7840 2018-01-11 Jakub Jelinek <jakub@redhat.com>
7843 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
7844 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
7845 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
7847 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
7848 integral modes instead of "ss" and "sd".
7849 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
7850 vectors with 32-bit and 64-bit elements.
7851 (vecdupssescalarmodesuffix): New mode attribute.
7852 (vec_dup<mode>): Use it.
7854 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
7857 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
7858 frame if argument is passed on stack.
7860 2018-01-11 Jakub Jelinek <jakub@redhat.com>
7863 * ree.c (combine_reaching_defs): Optimize also
7864 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
7865 reg2=any_extend(exp); reg1=reg2;, formatting fix.
7867 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
7870 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
7872 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
7875 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
7876 after they are computed.
7878 2018-01-11 Bin Cheng <bin.cheng@arm.com>
7880 PR tree-optimization/83695
7881 * gimple-loop-linterchange.cc
7882 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
7883 reset cached scev information after interchange.
7884 (pass_linterchange::execute): Remove call to scev_reset_htab.
7886 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7888 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
7889 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
7890 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
7891 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
7892 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
7893 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
7894 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
7895 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
7896 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
7897 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
7898 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
7899 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
7900 (V_lane_reg): Likewise.
7901 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
7903 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
7904 (vfmal_lane_low<mode>_intrinsic,
7905 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
7906 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
7907 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
7908 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
7909 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
7910 vfmsl_lane_high<mode>_intrinsic): New define_insns.
7912 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7914 * config/arm/arm-cpus.in (fp16fml): New feature.
7915 (ALL_SIMD): Add fp16fml.
7916 (armv8.2-a): Add fp16fml as an option.
7917 (armv8.3-a): Likewise.
7918 (armv8.4-a): Add fp16fml as part of fp16.
7919 * config/arm/arm.h (TARGET_FP16FML): Define.
7920 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
7922 * config/arm/arm-modes.def (V2HF): Define.
7923 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
7924 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
7925 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
7926 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
7927 vfmsl_low, vfmsl_high): New set of builtins.
7928 * config/arm/iterators.md (PLUSMINUS): New code iterator.
7929 (vfml_op): New code attribute.
7930 (VFMLHALVES): New int iterator.
7931 (VFML, VFMLSEL): New mode attributes.
7932 (V_reg): Define mapping for V2HF.
7933 (V_hi, V_lo): New mode attributes.
7934 (VF_constraint): Likewise.
7935 (vfml_half, vfml_half_selector): New int attributes.
7936 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
7938 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
7939 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
7941 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
7942 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
7943 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
7944 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
7946 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
7947 Document new effective target and option set.
7949 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7951 * config/arm/arm-cpus.in (armv8_4): New feature.
7952 (ARMv8_4a): New fgroup.
7953 (armv8.4-a): New arch.
7954 * config/arm/arm-tables.opt: Regenerate.
7955 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
7956 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
7957 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
7958 Add matching rules for -march=armv8.4-a and extensions.
7959 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
7961 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
7964 * config/rx/rx.md (BW): New mode attribute.
7965 (sync_lock_test_and_setsi): Add mode suffix to insn output.
7967 2018-01-11 Richard Biener <rguenther@suse.de>
7969 PR tree-optimization/83435
7970 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
7971 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
7972 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
7974 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7975 Alan Hayward <alan.hayward@arm.com>
7976 David Sherwood <david.sherwood@arm.com>
7978 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
7980 (aarch64_classify_address): Initialize it. Track polynomial offsets.
7981 (aarch64_print_address_internal): Use it to check for a zero offset.
7983 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7984 Alan Hayward <alan.hayward@arm.com>
7985 David Sherwood <david.sherwood@arm.com>
7987 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
7988 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
7989 Return a poly_int64 rather than a HOST_WIDE_INT.
7990 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
7991 rather than a HOST_WIDE_INT.
7992 * config/aarch64/aarch64.h (aarch64_frame): Protect with
7993 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
7994 hard_fp_offset, frame_size, initial_adjust, callee_offset and
7995 final_offset from HOST_WIDE_INT to poly_int64.
7996 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
7997 to_constant when getting the number of units in an Advanced SIMD
7999 (aarch64_builtin_vectorized_function): Check for a constant number
8001 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
8003 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
8004 attribute instead of GET_MODE_NUNITS.
8005 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
8006 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
8007 GET_MODE_SIZE for fixed-size registers.
8008 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
8009 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
8010 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
8011 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
8012 (aarch64_print_operand, aarch64_print_address_internal)
8013 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
8014 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
8015 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
8016 Handle polynomial GET_MODE_SIZE.
8017 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
8018 wider than SImode without modification.
8019 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
8020 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
8021 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
8022 passing and returning SVE modes.
8023 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
8024 rather than GEN_INT.
8025 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
8026 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
8027 (aarch64_allocate_and_probe_stack_space): Likewise.
8028 (aarch64_layout_frame): Cope with polynomial offsets.
8029 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
8030 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
8032 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
8033 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
8034 poly_int64 rather than a HOST_WIDE_INT.
8035 (aarch64_get_separate_components, aarch64_process_components)
8036 (aarch64_expand_prologue, aarch64_expand_epilogue)
8037 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
8038 (aarch64_anchor_offset): New function, split out from...
8039 (aarch64_legitimize_address): ...here.
8040 (aarch64_builtin_vectorization_cost): Handle polynomial
8041 TYPE_VECTOR_SUBPARTS.
8042 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
8044 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
8045 number of elements from the PARALLEL rather than the mode.
8046 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
8047 rather than GET_MODE_BITSIZE.
8048 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
8049 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
8050 (aarch64_expand_vec_perm_const_1): Handle polynomial
8051 d->perm.length () and d->perm elements.
8052 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
8053 Apply to_constant to d->perm elements.
8054 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
8055 polynomial CONST_VECTOR_NUNITS.
8056 (aarch64_move_pointer): Take amount as a poly_int64 rather
8058 (aarch64_progress_pointer): Avoid temporary variable.
8059 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
8060 the mode attribute instead of GET_MODE.
8062 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8063 Alan Hayward <alan.hayward@arm.com>
8064 David Sherwood <david.sherwood@arm.com>
8066 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
8067 x exists before using it.
8068 (aarch64_add_constant_internal): Rename to...
8069 (aarch64_add_offset_1): ...this. Replace regnum with separate
8070 src and dest rtxes. Handle the case in which they're different,
8071 including when the offset is zero. Replace scratchreg with an rtx.
8072 Use 2 additions if there is no spare register into which we can
8073 move a 16-bit constant.
8074 (aarch64_add_constant): Delete.
8075 (aarch64_add_offset): Replace reg with separate src and dest
8076 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
8077 Use aarch64_add_offset_1.
8078 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
8079 an rtx rather than an int. Take the delta as a poly_int64
8080 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
8081 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
8082 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
8083 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
8084 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
8086 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
8087 aarch64_add_constant.
8089 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8091 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
8092 Use scalar_float_mode.
8094 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8096 * config/aarch64/aarch64-simd.md
8097 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
8098 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
8099 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
8100 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
8101 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
8102 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
8103 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
8104 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
8105 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
8106 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
8108 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8111 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
8112 targ_options->x_arm_arch_string is non NULL.
8114 2018-01-11 Tamar Christina <tamar.christina@arm.com>
8116 * config/aarch64/aarch64.h
8117 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
8119 2018-01-11 Sudakshina Das <sudi.das@arm.com>
8122 * expmed.c (emit_store_flag_force): Swap if const op0
8123 and change VOIDmode to mode of op0.
8125 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8127 PR rtl-optimization/83761
8128 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
8129 than bytes to mode_for_size.
8131 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8134 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
8135 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
8138 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8141 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
8142 when in layout mode.
8143 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
8144 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
8147 2018-01-10 Michael Collison <michael.collison@arm.com>
8149 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
8150 * config/aarch64/aarch64-option-extension.def: Add
8151 AARCH64_OPT_EXTENSION of 'fp16fml'.
8152 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8153 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
8154 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
8155 * config/aarch64/constraints.md (Ui7): New constraint.
8156 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
8157 (VFMLA_SEL_W): Ditto.
8160 (VFMLA16_LOW): New int iterator.
8161 (VFMLA16_HIGH): Ditto.
8162 (UNSPEC_FMLAL): New unspec.
8163 (UNSPEC_FMLSL): Ditto.
8164 (UNSPEC_FMLAL2): Ditto.
8165 (UNSPEC_FMLSL2): Ditto.
8166 (f16mac): New code attribute.
8167 * config/aarch64/aarch64-simd-builtins.def
8168 (aarch64_fmlal_lowv2sf): Ditto.
8169 (aarch64_fmlsl_lowv2sf): Ditto.
8170 (aarch64_fmlalq_lowv4sf): Ditto.
8171 (aarch64_fmlslq_lowv4sf): Ditto.
8172 (aarch64_fmlal_highv2sf): Ditto.
8173 (aarch64_fmlsl_highv2sf): Ditto.
8174 (aarch64_fmlalq_highv4sf): Ditto.
8175 (aarch64_fmlslq_highv4sf): Ditto.
8176 (aarch64_fmlal_lane_lowv2sf): Ditto.
8177 (aarch64_fmlsl_lane_lowv2sf): Ditto.
8178 (aarch64_fmlal_laneq_lowv2sf): Ditto.
8179 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
8180 (aarch64_fmlalq_lane_lowv4sf): Ditto.
8181 (aarch64_fmlsl_lane_lowv4sf): Ditto.
8182 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
8183 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
8184 (aarch64_fmlal_lane_highv2sf): Ditto.
8185 (aarch64_fmlsl_lane_highv2sf): Ditto.
8186 (aarch64_fmlal_laneq_highv2sf): Ditto.
8187 (aarch64_fmlsl_laneq_highv2sf): Ditto.
8188 (aarch64_fmlalq_lane_highv4sf): Ditto.
8189 (aarch64_fmlsl_lane_highv4sf): Ditto.
8190 (aarch64_fmlalq_laneq_highv4sf): Ditto.
8191 (aarch64_fmlsl_laneq_highv4sf): Ditto.
8192 * config/aarch64/aarch64-simd.md:
8193 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
8194 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8195 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
8196 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8197 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
8198 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
8199 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
8200 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
8201 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
8202 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
8203 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
8204 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
8205 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
8206 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
8207 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
8208 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
8209 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
8210 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
8211 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
8212 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
8213 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
8214 (vfmlsl_low_u32): Ditto.
8215 (vfmlalq_low_u32): Ditto.
8216 (vfmlslq_low_u32): Ditto.
8217 (vfmlal_high_u32): Ditto.
8218 (vfmlsl_high_u32): Ditto.
8219 (vfmlalq_high_u32): Ditto.
8220 (vfmlslq_high_u32): Ditto.
8221 (vfmlal_lane_low_u32): Ditto.
8222 (vfmlsl_lane_low_u32): Ditto.
8223 (vfmlal_laneq_low_u32): Ditto.
8224 (vfmlsl_laneq_low_u32): Ditto.
8225 (vfmlalq_lane_low_u32): Ditto.
8226 (vfmlslq_lane_low_u32): Ditto.
8227 (vfmlalq_laneq_low_u32): Ditto.
8228 (vfmlslq_laneq_low_u32): Ditto.
8229 (vfmlal_lane_high_u32): Ditto.
8230 (vfmlsl_lane_high_u32): Ditto.
8231 (vfmlal_laneq_high_u32): Ditto.
8232 (vfmlsl_laneq_high_u32): Ditto.
8233 (vfmlalq_lane_high_u32): Ditto.
8234 (vfmlslq_lane_high_u32): Ditto.
8235 (vfmlalq_laneq_high_u32): Ditto.
8236 (vfmlslq_laneq_high_u32): Ditto.
8237 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
8238 (AARCH64_FL_FOR_ARCH8_4): New.
8239 (AARCH64_ISA_F16FML): New ISA flag.
8240 (TARGET_F16FML): New feature flag for fp16fml.
8241 (doc/invoke.texi): Document new fp16fml option.
8243 2018-01-10 Michael Collison <michael.collison@arm.com>
8245 * config/aarch64/aarch64-builtins.c:
8246 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
8247 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8248 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
8249 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
8250 (AARCH64_ISA_SHA3): New ISA flag.
8251 (TARGET_SHA3): New feature flag for sha3.
8252 * config/aarch64/iterators.md (sha512_op): New int attribute.
8253 (CRYPTO_SHA512): New int iterator.
8254 (UNSPEC_SHA512H): New unspec.
8255 (UNSPEC_SHA512H2): Ditto.
8256 (UNSPEC_SHA512SU0): Ditto.
8257 (UNSPEC_SHA512SU1): Ditto.
8258 * config/aarch64/aarch64-simd-builtins.def
8259 (aarch64_crypto_sha512hqv2di): New builtin.
8260 (aarch64_crypto_sha512h2qv2di): Ditto.
8261 (aarch64_crypto_sha512su0qv2di): Ditto.
8262 (aarch64_crypto_sha512su1qv2di): Ditto.
8263 (aarch64_eor3qv8hi): Ditto.
8264 (aarch64_rax1qv2di): Ditto.
8265 (aarch64_xarqv2di): Ditto.
8266 (aarch64_bcaxqv8hi): Ditto.
8267 * config/aarch64/aarch64-simd.md:
8268 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
8269 (aarch64_crypto_sha512su0qv2di): Ditto.
8270 (aarch64_crypto_sha512su1qv2di): Ditto.
8271 (aarch64_eor3qv8hi): Ditto.
8272 (aarch64_rax1qv2di): Ditto.
8273 (aarch64_xarqv2di): Ditto.
8274 (aarch64_bcaxqv8hi): Ditto.
8275 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
8276 (vsha512h2q_u64): Ditto.
8277 (vsha512su0q_u64): Ditto.
8278 (vsha512su1q_u64): Ditto.
8279 (veor3q_u16): Ditto.
8280 (vrax1q_u64): Ditto.
8282 (vbcaxq_u16): Ditto.
8283 * config/arm/types.md (crypto_sha512): New type attribute.
8284 (crypto_sha3): Ditto.
8285 (doc/invoke.texi): Document new sha3 option.
8287 2018-01-10 Michael Collison <michael.collison@arm.com>
8289 * config/aarch64/aarch64-builtins.c:
8290 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
8291 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8292 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
8293 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
8294 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
8295 (AARCH64_ISA_SM4): New ISA flag.
8296 (TARGET_SM4): New feature flag for sm4.
8297 * config/aarch64/aarch64-simd-builtins.def
8298 (aarch64_sm3ss1qv4si): Ditto.
8299 (aarch64_sm3tt1aq4si): Ditto.
8300 (aarch64_sm3tt1bq4si): Ditto.
8301 (aarch64_sm3tt2aq4si): Ditto.
8302 (aarch64_sm3tt2bq4si): Ditto.
8303 (aarch64_sm3partw1qv4si): Ditto.
8304 (aarch64_sm3partw2qv4si): Ditto.
8305 (aarch64_sm4eqv4si): Ditto.
8306 (aarch64_sm4ekeyqv4si): Ditto.
8307 * config/aarch64/aarch64-simd.md:
8308 (aarch64_sm3ss1qv4si): Ditto.
8309 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
8310 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
8311 (aarch64_sm4eqv4si): Ditto.
8312 (aarch64_sm4ekeyqv4si): Ditto.
8313 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
8314 (sm3part_op): Ditto.
8315 (CRYPTO_SM3TT): Ditto.
8316 (CRYPTO_SM3PART): Ditto.
8317 (UNSPEC_SM3SS1): New unspec.
8318 (UNSPEC_SM3TT1A): Ditto.
8319 (UNSPEC_SM3TT1B): Ditto.
8320 (UNSPEC_SM3TT2A): Ditto.
8321 (UNSPEC_SM3TT2B): Ditto.
8322 (UNSPEC_SM3PARTW1): Ditto.
8323 (UNSPEC_SM3PARTW2): Ditto.
8324 (UNSPEC_SM4E): Ditto.
8325 (UNSPEC_SM4EKEY): Ditto.
8326 * config/aarch64/constraints.md (Ui2): New constraint.
8327 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
8328 * config/arm/types.md (crypto_sm3): New type attribute.
8329 (crypto_sm4): Ditto.
8330 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
8331 (vsm3tt1aq_u32): Ditto.
8332 (vsm3tt1bq_u32): Ditto.
8333 (vsm3tt2aq_u32): Ditto.
8334 (vsm3tt2bq_u32): Ditto.
8335 (vsm3partw1q_u32): Ditto.
8336 (vsm3partw2q_u32): Ditto.
8337 (vsm4eq_u32): Ditto.
8338 (vsm4ekeyq_u32): Ditto.
8339 (doc/invoke.texi): Document new sm4 option.
8341 2018-01-10 Michael Collison <michael.collison@arm.com>
8343 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
8344 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
8345 (AARCH64_FL_FOR_ARCH8_4): New.
8346 (AARCH64_FL_V8_4): New flag.
8347 (doc/invoke.texi): Document new armv8.4-a option.
8349 2018-01-10 Michael Collison <michael.collison@arm.com>
8351 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8352 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
8353 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
8354 * config/aarch64/aarch64-option-extension.def: Add
8355 AARCH64_OPT_EXTENSION of 'sha2'.
8356 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
8357 (crypto): Disable sha2 and aes if crypto disabled.
8358 (crypto): Enable aes and sha2 if enabled.
8359 (simd): Disable sha2 and aes if simd disabled.
8360 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
8362 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
8363 (TARGET_SHA2): New feature flag for sha2.
8364 (TARGET_AES): New feature flag for aes.
8365 * config/aarch64/aarch64-simd.md:
8366 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
8367 conditional on TARGET_AES.
8368 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
8369 (aarch64_crypto_sha1hsi): Make pattern conditional
8371 (aarch64_crypto_sha1hv4si): Ditto.
8372 (aarch64_be_crypto_sha1hv4si): Ditto.
8373 (aarch64_crypto_sha1su1v4si): Ditto.
8374 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
8375 (aarch64_crypto_sha1su0v4si): Ditto.
8376 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
8377 (aarch64_crypto_sha256su0v4si): Ditto.
8378 (aarch64_crypto_sha256su1v4si): Ditto.
8379 (doc/invoke.texi): Document new aes and sha2 options.
8381 2018-01-10 Martin Sebor <msebor@redhat.com>
8383 PR tree-optimization/83781
8384 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
8387 2018-01-11 Martin Sebor <msebor@gmail.com>
8388 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8390 PR tree-optimization/83501
8391 PR tree-optimization/81703
8393 * tree-ssa-strlen.c (get_string_cst): Rename...
8394 (get_string_len): ...to this. Handle global constants.
8395 (handle_char_store): Adjust.
8397 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
8398 Jim Wilson <jimw@sifive.com>
8400 * config/riscv/riscv-protos.h (riscv_output_return): New.
8401 * config/riscv/riscv.c (struct machine_function): New naked_p field.
8402 (riscv_attribute_table, riscv_output_return),
8403 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
8404 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
8405 (riscv_compute_frame_info): Only compute frame->mask if not a naked
8407 (riscv_expand_prologue): Add early return for naked function.
8408 (riscv_expand_epilogue): Likewise.
8409 (riscv_function_ok_for_sibcall): Return false for naked function.
8410 (riscv_set_current_function): New.
8411 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
8412 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
8413 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
8414 * doc/extend.texi (RISC-V Function Attributes): New.
8416 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
8418 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
8419 check for 128-bit long double before checking TCmode.
8420 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
8421 128-bit long doubles before checking TFmode or TCmode.
8422 (FLOAT128_IBM_P): Likewise.
8424 2018-01-10 Martin Sebor <msebor@redhat.com>
8426 PR tree-optimization/83671
8427 * builtins.c (c_strlen): Unconditionally return zero for the empty
8429 Use -Warray-bounds for warnings.
8430 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
8431 for non-constant array indices with COMPONENT_REF, arrays of
8432 arrays, and pointers to arrays.
8433 (gimple_fold_builtin_strlen): Determine and set length range for
8434 non-constant character arrays.
8436 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
8439 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
8442 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
8444 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
8446 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8449 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
8450 VECTOR_MEM_ALTIVEC_OR_VSX_P.
8451 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
8452 indexed_or_indirect_operand predicate.
8453 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
8454 (*vsx_le_perm_load_v8hi): Likewise.
8455 (*vsx_le_perm_load_v16qi): Likewise.
8456 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
8457 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
8458 (*vsx_le_perm_store_v8hi): Likewise.
8459 (*vsx_le_perm_store_v16qi): Likewise.
8460 (eight unnamed splitters): Likewise.
8462 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8464 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
8465 * config/rs6000/emmintrin.h: Likewise.
8466 * config/rs6000/mmintrin.h: Likewise.
8467 * config/rs6000/xmmintrin.h: Likewise.
8469 2018-01-10 David Malcolm <dmalcolm@redhat.com>
8472 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
8474 * tree.c (tree_nop_conversion): Return true for location wrapper
8476 (maybe_wrap_with_location): New function.
8477 (selftest::check_strip_nops): New function.
8478 (selftest::test_location_wrappers): New function.
8479 (selftest::tree_c_tests): Call it.
8480 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
8481 (maybe_wrap_with_location): New decl.
8482 (EXPR_LOCATION_WRAPPER_P): New macro.
8483 (location_wrapper_p): New inline function.
8484 (tree_strip_any_location_wrapper): New inline function.
8486 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
8489 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
8490 stack_realign_offset for the largest alignment of stack slot
8492 (ix86_find_max_used_stack_alignment): New function.
8493 (ix86_finalize_stack_frame_flags): Use it. Set
8494 max_used_stack_alignment if we don't realign stack.
8495 * config/i386/i386.h (machine_function): Add
8496 max_used_stack_alignment.
8498 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
8500 * config/arm/arm.opt (-mbranch-cost): New option.
8501 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
8504 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
8507 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
8508 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
8510 2018-01-10 Richard Biener <rguenther@suse.de>
8513 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
8514 early out so it also covers the case where we have a non-NULL
8517 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8519 PR tree-optimization/83753
8520 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
8521 for non-strided grouped accesses if the number of elements is 1.
8523 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8526 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
8527 * i386.h (TARGET_USE_GATHER): Define.
8528 * x86-tune.def (X86_TUNE_USE_GATHER): New.
8530 2018-01-10 Martin Liska <mliska@suse.cz>
8533 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
8534 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
8536 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
8537 CLEANUP_NO_PARTITIONING is not set.
8539 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8541 * doc/rtl.texi: Remove documentation of (const ...) wrappers
8542 for vectors, as a partial revert of r254296.
8543 * rtl.h (const_vec_p): Delete.
8544 (const_vec_duplicate_p): Don't test for vector CONSTs.
8545 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
8546 * expmed.c (make_tree): Likewise.
8549 * common.md (E, F): Use CONSTANT_P instead of checking for
8551 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
8552 checking for CONST_VECTOR.
8554 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8557 * predict.c (force_edge_cold): Handle in more sane way edges
8560 2018-01-09 Carl Love <cel@us.ibm.com>
8562 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
8564 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
8565 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
8566 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
8567 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
8568 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
8569 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
8570 * config/rs6000/rs6000-protos.h: Add extern defition for
8571 rs6000_generate_float2_double_code.
8572 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
8574 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
8575 (float2_v2df): Add define_expand.
8577 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
8580 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
8581 op_mode in the force_to_mode call.
8583 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8585 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
8586 instead of checking each element individually.
8587 (aarch64_evpc_uzp): Likewise.
8588 (aarch64_evpc_zip): Likewise.
8589 (aarch64_evpc_ext): Likewise.
8590 (aarch64_evpc_rev): Likewise.
8591 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
8592 instead of checking each element individually. Return true without
8594 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
8595 whether all selected elements come from the same input, instead of
8596 checking each element individually. Remove calls to gen_rtx_REG,
8597 start_sequence and end_sequence and instead assert that no rtl is
8600 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8602 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
8603 order of HIGH and CONST checks.
8605 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8607 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
8608 if the destination isn't an SSA_NAME.
8610 2018-01-09 Richard Biener <rguenther@suse.de>
8612 PR tree-optimization/83668
8613 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
8615 (canonicalize_loop_form): ... here, renamed from ...
8616 (canonicalize_loop_closed_ssa_form): ... this and amended to
8617 swap successor edges for loop exit blocks to make us use
8618 the RPO order we need for initial schedule generation.
8620 2018-01-09 Joseph Myers <joseph@codesourcery.com>
8622 PR tree-optimization/64811
8623 * match.pd: When optimizing comparisons with Inf, avoid
8624 introducing or losing exceptions from comparisons with NaN.
8626 2018-01-09 Martin Liska <mliska@suse.cz>
8629 * asan.c (shadow_mem_size): Add gcc_assert.
8631 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
8633 Don't save registers in main().
8636 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
8637 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
8638 * config/avr/avr.c (avr_set_current_function): Don't error if
8639 naked, OS_task or OS_main are specified at the same time.
8640 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
8642 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
8644 * common/config/avr/avr-common.c (avr_option_optimization_table):
8645 Switch on -mmain-is-OS_task for optimizing compilations.
8647 2018-01-09 Richard Biener <rguenther@suse.de>
8649 PR tree-optimization/83572
8650 * graphite.c: Include cfganal.h.
8651 (graphite_transform_loops): Connect infinite loops to exit
8652 and remove fake edges at the end.
8654 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8656 * ipa-inline.c (edge_badness): Revert accidental checkin.
8658 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8661 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
8662 symbols; not inline clones.
8664 2018-01-09 Jakub Jelinek <jakub@redhat.com>
8667 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
8668 hard registers. Formatting fixes.
8670 PR preprocessor/83722
8671 * gcc.c (try_generate_repro): Pass
8672 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
8673 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
8676 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
8677 Kito Cheng <kito.cheng@gmail.com>
8679 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
8680 (riscv_leaf_function_p): Delete.
8681 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
8683 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8685 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
8687 (do_ifelse): New function.
8688 (do_isel): New function.
8689 (do_sub3): New function.
8690 (do_add3): New function.
8691 (do_load_mask_compare): New function.
8692 (do_overlap_load_compare): New function.
8693 (expand_compare_loop): New function.
8694 (expand_block_compare): Call expand_compare_loop() when appropriate.
8695 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
8697 (-mblock-compare-inline-loop-limit): New option.
8699 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8702 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
8703 Reverse order of second and third operands in first alternative.
8704 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
8705 of first and second elements in UNSPEC_VPERMR vector.
8706 (altivec_expand_vec_perm_le): Likewise.
8708 2018-01-08 Jeff Law <law@redhat.com>
8710 PR rtl-optimizatin/81308
8711 * tree-switch-conversion.c (cfg_altered): New file scoped static.
8712 (process_switch): If group_case_labels makes a change, then set
8714 (pass_convert_switch::execute): If a switch is converted, then
8715 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
8717 PR rtl-optimization/81308
8718 * recog.c (split_all_insns): Conditionally cleanup the CFG after
8721 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
8723 PR target/83663 - Revert r255946
8724 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
8725 generation for cases where splatting a value is not useful.
8726 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
8727 across a vec_duplicate and a paradoxical subreg forming a vector
8728 mode to a vec_concat.
8730 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8732 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
8733 -march=armv8.3-a variants.
8734 * config/arm/t-multilib: Likewise.
8735 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
8737 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8739 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
8741 (cceq_ior_compare_complement): Give it a name so I can use it, and
8742 change boolean_or_operator predicate to boolean_operator so it can
8743 be used to generate a crand.
8744 (eqne): New code iterator.
8745 (bd/bd_neg): New code_attrs.
8746 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
8747 a single define_insn.
8748 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
8749 decrement (bdnzt/bdnzf/bdzt/bdzf).
8750 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
8751 with the new names of the branch decrement patterns, and added the
8752 names of the branch decrement conditional patterns.
8754 2018-01-08 Richard Biener <rguenther@suse.de>
8756 PR tree-optimization/83563
8757 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
8760 2018-01-08 Richard Biener <rguenther@suse.de>
8763 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
8765 2018-01-08 Richard Biener <rguenther@suse.de>
8767 PR tree-optimization/83685
8768 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
8769 references to abnormals.
8771 2018-01-08 Richard Biener <rguenther@suse.de>
8774 * dwarf2out.c (output_indirect_strings): Handle empty
8775 skeleton_debug_str_hash.
8776 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
8778 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
8780 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
8781 (emit_store_direct): Likewise.
8782 (arc_trampoline_adjust_address): Likewise.
8783 (arc_asm_trampoline_template): New function.
8784 (arc_initialize_trampoline): Use asm_trampoline_template.
8785 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
8786 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
8787 * config/arc/arc.md (flush_icache): Delete pattern.
8789 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
8791 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
8792 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
8795 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
8798 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
8799 by not USED_FOR_TARGET.
8800 (make_pass_resolve_sw_modes): Likewise.
8802 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
8804 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
8807 2018-01-08 Richard Biener <rguenther@suse.de>
8810 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
8812 2018-01-08 Richard Biener <rguenther@suse.de>
8815 * match.pd ((t * 2) / 2) -> t): Add missing :c.
8817 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
8820 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
8821 basic blocks with a small number of successors.
8822 (convert_control_dep_chain_into_preds): Improve handling of
8824 (dump_predicates): Split apart into...
8825 (dump_pred_chain): ...here...
8826 (dump_pred_info): ...and here.
8827 (can_one_predicate_be_invalidated_p): Add debugging printfs.
8828 (can_chain_union_be_invalidated_p): Improve check for invalidation
8830 (uninit_uses_cannot_happen): Avoid unnecessary if
8831 convert_control_dep_chain_into_preds yielded nothing.
8833 2018-01-06 Martin Sebor <msebor@redhat.com>
8835 PR tree-optimization/83640
8836 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
8837 subtracting negative offset from size.
8838 (builtin_access::overlap): Adjust offset bounds of the access to fall
8839 within the size of the object if possible.
8841 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
8843 PR rtl-optimization/83699
8844 * expmed.c (extract_bit_field_1): Restrict the vector usage of
8845 extract_bit_field_as_subreg to cases in which the extracted
8846 value is also a vector.
8848 * lra-constraints.c (process_alt_operands): Test for the equivalence
8849 substitutions when detecting a possible reload cycle.
8851 2018-01-06 Jakub Jelinek <jakub@redhat.com>
8854 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
8855 by default if flag_selective_schedling{,2}. Formatting fixes.
8857 PR rtl-optimization/83682
8858 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
8859 if it has non-VECTOR_MODE element mode.
8860 (vec_duplicate_p): Likewise.
8863 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
8864 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
8866 2018-01-05 Jakub Jelinek <jakub@redhat.com>
8869 * config/i386/i386-builtin.def
8870 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
8871 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
8872 Require also OPTION_MASK_ISA_AVX512F in addition to
8873 OPTION_MASK_ISA_GFNI.
8874 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
8875 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
8876 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
8877 to OPTION_MASK_ISA_GFNI.
8878 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
8879 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
8880 OPTION_MASK_ISA_AVX512BW.
8881 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
8882 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
8883 addition to OPTION_MASK_ISA_GFNI.
8884 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
8885 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
8886 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
8887 to OPTION_MASK_ISA_GFNI.
8888 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
8889 a requirement for all ISAs rather than any of them with a few
8891 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
8893 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
8894 bitmasks to be enabled with 3 exceptions, instead of requiring any
8895 enabled ISA with lots of exceptions.
8896 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
8897 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
8898 Change avx512bw in isa attribute to avx512f.
8899 * config/i386/sgxintrin.h: Add license boilerplate.
8900 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
8901 to __AVX512F__ and __AVX512VL to __AVX512VL__.
8902 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
8903 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
8905 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
8906 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
8907 temporarily sse2 rather than sse if not enabled already.
8910 * config/i386/sse.md (VI248_VLBW): Rename to ...
8911 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
8912 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
8913 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
8914 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
8915 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
8916 mode iterator instead of VI248_VLBW.
8918 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
8920 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
8921 (record_modified): Skip clobbers; add debug output.
8922 (param_change_prob): Use sreal frequencies.
8924 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8926 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
8927 punt for user-aligned variables.
8929 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8931 * tree-chrec.c (chrec_contains_symbols): Return true for
8934 2018-01-05 Sudakshina Das <sudi.das@arm.com>
8937 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
8938 of (x|y) == x for BICS pattern.
8940 2018-01-05 Jakub Jelinek <jakub@redhat.com>
8942 PR tree-optimization/83605
8943 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
8944 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
8947 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
8949 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
8950 * config/epiphany/rtems.h: New file.
8952 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8953 Uros Bizjak <ubizjak@gmail.com>
8956 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
8957 QIreg_operand instead of register_operand predicate.
8958 * config/i386/i386.c (ix86_rop_should_change_byte_p,
8959 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
8960 comments instead of -fmitigate[-_]rop.
8962 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8965 * cgraphunit.c (symbol_table::compile): Switch to text_section
8966 before calling assembly_start debug hook.
8967 * run-rtl-passes.c (run_rtl_passes): Likewise.
8970 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8972 * tree-vrp.c (extract_range_from_binary_expr_1): Check
8973 range_int_cst_p rather than !symbolic_range_p before calling
8974 extract_range_from_multiplicative_op_1.
8976 2018-01-04 Jeff Law <law@redhat.com>
8978 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
8979 redundant test in assertion.
8981 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8983 * doc/rtl.texi: Document machine_mode wrapper classes.
8985 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8987 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
8990 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8992 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
8993 the VEC_PERM_EXPR fold to fail.
8995 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8998 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
8999 to switched_sections.
9001 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9004 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
9007 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
9010 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
9011 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
9013 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9016 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
9017 is BLKmode and bitpos not zero or mode change is needed.
9019 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9022 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
9025 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
9028 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
9029 instead of MULT rtx. Update all corresponding splitters.
9031 (*ssub<modesuffix>): Ditto.
9033 (*cmp_sadd_di): Update split patterns.
9034 (*cmp_sadd_si): Ditto.
9035 (*cmp_sadd_sidi): Ditto.
9036 (*cmp_ssub_di): Ditto.
9037 (*cmp_ssub_si): Ditto.
9038 (*cmp_ssub_sidi): Ditto.
9039 * config/alpha/predicates.md (const23_operand): New predicate.
9040 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
9041 Look for ASHIFT, not MULT inner operand.
9042 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
9044 2018-01-04 Martin Liska <mliska@suse.cz>
9046 PR gcov-profile/83669
9047 * gcov.c (output_intermediate_file): Add version to intermediate
9049 * doc/gcov.texi: Document new field 'version' in intermediate
9050 file format. Fix location of '-k' option of gcov command.
9052 2018-01-04 Martin Liska <mliska@suse.cz>
9055 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
9057 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9059 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
9061 2018-01-03 Martin Sebor <msebor@redhat.com>
9063 PR tree-optimization/83655
9064 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
9065 checking calls with invalid arguments.
9067 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9069 * tree-vect-stmts.c (vect_get_store_rhs): New function.
9070 (vectorizable_mask_load_store): Delete.
9071 (vectorizable_call): Return false for masked loads and stores.
9072 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
9073 instead of gimple_assign_rhs1.
9074 (vectorizable_load): Handle IFN_MASK_LOAD.
9075 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
9077 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9079 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
9081 (vectorizable_mask_load_store): ...here.
9082 (vectorizable_load): ...and here.
9084 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9086 * tree-vect-stmts.c (vect_build_all_ones_mask)
9087 (vect_build_zero_merge_argument): New functions, split out from...
9088 (vectorizable_load): ...here.
9090 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9092 * tree-vect-stmts.c (vect_check_store_rhs): New function,
9094 (vectorizable_mask_load_store): ...here.
9095 (vectorizable_store): ...and here.
9097 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9099 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
9101 (vectorizable_mask_load_store): ...here.
9103 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9105 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
9106 (vect_model_store_cost): Take a vec_load_store_type instead of a
9108 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
9109 (vect_model_store_cost): Take a vec_load_store_type instead of a
9111 (vectorizable_mask_load_store): Update accordingly.
9112 (vectorizable_store): Likewise.
9113 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
9115 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9117 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
9118 IFN_MASK_LOAD calls here rather than...
9119 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
9121 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9122 Alan Hayward <alan.hayward@arm.com>
9123 David Sherwood <david.sherwood@arm.com>
9125 * expmed.c (extract_bit_field_1): For vector extracts,
9126 fall back to extract_bit_field_as_subreg if vec_extract
9129 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9130 Alan Hayward <alan.hayward@arm.com>
9131 David Sherwood <david.sherwood@arm.com>
9133 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
9134 they are variable or constant sized.
9135 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
9136 slots for constant-sized data.
9138 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9139 Alan Hayward <alan.hayward@arm.com>
9140 David Sherwood <david.sherwood@arm.com>
9142 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
9143 handling COND_EXPRs with boolean comparisons, try to find a better
9144 basis for the mask type than the boolean itself.
9146 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9148 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
9149 is calculated and how it can be overridden.
9150 * genmodes.c (max_bitsize_mode_any_mode): New variable.
9151 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
9153 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
9156 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9157 Alan Hayward <alan.hayward@arm.com>
9158 David Sherwood <david.sherwood@arm.com>
9160 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
9161 Remove the mode argument.
9162 (aarch64_simd_valid_immediate): Remove the mode and inverse
9164 * config/aarch64/iterators.md (bitsize): New iterator.
9165 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
9166 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
9167 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
9168 aarch64_simd_valid_immediate.
9169 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
9170 (aarch64_reg_or_bic_imm): Likewise.
9171 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
9172 with an insn_type enum and msl with a modifier_type enum.
9173 Replace element_width with a scalar_mode. Change the shift
9174 to unsigned int. Add constructors for scalar_float_mode and
9175 scalar_int_mode elements.
9176 (aarch64_vect_float_const_representable_p): Delete.
9177 (aarch64_can_const_movi_rtx_p)
9178 (aarch64_simd_scalar_immediate_valid_for_move)
9179 (aarch64_simd_make_constant): Update call to
9180 aarch64_simd_valid_immediate.
9181 (aarch64_advsimd_valid_immediate_hs): New function.
9182 (aarch64_advsimd_valid_immediate): Likewise.
9183 (aarch64_simd_valid_immediate): Remove mode and inverse
9184 arguments. Rewrite to use the above. Use const_vec_duplicate_p
9185 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
9186 and aarch64_float_const_representable_p on the result.
9187 (aarch64_output_simd_mov_immediate): Remove mode argument.
9188 Update call to aarch64_simd_valid_immediate and use of
9189 simd_immediate_info.
9190 (aarch64_output_scalar_simd_mov_immediate): Update call
9193 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9194 Alan Hayward <alan.hayward@arm.com>
9195 David Sherwood <david.sherwood@arm.com>
9197 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
9198 (mode_nunits): Likewise CONST_MODE_NUNITS.
9199 * machmode.def (ADJUST_NUNITS): Document.
9200 * genmodes.c (mode_data::need_nunits_adj): New field.
9201 (blank_mode): Update accordingly.
9202 (adj_nunits): New variable.
9203 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
9205 (emit_mode_size_inline): Set need_bytesize_adj for all modes
9206 listed in adj_nunits.
9207 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
9208 listed in adj_nunits. Don't emit case statements for such modes.
9209 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
9210 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
9211 nothing if adj_nunits is nonnull.
9212 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
9213 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
9214 (emit_mode_fbit): Update use of print_maybe_const_decl.
9215 (emit_move_size): Likewise. Treat the array as non-const
9217 (emit_mode_adjustments): Handle adj_nunits.
9219 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9221 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
9222 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
9223 (VECTOR_MODES): Use it.
9224 (make_vector_modes): Take the prefix as an argument.
9226 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9227 Alan Hayward <alan.hayward@arm.com>
9228 David Sherwood <david.sherwood@arm.com>
9230 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
9231 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
9232 for MODE_VECTOR_BOOL.
9233 * machmode.def (VECTOR_BOOL_MODE): Document.
9234 * genmodes.c (VECTOR_BOOL_MODE): New macro.
9235 (make_vector_bool_mode): New function.
9236 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
9238 * lto-streamer-in.c (lto_input_mode_table): Likewise.
9239 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
9241 * stor-layout.c (int_mode_for_mode): Likewise.
9242 * tree.c (build_vector_type_for_mode): Likewise.
9243 * varasm.c (output_constant_pool_2): Likewise.
9244 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
9245 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
9246 for MODE_VECTOR_BOOL.
9247 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
9248 of mode class checks.
9249 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
9250 instead of a list of mode class checks.
9251 (expand_vector_scalar_condition): Likewise.
9252 (type_for_widest_vector_mode): Handle BImode as an inner mode.
9254 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9255 Alan Hayward <alan.hayward@arm.com>
9256 David Sherwood <david.sherwood@arm.com>
9258 * machmode.h (mode_size): Change from unsigned short to
9260 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
9261 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9262 or if measurement_type is not polynomial.
9263 (fixed_size_mode::includes_p): Check for constant-sized modes.
9264 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
9265 return a poly_uint16 rather than an unsigned short.
9266 (emit_mode_size): Change the type of mode_size from unsigned short
9267 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
9268 (emit_mode_adjustments): Cope with polynomial vector sizes.
9269 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9271 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9273 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
9274 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
9275 * caller-save.c (setup_save_areas): Likewise.
9276 (replace_reg_with_saved_mem): Likewise.
9277 * calls.c (emit_library_call_value_1): Likewise.
9278 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
9279 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
9280 (gen_lowpart_for_combine): Likewise.
9281 * convert.c (convert_to_integer_1): Likewise.
9282 * cse.c (equiv_constant, cse_insn): Likewise.
9283 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
9284 (cselib_subst_to_values): Likewise.
9285 * dce.c (word_dce_process_block): Likewise.
9286 * df-problems.c (df_word_lr_mark_ref): Likewise.
9287 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
9288 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
9289 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
9290 (rtl_for_decl_location): Likewise.
9291 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
9292 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
9293 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
9294 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
9295 (expand_expr_real_1): Likewise.
9296 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
9297 (pad_below): Likewise.
9298 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9299 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
9300 * ira.c (get_subreg_tracking_sizes): Likewise.
9301 * ira-build.c (ira_create_allocno_objects): Likewise.
9302 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
9303 (ira_sort_regnos_for_alter_reg): Likewise.
9304 * ira-costs.c (record_operand_costs): Likewise.
9305 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
9306 (resolve_simple_move): Likewise.
9307 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
9308 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
9309 (lra_constraints): Likewise.
9310 (CONST_POOL_OK_P): Reject variable-sized modes.
9311 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
9312 (add_pseudo_to_slot, lra_spill): Likewise.
9313 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9314 * optabs-query.c (get_best_extraction_insn): Likewise.
9315 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9316 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
9317 (expand_mult_highpart, valid_multiword_target_p): Likewise.
9318 * recog.c (offsettable_address_addr_space_p): Likewise.
9319 * regcprop.c (maybe_mode_change): Likewise.
9320 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
9321 * regrename.c (build_def_use): Likewise.
9322 * regstat.c (dump_reg_info): Likewise.
9323 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
9324 (find_reloads, find_reloads_subreg_address): Likewise.
9325 * reload1.c (eliminate_regs_1): Likewise.
9326 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
9327 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
9328 (simplify_binary_operation_1, simplify_subreg): Likewise.
9329 * targhooks.c (default_function_arg_padding): Likewise.
9330 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
9331 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
9332 (verify_gimple_assign_ternary): Likewise.
9333 * tree-inline.c (estimate_move_cost): Likewise.
9334 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9335 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
9336 (get_address_cost_ainc): Likewise.
9337 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
9338 (vect_supportable_dr_alignment): Likewise.
9339 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9340 (vectorizable_reduction): Likewise.
9341 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
9342 (vectorizable_operation, vectorizable_load): Likewise.
9343 * tree.c (build_same_sized_truth_vector_type): Likewise.
9344 * valtrack.c (cleanup_auto_inc_dec): Likewise.
9345 * var-tracking.c (emit_note_insn_var_location): Likewise.
9346 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
9347 (ADDR_VEC_ALIGN): Likewise.
9349 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9350 Alan Hayward <alan.hayward@arm.com>
9351 David Sherwood <david.sherwood@arm.com>
9353 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
9355 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9356 or if measurement_type is polynomial.
9357 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
9358 * combine.c (make_extraction): Likewise.
9359 * dse.c (find_shift_sequence): Likewise.
9360 * dwarf2out.c (mem_loc_descriptor): Likewise.
9361 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
9362 (extract_bit_field, extract_low_bits): Likewise.
9363 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
9364 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
9365 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
9366 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
9367 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9368 * reload.c (find_reloads): Likewise.
9369 * reload1.c (alter_reg): Likewise.
9370 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
9371 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
9372 * tree-if-conv.c (predicate_mem_writes): Likewise.
9373 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
9374 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
9375 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
9376 * valtrack.c (dead_debug_insert_temp): Likewise.
9377 * varasm.c (mergeable_constant_section): Likewise.
9378 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
9380 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9381 Alan Hayward <alan.hayward@arm.com>
9382 David Sherwood <david.sherwood@arm.com>
9384 * expr.c (expand_assignment): Cope with polynomial mode sizes
9385 when assigning to a CONCAT.
9387 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9388 Alan Hayward <alan.hayward@arm.com>
9389 David Sherwood <david.sherwood@arm.com>
9391 * machmode.h (mode_precision): Change from unsigned short to
9393 (mode_to_precision): Return a poly_uint16 rather than an unsigned
9395 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
9396 or if measurement_type is not polynomial.
9397 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
9398 in which the mode is already known to be a scalar_int_mode.
9399 * genmodes.c (emit_mode_precision): Change the type of mode_precision
9400 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
9402 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9403 for GET_MODE_PRECISION.
9404 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9405 for GET_MODE_PRECISION.
9406 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
9408 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
9409 (expand_field_assignment, make_extraction): Likewise.
9410 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
9411 (get_last_value): Likewise.
9412 * convert.c (convert_to_integer_1): Likewise.
9413 * cse.c (cse_insn): Likewise.
9414 * expr.c (expand_expr_real_1): Likewise.
9415 * lra-constraints.c (simplify_operand_subreg): Likewise.
9416 * optabs-query.c (can_atomic_load_p): Likewise.
9417 * optabs.c (expand_atomic_load): Likewise.
9418 (expand_atomic_store): Likewise.
9419 * ree.c (combine_reaching_defs): Likewise.
9420 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
9421 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
9422 * tree.h (type_has_mode_precision_p): Likewise.
9423 * ubsan.c (instrument_si_overflow): Likewise.
9425 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9426 Alan Hayward <alan.hayward@arm.com>
9427 David Sherwood <david.sherwood@arm.com>
9429 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
9430 polynomial numbers of units.
9431 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
9432 (valid_vector_subparts_p): New function.
9433 (build_vector_type): Remove temporary shim and take the number
9434 of units as a poly_uint64 rather than an int.
9435 (build_opaque_vector_type): Take the number of units as a
9436 poly_uint64 rather than an int.
9437 * tree.c (build_vector_from_ctor): Handle polynomial
9438 TYPE_VECTOR_SUBPARTS.
9439 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
9440 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
9441 (build_vector_from_val): If the number of units is variable,
9442 use build_vec_duplicate_cst for constant operands and
9443 VEC_DUPLICATE_EXPR otherwise.
9444 (make_vector_type): Remove temporary is_constant ().
9445 (build_vector_type, build_opaque_vector_type): Take the number of
9446 units as a poly_uint64 rather than an int.
9447 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
9449 * cfgexpand.c (expand_debug_expr): Likewise.
9450 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
9451 (store_constructor, expand_expr_real_1): Likewise.
9452 (const_scalar_mask_from_tree): Likewise.
9453 * fold-const-call.c (fold_const_reduction): Likewise.
9454 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
9455 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
9456 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
9457 (fold_relational_const): Likewise.
9458 (native_interpret_vector): Likewise. Change the size from an
9459 int to an unsigned int.
9460 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
9461 TYPE_VECTOR_SUBPARTS.
9462 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
9463 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
9464 duplicating a non-constant operand into a variable-length vector.
9465 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
9466 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
9467 * ipa-icf.c (sem_variable::equals): Likewise.
9468 * match.pd: Likewise.
9469 * omp-simd-clone.c (simd_clone_subparts): Likewise.
9470 * print-tree.c (print_node): Likewise.
9471 * stor-layout.c (layout_type): Likewise.
9472 * targhooks.c (default_builtin_vectorization_cost): Likewise.
9473 * tree-cfg.c (verify_gimple_comparison): Likewise.
9474 (verify_gimple_assign_binary): Likewise.
9475 (verify_gimple_assign_ternary): Likewise.
9476 (verify_gimple_assign_single): Likewise.
9477 * tree-pretty-print.c (dump_generic_node): Likewise.
9478 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9479 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
9480 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
9481 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
9482 (vect_shift_permute_load_chain): Likewise.
9483 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
9484 (expand_vector_condition, optimize_vector_constructor): Likewise.
9485 (lower_vec_perm, get_compute_type): Likewise.
9486 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9487 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
9488 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
9489 (vect_recog_mask_conversion_pattern): Likewise.
9490 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
9491 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
9492 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9493 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
9494 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
9495 (vectorizable_shift, vectorizable_operation, vectorizable_store)
9496 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
9497 (supportable_widening_operation): Likewise.
9498 (supportable_narrowing_operation): Likewise.
9499 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
9501 * varasm.c (output_constant): Likewise.
9503 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9504 Alan Hayward <alan.hayward@arm.com>
9505 David Sherwood <david.sherwood@arm.com>
9507 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
9508 so that both the length == 3 and length != 3 cases set up their
9509 own permute vectors. Add comments explaining why we know the
9510 number of elements is constant.
9511 (vect_permute_load_chain): Likewise.
9513 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9514 Alan Hayward <alan.hayward@arm.com>
9515 David Sherwood <david.sherwood@arm.com>
9517 * machmode.h (mode_nunits): Change from unsigned char to
9519 (ONLY_FIXED_SIZE_MODES): New macro.
9520 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
9521 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
9522 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
9524 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
9525 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
9526 or if measurement_type is not polynomial.
9527 * genmodes.c (ZERO_COEFFS): New macro.
9528 (emit_mode_nunits_inline): Make mode_nunits_inline return a
9530 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
9531 Use ZERO_COEFFS when emitting initializers.
9532 * data-streamer.h (bp_pack_poly_value): New function.
9533 (bp_unpack_poly_value): Likewise.
9534 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9535 for GET_MODE_NUNITS.
9536 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9537 for GET_MODE_NUNITS.
9538 * tree.c (make_vector_type): Remove temporary shim and make
9539 the real function take the number of units as a poly_uint64
9541 (build_vector_type_for_mode): Handle polynomial nunits.
9542 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
9543 * emit-rtl.c (const_vec_series_p_1): Likewise.
9544 (gen_rtx_CONST_VECTOR): Likewise.
9545 * fold-const.c (test_vec_duplicate_folding): Likewise.
9546 * genrecog.c (validate_pattern): Likewise.
9547 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
9548 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9549 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
9550 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
9551 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
9552 * rtlanal.c (subreg_get_info): Likewise.
9553 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9554 (vect_grouped_load_supported): Likewise.
9555 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
9556 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
9557 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
9558 (simplify_const_unary_operation, simplify_binary_operation_1)
9559 (simplify_const_binary_operation, simplify_ternary_operation)
9560 (test_vector_ops_duplicate, test_vector_ops): Likewise.
9561 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
9562 instead of CONST_VECTOR_NUNITS.
9563 * varasm.c (output_constant_pool_2): Likewise.
9564 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
9565 explicit-encoded elements in the XVEC for variable-length vectors.
9567 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9569 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
9571 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9572 Alan Hayward <alan.hayward@arm.com>
9573 David Sherwood <david.sherwood@arm.com>
9575 * coretypes.h (fixed_size_mode): Declare.
9576 (fixed_size_mode_pod): New typedef.
9577 * builtins.h (target_builtins::x_apply_args_mode)
9578 (target_builtins::x_apply_result_mode): Change type to
9579 fixed_size_mode_pod.
9580 * builtins.c (apply_args_size, apply_result_size, result_vector)
9581 (expand_builtin_apply_args_1, expand_builtin_apply)
9582 (expand_builtin_return): Update accordingly.
9584 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9586 * cse.c (hash_rtx_cb): Hash only the encoded elements.
9587 * cselib.c (cselib_hash_rtx): Likewise.
9588 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
9589 CONST_VECTOR encoding.
9591 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9592 Jeff Law <law@redhat.com>
9595 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
9596 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
9597 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
9598 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
9601 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
9602 explicitly probe *sp in a noreturn function if there were any callee
9603 register saves or frame pointer is needed.
9605 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9608 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
9609 BLKmode for ternary, binary or unary expressions.
9612 * var-tracking.c (delete_vta_debug_insn): New inline function.
9613 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
9614 insns from get_insns () to NULL instead of each bb separately.
9615 Use delete_vta_debug_insn. No longer static.
9616 (vt_debug_insns_local, variable_tracking_main_1): Adjust
9617 delete_vta_debug_insns callers.
9618 * rtl.h (delete_vta_debug_insns): Declare.
9619 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
9620 instead of variable_tracking_main.
9622 2018-01-03 Martin Sebor <msebor@redhat.com>
9624 PR tree-optimization/83603
9625 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
9626 arguments past the endof the argument list in functions declared
9627 without a prototype.
9628 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
9629 Avoid checking when arguments are null.
9631 2018-01-03 Martin Sebor <msebor@redhat.com>
9634 * doc/extend.texi (attribute const): Fix a typo.
9635 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
9636 issuing -Wsuggest-attribute for void functions.
9638 2018-01-03 Martin Sebor <msebor@redhat.com>
9640 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
9641 offset_int::from instead of wide_int::to_shwi.
9642 (maybe_diag_overlap): Remove assertion.
9643 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
9644 * gimple-ssa-sprintf.c (format_directive): Same.
9645 (parse_directive): Same.
9646 (sprintf_dom_walker::compute_format_length): Same.
9647 (try_substitute_return_value): Same.
9649 2018-01-03 Jeff Law <law@redhat.com>
9652 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
9653 non-constant residual for zero at runtime and avoid probing in
9654 that case. Reorganize code for trailing problem to mirror handling
9657 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9659 PR tree-optimization/83501
9660 * tree-ssa-strlen.c (get_string_cst): New.
9661 (handle_char_store): Call get_string_cst.
9663 2018-01-03 Martin Liska <mliska@suse.cz>
9665 PR tree-optimization/83593
9666 * tree-ssa-strlen.c: Include tree-cfg.h.
9667 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
9668 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
9669 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
9671 (strlen_dom_walker::before_dom_children): Call
9672 gimple_purge_dead_eh_edges. Dump tranformation with details
9674 (strlen_dom_walker::before_dom_children): Update call by adding
9675 new argument cleanup_eh.
9676 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
9678 2018-01-03 Martin Liska <mliska@suse.cz>
9681 * cif-code.def (VARIADIC_THUNK): New enum value.
9682 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
9685 2018-01-03 Jan Beulich <jbeulich@suse.com>
9687 * sse.md (mov<mode>_internal): Tighten condition for when to use
9688 vmovdqu<ssescalarsize> for TI and OI modes.
9690 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9692 Update copyright years.
9694 2018-01-03 Martin Liska <mliska@suse.cz>
9697 * ipa-visibility.c (function_and_variable_visibility): Skip
9698 functions with noipa attribure.
9700 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9702 * gcc.c (process_command): Update copyright notice dates.
9703 * gcov-dump.c (print_version): Ditto.
9704 * gcov.c (print_version): Ditto.
9705 * gcov-tool.c (print_version): Ditto.
9706 * gengtype.c (create_file): Ditto.
9707 * doc/cpp.texi: Bump @copying's copyright year.
9708 * doc/cppinternals.texi: Ditto.
9709 * doc/gcc.texi: Ditto.
9710 * doc/gccint.texi: Ditto.
9711 * doc/gcov.texi: Ditto.
9712 * doc/install.texi: Ditto.
9713 * doc/invoke.texi: Ditto.
9715 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9717 * vector-builder.h (vector_builder::m_full_nelts): Change from
9718 unsigned int to poly_uint64.
9719 (vector_builder::full_nelts): Update prototype accordingly.
9720 (vector_builder::new_vector): Likewise.
9721 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
9722 (vector_builder::operator ==): Likewise.
9723 (vector_builder::finalize): Likewise.
9724 * int-vector-builder.h (int_vector_builder::int_vector_builder):
9725 Take the number of elements as a poly_uint64 rather than an
9727 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
9728 from unsigned int to poly_uint64.
9729 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
9730 (vec_perm_indices::new_vector): Likewise.
9731 (vec_perm_indices::length): Likewise.
9732 (vec_perm_indices::nelts_per_input): Likewise.
9733 (vec_perm_indices::input_nelts): Likewise.
9734 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
9735 number of elements per input as a poly_uint64 rather than an
9736 unsigned int. Use the original encoding for variable-length
9737 vectors, rather than clamping each individual element.
9738 For the second and subsequent elements in each pattern,
9739 clamp the step and base before clamping their sum.
9740 (vec_perm_indices::series_p): Handle polynomial element counts.
9741 (vec_perm_indices::all_in_range_p): Likewise.
9742 (vec_perm_indices_to_tree): Likewise.
9743 (vec_perm_indices_to_rtx): Likewise.
9744 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
9745 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
9746 (tree_vector_builder::new_binary_operation): Handle polynomial
9747 element counts. Return false if we need to know the number
9748 of elements at compile time.
9749 * fold-const.c (fold_vec_perm): Punt if the number of elements
9750 isn't known at compile time.
9752 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9754 * vec-perm-indices.h (vec_perm_builder): Change element type
9755 from HOST_WIDE_INT to poly_int64.
9756 (vec_perm_indices::element_type): Update accordingly.
9757 (vec_perm_indices::clamp): Handle polynomial element_types.
9758 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9759 (vec_perm_indices::all_in_range_p): Likewise.
9760 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
9762 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
9763 polynomial vec_perm_indices element types.
9764 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
9765 * fold-const.c (fold_vec_perm): Likewise.
9766 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
9767 * tree-vect-generic.c (lower_vec_perm): Likewise.
9768 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9769 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
9770 element type to HOST_WIDE_INT.
9772 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9773 Alan Hayward <alan.hayward@arm.com>
9774 David Sherwood <david.sherwood@arm.com>
9776 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
9777 rather than an int. Use plus_constant.
9778 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
9779 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
9781 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9782 Alan Hayward <alan.hayward@arm.com>
9783 David Sherwood <david.sherwood@arm.com>
9785 * calls.c (emit_call_1, expand_call): Change struct_value_size from
9786 a HOST_WIDE_INT to a poly_int64.
9788 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9789 Alan Hayward <alan.hayward@arm.com>
9790 David Sherwood <david.sherwood@arm.com>
9792 * calls.c (load_register_parameters): Cope with polynomial
9793 mode sizes. Require a constant size for BLKmode parameters
9794 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
9795 forces a parameter to be padded at the lsb end in order to
9796 fill a complete number of words, require the parameter size
9797 to be ordered wrt UNITS_PER_WORD.
9799 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9800 Alan Hayward <alan.hayward@arm.com>
9801 David Sherwood <david.sherwood@arm.com>
9803 * reload1.c (spill_stack_slot_width): Change element type
9804 from unsigned int to poly_uint64_pod.
9805 (alter_reg): Treat mode sizes as polynomial.
9807 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9808 Alan Hayward <alan.hayward@arm.com>
9809 David Sherwood <david.sherwood@arm.com>
9811 * reload.c (complex_word_subreg_p): New function.
9812 (reload_inner_reg_of_subreg, push_reload): Use it.
9814 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9815 Alan Hayward <alan.hayward@arm.com>
9816 David Sherwood <david.sherwood@arm.com>
9818 * lra-constraints.c (process_alt_operands): Reject matched
9819 operands whose sizes aren't ordered.
9820 (match_reload): Refer to this check here.
9822 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9823 Alan Hayward <alan.hayward@arm.com>
9824 David Sherwood <david.sherwood@arm.com>
9826 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
9827 that the mode size is in the set {1, 2, 4, 8, 16}.
9829 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9830 Alan Hayward <alan.hayward@arm.com>
9831 David Sherwood <david.sherwood@arm.com>
9833 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
9834 Use plus_constant instead of gen_rtx_PLUS.
9836 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9837 Alan Hayward <alan.hayward@arm.com>
9838 David Sherwood <david.sherwood@arm.com>
9840 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
9841 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
9842 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
9843 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
9844 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
9845 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
9846 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
9847 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
9848 * config/i386/i386.c (ix86_push_rounding): ...this new function.
9849 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
9851 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
9852 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
9853 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
9854 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
9855 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
9856 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
9857 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
9858 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
9859 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
9860 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
9862 * expr.c (emit_move_resolve_push): Treat the input and result
9863 of PUSH_ROUNDING as a poly_int64.
9864 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
9865 (emit_push_insn): Likewise.
9866 * lra-eliminations.c (mark_not_eliminable): Likewise.
9867 * recog.c (push_operand): Likewise.
9868 * reload1.c (elimination_effects): Likewise.
9869 * rtlanal.c (nonzero_bits1): Likewise.
9870 * calls.c (store_one_arg): Likewise. Require the padding to be
9871 known at compile time.
9873 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9874 Alan Hayward <alan.hayward@arm.com>
9875 David Sherwood <david.sherwood@arm.com>
9877 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
9878 Use plus_constant instead of gen_rtx_PLUS.
9880 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9881 Alan Hayward <alan.hayward@arm.com>
9882 David Sherwood <david.sherwood@arm.com>
9884 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
9887 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9888 Alan Hayward <alan.hayward@arm.com>
9889 David Sherwood <david.sherwood@arm.com>
9891 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
9892 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
9893 via stack temporaries. Treat the mode size as polynomial too.
9895 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9896 Alan Hayward <alan.hayward@arm.com>
9897 David Sherwood <david.sherwood@arm.com>
9899 * expr.c (expand_expr_real_2): When handling conversions involving
9900 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
9901 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
9902 as a poly_uint64 too.
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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
9910 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9911 Alan Hayward <alan.hayward@arm.com>
9912 David Sherwood <david.sherwood@arm.com>
9914 * combine.c (can_change_dest_mode): Handle polynomial
9915 REGMODE_NATURAL_SIZE.
9916 * expmed.c (store_bit_field_1): Likewise.
9917 * expr.c (store_constructor): Likewise.
9918 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
9919 and polynomial REGMODE_NATURAL_SIZE.
9920 (gen_lowpart_common): Likewise.
9921 * reginfo.c (record_subregs_of_mode): Likewise.
9922 * rtlanal.c (read_modify_subreg_p): Likewise.
9924 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9925 Alan Hayward <alan.hayward@arm.com>
9926 David Sherwood <david.sherwood@arm.com>
9928 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
9929 numbers of elements.
9931 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9932 Alan Hayward <alan.hayward@arm.com>
9933 David Sherwood <david.sherwood@arm.com>
9935 * match.pd: Cope with polynomial numbers of vector elements.
9937 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9938 Alan Hayward <alan.hayward@arm.com>
9939 David Sherwood <david.sherwood@arm.com>
9941 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
9942 in a POINTER_PLUS_EXPR.
9944 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9945 Alan Hayward <alan.hayward@arm.com>
9946 David Sherwood <david.sherwood@arm.com>
9948 * omp-simd-clone.c (simd_clone_subparts): New function.
9949 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
9950 (ipa_simd_modify_function_body): Likewise.
9952 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9953 Alan Hayward <alan.hayward@arm.com>
9954 David Sherwood <david.sherwood@arm.com>
9956 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
9957 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
9958 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
9959 (expand_vector_condition, vector_element): Likewise.
9960 (subparts_gt): New function.
9961 (get_compute_type): Use subparts_gt.
9962 (count_type_subparts): Delete.
9963 (expand_vector_operations_1): Use subparts_gt instead of
9964 count_type_subparts.
9966 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9967 Alan Hayward <alan.hayward@arm.com>
9968 David Sherwood <david.sherwood@arm.com>
9970 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
9971 (vect_compile_time_alias): ...this new function. Do the calculation
9972 on poly_ints rather than trees.
9973 (vect_prune_runtime_alias_test_list): Update call accordingly.
9975 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9976 Alan Hayward <alan.hayward@arm.com>
9977 David Sherwood <david.sherwood@arm.com>
9979 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
9981 (vect_schedule_slp_instance): Likewise.
9983 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9984 Alan Hayward <alan.hayward@arm.com>
9985 David Sherwood <david.sherwood@arm.com>
9987 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
9988 constant and extern definitions for variable-length vectors.
9989 (vect_get_constant_vectors): Note that the number of units
9990 is known to be constant.
9992 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9993 Alan Hayward <alan.hayward@arm.com>
9994 David Sherwood <david.sherwood@arm.com>
9996 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
9997 of units as polynomial. Choose between WIDE and NARROW based
10000 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10001 Alan Hayward <alan.hayward@arm.com>
10002 David Sherwood <david.sherwood@arm.com>
10004 * tree-vect-stmts.c (simd_clone_subparts): New function.
10005 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
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-stmts.c (vectorizable_call): Treat the number of
10012 vectors as polynomial. Use build_index_vector for
10013 IFN_GOMP_SIMD_LANE.
10015 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10016 Alan Hayward <alan.hayward@arm.com>
10017 David Sherwood <david.sherwood@arm.com>
10019 * tree-vect-stmts.c (get_load_store_type): Treat the number of
10020 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
10021 for variable-length vectors.
10022 (vectorizable_mask_load_store): Treat the number of units as
10023 polynomial, asserting that it is constant if the condition has
10024 already been enforced.
10025 (vectorizable_store, vectorizable_load): Likewise.
10027 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10028 Alan Hayward <alan.hayward@arm.com>
10029 David Sherwood <david.sherwood@arm.com>
10031 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
10032 of units as polynomial. Punt if we can't tell at compile time
10033 which vector contains the final result.
10035 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10036 Alan Hayward <alan.hayward@arm.com>
10037 David Sherwood <david.sherwood@arm.com>
10039 * tree-vect-loop.c (vectorizable_induction): Treat the number
10040 of units as polynomial. Punt on SLP inductions. Use an integer
10041 VEC_SERIES_EXPR for variable-length integer reductions. Use a
10042 cast of such a series for variable-length floating-point
10045 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10046 Alan Hayward <alan.hayward@arm.com>
10047 David Sherwood <david.sherwood@arm.com>
10049 * tree.h (build_index_vector): Declare.
10050 * tree.c (build_index_vector): New function.
10051 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
10052 of units as polynomial, forcibly converting it to a constant if
10053 vectorizable_reduction has already enforced the condition.
10054 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
10055 to create a {1,2,3,...} vector.
10056 (vectorizable_reduction): Treat the number of units as polynomial.
10057 Choose vectype_in based on the largest scalar element size rather
10058 than the smallest number of units. Enforce the restrictions
10061 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10062 Alan Hayward <alan.hayward@arm.com>
10063 David Sherwood <david.sherwood@arm.com>
10065 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
10066 number of units as polynomial.
10068 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10069 Alan Hayward <alan.hayward@arm.com>
10070 David Sherwood <david.sherwood@arm.com>
10072 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
10073 * target.def (autovectorize_vector_sizes): Return the vector sizes
10074 by pointer, using vector_sizes rather than a bitmask.
10075 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
10076 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
10077 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
10079 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
10080 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
10081 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
10082 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
10083 * omp-general.c (omp_max_vf): Likewise.
10084 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10085 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
10086 * tree-vect-loop.c (vect_analyze_loop): Likewise.
10087 * tree-vect-slp.c (vect_slp_bb): Likewise.
10088 * doc/tm.texi: Regenerate.
10089 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
10091 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
10092 the vector size as a poly_uint64 rather than an unsigned int.
10093 (current_vector_size): Change from an unsigned int to a poly_uint64.
10094 (get_vectype_for_scalar_type): Update accordingly.
10095 * tree.h (build_truth_vector_type): Take the size and number of
10096 units as a poly_uint64 rather than an unsigned int.
10097 (build_vector_type): Add a temporary overload that takes
10098 the number of units as a poly_uint64 rather than an unsigned int.
10099 * tree.c (make_vector_type): Likewise.
10100 (build_truth_vector_type): Take the number of units as a poly_uint64
10101 rather than an unsigned int.
10103 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10104 Alan Hayward <alan.hayward@arm.com>
10105 David Sherwood <david.sherwood@arm.com>
10107 * target.def (get_mask_mode): Take the number of units and length
10108 as poly_uint64s rather than unsigned ints.
10109 * targhooks.h (default_get_mask_mode): Update accordingly.
10110 * targhooks.c (default_get_mask_mode): Likewise.
10111 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
10112 * doc/tm.texi: Regenerate.
10114 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10115 Alan Hayward <alan.hayward@arm.com>
10116 David Sherwood <david.sherwood@arm.com>
10118 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
10119 * omp-general.c (omp_max_vf): Likewise.
10120 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
10121 (expand_omp_simd): Handle polynomial safelen.
10122 * omp-low.c (omplow_simd_context): Add a default constructor.
10123 (omplow_simd_context::max_vf): Change from int to poly_uint64.
10124 (lower_rec_simd_input_clauses): Update accordingly.
10125 (lower_rec_input_clauses): Likewise.
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 * tree-vectorizer.h (vect_nunits_for_cost): New function.
10132 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
10133 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
10134 (vect_analyze_slp_cost): Likewise.
10135 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
10136 (vect_model_load_cost): Likewise.
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 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
10143 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
10144 from an unsigned int * to a poly_uint64_pod *.
10145 (calculate_unrolling_factor): New function.
10146 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
10148 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10149 Alan Hayward <alan.hayward@arm.com>
10150 David Sherwood <david.sherwood@arm.com>
10152 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
10153 from an unsigned int to a poly_uint64.
10154 (_loop_vec_info::slp_unrolling_factor): Likewise.
10155 (_loop_vec_info::vectorization_factor): Change from an int
10157 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
10158 (vect_get_num_vectors): New function.
10159 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
10160 (vect_get_num_copies): Use vect_get_num_vectors.
10161 (vect_analyze_data_ref_dependences): Change max_vf from an int *
10162 to an unsigned int *.
10163 (vect_analyze_data_refs): Change min_vf from an int * to a
10165 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10166 than an unsigned HOST_WIDE_INT.
10167 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
10168 (vect_analyze_data_ref_dependence): Change max_vf from an int *
10169 to an unsigned int *.
10170 (vect_analyze_data_ref_dependences): Likewise.
10171 (vect_compute_data_ref_alignment): Handle polynomial vf.
10172 (vect_enhance_data_refs_alignment): Likewise.
10173 (vect_prune_runtime_alias_test_list): Likewise.
10174 (vect_shift_permute_load_chain): Likewise.
10175 (vect_supportable_dr_alignment): Likewise.
10176 (dependence_distance_ge_vf): Take the vectorization factor as a
10177 poly_uint64 rather than an unsigned HOST_WIDE_INT.
10178 (vect_analyze_data_refs): Change min_vf from an int * to a
10180 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
10181 vfm1 as a poly_uint64 rather than an int. Make the same change
10182 for the returned bound_scalar.
10183 (vect_gen_vector_loop_niters): Handle polynomial vf.
10184 (vect_do_peeling): Likewise. Update call to
10185 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
10186 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
10188 * tree-vect-loop.c (vect_determine_vectorization_factor)
10189 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
10190 (vect_get_known_peeling_cost): Likewise.
10191 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
10192 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
10193 (vect_transform_loop): Likewise. Use the lowest possible VF when
10194 updating the upper bounds of the loop.
10195 (vect_min_worthwhile_factor): Make static. Return an unsigned int
10196 rather than an int.
10197 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
10198 polynomial unroll factors.
10199 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
10200 (vect_make_slp_decision): Likewise.
10201 (vect_supported_load_permutation_p): Likewise, and polynomial
10203 (vect_analyze_slp_cost): Handle polynomial vf.
10204 (vect_slp_analyze_node_operations): Likewise.
10205 (vect_slp_analyze_bb_1): Likewise.
10206 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10207 than an unsigned HOST_WIDE_INT.
10208 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
10209 (vectorizable_load): Handle polynomial vf.
10210 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
10212 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
10214 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10215 Alan Hayward <alan.hayward@arm.com>
10216 David Sherwood <david.sherwood@arm.com>
10218 * match.pd: Handle bit operations involving three constants
10219 and try to fold one pair.
10221 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10223 * tree-vect-loop-manip.c: Include gimple-fold.h.
10224 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
10225 niters_maybe_zero parameters. Handle other cases besides a step of 1.
10226 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
10227 Add a path that uses a step of VF instead of 1, but disable it
10229 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
10230 and niters_no_overflow parameters. Update calls to
10231 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
10232 Create a new SSA name if the latter choses to use a ste other
10233 than zero, and return it via niters_vector_mult_vf_var.
10234 * tree-vect-loop.c (vect_transform_loop): Update calls to
10235 vect_do_peeling, vect_gen_vector_loop_niters and
10236 slpeel_make_loop_iterate_ntimes.
10237 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
10238 (vect_gen_vector_loop_niters): Update declarations after above changes.
10240 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
10242 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
10243 128-bit round to integer instructions.
10244 (ceil<mode>2): Likewise.
10245 (btrunc<mode>2): Likewise.
10246 (round<mode>2): Likewise.
10248 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10250 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
10251 unaligned VSX load/store on P8/P9.
10252 (expand_block_clear): Allow the use of unaligned VSX
10253 load/store on P8/P9.
10255 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10257 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
10259 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
10260 swap associated with both a load and a store.
10262 2018-01-02 Andrew Waterman <andrew@sifive.com>
10264 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
10265 * config/riscv/riscv.md (clear_cache): Use it.
10267 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
10269 * web.c: Remove out-of-date comment.
10271 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10273 * expr.c (fixup_args_size_notes): Check that any existing
10274 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
10275 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
10276 (emit_single_push_insn): ...here.
10278 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10280 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
10281 (const_vector_encoded_nelts): New function.
10282 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
10283 (const_vector_int_elt, const_vector_elt): Declare.
10284 * emit-rtl.c (const_vector_int_elt_1): New function.
10285 (const_vector_elt): Likewise.
10286 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
10287 of CONST_VECTOR_ELT.
10289 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10291 * expr.c: Include rtx-vector-builder.h.
10292 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
10293 directly on the tree encoding.
10294 (const_vector_from_tree): Likewise.
10295 * optabs.c: Include rtx-vector-builder.h.
10296 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
10297 sequence of "u" values.
10298 * vec-perm-indices.c: Include rtx-vector-builder.h.
10299 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
10300 directly on the vec_perm_indices encoding.
10302 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10304 * doc/rtl.texi (const_vector): Describe new encoding scheme.
10305 * Makefile.in (OBJS): Add rtx-vector-builder.o.
10306 * rtx-vector-builder.h: New file.
10307 * rtx-vector-builder.c: Likewise.
10308 * rtl.h (rtx_def::u2): Add a const_vector field.
10309 (CONST_VECTOR_NPATTERNS): New macro.
10310 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
10311 (CONST_VECTOR_DUPLICATE_P): Likewise.
10312 (CONST_VECTOR_STEPPED_P): Likewise.
10313 (CONST_VECTOR_ENCODED_ELT): Likewise.
10314 (const_vec_duplicate_p): Check for a duplicated vector encoding.
10315 (unwrap_const_vec_duplicate): Likewise.
10316 (const_vec_series_p): Check for a non-duplicated vector encoding.
10317 Say that the function only returns true for integer vectors.
10318 * emit-rtl.c: Include rtx-vector-builder.h.
10319 (gen_const_vec_duplicate_1): Delete.
10320 (gen_const_vector): Call gen_const_vec_duplicate instead of
10321 gen_const_vec_duplicate_1.
10322 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
10323 (gen_const_vec_duplicate): Use rtx_vector_builder.
10324 (gen_const_vec_series): Likewise.
10325 (gen_rtx_CONST_VECTOR): Likewise.
10326 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
10327 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10328 Build a new vector rather than modifying a CONST_VECTOR in-place.
10329 (handle_special_swappables): Update call accordingly.
10330 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
10331 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10332 Build a new vector rather than modifying a CONST_VECTOR in-place.
10333 (handle_special_swappables): Update call accordingly.
10335 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10337 * simplify-rtx.c (simplify_const_binary_operation): Use
10338 CONST_VECTOR_ELT instead of XVECEXP.
10340 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10342 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
10343 the selector elements to be different from the data elements
10344 if the selector is a VECTOR_CST.
10345 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
10346 ssizetype for the selector.
10348 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10350 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
10351 before testing each element individually.
10352 * tree-vect-generic.c (lower_vec_perm): Likewise.
10354 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10356 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
10357 * selftest-run-tests.c (selftest::run_tests): Call it.
10358 * vector-builder.h (vector_builder::operator ==): New function.
10359 (vector_builder::operator !=): Likewise.
10360 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
10361 (vec_perm_indices::all_from_input_p): New function.
10362 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10363 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
10364 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
10365 instead of reading the VECTOR_CST directly. Detect whether both
10366 vector inputs are the same before constructing the vec_perm_indices,
10367 and update the number of inputs argument accordingly. Use the
10368 utility functions added above. Only construct sel2 if we need to.
10370 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10372 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
10373 the broadcast of the low byte.
10374 (expand_mult_highpart): Use an explicit encoding for the permutes.
10375 * optabs-query.c (can_mult_highpart_p): Likewise.
10376 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
10377 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10378 (vectorizable_bswap): Likewise.
10379 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
10380 explicit encoding for the power-of-2 permutes.
10381 (vect_permute_store_chain): Likewise.
10382 (vect_grouped_load_supported): Likewise.
10383 (vect_permute_load_chain): Likewise.
10385 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10387 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
10388 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
10389 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
10390 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10391 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
10392 (vect_gen_perm_mask_any): Likewise.
10394 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10396 * int-vector-builder.h: New file.
10397 * vec-perm-indices.h: Include int-vector-builder.h.
10398 (vec_perm_indices): Redefine as an int_vector_builder.
10399 (auto_vec_perm_indices): Delete.
10400 (vec_perm_builder): Redefine as a stand-alone class.
10401 (vec_perm_indices::vec_perm_indices): New function.
10402 (vec_perm_indices::clamp): Likewise.
10403 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
10404 (vec_perm_indices::new_vector): New function.
10405 (vec_perm_indices::new_expanded_vector): Update for new
10406 vec_perm_indices class.
10407 (vec_perm_indices::rotate_inputs): New function.
10408 (vec_perm_indices::all_in_range_p): Operate directly on the
10409 encoded form, without computing elided elements.
10410 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
10411 encoding. Update for new vec_perm_indices class.
10412 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
10413 the given vec_perm_builder.
10414 (expand_vec_perm_var): Update vec_perm_builder constructor.
10415 (expand_mult_highpart): Use vec_perm_builder instead of
10416 auto_vec_perm_indices.
10417 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
10418 vec_perm_indices instead of auto_vec_perm_indices. Use a single
10419 or double series encoding as appropriate.
10420 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
10421 vec_perm_indices instead of auto_vec_perm_indices.
10422 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10423 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10424 (vect_permute_store_chain): Likewise.
10425 (vect_grouped_load_supported): Likewise.
10426 (vect_permute_load_chain): Likewise.
10427 (vect_shift_permute_load_chain): Likewise.
10428 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10429 (vect_transform_slp_perm_load): Likewise.
10430 (vect_schedule_slp_instance): Likewise.
10431 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10432 (vectorizable_mask_load_store): Likewise.
10433 (vectorizable_bswap): Likewise.
10434 (vectorizable_store): Likewise.
10435 (vectorizable_load): Likewise.
10436 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
10437 vec_perm_indices instead of auto_vec_perm_indices. Use
10438 tree_to_vec_perm_builder to read the vector from a tree.
10439 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
10440 vec_perm_builder instead of a vec_perm_indices.
10441 (have_whole_vector_shift): Use vec_perm_builder and
10442 vec_perm_indices instead of auto_vec_perm_indices. Leave the
10443 truncation to calc_vec_perm_mask_for_shift.
10444 (vect_create_epilog_for_reduction): Likewise.
10445 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
10446 from auto_vec_perm_indices to vec_perm_indices.
10447 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10448 instead of changing individual elements.
10449 (aarch64_vectorize_vec_perm_const): Use new_vector to install
10450 the vector in d.perm.
10451 * config/arm/arm.c (expand_vec_perm_d::perm): Change
10452 from auto_vec_perm_indices to vec_perm_indices.
10453 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10454 instead of changing individual elements.
10455 (arm_vectorize_vec_perm_const): Use new_vector to install
10456 the vector in d.perm.
10457 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
10458 Update vec_perm_builder constructor.
10459 (rs6000_expand_interleave): Likewise.
10460 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
10461 (rs6000_expand_interleave): Likewise.
10463 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10465 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
10466 to qimode could truncate the indices.
10467 * optabs.c (expand_vec_perm_var): Likewise.
10469 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10471 * Makefile.in (OBJS): Add vec-perm-indices.o.
10472 * vec-perm-indices.h: New file.
10473 * vec-perm-indices.c: Likewise.
10474 * target.h (vec_perm_indices): Replace with a forward class
10476 (auto_vec_perm_indices): Move to vec-perm-indices.h.
10477 * optabs.h: Include vec-perm-indices.h.
10478 (expand_vec_perm): Delete.
10479 (selector_fits_mode_p, expand_vec_perm_var): Declare.
10480 (expand_vec_perm_const): Declare.
10481 * target.def (vec_perm_const_ok): Replace with...
10482 (vec_perm_const): ...this new hook.
10483 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
10484 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
10485 * doc/tm.texi: Regenerate.
10486 * optabs.def (vec_perm_const): Delete.
10487 * doc/md.texi (vec_perm_const): Likewise.
10488 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
10489 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
10490 expand_vec_perm for constant permutation vectors. Assert that
10491 the mode of variable permutation vectors is the integer equivalent
10492 of the mode that is being permuted.
10493 * optabs-query.h (selector_fits_mode_p): Declare.
10494 * optabs-query.c: Include vec-perm-indices.h.
10495 (selector_fits_mode_p): New function.
10496 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
10497 is defined, instead of checking whether the vec_perm_const_optab
10498 exists. Use targetm.vectorize.vec_perm_const instead of
10499 targetm.vectorize.vec_perm_const_ok. Check whether the indices
10500 fit in the vector mode before using a variable permute.
10501 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
10502 vec_perm_indices instead of an rtx.
10503 (expand_vec_perm): Replace with...
10504 (expand_vec_perm_const): ...this new function. Take the selector
10505 as a vec_perm_indices rather than an rtx. Also take the mode of
10506 the selector. Update call to shift_amt_for_vec_perm_mask.
10507 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
10508 Use vec_perm_indices::new_expanded_vector to expand the original
10509 selector into bytes. Check whether the indices fit in the vector
10510 mode before using a variable permute.
10511 (expand_vec_perm_var): Make global.
10512 (expand_mult_highpart): Use expand_vec_perm_const.
10513 * fold-const.c: Includes vec-perm-indices.h.
10514 * tree-ssa-forwprop.c: Likewise.
10515 * tree-vect-data-refs.c: Likewise.
10516 * tree-vect-generic.c: Likewise.
10517 * tree-vect-loop.c: Likewise.
10518 * tree-vect-slp.c: Likewise.
10519 * tree-vect-stmts.c: Likewise.
10520 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
10522 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
10523 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
10524 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
10525 (aarch64_vectorize_vec_perm_const): ...this new function.
10526 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10527 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10528 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
10529 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
10530 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10531 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10532 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
10534 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
10535 check for NEON modes.
10536 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
10537 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
10538 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
10539 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
10541 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
10542 the old VEC_PERM_CONST conditions.
10543 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
10544 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
10545 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
10546 (ia64_vectorize_vec_perm_const_ok): Merge into...
10547 (ia64_vectorize_vec_perm_const): ...this new function.
10548 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
10549 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
10550 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
10551 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
10552 * config/mips/mips.c (mips_expand_vec_perm_const)
10553 (mips_vectorize_vec_perm_const_ok): Merge into...
10554 (mips_vectorize_vec_perm_const): ...this new function.
10555 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
10556 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
10557 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
10558 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
10559 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
10560 (rs6000_expand_vec_perm_const): Delete.
10561 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
10563 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10564 (altivec_expand_vec_perm_const_le): Take each operand individually.
10565 Operate on constant selectors rather than rtxes.
10566 (altivec_expand_vec_perm_const): Likewise. Update call to
10567 altivec_expand_vec_perm_const_le.
10568 (rs6000_expand_vec_perm_const): Delete.
10569 (rs6000_vectorize_vec_perm_const_ok): Delete.
10570 (rs6000_vectorize_vec_perm_const): New function.
10571 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10572 an element count and rtx array.
10573 (rs6000_expand_extract_even): Update call accordingly.
10574 (rs6000_expand_interleave): Likewise.
10575 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
10576 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
10577 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
10578 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
10579 (rs6000_expand_vec_perm_const): Delete.
10580 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10581 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10582 (altivec_expand_vec_perm_const_le): Take each operand individually.
10583 Operate on constant selectors rather than rtxes.
10584 (altivec_expand_vec_perm_const): Likewise. Update call to
10585 altivec_expand_vec_perm_const_le.
10586 (rs6000_expand_vec_perm_const): Delete.
10587 (rs6000_vectorize_vec_perm_const_ok): Delete.
10588 (rs6000_vectorize_vec_perm_const): New function. Remove stray
10589 reference to the SPE evmerge intructions.
10590 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10591 an element count and rtx array.
10592 (rs6000_expand_extract_even): Update call accordingly.
10593 (rs6000_expand_interleave): Likewise.
10594 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
10595 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
10597 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10599 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10601 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
10602 vector mode and that that mode matches the mode of the data
10604 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
10605 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
10606 directly using expand_vec_perm_1 when forcing selectors into
10608 (expand_vec_perm_var): New function, split out from expand_vec_perm.
10610 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10612 * optabs-query.h (can_vec_perm_p): Delete.
10613 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
10614 * optabs-query.c (can_vec_perm_p): Split into...
10615 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
10616 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
10617 particular selector is valid.
10618 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10619 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10620 (vect_grouped_load_supported): Likewise.
10621 (vect_shift_permute_load_chain): Likewise.
10622 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10623 (vect_transform_slp_perm_load): Likewise.
10624 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10625 (vectorizable_bswap): Likewise.
10626 (vect_gen_perm_mask_checked): Likewise.
10627 * fold-const.c (fold_ternary_loc): Likewise. Don't take
10628 implementations of variable permutation vectors into account
10629 when deciding which selector to use.
10630 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
10631 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
10632 with a false third argument.
10633 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
10634 to test whether the constant selector is valid and can_vec_perm_var_p
10635 to test whether a variable selector is valid.
10637 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10639 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
10640 * optabs-query.c (can_vec_perm_p): Likewise.
10641 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
10642 instead of vec_perm_indices.
10643 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
10644 (vect_gen_perm_mask_checked): Likewise,
10645 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
10646 (vect_gen_perm_mask_checked): Likewise,
10648 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10650 * optabs-query.h (qimode_for_vec_perm): Declare.
10651 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
10652 (qimode_for_vec_perm): ...this new function.
10653 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
10655 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10657 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
10658 does not have a conditional at the top.
10660 2018-01-02 Richard Biener <rguenther@suse.de>
10662 * ipa-inline.c (big_speedup_p): Fix expression.
10664 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10667 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
10670 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10674 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
10675 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
10676 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
10677 cond_taken_branch_cost 3->4.
10679 2018-01-01 Jakub Jelinek <jakub@redhat.com>
10681 PR tree-optimization/83581
10682 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
10683 TODO_cleanup_cfg if any changes have been made.
10685 PR middle-end/83608
10686 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
10687 convert_modes if target mode has the right side, but different mode
10690 PR middle-end/83609
10691 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
10692 last argument when extracting from CONCAT. If either from_real or
10693 from_imag is NULL, use expansion through memory. If result is not
10694 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
10695 the parts directly to inner mode, if even that fails, use expansion
10698 PR middle-end/83623
10699 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
10700 check for bswap in mode rather than HImode and use that in expand_unop
10703 Copyright (C) 2018 Free Software Foundation, Inc.
10705 Copying and distribution of this file, with or without modification,
10706 are permitted in any medium without royalty provided the copyright
10707 notice and this notice are preserved.