1 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2 Tom de Vries <tom@codesourcery.com>
5 * omp-expand.c (expand_oacc_for): Add dummy false branch for
6 tiled basic blocks without omp continue statements.
8 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
11 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
12 vec_extract expression as having side effects to make sure it gets
15 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
18 * config/i386/i386.c (get_builtin_code_for_version): Check
21 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
24 * gcc/config.gcc: Support "skylake".
25 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
27 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
28 (processor_target_table): Add "skylake".
29 (ix86_option_override_internal): Add "skylake".
30 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
32 (get_builtin_code_for_version): Fix priority for
33 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
34 PROCESSOR_SKYLAKE-AVX512.
35 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
36 (processor_type): Add PROCESSOR_SKYLAKE.
38 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
39 Jason Merrill <jason@redhat.com>
42 * convert.c (convert_to_integer_1): Use direct recursion for
43 enumeral types and types with a precision less than the number
44 of bits in their mode.
46 2018-04-16 Julia Koval <julia.koval@intel.com>
49 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
50 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
52 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
55 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
56 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
59 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
62 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
63 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
64 Ensure negative shifts result in {0}.
66 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
68 PR rtl-optimization/79916
69 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
70 regs (if any) to define how to gnerate SD moves when LRA is in
73 2018-04-13 Jakub Jelinek <jakub@redhat.com>
75 PR rtl-optimization/85393
76 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
77 * except.c (expand_dw2_landing_pad_for_region): Make static.
78 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
79 a label and unconditional jump to old_bb, rather than
80 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
83 PR rtl-optimization/85376
84 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
85 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
86 instead of a specific value.
88 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
89 Bin Cheng <bin.cheng@arm.com>
91 PR tree-optimization/82965
92 PR tree-optimization/83991
93 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
94 by_profile_only parameter.
95 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
96 information if the loop was predicted to iterate too many times.
97 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
99 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
102 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
105 2018-04-13 Martin Liska <mliska@suse.cz>
106 Jakub Jelinek <jakub@redhat.com>
109 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
110 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
111 * builtins.c (expand_builtin_memory_copy_args): Use
112 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
113 handle dest_addr == pc_rtx.
115 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
118 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
119 asked to not generate direct moves.
120 (fix_trunc<mode>si2_stfiwx): Similar.
121 (fix_trunc<mode>si2_internal): Similar.
123 2018-04-12 Jakub Jelinek <jakub@redhat.com>
126 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
127 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
128 lookup if dest in some wider mode is known to be const0_rtx and
129 if so, record permanent equivalence for it to be ZERO_EXTEND of
130 the narrower mode destination.
132 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
134 * lto-streamer-out.c (output_function): Revert 259346.
135 * omp-expand.c (expand_oacc_for): Likewise.
137 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
139 PR rtl-optimization/85354
140 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
141 * sel-sched.c (sel_global_init): ... here.
143 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
146 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
147 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
148 mode for PE-COFF targets.
149 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
150 (i386_pe_asm_lto_end): Likewise.
151 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
152 (TARGET_ASM_LTO_END): Likewise.
153 * config/i386/winnt.c (saved_debug_info_level): New static variable.
154 (i386_pe_asm_lto_start): New function.
155 (i386_pe_asm_lto_end): Likewise.
157 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
158 Richard Biener <rguenther@suse.de>
161 * lto-streamer-out.c (output_function): Fix CFG loop state before
163 * omp-expand.c (expand_oacc_for): Handle calls to internal
164 functions like regular functions.
166 2018-04-12 Richard Biener <rguenther@suse.de>
169 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
170 for the early LTO debug to properly generate references to it
171 during DIE emission. Do not re-use that for the skeleton for
173 (dwarf2out_early_finish): Likewise.
175 2018-04-12 Jakub Jelinek <jakub@redhat.com>
179 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
180 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
181 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
182 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
183 and output is a reg, avoid creating invalid lowpart subreg, but
184 instead split into a 512-bit move. Don't split if not AVX512VL,
185 input is xmm16+ reg and output is a mem.
186 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
187 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
188 xmm16+ reg and output is a mem.
190 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
192 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
193 also for flag_dwarf2_cfi_asm.
195 2018-04-12 Jakub Jelinek <jakub@redhat.com>
197 PR rtl-optimization/85342
198 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
199 a bool scalar var inside of the loop instead. Don't try to update
200 recog_data.operand after failed apply_change_group.
202 2018-04-12 Tom de Vries <tom@codesourcery.com>
205 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
206 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
207 array with flexible array member as array without given dimension.
208 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
209 argument for undefined param to true.
211 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
214 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
215 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
216 from PowerPC section.
217 * config/rs6000/sysv4.opt (mcall-): Improve help text.
218 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
219 help text that is too long.
220 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
221 help text that is too long.
222 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
223 help text that is too long.
225 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
227 * config/alpha/alpha.md (stack_probe_internal): Rename
228 from "probe_stack". Update all callers.
230 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
232 PR rtl-optimization/84566
233 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
234 sched_macro_fuse_insns.
236 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
239 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
240 (compute_block_dependences): ... from here.
242 2018-04-11 Jakub Jelinek <jakub@redhat.com>
244 PR tree-optimization/85331
245 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
246 from int to HOST_WIDE_INT.
248 2018-04-11 Martin Jambor <mjambor@suse.cz>
251 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
252 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
253 not the same as the source val.
254 (cgraph_edge_brings_value_p): New parameter.
255 (gather_edges_for_value): Pass destination value to
256 cgraph_edge_brings_value_p.
257 (perhaps_add_new_callers): Likewise.
258 (get_info_about_necessary_edges): Likewise and exclude values brought
259 only by self-recursive edges.
260 (create_specialized_node): Redirect only clones of self-calling edges.
261 (+self_recursive_pass_through_p): New function.
262 (find_more_scalar_values_for_callers_subset): Use it.
263 (find_aggregate_values_for_callers_subset): Likewise.
264 (known_aggs_to_agg_replacement_list): Removed.
265 (decide_whether_version_node): Re-calculate known constants for all
266 remaining context clones.
268 2018-04-11 Richard Biener <rguenther@suse.de>
271 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
272 from early DWARF output.
273 (dwarf2out_early_finish): Output line info unconditionally into
274 early DWARF and add reference to it.
276 2018-04-11 Jakub Jelinek <jakub@redhat.com>
279 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
280 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
281 other than V2DFmode using iptr mode attribute.
282 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
284 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
286 PR rtl-optimization/84659
287 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
289 2018-04-11 Jakub Jelinek <jakub@redhat.com>
292 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
294 (output_loc_list): Pass address of a dummy size variable even in the
295 locview handling loop.
296 (index_location_lists): Add comment on why skip_loc_list_entry can't
299 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
302 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
305 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
308 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
309 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
310 and -mstring-compare-inline-limit.
312 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
315 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
316 for stack clash protection in a register whenever we need it to be in
319 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
321 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
322 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
324 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
327 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
329 (mlong-double-): Ditto.
330 * config/rs6000/sysv4.opt (msdata=): Ditto.
333 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
335 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
336 erroneous entries for
337 "vector int vec_ldl (int, long int *)", and
338 "vector unsigned int vec_ldl (int, unsigned long int *)".
339 Add comments and entries for
340 "vector bool char vec_ldl (int, bool char *)",
341 "vector bool short vec_ldl (int, bool short *)",
342 "vector bool int vec_ldl (int, bool int *)",
343 "vector bool long long vec_ldl (int, bool long long *)",
344 "vector pixel vec_ldl (int, pixel *)",
345 "vector long long vec_ldl (int, long long *)",
346 "vector unsigned long long vec_ldl (int, unsigned long long *)".
347 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
348 type tree bool_long_long_type_node and correct definition of
349 bool_V2DI_type_node to make reference to this new type tree.
350 (rs6000_mangle_type): Replace erroneous reference to
351 bool_long_type_node with bool_long_long_type_node.
352 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
353 comments to emphasize sign distinctions for char and int types and
354 replace RS6000_BTI_bool_long constant with
355 RS6000_BTI_bool_long_long constant. Also add comment to restrict
356 use of RS6000_BTI_pixel.
357 (bool_long_type_node): Remove this macro definition.
358 (bool_long_long_type_node): New macro definition
360 2018-04-10 Jakub Jelinek <jakub@redhat.com>
362 PR rtl-optimization/85300
363 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
364 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
365 simplify_unary_operation fails.
367 2018-04-10 Martin Liska <mliska@suse.cz>
369 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
370 cgraph_edge and ipa_ref.
372 2018-04-10 Jakub Jelinek <jakub@redhat.com>
377 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
378 computation of the VEC_MERGE selector from mask.
379 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
380 Fix decoding of the VEC_MERGE selector into mask.
382 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
384 PR tree-optimization/85286
385 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
387 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
389 * final.c (final_1): Set insn_last_address as well as
390 insn_current_address.
392 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
395 * explow.c (emit_stack_probe): Call validize_mem on memory location
396 before passing it to gen_probe_stack. Create address operand and
397 legitimize it for the probe_stack_address case.
399 2018-04-09 Jan Hubicka <jh@suse.cz>
402 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
403 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
404 * tree.c (free_lang_data_in_type): Fix handling of binfos;
406 (free_lang_data): Rebuild type inheritance graph.
408 2018-04-09 Martin Sebor <msebor@redhat.com>
410 * invoke.texi (-finline-small-functions): Mention other optimization
412 (-findirect-inlining, -fpartial-inlining): Same.
413 (-finline-functions-called-once): Same.
414 (-freorder-blocks-and-partition): Same.
416 2018-04-09 Jan Hubicka <jh@suse.cz>
419 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
420 jumps; choose last target that matches the criteria (i.e.
421 no partition changes for non-crossing jumps).
422 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
423 support for redirecting crossing jumps to non-crossing.
425 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
427 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
428 also for naked functions.
430 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
432 * config/arc/arc.md (add_shift): New pattern.
433 (add_shift2): Likewise.
434 (sub_shift): Likewise.
435 (sub_shift_cmp0_noout): Likewise.
436 (compare_si_ashiftsi): Likewise.
437 (xbfu_cmp0_noout): New combine pattern.
438 (xbfu_cmp0"): Likewise.
439 (movsi_set_cc_insn): Place the predicable variant first.
440 (commutative_binary_cmp0_noout): Remove clobber.
441 (commutative_binary_cmp0): New pattern.
442 (noncommutative_binary_cmp0): Likewise.
443 (noncommutative_binary_cmp0_noout): Likewise.
444 (noncommutative_binary_comparison_result_used): Removed.
445 (rsub_cmp0): New pattern.
446 (rsub_cmp0_noout): Likewise.
447 (extzvsi): Changed, keep only meaningful variants.
448 (SQH, SEZ): New iterators.
449 (SQH_postfix): New mode attribute.
450 (SEZ_prefix): New code attribute.
451 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
452 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
453 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
455 (noncommutative_operator): Check the availability of barrel
458 2018-04-09 Richard Biener <rguenther@suse.de>
460 PR tree-optimization/85284
461 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
462 Only use the niter constraining form of simple_iv when the exit
465 2018-04-09 Tom de Vries <tom@codesourcery.com>
468 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
469 (define_expand "*memory_barrier"): New define_expand.
470 (define_insn "memory_barrier"): New insn.
472 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
474 PR rtl-optimization/80463
475 PR rtl-optimization/83972
476 PR rtl-optimization/83480
478 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
479 correct producer for the insn.
480 (tidy_control_flow): Fixup seqnos in case of debug insns.
482 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
484 PR rtl-optimization/83913
486 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
487 different sched-times when merging exprs.
489 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
491 PR rtl-optimization/83962
493 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
494 tidy_fallthru_edge and tidy_control_flow.
496 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
498 PR rtl-optimization/83530
500 * sel-sched.c (force_next_insn): New global variable.
501 (remove_insn_for_debug): When force_next_insn is true, also leave only
502 next insn in the ready list.
503 (sel_sched_region): When the region wasn't scheduled, make another pass
504 over it with force_next_insn set to 1.
506 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
508 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
510 * config/nds32/constants.md (unspec_volatile_element): Add enum values
511 for interrupt control.
512 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
513 functions for interrupt control.
514 * config/nds32/nds32-intrinsic.md: Likewise.
515 * config/nds32/nds32_intrinsic.h: Likewise.
516 * config/nds32/nds32.h (nds32_builtins): Likewise.
518 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
520 * config/nds32/nds32.c (nds32_init_machine_status,
521 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
522 strict_aligned_p field.
523 (nds32_expand_to_rtl_hook): New function.
524 (TARGET_EXPAND_TO_RTL_HOOK): Define.
525 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
527 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
528 Chung-Ju Wu <jasonwucj@gmail.com>
530 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
531 * config/nds32/nds32-n7.md: New file.
532 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
533 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
535 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
536 * config/nds32/nds32.md (pipeline_model): Add n7.
537 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
538 * config/nds32/pipelines.md: Include n7 settings.
540 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
541 Chung-Ju Wu <jasonwucj@gmail.com>
543 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
544 * config/nds32/nds32-e8.md: New file.
545 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
546 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
548 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
549 * config/nds32/nds32.md (pipeline_model): Add e8.
550 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
551 * config/nds32/pipelines.md: Include e8 settings.
553 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
554 Chung-Ju Wu <jasonwucj@gmail.com>
556 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
557 * config/nds32/nds32-n8.md: New file.
558 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
559 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
561 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
562 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
563 * config/nds32/nds32.md (pipeline_model): Add n8.
564 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
565 * config/nds32/pipelines.md: Include n8 settings.
567 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
568 Chung-Ju Wu <jasonwucj@gmail.com>
570 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
571 * config/nds32/nds32-n9-2r1w.md: New file.
572 * config/nds32/nds32-n9-3r2w.md: New file.
573 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
574 nds32_register_ports): New or modify for cpu n9.
575 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
577 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
578 * config/nds32/nds32-utils.c: New file.
579 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
580 TARGET_MUL_SLOW): Define.
581 * config/nds32/nds32.md (pipeline_model): New attribute.
582 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
583 New options that support cpu n9.
584 * config/nds32/pipelines.md: Include n9 settings.
585 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
587 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
589 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
590 information if necessary.
591 (output_cond_branch_compare_zero): Likewise.
592 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
593 (nds32_target_alignment): Refine for alignment.
594 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
595 (FUNCTION_BOUNDARY): Modify.
596 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
598 * config/nds32/nds32.opt (malways-align, malign-functions): New.
600 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
602 * config/nds32/constants.md (unspec_volatile_element): Add values for
603 TLB operation and data prefetch.
604 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
605 functions for TLB operation and data prefetch.
606 * config/nds32/nds32-intrinsic.md: Likewise.
607 * config/nds32/nds32_intrinsic.h: Likewise.
608 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
609 (nds32_print_operand): Likewise.
610 * config/nds32/nds32.h (nds32_builtins): Likewise.
612 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
613 Andrew Pinski <pinsika@gcc.gnu.org>
616 * match.pd: Use constant_boolean_node of correct type instead of
617 boolean_true_node or boolean_false_node for simplifying
618 pointer comparisons to zero.
620 2018-04-07 Jakub Jelinek <jakub@redhat.com>
622 PR tree-optimization/80021
623 * tree.c (verify_type_variant): Make error call in verify_variant_match
624 translatable and remove final full stop.
626 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
628 * config/nds32/constants.md (unspec_volatile_element): Add
629 UNSPEC_VOLATILE_EH_RETURN.
630 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
631 nds32_output_stack_pop): Support dwarf exception handling process.
632 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
633 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
634 exception handling process.
635 (nds32_compute_stack_frame): Likewise.
636 (nds32_return_addr_rtx): Likewise.
637 (nds32_initial_elimination_offset): Likewise.
638 (nds32_expand_prologue): Likewise.
639 (nds32_expand_epilogue): Likewise.
640 (nds32_dynamic_chain_address): New function.
641 * config/nds32/nds32.h (machine_function): Add fields for dwarf
643 (DYNAMIC_CHAIN_ADDRESS): Define.
644 (EH_RETURN_DATA_REGNO): Define.
645 (EH_RETURN_STACKADJ_RTX): Define.
646 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
647 patterns for dwarf exception handling.
649 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
651 * config/nds32/nds32.h: Clean up obsolete macros.
653 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
655 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
656 Add enum values for particular instructions.
657 * config/nds32/nds32-intrinsic.c: Implementation of expanding
658 particular intrinsic functions.
659 * config/nds32/nds32-intrinsic.md: Likewise.
660 * config/nds32/nds32_intrinsic.h: Likewise.
661 * config/nds32/nds32.h (nds32_builtins): Likewise.
662 * config/nds32/nds32.md (type): Add pbsad and pbsada.
663 (btst, ave): New patterns for particular instructions.
665 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
667 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
668 Add enum values for atomic load/store and memory sync.
669 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
671 * config/nds32/nds32-intrinsic.md: Likewise.
672 * config/nds32/nds32_intrinsic.h: Likewise.
673 * config/nds32/nds32.h (nds32_builtins): Likewise.
675 2018-04-07 Jakub Jelinek <jakub@redhat.com>
677 PR tree-optimization/85257
678 * fold-const.c (native_encode_vector): If not all elts could fit
679 and off is -1, return 0 rather than offset.
680 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
681 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
682 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
683 adjust buffer in native_interpret_expr call.
685 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
687 * config/nds32/constants.md (unspec_volatile_element): Add cache
689 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
690 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
691 * config/nds32/nds32.c (nds32_cctl_names): New.
692 (nds32_print_operand): Handle cache control register names.
693 * config/nds32/nds32.h (nds32_builtins): New enum values.
694 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
696 * config/nds32/nds32.md (type): Add mmu.
697 * config/nds32/pipelines.md (simple_insn): Add mmu.
699 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
701 * config/nds32/nds32.md (type): Remove call.
702 * config/nds32/pipelines.md (simple_insn): Likewise.
704 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
706 * config/nds32/constants.md (unspec_volatile_element): Add
707 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
708 UNSPEC_VOLATILE_FMFCFG.
709 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
710 description for fmfcfg and fmfcsr.
711 (bdesc_1arg): Add fmtcsr.
712 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
713 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
714 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
715 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
716 unspec_fmfcfg): New patterns.
717 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
718 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
719 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
720 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
721 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
722 __nds32__fmfcfg): Define.
724 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
726 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
727 intrinsic register names.
728 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
729 intrinsic register enum values and macros.
731 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
733 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
734 for load/store addressing form.
735 (nds32_print_operand_address): Likewise.
737 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
740 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
741 based on LABEL_REF. Remove useless assertion.
742 (pic_address_needs_scratch): Fix formatting.
743 (sparc_legitimize_pic_address): Minor tweaks.
744 (sparc_delegitimize_address): Adjust assertion accordingly.
745 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
746 into symbolic_operand.
747 (movsi_high_pic_label_ref): Likewise.
748 (movsi_lo_sum_pic_label_ref): Likewise.
749 (movdi_pic_label_ref): Likewise.
750 (movdi_high_pic_label_ref): Likewise.
751 (movdi_lo_sum_pic_label_ref): Likewise.
753 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
755 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
756 custom LIB_SPEC setup.
758 2018-04-06 Ruslan Bukin <br@bsdpad.com>
759 Kito Cheng <kito.cheng@gmail.com>
761 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
762 * config/riscv/freebsd.h: New.
764 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
766 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
767 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
770 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
771 Kito Cheng <kito.cheng@gmail.com>
773 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
774 nds32_output_call, nds32_symbol_binds_local_p): New functions.
775 * config/nds32/nds32-protos.h (nds32_output_call,
776 nds32_output_return): Declare.
777 * config/nds32/nds32.md: Refine all the call and return patterns.
779 2018-04-06 Jakub Jelinek <jakub@redhat.com>
782 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
783 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
785 PR rtl-optimization/84872
786 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
787 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
790 2018-04-06 Tamar Christina <tamar.christina@arm.com>
792 * expr.c (copy_blkmode_to_reg): Revert 254862.
793 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
795 2018-04-06 Richard Biener <rguenther@suse.de>
798 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
799 after seeing a component reference with an adjacent field. Treat
800 refs to arrays at struct end of external decls similar to
801 refs to unconstrained commons.
803 2018-04-06 Jakub Jelinek <jakub@redhat.com>
806 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
807 look through SAVE_EXPRs with non-side-effects argument. Adjust
809 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
812 2018-04-06 Richard Biener <rguenther@suse.de>
815 * alias.c (find_base_term): New wrapper around find_base_term
816 unwinding CSELIB_VAL_PTR changes.
817 (find_base_term): Do not restore CSELIB_VAL_PTR during the
820 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
822 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
824 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
825 constant definitions.
826 ("nop"): lr 0,0 -> nopr r0
827 ("nop_lr0", "nop_lr1"): New insn definitions.
829 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
831 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
832 NDS32_V3PUSH_AVAILABLE_P macro.
834 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
835 Chung-Ju Wu <jasonwucj@gmail.com>
837 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
838 (nds32*-*-*): Add float and fpu_config into supported_defaults.
839 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
840 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
841 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
842 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
843 * config/nds32/constraints.md: New constraints and checking for hard
845 * config/nds32/iterators.md: New mode iterator and attribute for hard
847 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
849 * config/nds32/nds32-fpu.md: New file.
850 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
851 deal with hard float code generation.
852 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
854 (abi_type, float_reg_number): New enum type.
855 * config/nds32/nds32-predicates.c: New predicates for hard float.
856 * config/nds32/nds32-protos.h: Declare functions for hard float.
857 * config/nds32/nds32.c: Implementation for hard float configuration.
858 * config/nds32/nds32.h: Definitions for hard float configuration.
859 * config/nds32/nds32.md: Include hard float machine description and
860 modify patterns for hard float configuration.
861 * config/nds32/nds32.opt: New options for hard float configuration.
862 * config/nds32/predicates.md: New predicates for hard float
865 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
867 * common/config/nds32/nds32-common.c
868 (nds32_option_optimization_table): Enable -mreleax-hint by default.
870 2018-04-05 Jakub Jelinek <jakub@redhat.com>
873 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
874 CONSTRUCTOR_ELT (ctor, ...)->value.
876 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
879 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
881 2018-04-05 Tom de Vries <tom@codesourcery.com>
884 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
887 2018-04-05 Shiva Chen <shiva0217@gmail.com>
888 Kito Cheng <kito.cheng@gmail.com>
890 * config/nds32/constraints.md (U33): Fine-tune checking condition.
891 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
892 * config/nds32/nds32.h (nds32_16bit_address_type): Add
893 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
895 2018-04-05 Shiva Chen <shiva0217@gmail.com>
896 Kito Cheng <kito.cheng@gmail.com>
898 * config/nds32/constraints.md (Ufe): New memory constraint.
899 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
900 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
901 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
903 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
904 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
906 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
908 * config/nds32/nds32.md: Use optimize_size in the condition for
909 alu-shift instructions.
911 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
913 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
915 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
917 * config/nds32/nds32.md (negsi2): Refine pattern.
919 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
920 Chung-Ju Wu <jasonwucj@gmail.com>
922 * config/nds32/iterators.md (shift_rotate): New code iterator.
923 (shift): New code attribute.
924 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
925 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
926 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
927 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
929 (andsi3, *andsi3): Ditto.
930 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
931 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
932 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
933 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
934 nds32_ior_operand, nds32_xor_operand): New predicates.
936 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
938 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
939 (addsi3, subsi3): ... this.
941 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
943 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
945 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
947 * config/nds32/nds32.md: Adjust indention.
949 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
951 * config/nds32/nds32.md (feature): New attribute.
953 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
955 * config/nds32/nds32.md (subtype): New attribute.
957 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
960 * config/arm/arm-builtins.c (arm_expand_builtin): Change
961 expansion to perform a bitwise AND of the argument followed by a
962 boolean negation of the result.
964 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
966 PR rtl-optimization/84878
967 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
968 the basic block. Assert the use reference is not artificial and that
969 it has an associated insn.
971 2018-04-04 Michael Matz <matz@suse.de>
973 * builtins.c (compute_objsize): Pass correct operand
974 to array_at_struct_end_p.
976 2018-04-04 Richard Biener <rguenther@suse.de>
979 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
980 from contexts for DINFO_LEVEL_TERSE and below.
982 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
984 * config/nds32/nds32-doubleword.md (move_<mode>): Require
985 resiter_operand condition.
986 * config/nds32/nds32.md (*move<mode>): Ditto.
988 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
989 Monk Chiang <sh.chiang04@gmail.com>
991 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
993 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
995 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
997 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
998 Kito Cheng <kito.cheng@gmail.com>
1000 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
1001 nds32_cond_code_str, output_cond_branch,
1002 output_cond_branch_compare_zero, nds32_expand_cbranch,
1003 nds32_expand_cstore, nds32_expand_movcc,
1004 nds32_output_cbranchsi4_equality_zero,
1005 nds32_output_cbranchsi4_equality_reg,
1006 nds32_output_cbranchsi4_equality_reg_or_const_int,
1007 nds32_output_cbranchsi4_greater_less_zero: New functions.
1008 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
1009 nds32_expand_cstore, nds32_expand_movcc,
1010 nds32_output_cbranchsi4_equality_zero,
1011 nds32_output_cbranchsi4_equality_reg,
1012 nds32_output_cbranchsi4_equality_reg_or_const_int,
1013 nds32_output_cbranchsi4_greater_less_zero): Declare.
1014 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
1015 nds32_rimm11s_operand): New predicates.
1016 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
1017 * config/nds32/nds32.md: Rewrite all the branch and conditional move
1020 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1022 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
1023 * config/nds32/nds32.md: Ditto.
1024 * config/nds32/pipelines.md: Ditto.
1026 2018-04-04 Richard Biener <rguenther@suse.de>
1028 PR tree-optimization/85168
1029 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
1030 propagating abnormals.
1032 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1034 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
1036 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1037 Kito Cheng <kito.cheng@gmail.com>
1039 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
1040 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
1041 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
1042 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
1043 * config/nds32/nds32.md (sibcall_internal): New.
1044 (sibcall_register): Remove.
1045 (sibcall_immediate): Remove.
1046 (sibcall_value_internal): New.
1047 (sibcall_value_register): Remove.
1048 (sibcall_value_immediate): Remove.
1049 * config/nds32/predicates.md (nds32_general_register_operand): New.
1050 (nds32_call_address_operand): New.
1052 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1054 PR rtl-optimization/85167
1055 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
1056 bb_defs if *split_p, instead preinitialize it to NULL.
1058 PR tree-optimization/85156
1059 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
1060 evaluating the argument multiple times.
1062 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
1064 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
1066 (_mm_cvtpd_ps): Likewise.
1067 (_mm_cvttpd_epi32): Likewise.
1068 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
1069 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
1070 vector, pixel, and bool following altivec.h include.
1072 2018-04-03 Martin Sebor <msebor@redhat.com>
1074 * doc/extend.texi (Common Function Attributes): Clarify.
1075 (const attribute): Likewise.
1076 (pure attribute): Likewise.
1078 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1081 * config/i386/i386.c (ix86_expand_vector_set): Use
1082 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
1084 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
1086 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
1087 instructions when changing rounding bits to preserve precision bits
1088 in the x87 control word.
1090 2018-04-03 Martin Liska <mliska@suse.cz>
1092 PR tree-optimization/82491
1093 * rtl.h (strip_offset_and_add): Replace += suboffset with
1094 poly_uint64 () + suboffset.
1096 2018-03-29 Martin Liska <mliska@suse.cz>
1097 Martin Jambor <mjambor@suse.cz>
1100 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
1101 param_type is not an integral or pointer type.
1103 2018-04-03 Richard Biener <rguenther@suse.de>
1105 * sese.h (recompute_all_dominators): Remove.
1107 2018-04-02 Martin Sebor <msebor@redhat.com>
1109 * doc/invoke.texi (-Wrestrict): Fix typos.
1111 2018-04-02 Jim Wilson <jimw@sifive.com>
1113 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
1114 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
1115 (<optab>di3, <optab>si3_extend): Likewise.
1116 (<optab>si3_mask, <optab>si3_mask_1): New.
1117 (<optab>di3_mask, <optab>di3_mask_1): New.
1118 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
1119 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
1120 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
1122 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
1124 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
1127 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
1129 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
1130 (nds32_canonicalize_comparison): New function.
1132 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1133 Kito Cheng <kito.cheng@gmail.com>
1134 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1136 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
1137 * config/nds32/constants.md (unspec_volatile_element): Add
1138 UNSPEC_VOLATILE_RELAX_GROUP.
1139 * config/nds32/nds32-relax-opt.c: New file.
1140 * config/nds32/nds32-predicates.c
1141 (nds32_symbol_load_store_p): New function.
1142 * config/nds32/nds32-protos.h
1143 (nds32_symbol_load_store_p): Declare function.
1144 (make_pass_nds32_relax_opt): Declare new rtl pass function.
1145 * config/nds32/nds32.c
1146 (nds32_register_pass): New function to register pass.
1147 (nds32_register_passes): New function to register passes.
1148 * config/nds32/nds32.md (relax_group): New pattern.
1149 * config/nds32/nds32.opt (mrelax-hint): New option.
1150 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
1152 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
1154 * config/nds32/t-nds32: Modify files dependency.
1156 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1158 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
1159 (PROFILE_HOOK): Define its implementation.
1161 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1163 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
1164 type and 32-bit size.
1166 2018-04-01 Jakub Jelinek <jakub@redhat.com>
1169 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
1170 (V_128_256): New mode iterator.
1171 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
1172 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
1173 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
1175 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
1176 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
1178 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
1181 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
1182 NaN inputs correctly.
1184 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
1187 * config/rs6000/vsx.md (??r): New mode attribute.
1188 (*vsx_mov<mode>_64bit): Use it.
1189 (*vsx_mov<mode>_32bit): Likewise.
1191 2018-03-30 Martin Sebor <msebor@redhat.com>
1193 PR tree-optimization/84818
1194 * builtins.c (check_access): Use warning_n.
1196 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1199 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
1201 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
1204 2018-03-30 Julia Koval <julia.koval@intel.com>
1207 * x86-tune.def (movx, partial_reg_dependency): Enable for
1210 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
1213 * lra-constraints.c (process_alt_operands): Move setting
1214 this_alternative_matches below.
1216 2018-03-29 Martin Liska <mliska@suse.cz>
1219 * doc/invoke.texi: Document how LTO works with debug info.
1220 Describe auto-load support of binutils. Mention 'x86-64'
1221 as valid option value of -march option.
1223 2018-03-29 Jakub Jelinek <jakub@redhat.com>
1225 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
1228 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
1229 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
1230 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
1233 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1236 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
1237 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
1238 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
1239 for RS6000_BTM_POWERPC64.
1240 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
1241 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
1242 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
1247 2018-03-28 Carl Love <cel@us.ibm.com>
1250 2017-09-27 Carl Love <cel@us.ibm.com>
1252 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
1253 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
1254 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
1257 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1259 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
1260 instead of __vector bool.
1261 (_mm_max_pu8): Likewise.
1262 (_mm_min_pi16): Likewise.
1264 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1267 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
1268 (DIVWEUO): Likewise.
1270 (DIVDEUO): Likewise.
1271 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
1272 DIVWEUO and DIVDEUO.
1273 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
1274 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
1275 (div_extend): Likewise.
1276 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
1278 (__builtin_divweuo): Likewise.
1279 (__builtin_divdeo): Likewise.
1280 (__builtin_divdeuo): Likewise.
1282 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1285 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
1286 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
1288 PR tree-optimization/82004
1289 * gimple-match-head.c (optimize_pow_to_exp): New function.
1290 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
1291 Don't fold to exp if optimize_pow_to_exp is false.
1293 2018-03-28 Martin Liska <mliska@suse.cz>
1296 * calls.c (initialize_argument_information): Fix trailing space.
1297 * common.opt: Fix typo and provide better explanation for
1298 -fsanitize-coverage option.
1299 * config/i386/i386.opt: Fix typo.
1301 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1302 Martin Liska <mliska@suse.cz>
1305 * gimplify.c (asan_poison_variable): Don't do the check for
1306 gimplify_omp_ctxp here.
1307 (gimplify_decl_expr): Do it here.
1308 (gimplify_target_expr): Likewise.
1310 2018-03-28 Martin Liska <mliska@suse.cz>
1313 * config/i386/i386.c (ix86_function_arg_advance): Do not call
1314 chkp_type_bounds_count if MPX is not enabled.
1316 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
1318 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
1320 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
1323 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
1324 function to create the function decl for complex long double
1325 multiply and divide for -mabi=ieeelongdouble.
1326 (init_float128_ieee): Call it.
1328 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
1331 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
1332 -fcf-protection=branch -mibt.
1333 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
1335 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1338 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
1340 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
1343 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
1344 extern array declarations.
1346 2018-03-27 Richard Biener <rguenther@suse.de>
1349 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
1350 explicit single_use checks.
1352 2018-03-27 Richard Biener <rguenther@suse.de>
1354 PR tree-optimization/85082
1355 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
1358 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1360 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
1361 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
1362 Turn on fasynchronous-unwind-tables and funwind-tables.
1364 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
1367 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
1368 (*bmi_blsr_<mode>_ccz): Ditto.
1370 2018-03-26 Tom de Vries <tom@codesourcery.com>
1372 PR tree-optimization/85063
1373 * omp-general.c (offloading_function_p): New function. Factor out
1375 * omp-offload.c (pass_omp_target_link::gate): ... here.
1376 * omp-general.h (offloading_function_p): Declare.
1377 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
1378 with attribute omp declare target for offloading functions.
1380 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
1382 PR tree-optimization/84005
1383 * tree-data-ref.h (get_base_for_alignment): Declare.
1384 * tree-data-ref.c (get_base_for_alignment_1): New function.
1385 (get_base_for_alignment): Likewise.
1386 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
1387 get_base_for_alignment to find a suitable base object, instead
1388 of always using drb->base_address.
1390 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1393 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
1394 known size by default.
1396 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
1399 * lra-constraints.c (process_alt_operands): Don't match BLKmode
1400 and non BLKmode operands.
1402 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1405 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
1406 Clean up attributes.
1408 2018-03-23 Richard Biener <rguenther@suse.de>
1411 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
1412 we are going to emit early debug for LTO.
1414 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1417 * function.c (match_asm_constraints_1): Don't optimize if input
1418 doesn't satisfy general_operand predicate for output's mode.
1421 * alias.c (write_dependence_p): Don't require for x_canonicalized
1422 non-VOIDmode if x has VOIDmode.
1425 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
1426 just don't try to optimize it rather than assert it never happens.
1428 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1430 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
1431 macro expansions for definition of ST_INTERNAL_<mode> and
1432 LD_INTERNAL_<mode> builtins.
1433 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
1435 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
1437 (altivec_expand_st_builtin): Likewise.
1438 (altivec_expand_builtin): Remove calls to deleted functions.
1439 (rs6000_address_for_altivec): Delete this function.
1440 * config/rs6000/vector.md: Remove expands for
1441 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
1443 2018-03-22 Sudakshina Das <sudi.das@arm.com>
1446 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
1447 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
1448 re-computing once computed.
1449 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
1450 (arm_init_machine_status): Initialize
1451 machine->static_chain_stack_bytes.
1453 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1456 * doc/extend.texi: Add four new prototypes for vec_ld.
1457 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
1458 definitions for more logical presentation.
1459 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
1460 entries for V1TI variants of __builtin_altivec_ld builtin.
1461 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
1462 handling of V1TI variant of LVX icode pattern.
1463 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
1464 (rs6000_gimple_fold_builtin): Likewise.
1465 (altivec_init_builtins): Add code to define
1466 __builtin_altivec_lvx_v1ti function.
1468 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1471 * function.c (match_asm_constraints_1): Don't do the optimization
1472 if input isn't a REG, SUBREG, MEM or constant.
1474 2018-03-22 Tom de Vries <tom@codesourcery.com>
1476 PR tree-optimization/84956
1477 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
1478 bb_has_abnormal_pred.
1480 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1483 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
1484 DECL_INITIAL (decl) to decl at the end.
1485 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
1488 2018-03-21 Joseph Myers <joseph@codesourcery.com>
1490 * doc/extend.texi (__builtin_tgmath): Document when complex
1491 integer types are treated as _Complex _Float64.
1493 2018-03-21 Tom de Vries <tom@codesourcery.com>
1495 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
1497 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1499 PR tree-optimization/84960
1500 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
1501 if it is ENTRY block, move them into single succ of ENTRY in that case.
1503 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
1505 PR tree-optimization/84811
1506 * poly-int.h (poly_span_traits): Remove the T3 parameter and
1507 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
1508 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
1509 (known_subrange_p): Update accordingly. Cast each value involved
1510 in the size comparison, rather than casting the result of the
1513 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1515 PR tree-optimization/84982
1516 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
1517 by flipping the least significant bit rather than all bits from
1518 bitpos to bitpos + bitsize - 1.
1520 2018-03-21 Nathan Sidwell <nathan@acm.org>
1522 * doc/extend.texi (Deprecated Features): Remove mention of
1523 long-deleted deprecations.
1525 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1528 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
1529 * configure: Regenerate.
1531 2018-03-21 Tom de Vries <tom@codesourcery.com>
1533 PR tree-optimization/83126
1534 * tree-parloops.c (num_phis): New function.
1535 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
1537 2018-03-21 Nathan Sidwell <nathan@acm.org>
1539 * doc/extend.texi (Deprecated Features): Update deprecated flags,
1540 mention anon-struct/union members and trailing attributes.
1542 2018-03-21 Bin Cheng <bin.cheng@arm.com>
1544 PR tree-optimization/84969
1545 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
1546 builtin memset partitions if they set different rhs values.
1548 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1550 PR rtl-optimization/84989
1551 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
1552 VEC_DUPLICATE with scalar result mode.
1554 2018-03-21 Martin Liska <mliska@suse.cz>
1557 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
1558 not intended return statement.
1560 2018-03-21 Martin Liska <mliska@suse.cz>
1563 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
1564 (chkp_find_bound_slots_1): Limit number of iterations.
1566 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
1569 * Minor grammar fixes for x86 options.
1571 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1574 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
1575 holding REG_CFA_RESTORE notes, instead turn them into a USE.
1577 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
1580 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
1581 (altivec_lvx_<mode>_1op): Likewise.
1582 (altivec_stvx_<mode>_2op): Likewise.
1583 (altivec_stvx_<mode>_1op): Likewise.
1584 (altivec_lvx_<VM2:mode>): New define_expand.
1585 (altivec_stvx_<VM2:mode>): Likewise.
1586 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
1587 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1588 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
1589 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1590 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
1591 (rs6000_gen_lvx): Likewise.
1592 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
1593 (altivec_expand_stv_builtin): Likewise.
1594 (altivec_expand_builtin): Likewise.
1595 * config/rs6000/vector.md: Likewise.
1597 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1600 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
1603 2018-03-20 Richard Biener <rguenther@suse.de>
1606 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
1607 sign-conversions as zero, fall back to standard scalar_stmt
1610 2018-03-20 Martin Liska <mliska@suse.cz>
1613 * predict.c (rebuild_frequencies): Handle case when we have
1614 PROFILE_ABSENT, but flag_guess_branch_prob is false.
1616 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1619 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
1620 flag_section_anchors.
1621 * varasm.c (use_blocks_for_decl_p): Remove hack for
1622 dw2_force_const_mem.
1625 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
1627 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
1628 be created, use lowpart_subreg of operands[0] rather than operands[0]
1630 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
1631 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
1632 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
1633 and n constraint instead of aarch64_shift_imm_di and Usd.
1634 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
1635 (*aarch64_<optab>_reg_minus<mode>3): ... this.
1637 2018-03-20 Sudakshina Das <sudi.das@arm.com>
1640 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
1641 to favor GPR over NEON registers.
1642 (<shift>di3_neon): Likewise.
1644 2018-03-20 Tom de Vries <tom@codesourcery.com>
1647 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
1648 (nvptx_process_pars): Emit bar.sync asap and alap.
1650 2018-03-20 Tom de Vries <tom@codesourcery.com>
1653 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
1654 seen_label if seen_label is already set.
1656 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1659 * config/i386/i386.c (fold_builtin_cpu): For features above 31
1660 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
1661 Use 1U instead of 1. Formatting fixes.
1664 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
1665 instead of TREE_TYPE (s1) for the return value.
1667 2018-03-19 Jakub Jelinek <jakub@redhat.com>
1669 PR tree-optimization/84946
1670 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
1671 bitsize + bitsize in poly_uint64 rather than poly_int64.
1674 * dwarf2asm.c: Include fold-const.c.
1675 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
1676 of decl rather than decl itself.
1678 PR rtl-optimization/84643
1679 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
1681 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
1684 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
1685 calling assemble_variable.
1687 2018-03-19 Sudakshina Das <sudi.das@arm.com>
1690 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
1691 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
1693 2018-03-19 Jim Wilson <jimw@sifive.com>
1696 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
1697 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
1698 (riscv_first_stack_step): Likewise.
1699 (riscv_option_override): Use STACK_BOUNDARY instead of
1701 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
1703 (BIGGEST_ALIGNMENT): Set to 128.
1704 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
1705 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
1708 2018-03-19 Richard Biener <rguenther@suse.de>
1710 PR tree-optimization/84933
1711 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
1712 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
1714 2018-03-19 Richard Biener <rguenther@suse.de>
1716 PR tree-optimization/84859
1717 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
1718 (cond_if_else_store_replacement): Perform sinking operation on
1719 single-store BBs regardless of MAX_STORES_TO_SINK setting.
1720 Generalize what a BB with a single eligible store is.
1722 2018-03-19 Richard Biener <rguenther@suse.de>
1724 PR tree-optimization/84929
1725 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
1726 chrec_is_positive against non-chrec arg.
1728 2018-03-19 Tamar Christina <tamar.christina@arm.com>
1731 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
1733 2018-03-18 Martin Liska <mliska@suse.cz>
1735 PR rtl-optimization/84635
1736 * regrename.c (build_def_use): Use matches_mode only when
1739 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
1741 PR tree-optimization/84913
1742 * tree-vect-loop.c (vectorizable_reduction): Don't try to
1743 vectorize chains of COND_EXPRs.
1745 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1747 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
1749 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1751 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
1753 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1755 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
1757 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1758 Kito Cheng <kito.cheng@gmail.com>
1760 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
1761 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
1762 (nds32_adjust_reg_alloc_order): New function.
1763 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
1765 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
1767 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
1768 nds32_print_operand, nds32_print_operand_address): Use
1769 HOST_WIDE_INT_PRINT_DEC instead.
1771 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1773 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
1775 2018-03-17 Jakub Jelinek <jakub@redhat.com>
1778 * config/i386/i386.c (initial_ix86_tune_features,
1779 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
1781 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
1782 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
1783 rather than 1u << ix86_tune. Formatting fix.
1784 (ix86_option_override_internal): Change ix86_arch_mask from
1785 unsigned int to unsigned HOST_WIDE_INT, initialize to
1786 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
1787 (ix86_function_specific_restore): Likewise.
1789 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1792 * postreload.c (reload_combine_recognize_pattern): Perform
1793 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
1794 truncate_int_for_mode the result for the destination's mode.
1797 * hsa-gen.c (mem_type_for_type): Fix comment typo.
1798 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
1800 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1803 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
1806 * lra-assigns.c (lra_split_hard_reg_for): Don't use
1807 regno_allocno_class_array and sorted_pseudos.
1808 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
1809 insns where regno is used.
1811 2018-03-16 Martin Liska <mliska@suse.cz>
1814 * multiple_target.c (create_dispatcher_calls): Redirect
1815 reference in the symbol table.
1817 2018-03-16 Martin Liska <mliska@suse.cz>
1820 * multiple_target.c (create_dispatcher_calls): Redirect also
1823 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1827 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
1828 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
1829 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
1831 2018-03-16 Julia Koval <julia.koval@intel.com>
1833 * doc/invoke.texi (Skylake Server): Add CLWB.
1834 Cannonlake): Remove CLWB.
1836 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1838 PR tree-optimization/84841
1839 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
1841 (FLOAT_ONE_CONST_TYPE): Define.
1842 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
1843 (sort_by_operand_rank): Put entries with higher constant_type last
1844 rather than first to match comments.
1846 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
1848 * config/nios2/nios2.md (movsi_internal): Fix thinko in
1851 2018-03-15 Jakub Jelinek <jakub@redhat.com>
1854 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
1855 check and use address of target always.
1857 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
1860 * config/i386/i386.c (indirect_thunk_needed): Update comments.
1861 (indirect_thunk_bnd_needed): Likewise.
1862 (indirect_thunks_used): Likewise.
1863 (indirect_thunks_bnd_used): Likewise.
1864 (indirect_return_needed): New.
1865 (indirect_return_bnd_needed): Likewise.
1866 (output_indirect_thunk_function): Add a bool argument for
1868 (output_indirect_thunk_function): Don't generate alias for
1869 function return thunk.
1870 (ix86_code_end): Call output_indirect_thunk_function to generate
1871 function return thunks.
1872 (ix86_output_function_return): Set indirect_return_bnd_needed
1873 and indirect_return_needed instead of indirect_thunk_bnd_needed
1874 and indirect_thunk_needed.
1876 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
1878 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
1879 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
1880 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
1882 2018-03-15 David Malcolm <dmalcolm@redhat.com>
1883 Paul Hua <paul.hua.gm@gmail.com>
1886 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
1888 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
1890 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
1891 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
1894 2018-03-15 Tamar Christina <tamar.christina@arm.com>
1897 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
1898 instead of GET_MODE_SIZE when comparing Units.
1900 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1903 * varasm.c (hash_section): Return an unchangeble hash value
1904 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
1905 Return !aarch64_can_use_per_function_literal_pools_p ().
1907 2018-03-15 Jakub Jelinek <jakub@redhat.com>
1910 * optabs.c (emit_conditional_move): Pass address of cmode's copy
1911 rather than address of cmode as last argument to prepare_cmp_insn.
1913 2018-03-15 Julia Koval <julia.koval@intel.com>
1915 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
1916 F_AVX512VNNI, F_AVX512BITALG): New.
1918 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
1921 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
1922 insn for floating-point loads and stores.
1924 2018-03-14 Carl Love <cel@us.ibm.com>
1926 * config/rs6000/rs6000-c.c: Add macro definitions for
1927 ALTIVEC_BUILTIN_VEC_PERMXOR.
1928 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
1929 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
1930 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
1931 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
1933 * config/doc/extend.texi: Add prototypes for vec_permxor.
1935 2018-03-14 David Malcolm <dmalcolm@redhat.com>
1938 * diagnostic-show-locus.c (class layout_point): Convert m_line
1939 from int to linenum_type.
1940 (line_span::comparator): Use linenum "compare" function when
1941 comparing line numbers.
1942 (test_line_span): New function.
1943 (layout_range::contains_point): Convert param "row" from int to
1945 (layout_range::intersects_line_p): Likewise.
1946 (layout::will_show_line_p): Likewise.
1947 (layout::print_source_line): Likewise.
1948 (layout::should_print_annotation_line_p): Likewise.
1949 (layout::print_annotation_line): Likewise.
1950 (layout::print_leading_fixits): Likewise.
1951 (layout::annotation_line_showed_range_p): Likewise.
1952 (struct line_corrections): Likewise for field m_row.
1953 (line_corrections::line_corrections): Likewise for param "row".
1954 (layout::print_trailing_fixits): Likewise.
1955 (layout::get_state_at_point): Likewise.
1956 (layout::get_x_bound_for_row): Likewise.
1957 (layout::print_line): Likewise.
1958 (diagnostic_show_locus): Likewise for locals "last_line" and
1960 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
1961 * input.c (selftest::test_linenum_comparisons): New function.
1962 (selftest::input_c_tests): Call it.
1963 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
1964 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
1965 * selftest.h (ASSERT_GT): New macro.
1966 (ASSERT_GT_AT): New macro.
1967 (ASSERT_LT): New macro.
1968 (ASSERT_LT_AT): New macro.
1970 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
1972 PR rtl-optimization/84780
1973 * combine.c (distribute_links): Don't make a link based on pc_rtx.
1975 2018-03-14 Martin Liska <mliska@suse.cz>
1977 * tree.c (record_node_allocation_statistics): Use
1978 get_stats_node_kind.
1979 (get_stats_node_kind): New function extracted from
1980 record_node_allocation_statistics.
1981 (free_node): Use get_stats_node_kind.
1983 2018-03-14 Richard Biener <rguenther@suse.de>
1985 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
1986 that the value-set of ANTIC_IN doesn't grow.
1989 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1991 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1992 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1994 2018-03-14 Julia Koval <julia.koval@intel.com>
1996 * config.gcc (icelake-client, icelake-server): New.
1998 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
1999 (initial_ix86_arch_features): Ditto.
2000 (PTA_SKYLAKE): Add SGX.
2001 (PTA_ICELAKE): Remove.
2002 (PTA_ICELAKE_CLIENT): New.
2003 (PTA_ICELAKE_SERVER): New.
2004 (ix86_option_override_internal): Split up icelake on icelake client and
2006 (get_builtin_code_for_version): Ditto.
2007 (fold_builtin_cpu): Ditto.
2008 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
2009 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
2010 * config/i386/i386.h (processor_type): Ditto.
2011 * doc/invoke.texi: Ditto.
2013 2018-03-14 Jakub Jelinek <jakub@redhat.com>
2016 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
2017 INTEGER_CST offset, add it together with bitpos / 8 and
2018 sign extend based on POINTER_SIZE.
2022 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
2025 * config/i386/constraints.md (Yc): New register constraint.
2026 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
2027 Use Yc constraint for alternative 2 of operand 0. Remove
2028 preferred_for_speed attribute.
2030 2018-03-14 Richard Biener <rguenther@suse.de>
2032 PR tree-optimization/84830
2033 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
2034 with the old one to avoid oscillations.
2036 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
2039 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
2041 (assign_by_spills): Return a flag of reload assignment failure.
2042 Do not process the reload assignment failures. Do not spill other
2043 reload pseudos if they has the same reg class. Update n if
2045 (lra_assign): Add a return arg. Set up from the result of
2046 assign_by_spills call.
2047 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2048 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2049 usage_insns if it is not NULL.
2050 (spill_hard_reg_in_range): New function.
2051 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2052 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2053 function prototypes.
2054 (lra_assign): Change prototype.
2055 * lra.c (lra): Add code to deal with fails by splitting hard reg
2058 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
2060 * config/riscv/riscv.opt (mrelax): New option.
2061 * config/riscv/riscv.c (riscv_file_start): Emit ".option
2062 "norelax" when riscv_mrelax is disabled.
2063 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
2065 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2068 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
2069 reassociation for int modes.
2071 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2073 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
2074 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
2076 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
2077 * config/aarch64/aarch64-sve.md
2078 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
2079 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
2080 (*extend<mode><Vwide>2): Rename to...
2081 (aarch64_sve_extend<mode><Vwide>2): ...this.
2082 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
2083 renaming the old pattern to...
2084 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
2086 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
2087 define_expand, renaming the old pattern to...
2088 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
2089 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
2090 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
2091 account when deciding which SVE instruction the optab should use.
2092 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
2094 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2096 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
2097 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
2098 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
2099 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
2100 (tlsdesc_small_<mode>): Turn a define_expand and use
2101 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
2102 (tlsdesc_small_advsimd_<mode>): ...this.
2103 (tlsdesc_small_sve_<mode>): New pattern.
2105 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2107 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
2108 (UNSPEC_UMUL_HIGHPART): New constants.
2109 (MUL_HIGHPART): New int iteraor.
2110 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
2111 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
2113 (*<su>mul<mode>3_highpart): New define_insn.
2115 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2118 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2121 2018-03-13 Martin Liska <mliska@suse.cz>
2124 * (sem_item_optimizer::sem_item_optimizer): Initialize new
2126 (sem_item_optimizer::~sem_item_optimizer): Release it.
2127 (sem_item_optimizer::merge_classes): Register variable aliases.
2128 (sem_item_optimizer::fixup_pt_set): New function.
2129 (sem_item_optimizer::fixup_points_to_sets): Likewise.
2130 * ipa-icf.h: Declare new variables and functions.
2132 2018-03-13 Jakub Jelinek <jakub@redhat.com>
2135 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
2136 integer_pow2p@2 and test integer_pow2p in condition.
2137 (A < 0 ? C : 0): Similarly for @1.
2140 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
2141 characters starting at p contain '\0' character, don't look beyond
2145 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
2146 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
2149 * reg-stack.c (change_stack): Change update_end var from int to
2150 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
2151 also call set_block_for_insn on the newly added insns and rescan.
2154 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
2155 on the last operand.
2158 * tree.c (stabilize_reference_1): Return save_expr (e) for
2159 STATEMENT_LIST even if it doesn't have side-effects.
2161 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
2163 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
2165 2018-03-12 Renlin Li <renlin.li@arm.com>
2167 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
2168 aarch64_output_scalar_simd_mov_immediate.
2170 2018-03-12 Martin Sebor <msebor@redhat.com>
2172 PR tree-optimization/83456
2173 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
2174 for perfectly overlapping calls to memcpy.
2175 (gimple_fold_builtin_memory_chk): Same.
2176 (gimple_fold_builtin_strcpy): Handle no-warning.
2177 (gimple_fold_builtin_stxcpy_chk): Same.
2178 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
2180 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2182 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
2183 parameter. Use it for SFmode.
2184 (rs6000_function_arg_advance_1): Adjust.
2185 (rs6000_function_arg): Adjust.
2186 (rs6000_gimplify_va_arg): Pass false for that new parameter.
2188 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2190 PR rtl-optimization/84169
2191 PR rtl-optimization/84780
2192 * combine.c (can_combine_p): Check for a 2-insn combination whether
2193 the destination register is used between the two insns, too.
2195 2018-03-12 Richard Biener <rguenther@suse.de>
2197 PR tree-optimization/84803
2198 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
2199 for refs DR analysis didn't process.
2201 2018-03-12 Richard Biener <rguenther@suse.de>
2203 PR tree-optimization/84777
2204 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
2205 force-vectorize loops ignore whether we are optimizing for size.
2207 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
2209 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
2210 (TARGET_MD_ASM_ADJUST): Define.
2212 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
2213 Kito Cheng <kito.cheng@gmail.com>
2214 Chung-Ju Wu <jasonwucj@gmail.com>
2216 * config/nds32/nds32.c (nds32_compute_stack_frame,
2217 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
2218 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
2219 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
2220 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
2221 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
2222 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
2223 * config/nds32/nds32.md (prologue, epilogue): Use macro
2224 NDS32_V3PUSH_AVAILABLE_P to do checking.
2226 2018-03-11 Jakub Jelinek <jakub@redhat.com>
2229 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
2230 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
2231 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
2232 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
2233 addition of most attributes on !orig_type_die or the attribute not
2234 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
2236 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2237 Chung-Ju Wu <jasonwucj@gmail.com>
2239 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
2241 * config/nds32/nds32.opt (mvh): New option.
2243 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2244 Chung-Ju Wu <jasonwucj@gmail.com>
2246 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
2248 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
2249 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
2252 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2253 Chung-Ju Wu <jasonwucj@gmail.com>
2255 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
2257 * config/nds32/nds32-multiple.md (strlensi): New pattern.
2258 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
2260 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
2261 Kito Cheng <kito.cheng@gmail.com>
2262 Chung-Ju Wu <jasonwucj@gmail.com>
2264 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
2265 UNSPEC_FFMISM and UNSPEC_FLMISM.
2266 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
2267 for ffb, ffmism and flmism.
2268 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
2269 (unspec_ffmism): Ditto.
2270 (unspec_flmism): Ditto.
2271 (nds32_expand_builtin_impl): Check if string extension is available.
2272 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
2273 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
2275 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
2278 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2281 * lra-assigns.c (assign_by_spills): Return a flag of reload
2282 assignment failure. Do not process the reload assignment
2283 failures. Do not spill other reload pseudos if they has the same
2285 (lra_assign): Add a return arg. Set up from the result of
2286 assign_by_spills call.
2287 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2288 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2289 usage_insns if it is not NULL.
2290 (spill_hard_reg_in_range): New function.
2291 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2292 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2293 function prototypes.
2294 (lra_assign): Change prototype.
2295 * lra.c (lra): Add code to deal with fails by splitting hard reg
2298 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
2301 * config/i386/i386.opt: Replace Enforcment with Enforcement.
2303 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
2306 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
2307 (dw_val_node): Add val_symbolic_view.
2308 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
2309 (symview_upper_bound): New.
2310 (new_line_info_table): Initialize symviews_since_reset.
2311 (dwarf2out_source_line): Count symviews_since_reset and set
2312 symview_upper_bound.
2313 (dw_val_equal_p): Handle symview.
2314 (add_AT_symview): New.
2315 (print_dw_val): Handle symview.
2316 (attr_checksum, attr_checksum_ordered): Likewise.
2317 (same_dw_val_p, size_of_die): Likewise.
2318 (value_format, output_die): Likewise.
2319 (add_high_low_attributes): Use add_AT_symview for entry_view.
2320 (dwarf2out_finish): Reset symview_upper_bound, clear
2323 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
2326 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
2327 Add strict argument and use it.
2328 (rs6000_split_multireg_move): Update for new strict argument.
2329 (mem_operand_gpr): Disallow all non-offsettable addresses.
2330 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
2332 2018-03-09 Jakub Jelinek <jakub@redhat.com>
2335 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
2336 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
2337 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
2340 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
2341 decl, use remap_type if we want to use the type.
2343 2018-03-09 Martin Sebor <msebor@redhat.com>
2345 PR tree-optimization/84526
2346 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2348 (builtin_access::generic_overlap): Be prepared to handle non-array
2351 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
2353 PR rtl-optimization/84682
2354 * lra-constraints.c (process_address_1): Check is_address flag
2355 for address constraints.
2356 (process_alt_operands): Likewise.
2357 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
2358 preprocess_constraints.
2359 * recog.h (preprocess_constraints): Add oploc parameter.
2361 * recog.c (preprocess_constraints): Test address_operand for
2362 CT_ADDRESS constraints.
2364 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2367 * lra-assigns.c (assign_by_spills): Return a flag of reload
2368 assignment failure. Do not process the reload assignment
2369 failures. Do not spill other reload pseudos if they has the same
2371 (lra_assign): Add a return arg. Set up from the result of
2372 assign_by_spills call.
2373 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2374 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2375 usage_insns if it is not NULL.
2376 (spill_hard_reg_in_range): New function.
2377 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2378 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2379 function prototypes.
2380 (lra_assign): Change prototype.
2381 * lra.c (lra): Add code to deal with fails by splitting hard reg
2384 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2387 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
2388 Accept complain bool parameter. Only emit errors if it is true.
2389 (arm_parse_cpu_option_name): Likewise.
2390 (arm_target_thumb_only): Adjust callers of the above.
2391 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
2392 prototype to take a default true bool parameter.
2393 (arm_parse_arch_option_name): Likewise.
2395 2018-03-09 David Malcolm <dmalcolm@redhat.com>
2396 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
2400 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
2401 * configure: Regenerate.
2402 * configure.ac ("linker --version-script option"): New.
2403 ("linker soname option"): New.
2405 2018-03-09 Richard Biener <rguenther@suse.de>
2407 PR tree-optimization/84775
2408 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
2409 immediate uses of predicate stmts and mark them modified.
2412 PR tree-optimization/84178
2413 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2415 (version_loop_for_if_conversion): Delay update_ssa call.
2416 (tree_if_conversion): Delay update_ssa until after predicate
2419 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
2422 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
2423 when the function accesses prior frames.
2425 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2428 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
2429 gen_llsym, otherwise call maybe_gen_llsym.
2432 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
2433 has ',' character inside of it.
2435 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2438 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
2439 as clobbering CC_REGNUM.
2441 2018-03-08 Richard Biener <rguenther@suse.de>
2444 * tree-scalar-evolution.c: Include tree-into-ssa.h.
2445 (follow_copies_to_constant): Do not follow SSA names registered
2448 2018-03-08 Richard Biener <rguenther@suse.de>
2450 PR tree-optimization/84178
2451 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2453 (version_loop_for_if_conversion): Delay update_ssa call.
2454 (tree_if_conversion): Delay update_ssa until after predicate
2457 2018-03-08 David Malcolm <dmalcolm@redhat.com>
2459 PR tree-optimization/84178
2460 * tree-if-conv.c (release_bb_predicate): Remove the
2461 the assertion that the stmts have NULL use_ops.
2462 Discard the statements, asserting that they haven't
2463 yet been added to a BB.
2465 2018-03-08 Richard Biener <rguenther@suse.de>
2467 PR tree-optimization/84746
2468 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
2469 (phi_translate): Pass in destination ANTIC_OUT set.
2470 (phi_translate_1): Likewise. For a simplified result lookup
2471 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
2472 (phi_translate_set): Adjust.
2473 (do_pre_regular_insertion): Likewise.
2474 (do_pre_partial_partial_insertion): Likewise.
2476 2018-03-08 Martin Liska <mliska@suse.cz>
2478 PR gcov-profile/84735
2479 * doc/gcov.texi: Document usage of profile files.
2480 * gcov-io.h: Document changes in the format.
2482 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
2486 * dwarf2out.c (struct dw_line_info_table): Update comments for
2488 (FORCE_RESET_NEXT_VIEW): New.
2489 (FORCE_RESETTING_VIEW_P): New.
2490 (RESETTING_VIEW_P): Check for -1 too.
2491 (ZERO_VIEW_P): Likewise.
2492 (new_line_info_table): Force-reset next view.
2493 (dwarf2out_begin_function): Likewise.
2494 (dwarf2out_source_line): Simplify zero_view_p initialization.
2495 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
2496 view directly. Omit view when omitting .loc at line 0.
2498 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2500 PR tree-optimization/84740
2501 * tree-switch-conversion.c (process_switch): Call build_constructors
2502 only if info.phi_count is non-zero.
2504 PR tree-optimization/84739
2505 * tree-tailcall.c (find_tail_calls): Check call arguments against
2506 DECL_ARGUMENTS (current_function_decl) rather than
2507 DECL_ARGUMENTS (func) when checking for tail recursion.
2509 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2511 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
2512 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
2513 Volker Reichelt's entry and add entries for people that perform
2514 GCC fuzzy testing and report numerous bugs.
2516 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
2519 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
2520 readonly data in sdata, if that is disabled.
2521 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
2522 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
2523 -mreadonly-in-sdata option.
2525 2018-03-07 Martin Sebor <msebor@redhat.com>
2527 PR tree-optimization/84468
2528 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
2529 basic block when looking for nul assignment.
2531 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
2534 * except.h (output_function_exception_table): Adjust prototype.
2535 * except.c (output_function_exception_table): Remove FNNAME parameter
2536 and add SECTION parameter. Ouput one part of the table at a time.
2537 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
2538 the first part of the exception table and emit unwind directives.
2539 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
2540 (i386_pe_seh_cold_init): Likewise.
2541 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
2542 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
2543 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
2544 (ix86_output_call_insn): Emit a nop in one more case for SEH.
2545 * config/i386/winnt.c: Include except.h.
2546 (struct seh_frame_state): Add reg_offset, after_prologue and
2547 in_cold_section fields.
2548 (i386_pe_seh_end_prologue): Set seh->after_prologue.
2549 (i386_pe_seh_cold_init): New function.
2550 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
2551 to seh->in_cold_section.
2552 (seh_emit_push): Record the offset of the push.
2553 (seh_emit_save): Record the offet of the save.
2554 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
2555 Test seh->after_prologue to disregard the epilogue.
2556 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
2557 (i386_pe_end_cold_function): New function.
2559 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2562 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
2563 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
2566 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
2568 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
2569 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
2572 * multiple_target.c: Include tree-inline.h and intl.h.
2573 (expand_target_clones): Diagnose and fail if node->definition and
2574 !tree_versionable_function_p (node->decl).
2576 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
2578 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
2580 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
2581 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
2582 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
2584 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2587 * combine.c (try_combine): Use reg_or_subregno instead of handling
2588 just paradoxical SUBREGs and REGs.
2590 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
2592 * config/arc/arc.c (arc_finalize_pic): Remove function.
2593 (arc_must_save_register): We use single base PIC register, remove
2594 checks to save/restore the PIC register.
2595 (arc_expand_prologue): Likewise.
2596 * config/arc/arc-protos.h (arc_set_default_type_attributes):
2598 (arc_verify_short): Likewise.
2599 (arc_attr_type): Likewise.
2600 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
2601 (walk_stores): Likewise.
2602 (arc_address_cost): Make it static.
2603 (arc_verify_short): Likewise.
2604 (branch_dest): Likewise.
2605 (arc_attr_type): Likewise.
2606 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
2607 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
2608 (arc_final_prescan_insn): Remove inserting the nops due to
2609 hardware hazards. It is done in reorg step.
2610 (insn_length_variant_t): Remove.
2611 (insn_length_parameters_t): Likewise.
2612 (arc_insn_length_parameters): Likewise.
2613 (arc_get_insn_variants): Likewise.
2614 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
2616 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2619 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
2622 PR tree-optimization/84687
2623 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
2625 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
2627 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2629 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
2630 Rename to ppc_speculation_barrier.
2631 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
2632 __builtin_ppc_speculation_barrier.
2634 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2637 * combine.c (combine_simplify_rtx): Don't try to simplify if
2638 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
2641 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
2643 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
2644 to 32 bytes when compiling for POWER9.
2646 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2649 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
2650 regparm >= 3 with no arg reg available also for calls with
2651 flag_force_indirect_call. Pass decl to ix86_function_regparm.
2654 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
2656 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
2658 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
2661 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
2663 2018-03-05 Richard Biener <rguenther@suse.de>
2665 PR tree-optimization/84486
2666 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
2667 When inserting a __builtin_assume_aligned call set the LHS
2668 SSA name alignment info accordingly.
2670 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
2672 PR tree-optimization/84114
2673 * config/aarch64/aarch64.c (aarch64_reassociation_width)
2674 Avoid reassociation of FLOAT_MODE addition.
2676 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
2678 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
2679 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
2680 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
2681 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
2682 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
2683 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
2684 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
2686 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
2687 __builtin_ia32_wbinvd): New builtins.
2688 (SPECIAL_ARGS2): New.
2689 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
2690 (SPECIAL_ARGS2): New.
2691 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
2692 (ix86_valid_target_attribute_inner_p): Ditto.
2693 (ix86_init_mmx_sse_builtins): Add special_args2.
2694 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
2695 TARGET_WBNOINVD_P): New.
2696 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
2697 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
2698 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
2699 * config/i386/immintrin.h (_wbinvd): New intrinsic.
2700 * config/i386/pconfigintrin.h: New file.
2701 * config/i386/wbnoinvdintrin.h: Ditto.
2702 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
2703 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
2705 2018-03-05 Richard Biener <rguenther@suse.de>
2707 PR tree-optimization/84670
2708 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2710 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2711 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2712 (compute_antic_aux): Only assert the number of values in ANTIC_IN
2713 doesn't grow if all successors (recursively) were visited at least
2716 2018-03-05 Richard Biener <rguenther@suse.de>
2718 PR tree-optimization/84650
2719 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
2720 if executed in the loop pipeline.
2722 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
2724 * doc/configfiles.texi (Configuration Files): Move info about
2725 conditionalizing $target-protos.h to...
2726 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
2727 differs from $target-protos.h.
2729 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
2730 Chung-Ju Wu <jasonwucj@gmail.com>
2732 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
2733 * config/nds32/nds32-multiple.md (setmemsi): Define.
2734 * config/nds32/nds32-memory-manipulation.c
2735 (nds32_gen_dup_4_byte_to_word_value): New.
2736 (emit_setmem_word_loop): New.
2737 (emit_setmem_byte_loop): New.
2738 (nds32_expand_setmem_loop): New.
2739 (nds32_expand_setmem_loop_v3m): New.
2740 (nds32_expand_setmem_unroll): New.
2741 (nds32_expand_setmem): New.
2743 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2744 Chung-Ju Wu <jasonwucj@gmail.com>
2746 * config/nds32/nds32-memory-manipulation.c
2747 (nds32_emit_load_store): New.
2748 (nds32_emit_post_inc_load_store): New.
2749 (nds32_emit_mem_move): New.
2750 (nds32_emit_mem_move_block): New.
2751 (nds32_expand_movmemsi_loop_unknown_size): New.
2752 (nds32_expand_movmemsi_loop_known_size): New.
2753 (nds32_expand_movmemsi_loop): New.
2754 (nds32_expand_movmemsi_unroll): New.
2755 (nds32_expand_movmemqi): Rename ...
2756 (nds32_expand_movmemsi): ... to this.
2757 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
2758 (movmemsi): ... to this.
2759 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
2760 (nds32_expand_movmemsi): ... to this.
2762 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2763 Monk Chiang <sh.chiang04@gmail.com>
2764 Chung-Ju Wu <jasonwucj@gmail.com>
2766 * config/nds32/nds32-protos.h
2767 (nds32_expand_load_multiple): New arguments.
2768 (nds32_expand_store_multiple): Ditto.
2769 (nds32_valid_multiple_load_store): Rename ...
2770 (nds32_valid_multiple_load_store_p): ... to this.
2771 * config/nds32/nds32-memory-manipulation.c
2772 (nds32_expand_load_multiple): Refine implementation.
2773 (nds32_expand_store_multiple): Ditto.
2774 * config/nds32/nds32-multiple.md
2775 (load_multiple): Update nds32_expand_load_multiple interface.
2776 (store_multiple): Update nds32_expand_store_multiple interface.
2777 * config/nds32/nds32-predicates.c
2778 (nds32_valid_multiple_load_store): Rename ...
2779 (nds32_valid_multiple_load_store_p): ... to this and refine
2781 * config/nds32/predicates.md
2782 (nds32_load_multiple_and_update_address_operation): New predicate.
2783 (nds32_store_multiple_and_update_address_operation): New predicate.
2785 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2786 Chung-Ju Wu <jasonwucj@gmail.com>
2788 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
2789 (combo): New attribute.
2790 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
2792 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
2794 * config/nds32/nds32.opt: Change -mcmodel= default value.
2796 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
2797 Monk Chiang <sh.chiang04@gmail.com>
2798 Chung-Ju Wu <jasonwucj@gmail.com>
2800 * config/nds32/constants.md (unspec_element): New enum.
2801 * config/nds32/constraints.md (Umw): New constraint.
2802 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
2803 * config/nds32/nds32-intrinsic.md: Likewise.
2804 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
2805 (nds32_valid_smw_lwm_base_p): New.
2806 (nds32_output_smw_single_word): New.
2807 (nds32_output_lmw_single_word): New.
2808 (nds32_expand_unaligned_load): New.
2809 (nds32_expand_unaligned_store): New.
2810 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
2811 (nds32_output_smw_single_word): Declare.
2812 (nds32_output_lmw_single_word): Declare.
2813 (nds32_expand_unaligned_load): Declare.
2814 (nds32_expand_unaligned_store): Declare.
2815 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
2816 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
2817 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
2818 NDS32_BUILTIN_UASTORE_DW.
2819 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
2822 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
2823 Kito Cheng <kito.cheng@gmail.com>
2824 Chung-Ju Wu <jasonwucj@gmail.com>
2826 * config/nds32/nds32-intrinsic.c
2827 (nds32_expand_builtin_null_ftype_reg): Delete.
2828 (nds32_expand_builtin_reg_ftype_imm): Ditto.
2829 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
2830 (nds32_read_argument): New.
2831 (nds32_legitimize_target): Ditto.
2832 (nds32_legitimize_argument): Ditto.
2833 (nds32_check_constant_argument): Ditto.
2834 (nds32_expand_unop_builtin): Ditto.
2835 (nds32_expand_unopimm_builtin): Ditto.
2836 (nds32_expand_binop_builtin): Ditto.
2837 (nds32_builtin_decl_impl): Ditto.
2838 (builtin_description): Ditto.
2839 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
2840 (nds32_init_builtins_impl): Ditto.
2841 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
2842 (nds32_builtin_decl): New.
2843 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
2844 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
2846 2018-03-02 Jeff Law <law@redhat.com>
2848 * reorg.c (stop_search_p): Handle DEBUG_INSN.
2849 (redundant_insn, fill_simple_delay_slots): Likewise.
2850 (fill_slots_from_thread): Likewise.
2851 * resource.c (mark_referenced_resources): Likewise.
2852 (mark_set_resources, find_dead_or_set_registers): Likewise.
2854 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2856 * substring-locations.h (format_warning_va): Formatting fix for
2858 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
2860 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
2861 * substring-locations.c: Include intl.h.
2862 (format_warning_va): Turned into small wrapper around
2863 format_warning_n_va, renamed to ...
2864 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
2865 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
2867 (format_warning_at_substring_n): New function.
2868 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
2869 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
2870 format_warning_at_substring with just a shorter name instead of
2871 const function pointer.
2872 (fmtwarn_n): New function.
2873 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
2874 appropriate, get rid of all the fmtstr temporaries, move conditionals
2875 with G_() wrapped string literals directly into fmtwarn arguments,
2876 cast dir.len to (int), formatting fixes.
2878 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
2880 * doc/invoke.texi: Remove "Cilk Plus" references.
2882 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2883 Richard Biener <rguenther@suse.de>
2886 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
2887 for error or warning attributes if CALL_FROM_THUNK_P is set.
2890 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2893 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
2894 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
2897 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
2898 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
2900 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
2901 instead of -1U in last predictors element's probability member.
2903 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2906 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2907 arguments if they are comparable.
2909 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
2911 PR tree-optimization/84634
2912 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
2913 masks and masked_loop_p with a single loop_masks, making sure it's
2914 null for bb vectorization.
2916 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
2918 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
2919 (vect_analyze_data_ref_access): Use loop->safe_len rather than
2920 loop->force_vectorize to check whether there is no alias.
2922 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2925 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
2927 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
2929 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
2930 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
2931 instead of a loop around prev_real_insn.
2932 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
2936 * config/i386/i386.c (ix86_print_operand): Use conditional
2937 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
2940 2018-03-02 Richard Biener <rguenther@suse.de>
2942 PR tree-optimization/84427
2943 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
2944 (bitmap_set_subtract_values): Rewrite to handle multiple
2947 (prune_clobbered_mems): Likewise.
2948 (phi_translate): Take edge instead of pred/phiblock.
2949 (phi_translate_1): Likewise.
2950 (phi_translate_set): Likewise. Insert all translated
2951 exprs for a value into the set, keeping possibly multiple
2952 expressions per value.
2953 (compute_antic_aux): Adjust for phi_translate changes.
2954 When intersecting union the expressions and prune those
2955 not in the final value set, keeping possibly multiple
2956 expressions per value. Do not use value-insertion
2957 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
2958 all expressions. Add verification that the value-sets
2959 only shrink during iteration.
2960 (compute_partial_antic_aux): Adjust for the phi_translate changes.
2961 (do_pre_regular_insertion): Likewise.
2962 (do_pre_partial_partial_insertion): Likewise.
2964 2018-03-02 Richard Biener <rguenther@suse.de>
2967 * config/darwin.c (saved_debug_info_level): New static global.
2968 (darwin_asm_lto_start): Disable debug info generation for LTO out.
2969 (darwin_asm_lto_end): Restore debug info generation settings.
2971 2018-03-01 Martin Liska <mliska@suse.cz>
2974 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
2977 2018-03-01 Richard Biener <rguenther@suse.de>
2980 * dwarf2out.c (gen_variable_die): Properly handle late VLA
2981 type annotation with LTO when debug was disabled at compile-time.
2983 2018-03-01 Matthew Fortune <mfortune@gmail.com>
2985 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
2987 (mips_final_postscan_insn): Likewise.
2989 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
2991 PR rtl-optimization/84528
2992 * alias.c (init_alias_target): Add commentary.
2993 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
2994 a unique base value if the frame pointer is not eliminated
2995 to the stack pointer.
2997 2018-03-01 Tom de Vries <tom@codesourcery.com>
2999 PR rtl-optimization/83327
3000 * lra-int.h (hard_regs_spilled_into): Declare.
3001 * lra.c (hard_regs_spilled_into): Define.
3002 (init_reg_info): Init hard_regs_spilled_into.
3003 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
3004 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
3005 (process_bb_lives): Handle hard_regs_spilled_into.
3006 (lra_create_live_ranges_1): Before doing liveness propagation, clear
3007 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
3009 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
3011 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
3012 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
3013 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
3014 * config/rs6000/aix72.h: New file.
3016 2018-02-28 Jakub Jelinek <jakub@redhat.com>
3018 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
3019 instead of warning_at with conditional singular and plural messages
3023 * stor-layout.c (update_alignment_for_field): For
3024 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
3025 && !DECL_PACKED (field), do the alignment update, just use
3026 only desired_align instead of MAX (type_align, desired_align)
3028 (place_field): Don't do known_align < desired_align handling
3029 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
3030 is non-NULL, instead do it after rli->prev_field handling and
3031 only if not within a bitfield word. For DECL_PACKED (field)
3032 use type_align of BITS_PER_UNIT.
3034 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
3036 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
3037 superfluous parentheses and trailing spaces.
3039 2018-02-28 Richard Biener <rguenther@suse.de>
3041 PR tree-optimization/84584
3042 * graphite-scop-detection.c (scop_detection::add_scop): Discard
3043 SCoPs with fake exit edge.
3045 2018-02-28 Martin Liska <mliska@suse.cz>
3048 * timevar.c (timer::print): Fix format to properly print 100%
3051 2018-02-28 Richard Biener <rguenther@suse.de>
3054 * genmatch.c (capture_info::walk_match): Do not mark
3055 captured expressions without operands as expr_p given
3056 they act more like predicates and should be subject to
3057 "lost tail" side-effect preserving.
3059 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
3061 PR rtl-optimization/81611
3062 * auto-inc-dec.c (attempt_change): Move dead note from
3063 mem_insn if it's the next use of regno
3064 (find_address): Take address use of reg holding
3065 non-incremented value. Add parm to limit search to the named
3067 (merge_in_block): Attempt to use a mem insn that is the next
3068 use of the original regno.
3070 2018-02-27 Martin Sebor <msebor@redhat.com>
3073 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
3074 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
3076 2018-02-27 Martin Sebor <msebor@redhat.com>
3078 PR translation/84207
3079 * diagnostic-core.h (warning_n, error_n, inform_n): Change
3080 n argument to unsigned HOST_WIDE_INT.
3081 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
3082 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
3083 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
3084 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
3086 2018-02-27 Richard Biener <rguenther@suse.de>
3088 PR tree-optimization/84512
3089 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
3090 Do not use the estimate returned from record_stmt_cost for
3091 the scalar iteration cost but sum properly using add_stmt_cost.
3093 2018-02-27 Richard Biener <rguenther@suse.de>
3095 PR tree-optimization/84466
3096 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3097 Adjust last change to less strictly validate use operands.
3099 2018-02-27 Martin Liska <mliska@suse.cz>
3101 PR gcov-profile/84548
3102 * gcov.c (process_file): Allow partial overlap and consider it
3103 also as group functions.
3104 (output_lines): Properly calculate range of lines for a group.
3106 2018-02-27 Martin Liska <mliska@suse.cz>
3108 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
3109 'ggc' suffixes. Change first column width.
3110 (timer::print): Fix formatting of the column.
3112 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
3114 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
3115 preserve inline entry blocks for the sake of debug inline
3116 entry point markers alone.
3117 (remove_unused_locals): Suggest in comments a better place to
3118 force the preservation of inline entry blocks that are
3119 otherwise unused, but do not preserve them.
3121 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3123 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
3125 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3128 * config/i386/constraints.md (Bs): Replace
3129 ix86_indirect_branch_register with
3130 TARGET_INDIRECT_BRANCH_REGISTER.
3132 * config/i386/i386.md (indirect_jump): Likewise.
3133 (tablejump): Likewise.
3134 (*sibcall_memory): Likewise.
3135 (*sibcall_value_memory): Likewise.
3136 Peepholes of indirect call and jump via memory: Likewise.
3137 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
3138 (*sibcall_value_GOT_32): Likewise.
3139 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3140 (GOT_memory_operand): Likewise.
3141 (call_insn_operand): Likewise.
3142 (sibcall_insn_operand): Likewise.
3143 (GOT32_symbol_operand): Likewise.
3144 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
3146 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3148 PR rtl-optimization/83496
3149 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
3150 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
3151 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
3152 redundant insn, if any.
3153 (relax_delay_slots): Likewise.
3154 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
3156 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
3158 PR tree-optimization/83965
3159 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
3160 that grouped statements are part of a reduction chain. Return
3161 true if the statement is not marked as a reduction itself but
3163 (vect_recog_dot_prod_pattern): Don't check whether the statement
3164 is part of a group here.
3165 (vect_recog_sad_pattern): Likewise.
3166 (vect_recog_widen_sum_pattern): Likewise.
3168 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3171 * final.c (rest_of_clean_state): Also look for calls inside sequences.
3173 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3176 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
3178 (ix86_output_indirect_function_return): New prototype.
3179 (ix86_split_simple_return_pop_internal): Likewise.
3180 * config/i386/i386.c (indirect_return_via_cx): New.
3181 (indirect_return_via_cx_bnd): Likewise.
3182 (indirect_thunk_name): Handle return va CX_REG.
3183 (output_indirect_thunk_function): Create alias for
3184 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
3185 (ix86_output_indirect_jmp): Remove the bool argument.
3186 (ix86_output_indirect_function_return): New function.
3187 (ix86_split_simple_return_pop_internal): Likewise.
3188 * config/i386/i386.md (*indirect_jump): Don't pass false
3189 to ix86_output_indirect_jmp.
3190 (*tablejump_1): Likewise.
3191 (simple_return_pop_internal): Change it to define_insn_and_split.
3192 Call ix86_split_simple_return_pop_internal to split it for
3194 (simple_return_indirect_internal): Call
3195 ix86_output_indirect_function_return instead of
3196 ix86_output_indirect_jmp.
3198 2018-02-26 Jakub Jelinek <jakub@redhat.com>
3201 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
3202 memset and value initialization afterwards.
3204 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
3206 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
3208 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3211 * common/config/aarch64/aarch64-common.c
3212 (aarch_option_optimization_table[]): Switch
3213 off fomit-frame-pointer
3215 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3216 Chung-Ju Wu <jasonwucj@gmail.com>
3218 * config/nds32/nds32-multiple.md (load_multiple): Disallow
3220 (store_multiple): Ditto.
3222 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3224 * config.gcc: Add --with-cpu support for nds32 target.
3225 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
3226 * config/nds32/nds32.opt: Add -mcpu= option.
3228 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
3230 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
3231 isel=yes): Warn for these deprecated options.
3233 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
3235 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
3236 ISA_2_5_MASKS_EMBEDDED.
3238 2018-02-23 Jakub Jelinek <jakub@redhat.com>
3240 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
3241 p->max as pointers rather than using iterative_hash_expr.
3243 2018-02-23 Carl Love <cel@us.ibm.com>
3245 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
3246 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
3248 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
3249 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
3250 P8V_BUILTIN_VEC_VUNSIGNED2.
3252 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
3255 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
3256 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
3258 * lra-constraints.c (curr_insn_transform): Set up
3259 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
3261 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
3262 * lra-eliminations.c (spill_pseudos): Ditto.
3263 (process_insn_for_elimination): Ditto.
3264 * lra-lives.c (reg_early_clobber_p): Use the new macros.
3265 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
3266 LRA_NON_CLOBBERED_ALT.
3268 2018-02-22 Martin Sebor <msebor@redhat.com>
3270 PR tree-optimization/84480
3271 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
3272 to maybe_diag_stxncpy_trunc. Call it.
3273 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
3274 from gimple_fold_builtin_strcpy. Print inlining stack.
3275 (handle_builtin_stxncpy): Print inlining stack.
3276 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
3278 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
3281 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
3282 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
3283 and -fcheck-pointer-bounds are used together.
3284 (indirect_thunk_prefix): New enum.
3285 (indirect_thunk_need_prefix): New function.
3286 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
3287 "_nt" instead of "_bnd" for NOTRACK prefix.
3288 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
3289 (output_indirect_thunk_function): Likewise.
3291 (ix86_code_end): Update output_indirect_thunk_function calls.
3292 (ix86_output_indirect_branch_via_reg): Replace
3293 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
3294 (ix86_output_indirect_branch_via_push): Likewise.
3295 (ix86_output_function_return): Likewise.
3296 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
3297 incompatible with -fcf-protection=branch and
3298 -fcheck-pointer-bounds.
3300 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3303 * config/aarch64/aarch64.c (aarch64_print_address_internal):
3304 Change gcc_assert call to output_operand_lossage.
3306 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3308 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
3310 2018-02-22 DJ Delorie <dj@redhat.com>
3311 Sebastian Perta <sebastian.perta@renesas.com>
3312 Oleg Endo <olegendo@gcc.gnu.org>
3314 * config/rx/rx.c (rx_rtx_costs): New function.
3315 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
3317 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
3319 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
3321 2018-02-22 Martin Liska <mliska@suse.cz>
3324 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
3325 Add "native" as a possible value.
3327 2018-02-22 Martin Liska <mliska@suse.cz>
3330 * config/i386/i386.c (ix86_option_override_internal):
3331 Add "native" as a possible value for -march and -mtune.
3333 2018-02-22 Jakub Jelinek <jakub@redhat.com>
3336 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
3337 to all type variants.
3339 PR tree-optimization/84503
3340 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
3341 width as info->bitpos + info->bitsize - start.
3342 (merged_store_group::merge_overlapping): Simplify width computation.
3343 (check_no_overlap): New function.
3344 (imm_store_chain_info::try_coalesce_bswap): Compute expected
3345 start + width and last_order of the group, fail if check_no_overlap
3347 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
3348 to group if check_no_overlap fails.
3350 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3352 * config/rs6000/altivec.md: Delete contraint arguments to
3353 define_expand, define_split, and define_peephole2, and in
3354 define_insn_and_split if always unused.
3355 * config/rs6000/darwin.md: Ditto.
3356 * config/rs6000/dfp.md: Ditto.
3357 * config/rs6000/rs6000.md: Ditto.
3358 * config/rs6000/sync.md: Ditto.
3359 * config/rs6000/vector.md: Ditto.
3360 * config/rs6000/vsx.md: Ditto.
3362 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3364 * config/rs6000/altivec.md: Write output control strings as braced
3365 blocks instead of double-quoted strings.
3366 * config/rs6000/darwin.md: Ditto.
3367 * config/rs6000/rs6000.md: Ditto.
3368 * config/rs6000/vector.md: Ditto.
3369 * config/rs6000/vsx.md: Ditto.
3371 2018-02-21 Jason Merrill <jason@redhat.com>
3373 PR c++/84314 - ICE with templates and fastcall attribute.
3374 * attribs.c (build_type_attribute_qual_variant): Remove assert.
3376 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3378 * ipa-cp.c (determine_versionability): Fix comment typos.
3380 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3383 * ipa-cp.c (determine_versionability): Do not version functions caling
3386 2018-02-21 Martin Liska <mliska@suse.cz>
3389 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
3390 Add "native" as a possible value.
3391 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
3392 the macro when native cpu detection is available.
3394 2018-02-21 Martin Liska <mliska@suse.cz>
3397 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
3398 Add "native" as a possible value.
3399 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
3400 when native cpu detection is available.
3402 2018-02-21 Jakub Jelinek <jakub@redhat.com>
3403 Martin Sebor <msebor@redhat.com>
3405 PR tree-optimization/84478
3406 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
3408 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
3409 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
3410 support which is conservatively correct, for 2 only stay conservative
3411 for maxlen. Formatting and comment capitalization fixes. Add STRICT
3412 argument to the 2 argument get_range_strlen, adjust 6 arg
3413 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
3415 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
3416 (gimple_fold_builtin_strlen): Pass true as last argument to
3419 2018-02-20 Martin Sebor <msebor@redhat.com>
3422 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
3423 (builtin_memref::set_base_and_offset): Same. Handle inner references.
3424 (builtin_memref::builtin_memref): Factor out parts into
3425 set_base_and_offset and call it.
3427 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3430 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
3431 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
3432 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
3433 search at the associated MODE_INT.
3435 2018-02-20 Jeff Law <law@redhat.com>
3438 PR tree-optimization/81592
3440 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
3441 for range data rather than using global data.
3442 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
3443 range data rather than using global data.
3444 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
3445 pass it to children as needed.
3446 (struct directive::fmtresult): Similarly.
3447 (struct directive::set_width): Similarly.
3448 (struct directive::set_precision): Similarly.
3449 (format_integer, format_directive, parse_directive): Similarly.
3450 (format_none): Accept unnamed vr_values parameter.
3451 (format_percent, format_floating, format_character): Similarly.
3452 (format_string, format_plain): Similarly.
3453 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
3454 the EVRP range analyzer for range data rather than using global data.
3455 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
3456 gimple-ssa-evrp-analyze.h
3457 (class sprintf_dom_walker): Add after_dom_children member function.
3458 Add evrp_range_analyzer member.
3459 (sprintf_dom_walker::before_dom_children): Call into the EVRP
3460 range analyzer as needed.
3461 (sprintf_dom_walker::after_dom_children): New member function.
3462 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
3464 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
3465 (evrp_range_analyzer::pop_to_marker): Likewise.
3467 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3469 PR tree-optimization/84419
3470 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
3471 with the required type if its current type is compatible but
3474 2018-02-20 Jakub Jelinek <jakub@redhat.com>
3477 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
3478 after vectorization.
3480 2018-02-20 Martin Liska <mliska@suse.cz>
3483 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
3484 possible values if we don't have a hint.
3486 2018-02-20 Martin Liska <mliska@suse.cz>
3490 * final.c (shorten_branches): Build align_tab array with one
3492 * opts.c (finish_options): Add alignment option limit check.
3493 (MAX_CODE_ALIGN): Likewise.
3494 (MAX_CODE_ALIGN_VALUE): Likewise.
3495 * doc/invoke.texi: Document maximum allowed option value for
3496 all -falign-* options.
3498 2018-02-19 Jakub Jelinek <jakub@redhat.com>
3501 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
3502 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
3503 * var-tracking.c (emit_note_insn_var_location): Remove all references
3504 to NOTE_INSN_CALL_ARG_LOCATION.
3505 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
3506 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
3507 Use copy_rtx_if_shared.
3508 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
3509 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
3510 (dwarf2out_var_location): Remove handling of
3511 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
3513 * final.c (final_scan_insn): Remove all references to
3514 NOTE_INSN_CALL_ARG_LOCATION.
3515 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
3516 before dumping final insns.
3517 * except.c (emit_note_eh_region_end): Remove all references to
3518 NOTE_INSN_CALL_ARG_LOCATION.
3519 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
3520 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
3521 * config/arc/arc.c (hwloop_optimize): Likewise.
3522 * config/arm/arm.c (create_fix_barrier): Likewise.
3523 * config/s390/s390.c (s390_chunkify_start): Likewise.
3524 * config/sh/sh.c (find_barrier): Likewise.
3525 * config/i386/i386.c (rest_of_insert_endbranch,
3526 ix86_seh_fixup_eh_fallthru): Likewise.
3527 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
3528 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
3529 * config/frv/frv.c (frv_function_prologue): Likewise.
3530 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
3532 (note_outside_basic_block_p): Remove all references to
3533 NOTE_INSN_CALL_ARG_LOCATION.
3534 * gengtype.c (adjust_field_rtx_def): Likewise.
3535 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
3537 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
3538 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
3541 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
3544 PR tree-optimization/84452
3545 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
3546 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
3549 2018-02-19 Martin Liska <mliska@suse.cz>
3552 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
3554 2018-02-19 Martin Liska <mliska@suse.cz>
3555 Richard Sandiford <richard.sandiford@linaro.org>
3557 PR tree-optimization/82491
3558 * gimple-fold.c (get_base_constructor): Make earlier bail out
3561 2018-02-19 Carl Love <cel@us.ibm.com>
3563 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
3564 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
3566 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
3567 P8V_BUILTIN_VEC_NEG.
3569 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
3571 * config/rl78/rl78.md (movdf): New define expand.
3573 2018-02-19 Martin Liska <mliska@suse.cz>
3576 * doc/invoke.texi: Fix typo.
3577 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
3579 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
3581 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
3582 handle rs6000_single_float and rs6000_double_float specially for
3585 2018-02-16 Jeff Law <law@redhat.com>
3587 * config/rx/rx.c (add_pop_cfi_notes): New function.;
3590 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3593 * ipa-inline.c (inline_small_functions): Fix a typo.
3595 2018-02-16 Nathan Sidwell <nathan@acm.org>
3597 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
3599 2018-02-16 Carl Love <cel@us.ibm.com>
3601 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
3602 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
3603 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
3604 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
3605 expansion to P8V_BUILTIN_VEC_FLOAT2.
3607 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
3609 PR rtl-optimization/70023
3610 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
3611 src_regno into account.
3613 2018-02-16 Carl Love <cel@us.ibm.com>
3615 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
3616 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
3617 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
3618 * config/rs6000/rs6000.c: Remove case statements for
3619 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
3620 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
3621 and P9V_BUILTIN_VEC_VINSERT4B.
3622 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
3623 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
3624 * config/rs6000/vsx.md:
3625 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
3628 2018-02-16 Carl Love <cel@us.ibm.com>
3630 * config/rs6000/altivec.h: Add builtin names vec_extract4b
3632 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
3634 * config/rs6000/rs6000-c.c: Add the definitions for
3635 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
3636 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
3637 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
3638 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
3639 definition for insert4b and define insn *insert3b_internal.
3640 * doc/extend.texi: Add documentation for vec_extract4b.
3642 2018-02-16 Nathan Sidwell <nathan@acm.org>
3644 * doc/extend.texi (Backwards Compatibility): Mention friend
3645 injection. Note for-scope is deprecated.
3646 * doc/invoke.texi (-ffriend-injection): Deprecate.
3648 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
3650 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
3651 that moved to I2, also allow destinations that are a paradoxical
3652 subreg (instead of a normal reg).
3654 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
3657 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
3660 2018-02-16 Richard Biener <rguenther@suse.de>
3662 PR tree-optimization/84037
3663 PR tree-optimization/84016
3665 * config/i386/i386.c (ix86_builtin_vectorization_cost):
3666 Adjust vec_construct for the fact we need additional higher latency
3667 128bit inserts for AVX256 and AVX512 vector builds.
3668 (ix86_add_stmt_cost): Scale vector construction cost for
3671 2018-02-16 Richard Biener <rguenther@suse.de>
3673 PR tree-optimization/84417
3674 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
3675 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
3676 (non_rewritable_lvalue_p): Likewise, use poly-ints.
3678 2018-02-16 Martin Liska <mliska@suse.cz>
3681 * internal-fn.def (ASAN_CHECK): Set proper flags.
3682 (ASAN_MARK): Likewise.
3684 2018-02-16 Julia Koval <julia.koval@intel.com>
3686 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
3687 from PTA_CANNONLAKE.
3689 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3692 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
3693 Use ++iter rather than iter++ for std::list iterators.
3694 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
3695 defer deleting them until all nodes in the forest are processed. Do
3696 free even leaf nodes. Change to_process into auto_vec.
3699 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
3700 * vec.h (vec_default_construct): Use memset instead of placement new
3701 if BROKEN_VALUE_INITIALIZATION is defined.
3702 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
3703 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
3706 PR rtl-optimization/83723
3707 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
3708 * lra.c (lra_substitute_pseudo): Likewise. If true, use
3709 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
3711 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
3713 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
3715 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
3717 PR rtl-optimization/81443
3718 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
3719 from inner REGs to paradoxical SUBREGs.
3721 2018-02-16 Richard Biener <rguenther@suse.de>
3723 PR tree-optimization/84399
3724 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3725 For operands we can analyze at their definition make sure we can
3726 analyze them at each use as well.
3728 2018-02-16 Richard Biener <rguenther@suse.de>
3730 PR tree-optimization/84190
3731 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
3732 volatile accesses if the decl isn't volatile.
3734 2018-02-15 Jason Merrill <jason@redhat.com>
3736 PR c++/84314 - ICE with templates and fastcall attribute.
3737 * attribs.c (build_type_attribute_qual_variant): Don't clobber
3738 TYPE_CANONICAL on an existing type.
3740 2018-02-15 Jakub Jelinek <jakub@redhat.com>
3742 PR tree-optimization/84383
3743 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
3744 dstoff nor call operand_equal_p if dstbase is NULL.
3746 PR tree-optimization/84334
3747 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
3748 also a CONSTANT_CLASS_P, punt.
3750 2018-02-14 Jim Wilson <jimw@sifive.com>
3752 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
3753 first SMALL_OPERAND check. New local min_second_step. Move assert
3754 to where locals are set. Add TARGET_RVC support.
3755 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
3757 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
3759 * doc/invoke.texi: Correct -Wformat-overflow code sample.
3761 2018-02-14 Martin Sebor <msebor@redhat.com>
3763 PR tree-optimization/83698
3764 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
3765 arrays constrain the offset range to their bounds.
3766 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
3767 (builtin_access::overlap): Avoid setting the size of overlap if it's
3769 (maybe_diag_overlap): Also consider arrays when deciding what values
3770 of offsets to include in diagnostics.
3772 2018-02-14 Martin Sebor <msebor@redhat.com>
3775 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
3776 that correspond to the kind of a declaration.
3778 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
3781 * config/pa/pa.md: Load address of PIC label using the linkage table
3782 if the label is nonlocal.
3784 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
3786 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
3787 warning message if user requests -maltivec=be.
3788 * doc/invoke.texi: Document deprecation of -maltivec=be.
3790 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
3793 * config/rs6000/rs6000-c.c: Update definitions for
3794 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
3795 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
3797 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3800 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
3801 add _get_ssp intrinsics. Remove argument from
3802 __builtin_ia32_rdssp[d|q].
3803 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
3804 * config/i386/i386-builtin.def: Remove argument from
3805 __builtin_ia32_rdssp[d|q].
3806 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
3807 ix86_expand_special_args_builtin for _rdssp[d|q].
3808 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
3809 Clear register before usage.
3810 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
3811 Add documentation for new _get_ssp and _inc_ssp intrinsics.
3813 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
3815 PR tree-optimization/84357
3816 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
3817 operand 1 of an ARRAY_REF too.
3819 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
3822 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
3823 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3825 (set_of_reg): New struct.
3826 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
3827 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
3828 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3830 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
3831 Split into bitclr, bitset, bitinvert patterns if appropriate.
3832 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
3833 use rx_fuse_in_memory_bitop.
3834 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
3835 to named insn, correct maximum insn length.
3837 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
3840 * machmode.def: Define a complex mode for PARTIAL_INT.
3841 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
3843 * doc/rtl.texi: Document CSPImode.
3844 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
3846 (msp430_hard_regno_nregs_with_padding): Likewise.
3848 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
3851 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
3853 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
3855 PR rtl-optimization/84169
3856 * combine.c (try_combine): New variable split_i2i3. Set it to true if
3857 we generated a parallel as new i3 and we split that to new i2 and i3
3858 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
3859 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
3860 those to i2, not i1. Partially rewrite this scan code.
3862 2018-02-13 Jakub Jelinek <jakub@redhat.com>
3865 * stor-layout.c (place_field): For variable length fields, adjust
3866 offset_align afterwards not just based on the field's alignment,
3867 but also on the size.
3870 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
3871 of exps and logs in the use_exp2 case.
3873 2018-02-13 Jeff Law <law@redhat.com>
3875 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
3878 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
3881 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
3885 * common.opt (gas-loc-support, gas-locview-support): New.
3886 (ginline-points, ginternal-reset-location-views): New.
3887 * doc/invoke.texi: Document them. Use @itemx where intended.
3888 (gvariable-location-views): Adjust.
3889 * target.def (reset_location_view): New.
3890 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
3891 (TARGET_RESET_LOCATION_VIEW): New.
3892 * doc/tm.texi: Rebuilt.
3893 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
3894 (dwarf2out_default_as_locview_support): New.
3895 (output_asm_line_debug_info): Use option variables.
3896 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
3897 (output_loc_list): Likewise.
3898 (add_high_low_attributes): Check option variables.
3899 Don't output entry view attribute in strict mode.
3900 (gen_inlined_subroutine_die): Check option variables.
3901 (dwarf2out_inline_entry): Likewise.
3902 (init_sections_and_labels): Likewise.
3903 (dwarf2out_early_finish): Likewise.
3904 (maybe_reset_location_view): New, from...
3905 (dwarf2out_var_location): ... here. Call it.
3906 * debug.h (dwarf2out_default_as_loc_support): Declare.
3907 (dwarf2out_default_as_locview_support): Declare.
3908 * hooks.c (hook_int_rtx_insn_0): New.
3909 * hooks.h (hook_int_rtx_insn_0): Declare.
3910 * toplev.c (process_options): Take -gas-loc-support and
3911 -gas-locview-support from dwarf2out. Enable
3912 -gvariable-location-views by default only with locview
3913 assembler support. Enable -ginternal-reset-location-views by
3914 default only if the target defines the corresponding hook.
3915 Enable -ginline-points by default if location views are
3916 enabled; force it disabled if statement frontiers are
3918 * tree-inline.c (expand_call_inline): Check option variables.
3919 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
3921 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
3923 PR tree-optimization/84321
3924 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
3925 handling. Also check whether the anti-range contains any values
3926 that satisfy the mask; switch to a VR_RANGE if not.
3928 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
3931 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
3933 2018-02-13 Martin Jambor <mjambor@suse.cz>
3936 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
3937 of call statements, also set location of a load to a temporary.
3939 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
3941 * config/rl78/rl78.c (add_vector_labels): New function.
3942 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
3943 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
3944 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
3945 which checks that no arguments are passed.
3946 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
3947 * doc/extend.texi: Documentation for the new attribute.
3949 2018-02-13 Andreas Schwab <schwab@suse.de>
3951 * config/riscv/linux.h (CPP_SPEC): Define.
3953 2018-02-13 Jakub Jelinek <jakub@redhat.com>
3956 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
3957 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
3958 OPTION_MASK_ISA_AES as first argument to def_builtin_const
3959 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
3960 instead of OPTION_MASK_ISA_PCLMUL as first argument to
3961 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
3962 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
3963 temporarily for AES and PCLMUL builtins.
3965 PR tree-optimization/84339
3966 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
3967 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
3971 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
3972 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
3973 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
3975 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
3977 * omp-simd-clone.h: New file.
3978 * omp-simd-clone.c: Include omp-simd-clone.h.
3979 (expand_simd_clones): No longer static.
3980 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
3981 cgraph.h and omp-simd-clone.h.
3982 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
3983 (vect_recog_widen_shift_pattern): Formatting fix.
3984 (vect_pattern_recog_1): Don't check optab for calls.
3987 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
3988 operands[2] into a REG before using gen_lowpart on it.
3990 2018-02-12 Jeff Law <law@redhat.com>
3993 * config/sh/sh.c (find_barrier): Consider a sibling call
3996 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
3997 successfully back substituting a reg.
3999 2018-02-12 Richard Biener <rguenther@suse.de>
4001 PR tree-optimization/84037
4002 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
4003 parameter, move visited init to caller.
4004 (vect_slp_analyze_operations): Separate cost from validity
4005 check, initialize visited once for all instances.
4006 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
4008 * tree-vect-stmts.c (vect_model_simple_cost): Make early
4010 (vect_model_promotion_demotion_cost): Likewise.
4011 (vectorizable_bswap): Guard cost modeling with !slp_node
4012 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
4014 (vectorizable_call): Likewise.
4015 (vectorizable_conversion): Likewise.
4016 (vectorizable_assignment): Likewise.
4017 (vectorizable_shift): Likewise.
4018 (vectorizable_operation): Likewise.
4019 (vectorizable_store): Likewise.
4020 (vectorizable_load): Likewise.
4021 (vectorizable_condition): Likewise.
4022 (vectorizable_comparison): Likewise.
4024 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
4027 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
4028 (ASAN_MARK): Fix fnspec to account for return value, change pointer
4029 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
4031 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4034 * params.def (inline-min-speedup): Increase from 8 to 15.
4035 (max-inline-insns-auto): Decrease from 40 to 30.
4036 * ipa-split.c (consider_split): Add some buffer for function to
4037 be considered inlining candidate.
4038 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
4041 2018-02-12 Richard Biener <rguenther@suse.de>
4043 PR tree-optimization/84037
4044 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
4045 matched stmts if we cannot swap the non-matched ones.
4047 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4049 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
4050 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
4051 _mm_maskz_scalef_round_ss): New intrinsics.
4052 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
4053 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
4054 __builtin_ia32_scalefss_round): Remove.
4055 (__builtin_ia32_scalefsd_mask_round,
4056 __builtin_ia32_scalefss_mask_round): New intrinsics.
4057 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
4058 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
4059 ((match_operand:VF_128 2 "<round_nimm_predicate>"
4060 "<round_constraint>")): Changed to ...
4061 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
4062 "<round_scalar_constraint>")): ... this.
4063 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
4064 %0, %1, %2<round_op3>}"): Changed to ...
4065 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
4066 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
4067 %2<round_scalar_mask_op3>}"): ... this.
4068 * config/i386/subst.md (round_scalar_nimm_predicate): New.
4070 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4072 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
4073 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
4074 (_mm_maskz_sqrt_round_ss): New intrinsics.
4075 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
4076 (__builtin_ia32_sqrtsd_mask_round)
4077 (__builtin_ia32_sqrtss_mask_round): New builtins.
4078 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
4079 (__builtin_ia32_sqrtss_round): Remove.
4080 (__builtin_ia32_sqrtsd_mask_round)
4081 (__builtin_ia32_sqrtss_mask_round): New builtins.
4082 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
4083 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
4084 ((match_operand:VF_128 1 "vector_operand"
4085 "xBm,<round_constraint>")): Changed to ...
4086 ((match_operand:VF_128 1 "vector_operand"
4087 "xBm,<round_scalar_constraint>")): ... this.
4088 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
4089 %0, %2, %<iptr>1<round_op3>}): Changed to ...
4090 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
4091 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
4092 %<iptr>1<round_scalar_mask_op3>}): ... this.
4093 ((set_attr "prefix" "<round_prefix>")): Changed to ...
4094 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
4096 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
4099 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
4100 Cast vec_cmpeq result to correct type.
4101 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
4102 Cast vec_cmpgt result to correct type.
4104 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
4106 * final.c (final_scan_insn_1): Renamed from...
4107 (final_scan_insn): ... this. New wrapper, to recover
4108 seen from the outermost call in recursive ones.
4109 * config/sparc/sparc.c (output_return): Drop seen from call.
4110 (output_sibcall): Likewise.
4111 * config/visium/visium.c (output_branch): Likewise.
4113 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
4115 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
4118 2018-02-10 Alan Modra <amodra@gmail.com>
4121 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
4122 Specify LR as an input.
4124 2018-02-10 Jakub Jelinek <jakub@redhat.com>
4127 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
4128 remove_member_access_dummy_vars): New functions.
4129 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
4130 lower_omp_1, execute_lower_omp): Use them.
4132 PR rtl-optimization/84308
4133 * shrink-wrap.c (spread_components): Release todo vector.
4135 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
4137 PR rtl-optimization/57193
4138 * ira-color.c (struct allocno_color_data): Add member
4139 conflict_allocno_hard_prefs.
4140 (update_conflict_allocno_hard_prefs): New.
4141 (bucket_allocno_compare_func): Add a preference based on
4142 conflict_allocno_hard_prefs.
4143 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
4144 (color_allocnos): Remove a dead code. Initiate
4145 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
4147 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4150 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
4151 constraint from =wa to wa. Avoid a subreg on the output operand,
4152 instead use a pseudo and subreg it in a move.
4153 (p9_xxbrd_<mode>): Changed to ...
4154 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
4155 (p9_xxbrd_v2df): New expander.
4156 (p9_xxbrw_<mode>): Changed to ...
4157 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
4158 (p9_xxbrw_v4sf): New expander.
4160 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4162 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
4164 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
4167 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
4168 multiply in 32-bit mode.
4169 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
4170 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
4173 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4175 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
4176 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
4177 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
4178 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
4180 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
4183 * dwarf2out.c (is_trivial_indirect_ref): New function.
4184 (dwarf2out_late_global_decl): Do not generate a location
4185 attribute for variables that have a non-trivial DECL_VALUE_EXPR
4186 and that are not defined in the current unit.
4188 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4190 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
4191 instead of a libcall for UNORDERED.
4193 2018-02-09 Tamar Christina <tamar.christina@arm.com>
4196 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
4197 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
4199 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4202 * config/s390/s390.c (s390_set_current_function): Invoke
4203 s390_indirect_branch_settings also if fndecl didn't change.
4205 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4207 * config/rs6000/rs6000.md (blockage): Set length to zero.
4209 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4211 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
4213 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4216 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
4217 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
4221 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
4222 PARALLEL incoming that failed vt_get_decl_and_offset check.
4225 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
4226 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
4228 (get_variable_section): For decls in named .bss* sections pass true as
4229 second argument to bss_initializer_p.
4231 2018-02-09 Marek Polacek <polacek@redhat.com>
4232 Jakub Jelinek <jakub@redhat.com>
4235 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
4236 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
4237 Sync some changes from cxx_fold_indirect_ref.
4239 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4241 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
4243 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
4244 (BLOCK_INLINE_ENTRY_LABEL): New.
4245 (dwarf2out_var_location): Disregard inline entry markers.
4246 (inline_entry_data): New struct.
4247 (inline_entry_data_hasher): New hashtable type.
4248 (inline_entry_data_hasher::hash): New.
4249 (inline_entry_data_hasher::equal): New.
4250 (inline_entry_data_table): New variable.
4251 (add_high_low_attributes): Add DW_AT_entry_pc and
4252 DW_AT_GNU_entry_view attributes if a pending entry is found
4253 in inline_entry_data_table. Add old entry_pc attribute only
4254 if debug nonbinding markers are disabled.
4255 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
4256 markers are enabled.
4257 (block_within_block_p, dwarf2out_inline_entry): New.
4258 (dwarf2out_finish): Check that no entries remained in
4259 inline_entry_data_table.
4260 * final.c (reemit_insn_block_notes): Handle inline entry notes.
4261 (final_scan_insn, notice_source_line): Likewise.
4262 (rest_of_clean_state): Skip inline entry markers.
4263 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
4265 * gimple.c (gimple_build_debug_inline_entry): New.
4266 * gimple.h (enum gimple_debug_subcode): Add
4267 GIMPLE_DEBUG_INLINE_ENTRY.
4268 (gimple_build_debug_inline_entry): Declare.
4269 (gimple_debug_inline_entry_p): New.
4270 (gimple_debug_nonbind_marker_p): Adjust.
4271 * insn-notes.def (INLINE_ENTRY): New.
4272 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
4273 inline entry marker notes.
4274 (print_insn): Likewise.
4275 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
4276 (INSN_DEBUG_MARKER_KIND): Likewise.
4277 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
4278 * tree-inline.c (expand_call_inline): Build and insert
4279 debug_inline_entry stmt.
4280 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
4281 inline entry blocks early, if nonbind markers are enabled.
4282 (dump_scope_block): Dump fragment info.
4283 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
4284 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
4285 (gimple_build_debug_inline_entry): New.
4286 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
4287 Enable/disable inline entry points too.
4288 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
4289 (DEBUG_INSN): Describe inline entry markers.
4291 * common.opt (gvariable-location-views): New.
4292 (gvariable-location-views=incompat5): New.
4293 * config.in: Rebuilt.
4294 * configure: Rebuilt.
4295 * configure.ac: Test assembler for view support.
4296 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
4297 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
4298 * dwarf2out.c (var_loc_view): New typedef.
4299 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
4300 (dwarf2out_locviews_in_attribute): New.
4301 (dwarf2out_locviews_in_loclist): New.
4302 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
4303 (enum dw_line_info_opcode): Add LI_adv_address.
4304 (struct dw_line_info_table): Add view.
4305 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
4306 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
4307 (zero_view_p): New variable.
4308 (ZERO_VIEW_P): New macro.
4309 (output_asm_line_debug_info): New.
4310 (struct var_loc_node): Add view.
4311 (add_AT_view_list, AT_loc_list): New.
4312 (add_var_loc_to_decl): Add view param. Test it against last.
4313 (new_loc_list): Add view params. Record them.
4314 (AT_loc_list_ptr): Handle loc and view lists.
4315 (view_list_to_loc_list_val_node): New.
4316 (print_dw_val): Handle dw_val_class_view_list.
4317 (size_of_die): Likewise.
4318 (value_format): Likewise.
4319 (loc_list_has_views): New.
4320 (gen_llsym): Set vl_symbol too.
4321 (maybe_gen_llsym, skip_loc_list_entry): New.
4322 (dwarf2out_maybe_output_loclist_view_pair): New.
4323 (output_loc_list): Output view list or entries too.
4324 (output_view_list_offset): New.
4325 (output_die): Handle dw_val_class_view_list.
4326 (output_dwarf_version): New.
4327 (output_compilation_unit_header): Use it.
4328 (output_skeleton_debug_sections): Likewise.
4329 (output_rnglists, output_line_info): Likewise.
4330 (output_pubnames, output_aranges): Update version comments.
4331 (output_one_line_info_table): Output view numbers in asm comments.
4332 (dw_loc_list): Determine current endview, pass it to new_loc_list.
4333 Call maybe_gen_llsym.
4334 (loc_list_from_tree_1): Adjust.
4335 (add_AT_location_description): Create view list attribute if
4336 needed, check it's absent otherwise.
4337 (convert_cfa_to_fb_loc_list): Adjust.
4338 (maybe_emit_file): Call output_asm_line_debug_info for test.
4339 (dwarf2out_var_location): Reset views as needed. Precompute
4340 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
4341 attribute. Set view.
4342 (new_line_info_table): Reset next view.
4343 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
4344 (dwarf2out_source_line): Likewise. Output view resets and labels to
4345 the assembler, or select appropriate line info opcodes.
4346 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
4347 (optimize_string_length): Catch it. Adjust.
4348 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
4349 dw_val_class_view_list, and remove it if no longer needed.
4350 (hash_loc_list): Hash view numbers.
4351 (loc_list_hasher::equal): Compare them.
4352 (optimize_location_lists): Check whether a view list symbol is
4353 needed, and whether the locview attribute is present, and
4354 whether they match. Remove the locview attribute if no longer
4356 (index_location_lists): Call skip_loc_list_entry for test.
4357 (dwarf2out_finish): Call output_asm_line_debug_info for test.
4358 Use output_dwarf_version.
4359 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
4360 (struct dw_val_node): Add val_view_list.
4361 * final.c (SEEN_NEXT_VIEW): New.
4362 (set_next_view_needed): New.
4363 (clear_next_view_needed): New.
4364 (maybe_output_next_view): New.
4365 (final_start_function): Rename to...
4366 (final_start_function_1): ... this. Take pointer to FIRST,
4367 add SEEN parameter. Emit param bindings in the initial view.
4368 (final_start_function): Reintroduce SEEN-less interface.
4369 (final): Rename to...
4370 (final_1): ... this. Take SEEN parameter. Output final pending
4371 next view at the end.
4372 (final): Reintroduce seen-less interface.
4373 (final_scan_insn): Output pending next view before switching
4374 sections or ending a block. Mark the next view as needed when
4375 outputting variable locations. Notify debug backend of section
4376 changes, and of location view changes.
4377 (rest_of_handle_final): Adjust.
4378 * toplev.c (process_options): Autodetect value for debug variable
4379 location views option. Warn on incompat5 without -gdwarf-5.
4380 * doc/invoke.texi (gvariable-location-views): New.
4381 (gvariable-location-views=incompat5): New.
4382 (gno-variable-location-views): New.
4384 2018-02-08 David Malcolm <dmalcolm@redhat.com>
4386 PR tree-optimization/84136
4387 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
4388 that the result of find_edge is non-NULL.
4390 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4393 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
4394 storing integer register in SImode. Fix cost of 256 and 512
4395 byte aligned SSE register store.
4397 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4399 * config/i386/i386.c (ix86_multiplication_cost): Fix
4400 multiplication cost for TARGET_AVX512DQ.
4402 2018-02-08 Marek Polacek <polacek@redhat.com>
4404 PR tree-optimization/84238
4405 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
4408 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4410 PR tree-optimization/84265
4411 * tree-vect-stmts.c (vectorizable_store): Don't treat
4412 VMAT_CONTIGUOUS accesses as grouped.
4413 (vectorizable_load): Likewise.
4415 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4417 PR tree-optimization/81635
4418 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
4419 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
4420 (test_round_for_mask): New functions.
4421 (wide_int_cc_tests): Call test_round_for_mask.
4422 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
4423 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
4424 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
4425 range returned by get_range_info.
4427 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4430 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
4431 * symtab.c: Include builtins.h
4432 (symtab_node::output_to_lto_symbol_table_p): Move here
4433 from lto-streamer-out.c:output_symbol_p.
4434 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
4435 (output_symbol_p): Move all logic to symtab.c
4436 (produce_symtab): Update.
4438 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4440 * config/s390/s390-opts.h (enum indirect_branch): Define.
4441 * config/s390/s390-protos.h (s390_return_addr_from_memory)
4442 (s390_indirect_branch_via_thunk)
4443 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
4444 (enum s390_indirect_branch_type): Define.
4445 * config/s390/s390.c (struct s390_frame_layout, struct
4446 machine_function): Remove.
4447 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
4448 (indirect_branch_table_label_no, indirect_branch_table_name):
4450 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
4451 (enum s390_indirect_branch_option): Define.
4452 (s390_return_addr_from_memory): New function.
4453 (s390_handle_string_attribute): New function.
4454 (s390_attribute_table): Add new attribute handler.
4455 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
4456 (s390_indirect_branch_via_thunk): New function.
4457 (s390_indirect_branch_via_inline_thunk): New function.
4458 (s390_function_ok_for_sibcall): When jumping via thunk disallow
4459 sibling call optimization for non z10 compiles.
4460 (s390_emit_call): Force indirect branch target to be a single
4461 register. Add r1 clobber for non-z10 compiles.
4462 (s390_emit_epilogue): Emit return jump via return_use expander.
4463 (s390_reorg): Handle JUMP_INSNs as execute targets.
4464 (s390_option_override_internal): Perform validity checks for the
4465 new command line options.
4466 (s390_indirect_branch_attrvalue): New function.
4467 (s390_indirect_branch_settings): New function.
4468 (s390_set_current_function): Invoke s390_indirect_branch_settings.
4469 (s390_output_indirect_thunk_function): New function.
4470 (s390_code_end): Implement target hook.
4471 (s390_case_values_threshold): Implement target hook.
4472 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
4474 * config/s390/s390.h (struct s390_frame_layout)
4475 (struct machine_function): Move here from s390.c.
4476 (TARGET_INDIRECT_BRANCH_NOBP_RET)
4477 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
4478 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
4479 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
4480 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
4481 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
4482 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
4483 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
4484 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
4485 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
4486 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
4487 (mnemonic attribute): Add values which aren't recognized
4489 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
4490 pattern for branch conversion. Fix mnemonic attribute.
4491 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
4492 indirect branch via thunk if requested.
4493 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
4494 ("*indirect_jump"): Disable for branch conversion using out of
4496 ("indirect_jump_via_thunk<mode>_z10")
4497 ("indirect_jump_via_thunk<mode>")
4498 ("indirect_jump_via_inlinethunk<mode>_z10")
4499 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
4500 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
4501 ("casesi_jump_via_inlinethunk<mode>_z10")
4502 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
4503 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
4504 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
4505 ("*indirect2_jump"): Disable for branch conversion.
4506 ("casesi_jump"): Turn into expander and expand patterns for branch
4508 ("return_use"): New expander.
4509 ("*return"): Emit return via thunk and rename it to ...
4510 ("*return<mode>"): ... this one.
4511 * config/s390/s390.opt: Add new options and and enum for the
4514 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4516 * lra-constraints.c (match_reload): Unconditionally use
4517 gen_lowpart_SUBREG, rather than selecting between that
4518 and equivalent gen_rtx_SUBREG code.
4520 2018-02-08 Richard Biener <rguenther@suse.de>
4522 PR tree-optimization/84233
4523 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
4524 changed flag instead of boguously re-using phi_inserted.
4526 2018-02-08 Martin Jambor <mjambor@suse.cz>
4528 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
4529 static local variables.
4531 2018-02-08 Richard Biener <rguenther@suse.de>
4533 PR tree-optimization/84278
4534 * tree-vect-stmts.c (vectorizable_store): When looking for
4535 smaller vector types to perform grouped strided loads/stores
4536 make sure the mode is supported by the target.
4537 (vectorizable_load): Likewise.
4539 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4541 * config/aarch64/aarch64.c (aarch64_components_for_bb):
4542 Increase LDP/STP opportunities by adding adjacent callee-saves.
4544 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4546 PR rtl-optimization/84068
4547 PR rtl-optimization/83459
4548 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
4550 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
4552 PR tree-optimization/84224
4553 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
4554 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
4557 2018-02-07 Iain Sandoe <iain@codesourcery.com>
4560 * config/rs6000/altivec.md (*restore_world): Remove LR use.
4561 * config/rs6000/predicates.md (restore_world_operation): Adjust op
4562 count, remove one USE.
4564 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
4566 * doc/install.texi (Configuration): Document the
4567 --with-long-double-format={ibm,ieee} PowerPC configuration
4571 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
4572 Convert from define_expand to be define_insn_and_split. Rework
4573 float/double/_Float128 conversions to QI/HI/SImode to work with
4574 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
4575 conversions to QI/HImode types did a store and then a load to
4576 truncate the value. For conversions to VSX registers, don't split
4577 the insn, instead emit the code directly. Use the code iterator
4578 any_fix to combine signed and unsigned conversions.
4579 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
4580 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
4581 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4582 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
4583 (fix_<mode>di2_hw): Likewise.
4584 (fixuns_<mode>di2_hw): Likewise.
4585 (fix_<mode>si2_hw): Likewise.
4586 (fixuns_<mode>si2_hw): Likewise.
4587 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
4588 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4589 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
4590 fix<uns>_trunc<SFDF:mode>si2_p8.
4591 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
4593 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
4594 (fix<uns>_<mode>_mem): Likewise.
4595 (fctiw<u>z_<mode>_mem): Likewise.
4596 (fix<uns>_<mode>_mem): Likewise.
4597 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
4598 the register allocator from doing a direct move to the GPRs to do
4599 a store, and instead use the ISA 3.0 store byte/half-word from
4600 vector register instruction. For IEEE 128-bit floating point,
4601 also optimize stores of 32-bit ints.
4602 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
4604 2018-02-07 Alan Hayward <alan.hayward@arm.com>
4606 * genextract.c (push_pathstr_operand): New function to support
4608 (walk_rtx): Call push_pathstr_operand.
4609 (print_path): Support [a-zA-Z].
4611 2018-02-07 Richard Biener <rguenther@suse.de>
4613 PR tree-optimization/84037
4614 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
4615 (cse_and_gimplify_to_preheader): Declare.
4616 (vect_get_place_in_interleaving_chain): Likewise.
4617 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4619 (_loop_vec_info::~_loop_vec_info): Delete it.
4620 (cse_and_gimplify_to_preheader): New function.
4621 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
4622 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
4623 (vectorizable_load): Likewise. For grouped stores always base
4624 the IV on the first element.
4625 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
4626 condition before gimplifying.
4628 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4630 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
4631 *DIV_EXPR and *MOD_EXPR.
4633 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
4636 * config/i386/i386.c (ix86_option_override_internal): Mask out
4637 the CF_SET bit when checking -fcf-protection.
4639 2018-02-07 Tom de Vries <tom@codesourcery.com>
4642 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
4645 2018-02-07 Richard Biener <rguenther@suse.de>
4647 PR tree-optimization/84204
4648 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
4651 PR tree-optimization/84205
4652 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
4653 special-case isl_ast_op_zdiv_r.
4655 PR tree-optimization/84223
4656 * graphite-scop-detection.c (gather_bbs::before_dom_children):
4657 Only add conditions from within the region.
4658 (gather_bbs::after_dom_children): Adjust.
4660 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
4663 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
4664 * config/avr/avr.md: Only post-reload split REG-REG moves if
4665 either register is GENERAL_REG_P.
4667 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4669 PR tree-optimization/84235
4670 * tree-ssa-scopedtables.c
4671 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
4672 if the subtraction is performed in floating point type where NaNs are
4673 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
4674 build 1. Formatting fix.
4676 2018-02-06 Jakub Jelinek <jakub@redhat.com>
4679 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
4680 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
4681 and skip it regardless of bb boundaries. Use CALL_P macro,
4682 don't test INSN_P (insn) together with CALL_P or JUMP_P check
4683 unnecessarily, formatting fix.
4685 2018-02-06 Michael Collison <michael.collison@arm.com>
4687 * config/arm/thumb2.md:
4688 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
4689 (*thumb_mov_notscc): Ditto.
4691 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
4694 * config/rs6000/rs6000.md (su code attribute): Use "u" for
4695 unsigned_fix, not "s".
4697 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4699 * configure.ac (gcc_fn_eh_frame_ro): New function.
4700 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
4701 correct .eh_frame permissions.
4702 * configure: Regenerate.
4704 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
4706 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
4709 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4711 * config/rs6000/rs6000.c (rs6000_option_override_internal):
4712 Display warning message for -mno-speculate-indirect-jumps.
4714 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
4716 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
4718 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
4720 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
4722 PR tree-optimization/84225
4723 * tree-eh.c (find_trapping_overflow): Only call
4724 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
4726 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4729 * config/i386/i386.c: Reimplement the check of possible options
4730 -mibt/-mshstk conbination. Change error messages.
4731 * doc/invoke.texi: Fix a typo: remove extra '='.
4733 2018-02-06 Marek Polacek <polacek@redhat.com>
4735 PR tree-optimization/84228
4736 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
4738 2018-02-06 Tamar Christina <tamar.christina@arm.com>
4741 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
4742 emitted arch directives.
4743 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
4744 __ARM_FEATURE_COPROC before changing architectures.
4746 2018-02-06 Richard Biener <rguenther@suse.de>
4748 * config/i386/i386.c (print_reg): Fix typo.
4749 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
4751 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
4753 * configure: Regenerate.
4755 2018-02-05 Martin Sebor <msebor@redhat.com>
4757 PR tree-optimization/83369
4758 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
4761 2018-02-05 Martin Liska <mliska@suse.cz>
4763 * doc/invoke.texi: Cherry-pick upstream r323995.
4765 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
4767 * ira.c (ira_init_register_move_cost): Adjust comment.
4769 2018-02-05 Martin Liska <mliska@suse.cz>
4771 PR gcov-profile/84137
4772 * doc/gcov.texi: Fix typo in documentation.
4774 2018-02-05 Martin Liska <mliska@suse.cz>
4776 PR gcov-profile/83879
4777 * doc/gcov.texi: Document necessity of --dynamic-list-data when
4778 using dlopen functionality.
4780 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
4782 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
4783 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
4784 _mm_maskz_range_ss, _mm_mask_range_round_ss,
4785 _mm_maskz_range_round_ss): New intrinsics.
4786 (__builtin_ia32_rangesd128_round)
4787 (__builtin_ia32_rangess128_round): Remove.
4788 (__builtin_ia32_rangesd128_mask_round,
4789 __builtin_ia32_rangess128_mask_round): New builtins.
4790 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
4791 __builtin_ia32_rangess128_round): Remove.
4792 (__builtin_ia32_rangesd128_mask_round,
4793 __builtin_ia32_rangess128_mask_round): New builtins.
4794 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
4795 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
4796 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
4797 "<round_saeonly_constraint>")): Changed to ...
4798 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
4799 "<round_saeonly_scalar_constraint>")): ... this.
4800 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
4801 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
4802 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
4803 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
4804 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
4806 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
4808 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
4810 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
4811 Remove all values except native, 8540 and 8548.
4813 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
4815 * config/i386/i386.c (ix86_output_function_return): Pass
4816 INVALID_REGNUM, instead of -1, as invalid register number to
4817 indirect_thunk_name and output_indirect_thunk.
4819 2018-02-02 Julia Koval <julia.koval@intel.com>
4821 * config.gcc: Add -march=icelake.
4822 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
4823 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
4824 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
4825 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
4826 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
4827 (processor_target_table): Add icelake.
4828 (ix86_option_override_internal): Handle new PTAs.
4829 (get_builtin_code_for_version): Handle icelake.
4830 (M_INTEL_COREI7_ICELAKE): New.
4831 (fold_builtin_cpu): Handle icelake.
4832 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
4833 * doc/invoke.texi: Add -march=icelake.
4835 2018-02-02 Julia Koval <julia.koval@intel.com>
4837 * config/i386/i386.c (ix86_option_override_internal): Change flags type
4838 to wide_int_bitmask.
4839 * wide-int-bitmask.h: New.
4841 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4844 * config/i386/i386.md: Replace Pmode with word_mode in
4845 builtin_setjmp_setup and builtin_longjmp to support x32.
4847 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
4851 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
4853 (rs6000_supported_cpu_names): New static variable.
4854 (linux_cpu_translation_table): Likewise.
4855 (elf_platform) <cpu>: Define new static variable and use it.
4856 Translate kernel AT_PLATFORM name to canonical name if needed.
4857 Error if platform name is unknown.
4859 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
4862 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
4864 2018-02-01 Jeff Law <law@redhat.com>
4867 * config/i386/i386.c (release_scratch_register_on_entry): Add new
4868 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
4869 the scratch if RELEASE_VIA_POP is false.
4870 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
4871 If we have to save a temporary register, decrement SIZE appropriately.
4872 Pass new arguments to release_scratch_register_on_entry.
4873 (ix86_adjust_stack_and_probe): Likewise.
4874 (ix86_emit_probe_stack_range): Pass new arguments to
4875 release_scratch_register_on_entry.
4877 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
4879 PR rtl-optimization/84157
4880 * combine.c (change_zero_ext): Use REG_P predicate in
4881 front of HARD_REGISTER_P predicate.
4883 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
4885 * config/avr/avr.c (avr_option_override): Move disabling of
4886 -fdelete-null-pointer-checks to...
4887 * common/config/avr/avr-common.c (avr_option_optimization_table):
4890 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4892 PR tree-optimization/81635
4893 * tree-data-ref.c (split_constant_offset_1): For types that
4894 wrap on overflow, try to use range info to prove that wrapping
4897 2018-02-01 Renlin Li <renlin.li@arm.com>
4900 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
4902 (aarch64_register_move_cost): Likewise.
4903 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
4905 (REG_CLASS_NAMES): Likewise.
4906 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
4907 TAILCALL_ADDR_REGS. Remove IP registers.
4908 * config/aarch64/aarch64.md (Ucs): Update register constraint.
4910 2018-02-01 Richard Biener <rguenther@suse.de>
4912 * domwalk.h (dom_walker::dom_walker): Add additional constructor
4913 for specifying RPO order and allow NULL for that.
4914 * domwalk.c (dom_walker::dom_walker): Likewise.
4915 (dom_walker::walk): Handle NULL RPO order.
4916 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
4918 (rewrite_update_dom_walker): Likewise.
4919 (mark_def_dom_walker): Likewise.
4921 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4923 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
4924 (aarch64_maybe_expand_sve_subreg_move): Declare.
4925 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
4926 * config/aarch64/predicates.md (aarch64_any_register_operand): New
4928 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
4929 that are semantically a reverse operation.
4930 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
4931 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
4932 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
4934 (aarch64_can_change_mode_class): For big-endian, forbid changes
4935 between two SVE modes if they have different element sizes.
4937 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4939 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
4940 the TImode handling for big-endian targets.
4942 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4944 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
4945 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
4947 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
4948 Remove BSWAP handing for big-endian targets and use the form of
4949 LD1RQ appropariate for the mode.
4951 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4953 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
4954 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
4957 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4960 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
4961 check for operands that need to go through aarch64_sve_reload_be.
4963 2018-02-01 Jakub Jelinek <jakub@redhat.com>
4965 PR tree-optimization/81661
4966 PR tree-optimization/84117
4967 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
4968 * tree-eh.c: Include gimplify.h.
4969 (find_trapping_overflow, replace_trapping_overflow,
4970 rewrite_to_non_trapping_overflow): New functions.
4971 * tree-vect-loop.c: Include tree-eh.h.
4972 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
4973 * tree-data-ref.c: Include tree-eh.h.
4974 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
4976 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
4978 PR rtl-optimization/84123
4979 * combine.c (change_zero_ext): Check if hard register satisfies
4980 can_change_dest_mode before calling gen_lowpart_SUBREG.
4982 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
4985 * ira.c (ira_init_register_move_cost): Remove assert.
4987 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4989 PR rtl-optimization/84071
4990 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
4991 * doc/tm.texi: Regenerate.
4993 2018-01-31 Richard Biener <rguenther@suse.de>
4995 PR tree-optimization/84132
4996 * tree-data-ref.c (analyze_miv_subscript): Properly
4997 check whether evolution_function_is_affine_multivariate_p
4998 before calling gcd_of_steps_may_divide_p.
5000 2018-01-31 Julia Koval <julia.koval@intel.com>
5003 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
5004 * config/i386/i386.md (rdpid_rex64) New.
5005 (rdpid): Make 32bit only.
5007 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
5010 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
5011 an IDENTIFIER_NODE for FUNCTION_TYPE's.
5013 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5016 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5018 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5020 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5022 PR rtl-optimization/84071
5023 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
5024 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
5026 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5028 * config/arc/arc.c (arc_handle_aux_attribute): New function.
5029 (arc_attribute_table): Add 'aux' attribute.
5030 (arc_in_small_data_p): Consider aux like variables.
5031 (arc_is_aux_reg_p): New function.
5032 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
5033 (arc_get_aux_arg): New function.
5034 (prepare_move_operands): Handle aux-register access.
5035 (arc_handle_aux_attribute): New function.
5036 * doc/extend.texi (ARC Variable attributes): Add subsection.
5038 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5040 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
5041 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
5042 (arc_attribute_table): Add 'uncached' attribute.
5043 (arc_print_operand): Print '.di' flag for uncached memory
5045 (arc_in_small_data_p): Do not consider for small data the uncached
5047 (arc_is_uncached_mem_p): New function.
5048 * config/arc/predicates.md (compact_store_memory_operand): Check
5049 for uncached memory accesses.
5050 (nonvol_nonimm_operand): Likewise.
5051 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
5053 2018-01-31 Jakub Jelinek <jakub@redhat.com>
5056 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
5057 falign-loops=): Add Optimization flag.
5059 2018-01-30 Jeff Law <law@redhat.com>
5062 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
5063 INT_REGISTERS_SAVED. Check it prior to calling
5064 get_scratch_register_on_entry.
5065 (ix86_adjust_stack_and_probe): Similarly.
5066 (ix86_emit_probe_stack_range): Similarly.
5067 (ix86_expand_prologue): Corresponding changes.
5069 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5072 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
5073 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
5075 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
5078 * lra-constraints.c (curr_insn_transform): Process AND in the
5081 2018-01-30 Jakub Jelinek <jakub@redhat.com>
5083 PR rtl-optimization/83986
5084 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
5085 dependence against last_pending_memory_flush in addition to
5088 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
5090 PR tree-optimization/81611
5091 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
5094 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5097 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
5100 2018-01-30 Richard Biener <rguenther@suse.de>
5101 Jakub Jelinek <jakub@redhat.com>
5103 PR tree-optimization/84111
5104 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
5105 inner loops added during recursion, as they don't have up-to-date
5108 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5111 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
5112 (can_inline_edge_by_limits_p): ... here.
5113 (can_early_inline_edge_p, check_callers,
5114 update_caller_keys, update_callee_keys, recursive_inlining,
5115 add_new_edges_to_heap, speculation_useful_p,
5116 inline_small_functions,
5117 inline_small_functions, flatten_function,
5118 inline_to_all_callers_1): Update.
5120 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5122 * profile-count.c (profile_count::combine_with_ipa_count): Handle
5125 2018-01-30 Richard Biener <rguenther@suse.de>
5127 PR tree-optimization/83008
5128 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
5129 invariant and constant vector uses in stmts when they need
5132 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5135 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
5136 * configure: Regenerate.
5138 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5140 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
5142 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
5143 Use gen_rtx_REG rather than gen_lowpart.
5145 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5147 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
5148 rather than 0 when creating partial subregs.
5150 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5152 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
5155 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
5158 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
5159 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
5160 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
5161 flags. This restores the settings used before the 2017-07-24.
5162 Turning off pre increment/decrement/modify allows IVOPTS to
5163 optimize DF/SF loops where the index is an int.
5165 2018-01-29 Richard Biener <rguenther@suse.de>
5166 Kelvin Nilsen <kelvin@gcc.gnu.org>
5169 * tree-vect-stmts.c (vectorizable_call): Don't call
5170 targetm.vectorize_builtin_md_vectorized_function if callee is
5173 2018-01-22 Carl Love <cel@us.ibm.com>
5175 * doc/extend.tex: Fix typo in second arg in
5176 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
5178 2018-01-29 Richard Biener <rguenther@suse.de>
5180 PR tree-optimization/84086
5181 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
5182 (flush_ssaname_freelist): When SSA names were released reset
5183 the SCEV hash table.
5185 2018-01-29 Richard Biener <rguenther@suse.de>
5187 PR tree-optimization/84057
5188 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
5189 removed paths when removing edges.
5191 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
5193 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
5194 -mfunction-return=@var{choice}.
5196 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
5199 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
5201 (layout::print_source_line): Likewise.
5202 (test_get_line_width_without_trailing_whitespace): Add test cases.
5204 2018-01-27 Jakub Jelinek <jakub@redhat.com>
5207 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
5210 2018-01-26 Jim Wilson <jimw@sifive.com>
5212 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
5214 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
5217 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5219 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
5220 and CMP + SUB-immediate -> SUBS.
5222 2018-01-26 Martin Sebor <msebor@redhat.com>
5224 PR tree-optimization/83896
5225 * tree-ssa-strlen.c (get_string_len): Rename...
5226 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
5227 Avoid assuming length is constant.
5228 (handle_char_store): Use HOST_WIDE_INT for string length.
5230 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
5233 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
5234 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
5236 2018-01-26 Richard Biener <rguenther@suse.de>
5238 PR rtl-optimization/84003
5239 * dse.c (record_store): Only record redundant stores when
5240 the earlier store aliases at least all accesses the later one does.
5242 2018-01-26 Jakub Jelinek <jakub@redhat.com>
5244 PR rtl-optimization/83985
5245 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
5246 REG_CFA_RESTORE insns.
5247 (delete_unmarked_insns): Don't ignore separate shrink wrapping
5248 REG_CFA_RESTORE insns here.
5251 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
5252 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
5254 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5256 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
5257 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
5258 (arc_init): Likewise.
5259 (arc_override_options): Likewise.
5260 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
5262 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
5264 * config/arc/arc.h (TARGET_DBNZ): Define.
5265 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
5266 properly set the tune attribute.
5267 (dbnz): Use TARGET_DBNZ guard.
5268 * config/arc/arc.opt (mtune): Add core3 option.
5270 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5272 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
5273 recognize new pic like addresses.
5274 (arc_delegitimize_address): Clean up.
5276 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5278 * config/arc/arc-arches.def: Option mrf16 valid for all
5280 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
5281 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
5282 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
5283 * config/arc/arc-tables.opt: Regenerate.
5284 * config/arc/arc.c (arc_conditional_register_usage): Handle
5285 reduced register file case.
5286 (arc_file_start): Set must have build attributes.
5287 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
5289 * config/arc/arc.opt (mrf16): Add new option.
5290 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
5291 * config/arc/genmultilib.awk: Handle new mrf16 option.
5292 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
5293 * config/arc/t-multilib: Regenerate.
5294 * doc/invoke.texi (ARC Options): Document mrf16 option.
5296 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5298 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
5299 * config/arc/arc.c (arc_handle_secure_attribute): New function.
5300 (arc_attribute_table): Add 'secure_call' attribute.
5301 (arc_print_operand): Print secure call operand.
5302 (arc_function_ok_for_sibcall): Don't optimize tail calls when
5304 (arc_is_secure_call_p): New function. * config/arc/arc.md
5305 (call_i): Add support for sjli instruction.
5306 (call_value_i): Likewise.
5307 * config/arc/constraints.md (Csc): New constraint.
5309 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5310 John Eric Martin <John.Martin@emmicro-us.com>
5312 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
5313 * config/arc/arc.c (_arc_jli_section): New struct.
5314 (arc_jli_section): New type.
5315 (rc_jli_sections): New static variable.
5316 (arc_handle_jli_attribute): New function.
5317 (arc_attribute_table): Add jli_always and jli_fixed attribute.
5318 (arc_file_end): New function.
5319 (TARGET_ASM_FILE_END): Define.
5320 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
5321 (arc_add_jli_section): New function.
5322 (jli_call_scan): Likewise.
5323 (arc_reorg): Call jli_call_scan.
5324 (arc_output_addsi): Remove 'S' from printing asm operand.
5325 (arc_is_jli_call_p): New function.
5326 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
5328 (movhi_insn): Likewise.
5329 (movsi_insn): Likewise.
5330 (movsi_set_cc_insn): Likewise.
5331 (loadqi_update): Likewise.
5332 (load_zeroextendqisi_update): Likewise.
5333 (load_signextendqisi_update): Likewise.
5334 (loadhi_update): Likewise.
5335 (load_zeroextendhisi_update): Likewise.
5336 (load_signextendhisi_update): Likewise.
5337 (loadsi_update): Likewise.
5338 (loadsf_update): Likewise.
5339 (movsicc_insn): Likewise.
5340 (bset_insn): Likewise.
5341 (bxor_insn): Likewise.
5342 (bclr_insn): Likewise.
5343 (bmsk_insn): Likewise.
5344 (bicsi3_insn): Likewise.
5345 (cmpsi_cc_c_insn): Likewise.
5346 (movsi_ne): Likewise.
5347 (movsi_cond_exec): Likewise.
5348 (clrsbsi2): Likewise.
5359 (call_i): Remove 'S' asm letter, add jli instruction.
5360 (call_value_i): Likewise.
5361 * config/arc/arc.op (mjli-always): New option.
5362 * config/arc/constraints.md (Cji): New constraint.
5363 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
5365 (subsf3_fpx): Likewise.
5366 (mulsf3_fpx): Likewise.
5367 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
5369 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
5371 * doc/invoke.texi (ARC): Document mjli-always option.
5373 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
5375 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
5376 avoid addition with 0 and use incw and decw where possible.
5378 2018-01-26 Richard Biener <rguenther@suse.de>
5380 PR tree-optimization/81082
5381 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
5382 association if it requires casting to unsigned.
5383 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
5384 from fold_plusminus_mult_expr to catch important cases late when
5385 range info is available.
5387 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5389 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
5390 * configure.ac (hidden_linkonce): New test.
5391 * configure: Regenerate.
5392 * config.in: Regenerate.
5394 2018-01-26 Julia Koval <julia.koval@intel.com>
5396 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
5397 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
5398 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
5399 _mm_mask_bitshuffle_epi64_mask): Fix type.
5400 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
5401 USI_FTYPE_V4DI_V4DI_USI): Remove.
5402 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
5403 __builtin_ia32_vpshufbitqmb256_mask,
5404 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
5405 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
5406 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
5408 2018-01-26 Alan Modra <amodra@gmail.com>
5411 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
5412 UNSPEC_VBPERMQ. Sort other unspecs.
5414 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
5416 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
5418 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
5421 * predict.c (drop_profile): Do not push/pop cfun; update also
5423 (handle_missing_profiles): Fix logic looking for zero profiles.
5425 2018-01-25 Jakub Jelinek <jakub@redhat.com>
5428 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
5429 on functions with #pragma omp declare simd or functions with simd
5431 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
5432 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
5433 Remove trailing \n from warning_at calls.
5435 2018-01-25 Tom de Vries <tom@codesourcery.com>
5438 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5439 for neutered workers.
5441 2018-01-24 Joseph Myers <joseph@codesourcery.com>
5444 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
5445 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
5447 2018-01-24 Jeff Law <law@redhat.com>
5450 * i386.c (get_probe_interval): Move to earlier point.
5451 (ix86_compute_frame_layout): If -fstack-clash-protection and
5452 the frame is larger than the probe interval, then use pushes
5453 to save registers rather than reg->mem moves.
5454 (ix86_expand_prologue): Remove conditional for int_registers_saved
5457 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
5460 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
5461 min/max for never referenced object.
5463 2018-01-24 Jakub Jelinek <jakub@redhat.com>
5466 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
5468 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
5469 attributes from DECL_ATTRIBUTES (decl) without affecting
5470 DECL_ATTRIBUTES (current_function_decl).
5471 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
5472 functions with non-NULL DECL_ABSTRACT_ORIGIN.
5474 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
5476 PR tree-optimization/83979
5477 * fold-const.c (fold_comparison): Use constant_boolean_node
5478 instead of boolean_{true,false}_node.
5480 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
5482 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
5485 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5487 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5488 Simplify the clause that sets the length attribute.
5489 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5490 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
5491 clause that sets the length attribute.
5492 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5494 2018-01-24 Tom de Vries <tom@codesourcery.com>
5497 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
5498 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
5499 Add strict parameter.
5500 (prevent_branch_around_nothing): Insert dummy insn between branch to
5501 label and label with no ptx insn inbetween.
5502 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
5504 2018-01-24 Tom de Vries <tom@codesourcery.com>
5507 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5508 for neutered threads in warp.
5509 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
5511 2018-01-24 Richard Biener <rguenther@suse.de>
5513 PR tree-optimization/83176
5514 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
5517 2018-01-24 Richard Biener <rguenther@suse.de>
5519 PR tree-optimization/82819
5520 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
5521 code generating pluses that are no-ops in the target precision.
5523 2018-01-24 Richard Biener <rguenther@suse.de>
5526 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
5528 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5530 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
5531 to merge probabilities.
5532 * predict.c (probably_never_executed): Also mark as cold functions
5533 with global 0 profile and guessed local profile.
5534 * profile-count.c (profile_probability::combine_with_count): New
5536 * profile-count.h (profile_probability::operator*,
5537 profile_probability::operator*=, profile_probability::operator/,
5538 profile_probability::operator/=): Reduce precision to adjusted
5539 and set value to guessed on contradictory divisions.
5540 (profile_probability::combine_with_freq): Remove.
5541 (profile_probability::combine_wiht_count): Declare.
5542 (profile_count::force_nonzero):: Set to adjusted.
5543 (profile_count::probability_in):: Set quality to adjusted.
5544 * tree-ssa-tail-merge.c (replace_block_by): Use
5547 2018-01-23 Andrew Waterman <andrew@sifive.com>
5548 Jim Wilson <jimw@sifive.com>
5550 * config/riscv/riscv.c (riscv_stack_boundary): New.
5551 (riscv_option_override): Set riscv_stack_boundary. Handle
5552 riscv_preferred_stack_boundary_arg.
5553 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
5554 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
5555 (STACK_BOUNDARY): Set to riscv_stack_boundary.
5556 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
5557 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
5558 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
5560 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
5563 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
5564 of struct ix86_frame.
5565 (ix86_expand_epilogue): Likewise. Add a local variable for
5566 the reg_save_offset field in struct ix86_frame.
5568 2018-01-23 Bin Cheng <bin.cheng@arm.com>
5570 PR tree-optimization/82604
5571 * tree-loop-distribution.c (enum partition_kind): New enum item
5572 PKIND_PARTIAL_MEMSET.
5573 (partition_builtin_p): Support above new enum item.
5574 (generate_code_for_partition): Ditto.
5575 (compute_access_range): Differentiate cases that equality can be
5576 proven at all loops, the innermost loops or no loops.
5577 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
5578 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
5579 (finalize_partitions, distribute_loop): Don't fuse partition of
5580 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
5581 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
5584 2018-01-23 Martin Liska <mliska@suse.cz>
5586 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
5587 order to ignore the predictor.
5588 (PRED_POLYMORPHIC_CALL): Likewise.
5589 (PRED_RECURSIVE_CALL): Likewise.
5591 2018-01-23 Martin Liska <mliska@suse.cz>
5593 * tree-profile.c (tree_profiling): Print function header to
5594 aware reader which function we are working on.
5595 * value-prof.c (gimple_find_values_to_profile): Do not print
5596 not interesting value histograms.
5598 2018-01-23 Martin Liska <mliska@suse.cz>
5600 * profile-count.h (enum profile_quality): Add
5601 profile_uninitialized as the first value. Do not number values
5602 as they are zero based.
5603 (profile_count::verify): Update sanity check.
5604 (profile_probability::verify): Likewise.
5606 2018-01-23 Nathan Sidwell <nathan@acm.org>
5608 * doc/invoke.texi (ffor-scope): Deprecate.
5610 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5612 PR tree-optimization/83510
5613 * domwalk.c (set_all_edges_as_executable): New function.
5614 (dom_walker::dom_walker): Convert bool param
5615 "skip_unreachable_blocks" to enum reachability. Move setup of
5616 edge flags to set_all_edges_as_executable and only do it when
5617 reachability is REACHABLE_BLOCKS.
5618 * domwalk.h (enum dom_walker::reachability): New enum.
5619 (dom_walker::dom_walker): Convert bool param
5620 "skip_unreachable_blocks" to enum reachability.
5621 (set_all_edges_as_executable): New decl.
5622 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
5623 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
5625 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
5626 but converting true to REACHABLE_BLOCKS.
5627 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
5629 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
5630 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
5631 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
5633 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
5634 if check_all_array_refs will be called.
5636 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5638 * tree.c (selftest::test_location_wrappers): Add more test
5641 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5643 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
5644 (selftest::test_bit_in_range): Likewise.
5646 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5649 * doc/sourcebuild.texi (vect_float): Say that the selector
5650 only describes the situation when -funsafe-math-optimizations is on.
5651 (vect_float_strict): Document.
5653 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5655 PR tree-optimization/83965
5656 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
5657 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
5658 instead of checking only for a reduction.
5659 (vect_recog_widen_sum_pattern): Likewise.
5661 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5663 * predict.c (probably_never_executed): Only use precise profile info.
5664 (compute_function_frequency): Skip after inlining hack since we now
5665 have quality checking.
5667 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5669 * profile-count.h (profile_probability::very_unlikely,
5670 profile_probability::unlikely, profile_probability::even): Set
5671 precision to guessed.
5673 2018-01-23 Richard Biener <rguenther@suse.de>
5675 PR tree-optimization/83963
5676 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
5677 Properly terminate dominator walk when crossing the exit edge not
5678 when visiting its source block.
5680 2018-01-23 Jakub Jelinek <jakub@redhat.com>
5683 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
5684 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
5686 2018-01-22 Jakub Jelinek <jakub@redhat.com>
5688 PR tree-optimization/83957
5689 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
5690 semicolon after for body surrounded by braces.
5692 PR tree-optimization/83081
5693 * profile-count.h (profile_probability::split): New method.
5694 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
5695 Use profile_probability::split.
5696 (do_compare_rtx_and_jump): Fix adjustment of probabilities
5697 when splitting a single conditional jump into 2.
5699 2018-01-22 David Malcolm <dmalcolm@redhat.com>
5701 PR tree-optimization/69452
5702 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
5705 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5707 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
5708 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
5709 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
5711 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5713 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
5714 * config/rl78/rl78.md (movdi): New define_expand.
5715 * config/rl78/rl78.c (rl78_split_movdi): New function.
5717 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
5720 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
5722 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
5723 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
5724 128-bit to produce an UNSPEC move to get the double word with the
5725 signbit and then a shift directly to do signbit.
5726 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
5727 implementation with a new version that just does either a direct
5728 move or a regular move. Move memory interface to separate insns.
5729 Move insns so they are next to the expander.
5730 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
5731 with signbit move. Split big and little endian case.
5732 (signbit<mode>2_dm_mem_le): Likewise.
5733 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
5734 (signbit<mode>2_dm2): Likewise.
5736 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5738 * config/rl78/rl78.md (anddi3): New define_expand.
5740 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5742 * config/rl78/rl78.md (umindi3): New define_expand.
5744 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5746 * config/rl78/rl78.md (smindi3): New define_expand.
5748 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5750 * config/rl78/rl78.md (smaxdi3): New define_expand.
5752 2018-01-22 Carl Love <cel@us.ibm.com>
5754 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
5755 LVX_V1TI): Add macro expansion.
5756 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
5757 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
5758 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
5759 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
5760 Change check to determine if the instruction is a byte reversing
5761 entry. Fix typo in comment.
5762 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
5763 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
5764 Add def_builtin calls for new builtins.
5765 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
5766 Add define_insn expansion.
5768 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5770 * config/rl78/rl78.md (umaxdi3): New define_expand.
5772 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5774 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
5775 for non-QImode registers.
5777 2018-01-22 Richard Biener <rguenther@suse.de>
5779 PR tree-optimization/83963
5780 * graphite-scop-detection.c (scop_detection::get_sese): Delay
5781 including the loop exit block.
5782 (scop_detection::merge_sese): Likewise.
5783 (scop_detection::add_scop): Do it here instead.
5785 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5787 * doc/sourcebuild.texi (arm_softfloat): Document.
5789 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
5792 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
5793 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
5794 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
5796 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5797 David Edelsohn <dje.gcc@gmail.com>
5800 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5801 Change "crset eq" to "crset 2".
5802 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5803 (*call_indirect_aix<mode>_nospec): Likewise.
5804 (*call_value_indirect_aix<mode>_nospec): Likewise.
5805 (*call_indirect_elfv2<mode>_nospec): Likewise.
5806 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
5807 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
5808 change assembly output from . to $.
5809 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5810 (indirect_jump<mode>_nospec): Change assembly output from . to $.
5811 (*tablejump<mode>_internal1_nospec): Likewise.
5813 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
5816 * config/sh/sh_optimize_sett_clrt.cc:
5817 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
5819 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
5821 PR tree-optimization/83940
5822 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
5823 offset_dt to vect_constant_def rather than vect_unknown_def_type.
5824 (vect_check_load_store_mask): Add a mask_dt_out parameter and
5825 use it to pass back the definition type.
5826 (vect_check_store_rhs): Likewise rhs_dt_out.
5827 (vect_build_gather_load_calls): Add a mask_dt argument and use
5828 it instead of a call to vect_is_simple_use.
5829 (vectorizable_store): Update calls to vect_check_load_store_mask
5830 and vect_check_store_rhs. Use the dt returned by the latter instead
5831 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
5832 instead of calls to vect_is_simple_use. Pass the scalar rather
5833 than the vector operand to vect_is_simple_use when handling
5834 second and subsequent copies of an rhs value.
5835 (vectorizable_load): Update calls to vect_check_load_store_mask
5836 and vect_build_gather_load_calls. Use the cached mask_dt and
5837 gs_info.offset_dt instead of calls to vect_is_simple_use.
5839 2018-01-20 Jakub Jelinek <jakub@redhat.com>
5842 * tree-emutls.c: Include gimplify.h.
5843 (lower_emutls_2): New function.
5844 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
5845 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
5846 it before further processing.
5849 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
5850 UINTVAL (trueop1) instead of INTVAL (op1).
5852 2018-01-19 Jakub Jelinek <jakub@redhat.com>
5856 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
5857 INCOMING_FRAME_SP_OFFSET if not defined.
5858 (scan_trace): Add ENTRY argument. If true and
5859 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
5860 emit a note to adjust the CFA offset.
5861 (create_cfi_notes): Adjust scan_trace callers.
5862 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
5863 INCOMING_FRAME_SP_OFFSET in the CIE.
5864 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
5865 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
5867 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
5868 * doc/tm.texi: Regenerated.
5870 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5872 PR rtl-optimization/83147
5873 * lra-constraints.c (remove_inheritance_pseudos): Use
5874 lra_substitute_pseudo_within_insn.
5876 2018-01-19 Tom de Vries <tom@codesourcery.com>
5877 Cesar Philippidis <cesar@codesourcery.com>
5880 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
5882 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
5885 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
5886 spaces for function labels.
5888 2018-01-19 Martin Liska <mliska@suse.cz>
5890 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
5891 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
5892 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
5893 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
5894 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
5895 (PRED_CONST_RETURN): Change from 69 to 65.
5896 (PRED_NULL_RETURN): Change from 91 to 71.
5897 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
5898 (PRED_LOOP_GUARD): Change from 66 to 73.
5900 2018-01-19 Martin Liska <mliska@suse.cz>
5902 * predict.c (predict_insn_def): Add new assert.
5903 (struct branch_predictor): Change type to signed integer.
5904 (test_prediction_value_range): Amend test to cover
5906 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
5907 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
5908 (PRED_LOOP_ITERATIONS_MAX): Likewise.
5909 (PRED_LOOP_IV_COMPARE): Likewise.
5910 * predict.h (PROB_UNINITIALIZED): Define new constant.
5912 2018-01-19 Martin Liska <mliska@suse.cz>
5914 * predict.c (dump_prediction): Add new format for
5915 analyze_brprob.py script which is enabled with -details
5917 * profile-count.h (precise_p): New function.
5919 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
5921 PR tree-optimization/83922
5922 * tree-vect-loop.c (vect_verify_full_masking): Return false if
5923 there are no statements that need masking.
5924 (vect_active_double_reduction_p): New function.
5925 (vect_analyze_loop_operations): Use it when handling phis that
5926 are not in the loop header.
5928 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
5930 PR tree-optimization/83914
5931 * tree-vect-loop.c (vectorizable_induction): Don't convert
5932 init_expr or apply the peeling adjustment for inductions
5933 that are nested within the vectorized loop.
5935 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5937 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
5940 2018-01-18 Jakub Jelinek <jakub@redhat.com>
5944 * function.h (gimplify_parameters): Add gimple_seq * argument.
5945 * function.c: Include gimple.h and options.h.
5946 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
5947 for the added local temporaries if needed.
5948 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
5949 if there are any parameter cleanups, wrap whole body into a
5950 try/finally with the cleanups.
5952 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
5955 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
5956 Use GET_MODE_CLASS for scalar floating point.
5958 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
5962 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
5963 Fix call of call_cgraph_insertion_hooks.
5965 2018-01-18 Martin Sebor <msebor@redhat.com>
5967 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
5969 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
5972 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
5975 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
5978 * common.opt: (-ffile-prefix-map): New option.
5979 * opts.c (common_handle_option): Defer it.
5980 * opts-global.c (handle_common_deferred_options): Handle it.
5981 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
5982 * file-prefix-map.h: New file.
5983 (remap_debug_filename, add_debug_prefix_map): ...here.
5984 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
5985 * final.c (debug_prefix_map, add_debug_prefix_map
5986 remap_debug_filename): Move to...
5987 * file-prefix-map.c: New file.
5988 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
5989 generalize, get rid of alloca(), use strrchr() instead of strchr().
5990 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
5991 Implement in terms of add_prefix_map().
5992 (remap_macro_filename, remap_debug_filename): Implement in term of
5994 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
5995 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
5996 * dbxout.c: Include file-prefix-map.h.
5997 * varasm.c: Likewise.
5998 * vmsdbgout.c: Likewise.
5999 * xcoffout.c: Likewise.
6000 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
6001 * doc/cppopts.texi (-fmacro-prefix-map): Document.
6002 * doc/invoke.texi (-ffile-prefix-map): Document.
6003 (-fdebug-prefix-map): Update description.
6005 2018-01-18 Martin Liska <mliska@suse.cz>
6007 * config/i386/i386.c (indirect_thunk_name): Document that also
6009 (output_indirect_thunk): Document why both instructions
6010 (pause and lfence) are generated.
6012 2018-01-18 Richard Biener <rguenther@suse.de>
6014 PR tree-optimization/83887
6015 * graphite-scop-detection.c
6016 (scop_detection::get_nearest_dom_with_single_entry): Remove.
6017 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
6018 (scop_detection::merge_sese): Re-implement with a flood-fill
6019 algorithm that properly finds a SESE region if it exists.
6021 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6024 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
6025 pointer_diff optimizations use view_convert instead of convert.
6027 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6029 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6030 Generate different code for -mno-speculate-indirect-jumps.
6031 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6032 (*call_indirect_aix<mode>): Disable for
6033 -mno-speculate-indirect-jumps.
6034 (*call_indirect_aix<mode>_nospec): New define_insn.
6035 (*call_value_indirect_aix<mode>): Disable for
6036 -mno-speculate-indirect-jumps.
6037 (*call_value_indirect_aix<mode>_nospec): New define_insn.
6038 (*sibcall_nonlocal_sysv<mode>): Generate different code for
6039 -mno-speculate-indirect-jumps.
6040 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6042 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6044 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
6045 long double type, set the flags for noting the default long double
6046 type, even if we don't pass or return a long double type.
6048 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
6051 * ipa-inline.c (flatten_function): Do not overwrite final inlining
6054 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
6056 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
6057 support for merge[hl].
6058 (fold_mergehl_helper): New helper function.
6059 (tree-vector-builder.h): New #include for tree_vector_builder usage.
6060 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
6061 (altivec_vmrglw_direct): Add xxmrglw insn.
6063 2018-01-17 Andrew Waterman <andrew@sifive.com>
6065 * config/riscv/riscv.c (riscv_conditional_register_usage): If
6066 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
6068 2018-01-17 David Malcolm <dmalcolm@redhat.com>
6071 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
6072 call the lto_location_cache before reading the
6073 DECL_SOURCE_LOCATION of the types.
6075 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
6076 Richard Sandiford <richard.sandiford@linaro.org>
6078 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
6079 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
6080 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
6081 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
6082 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
6084 * config/aarch64/constraints.md (aarch64_movti_operand):
6086 * config/aarch64/predicates.md (Uti): Add new constraint.
6088 2018-01-17 Carl Love <cel@us.ibm.com>
6090 * config/rs6000/vsx.md (define_expand xl_len_r,
6091 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
6092 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
6094 (define_expand, define_insn): Move the shift left from the
6095 define_insn to the define_expand for lxvl and stxvl instructions.
6096 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
6097 and XL_LEN_R definitions to PURE.
6099 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
6101 * config/i386/i386.c (indirect_thunk_name): Declare regno
6102 as unsigned int. Compare regno with INVALID_REGNUM.
6103 (output_indirect_thunk): Ditto.
6104 (output_indirect_thunk_function): Ditto.
6105 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
6106 in the call to output_indirect_thunk_function.
6108 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
6111 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
6112 rather than the size of inner_type to determine the stack slot size
6113 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
6115 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
6118 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
6121 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
6123 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
6124 endian Linux systems to optionally enable multilibs for selecting
6125 the long double type if the user configured an explicit type.
6126 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
6127 have no long double multilibs if not defined.
6128 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
6129 warn if the user used -mabi={ieee,ibm}longdouble and we built
6130 multilibs for long double.
6131 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
6132 appropriate multilib option.
6133 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
6135 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
6136 for building long double multilibs.
6137 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
6139 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
6141 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
6144 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
6146 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
6149 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
6152 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
6155 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6157 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
6158 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
6160 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
6162 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
6163 different rtl trees depending on TARGET_64BIT.
6164 (rs6000_gen_lvx): Likewise.
6166 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6168 * config/visium/visium.md (nop): Tweak comment.
6169 (hazard_nop): Likewise.
6171 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6173 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
6174 -mspeculate-indirect-jumps.
6175 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
6176 for -mno-speculate-indirect-jumps.
6177 (*call_indirect_elfv2<mode>_nospec): New define_insn.
6178 (*call_value_indirect_elfv2<mode>): Disable for
6179 -mno-speculate-indirect-jumps.
6180 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
6181 (indirect_jump): Emit different RTL for
6182 -mno-speculate-indirect-jumps.
6183 (*indirect_jump<mode>): Disable for
6184 -mno-speculate-indirect-jumps.
6185 (*indirect_jump<mode>_nospec): New define_insn.
6186 (tablejump): Emit different RTL for
6187 -mno-speculate-indirect-jumps.
6188 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
6189 (tablejumpsi_nospec): New define_expand.
6190 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
6191 (tablejumpdi_nospec): New define_expand.
6192 (*tablejump<mode>_internal1): Disable for
6193 -mno-speculate-indirect-jumps.
6194 (*tablejump<mode>_internal1_nospec): New define_insn.
6195 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
6198 2018-01-16 Artyom Skrobov tyomitch@gmail.com
6200 * caller-save.c (insert_save): Drop unnecessary parameter. All
6203 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6204 Richard Biener <rguenth@suse.de>
6207 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
6208 return early, inline manually is_gimple_sizepos. Make sure if we
6209 call gimplify_expr we don't end up with a gimple constant.
6210 * tree.c (variably_modified_type_p): Don't return true for
6211 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
6212 * gimplify.h (is_gimple_sizepos): Remove.
6214 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6216 PR tree-optimization/83857
6217 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
6218 vectorizable_live_operation for pure SLP statements.
6219 (vectorizable_live_operation): Handle PHIs.
6221 2018-01-16 Richard Biener <rguenther@suse.de>
6223 PR tree-optimization/83867
6224 * tree-vect-stmts.c (vect_transform_stmt): Precompute
6225 nested_in_vect_loop_p since the scalar stmt may get invalidated.
6227 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6230 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
6231 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
6232 If off is not INTEGER_CST, issue a may not be aligned warning
6233 rather than isn't aligned. Use isn%'t rather than isn't.
6234 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
6236 <case MULT_EXPR>: Improve the case when bottom and one of the
6237 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
6238 operand, in that case check if the other operand is multiple of
6239 bottom divided by the INTEGER_CST operand.
6241 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6244 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
6245 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
6246 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
6247 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
6248 * config/pa/pa.c (pa_function_arg_advance): Likewise.
6249 (pa_function_arg, pa_arg_partial_bytes): Likewise.
6250 (pa_function_arg_size): New function.
6252 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6254 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
6255 in a separate statement.
6257 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6259 PR tree-optimization/83847
6260 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
6261 group gathers and scatters.
6263 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6265 PR rtl-optimization/86620
6266 * params.def (max-sched-ready-insns): Bump minimum value to 1.
6268 PR rtl-optimization/83213
6269 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
6270 to last if both are JUMP_INSNs.
6272 PR tree-optimization/83843
6273 * gimple-ssa-store-merging.c
6274 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
6275 store_immediate_info for bswap/nop orig_stores.
6277 2018-01-15 Andrew Waterman <andrew@sifive.com>
6279 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
6281 <UDIV>: Increase cost if !TARGET_DIV.
6283 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
6285 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
6286 (define_attr "cr_logical_3op"): New.
6287 (cceq_ior_compare): Adjust.
6288 (cceq_ior_compare_complement): Adjust.
6289 (*cceq_rev_compare): Adjust.
6290 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
6291 (is_cracked_insn): Adjust.
6292 (insn_must_be_first_in_group): Adjust.
6293 * config/rs6000/40x.md: Adjust.
6294 * config/rs6000/440.md: Adjust.
6295 * config/rs6000/476.md: Adjust.
6296 * config/rs6000/601.md: Adjust.
6297 * config/rs6000/603.md: Adjust.
6298 * config/rs6000/6xx.md: Adjust.
6299 * config/rs6000/7450.md: Adjust.
6300 * config/rs6000/7xx.md: Adjust.
6301 * config/rs6000/8540.md: Adjust.
6302 * config/rs6000/cell.md: Adjust.
6303 * config/rs6000/e300c2c3.md: Adjust.
6304 * config/rs6000/e500mc.md: Adjust.
6305 * config/rs6000/e500mc64.md: Adjust.
6306 * config/rs6000/e5500.md: Adjust.
6307 * config/rs6000/e6500.md: Adjust.
6308 * config/rs6000/mpc.md: Adjust.
6309 * config/rs6000/power4.md: Adjust.
6310 * config/rs6000/power5.md: Adjust.
6311 * config/rs6000/power6.md: Adjust.
6312 * config/rs6000/power7.md: Adjust.
6313 * config/rs6000/power8.md: Adjust.
6314 * config/rs6000/power9.md: Adjust.
6315 * config/rs6000/rs64.md: Adjust.
6316 * config/rs6000/titan.md: Adjust.
6318 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6320 * config/i386/predicates.md (indirect_branch_operand): Rewrite
6321 ix86_indirect_branch_register logic.
6323 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6325 * config/i386/constraints.md (Bs): Update
6326 ix86_indirect_branch_register check. Don't check
6327 ix86_indirect_branch_register with GOT_memory_operand.
6329 * config/i386/predicates.md (GOT_memory_operand): Don't check
6330 ix86_indirect_branch_register here.
6331 (GOT32_symbol_operand): Likewise.
6333 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6335 * config/i386/predicates.md (constant_call_address_operand):
6336 Rewrite ix86_indirect_branch_register logic.
6337 (sibcall_insn_operand): Likewise.
6339 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6341 * config/i386/constraints.md (Bs): Replace
6342 ix86_indirect_branch_thunk_register with
6343 ix86_indirect_branch_register.
6345 * config/i386/i386.md (indirect_jump): Likewise.
6346 (tablejump): Likewise.
6347 (*sibcall_memory): Likewise.
6348 (*sibcall_value_memory): Likewise.
6349 Peepholes of indirect call and jump via memory: Likewise.
6350 * config/i386/i386.opt: Likewise.
6351 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6352 (GOT_memory_operand): Likewise.
6353 (call_insn_operand): Likewise.
6354 (sibcall_insn_operand): Likewise.
6355 (GOT32_symbol_operand): Likewise.
6357 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6360 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
6361 type rather than type addr's type points to.
6362 (expand_omp_atomic_mutex): Likewise.
6363 (expand_omp_atomic): Likewise.
6365 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6368 * config/i386/i386.c (output_indirect_thunk_function): Use
6369 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
6370 for __x86_return_thunk.
6372 2018-01-15 Richard Biener <rguenther@suse.de>
6375 * expmed.c (extract_bit_field_1): Fix typo.
6377 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6380 * config/arm/iterators.md (VF): New mode iterator.
6381 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
6382 Remove integer-related logic from pattern.
6383 (neon_vabd<mode>_3): Likewise.
6385 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6388 * common.opt (fstrict-overflow): No longer an alias.
6389 (fwrapv-pointer): New option.
6390 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
6391 also for pointer types based on flag_wrapv_pointer.
6392 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
6393 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
6394 opts->x_flag_wrapv got set.
6395 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
6396 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
6397 POINTER_TYPE_OVERFLOW_UNDEFINED.
6398 * match.pd: Likewise in address comparison pattern.
6399 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
6401 2018-01-15 Richard Biener <rguenther@suse.de>
6404 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
6405 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
6406 Reset type names to their identifier if their TYPE_DECL doesn't
6407 have linkage (and thus is used for ODR and devirt).
6408 (save_debug_info_for_decl): Remove.
6409 (save_debug_info_for_type): Likewise.
6410 (add_tree_to_fld_list): Adjust.
6411 * tree-pretty-print.c (dump_generic_node): Make dumping of
6412 type names more robust.
6414 2018-01-15 Richard Biener <rguenther@suse.de>
6416 * BASE-VER: Bump to 8.0.1.
6418 2018-01-14 Martin Sebor <msebor@redhat.com>
6421 * builtins.c (check_access): Avoid warning when the no-warning bit
6424 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
6426 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
6427 * ira-color (allocno_hard_regs_compare): Likewise.
6429 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
6432 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
6433 Use .pushsection/.popsection.
6435 2018-01-14 Martin Sebor <msebor@redhat.com>
6438 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
6440 2018-01-14 Jakub Jelinek <jakub@redhat.com>
6442 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
6443 entry from extra_headers.
6444 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
6445 extra_headers, make the list bitwise identical to the i?86-*-* one.
6447 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6449 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
6450 -mcmodel=large with -mindirect-branch=thunk,
6451 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
6452 -mfunction-return=thunk-extern.
6453 * doc/invoke.texi: Document -mcmodel=large is incompatible with
6454 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
6455 -mfunction-return=thunk and -mfunction-return=thunk-extern.
6457 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6459 * config/i386/i386.c (print_reg): Print the name of the full
6460 integer register without '%'.
6461 (ix86_print_operand): Handle 'V'.
6462 * doc/extend.texi: Document 'V' modifier.
6464 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6466 * config/i386/constraints.md (Bs): Disallow memory operand for
6467 -mindirect-branch-register.
6469 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6470 (GOT_memory_operand): Likewise.
6471 (call_insn_operand): Likewise.
6472 (sibcall_insn_operand): Likewise.
6473 (GOT32_symbol_operand): Likewise.
6474 * config/i386/i386.md (indirect_jump): Call convert_memory_address
6475 for -mindirect-branch-register.
6476 (tablejump): Likewise.
6477 (*sibcall_memory): Likewise.
6478 (*sibcall_value_memory): Likewise.
6479 Disallow peepholes of indirect call and jump via memory for
6480 -mindirect-branch-register.
6481 (*call_pop): Replace m with Bw.
6482 (*call_value_pop): Likewise.
6483 (*sibcall_pop_memory): Replace m with Bs.
6484 * config/i386/i386.opt (mindirect-branch-register): New option.
6485 * doc/invoke.texi: Document -mindirect-branch-register option.
6487 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6489 * config/i386/i386-protos.h (ix86_output_function_return): New.
6490 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
6491 set function_return_type.
6492 (indirect_thunk_name): Add ret_p to indicate thunk for function
6494 (output_indirect_thunk_function): Pass false to
6495 indirect_thunk_name.
6496 (ix86_output_indirect_branch_via_reg): Likewise.
6497 (ix86_output_indirect_branch_via_push): Likewise.
6498 (output_indirect_thunk_function): Create alias for function
6499 return thunk if regno < 0.
6500 (ix86_output_function_return): New function.
6501 (ix86_handle_fndecl_attribute): Handle function_return.
6502 (ix86_attribute_table): Add function_return.
6503 * config/i386/i386.h (machine_function): Add
6504 function_return_type.
6505 * config/i386/i386.md (simple_return_internal): Use
6506 ix86_output_function_return.
6507 (simple_return_internal_long): Likewise.
6508 * config/i386/i386.opt (mfunction-return=): New option.
6509 (indirect_branch): Mention -mfunction-return=.
6510 * doc/extend.texi: Document function_return function attribute.
6511 * doc/invoke.texi: Document -mfunction-return= option.
6513 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6515 * config/i386/i386-opts.h (indirect_branch): New.
6516 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
6517 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
6518 with local indirect jump when converting indirect call and jump.
6519 (ix86_set_indirect_branch_type): New.
6520 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
6521 (indirectlabelno): New.
6522 (indirect_thunk_needed): Likewise.
6523 (indirect_thunk_bnd_needed): Likewise.
6524 (indirect_thunks_used): Likewise.
6525 (indirect_thunks_bnd_used): Likewise.
6526 (INDIRECT_LABEL): Likewise.
6527 (indirect_thunk_name): Likewise.
6528 (output_indirect_thunk): Likewise.
6529 (output_indirect_thunk_function): Likewise.
6530 (ix86_output_indirect_branch_via_reg): Likewise.
6531 (ix86_output_indirect_branch_via_push): Likewise.
6532 (ix86_output_indirect_branch): Likewise.
6533 (ix86_output_indirect_jmp): Likewise.
6534 (ix86_code_end): Call output_indirect_thunk_function if needed.
6535 (ix86_output_call_insn): Call ix86_output_indirect_branch if
6537 (ix86_handle_fndecl_attribute): Handle indirect_branch.
6538 (ix86_attribute_table): Add indirect_branch.
6539 * config/i386/i386.h (machine_function): Add indirect_branch_type
6540 and has_local_indirect_jump.
6541 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
6543 (tablejump): Likewise.
6544 (*indirect_jump): Use ix86_output_indirect_jmp.
6545 (*tablejump_1): Likewise.
6546 (simple_return_indirect_internal): Likewise.
6547 * config/i386/i386.opt (mindirect-branch=): New option.
6548 (indirect_branch): New.
6551 (thunk-inline): Likewise.
6552 (thunk-extern): Likewise.
6553 * doc/extend.texi: Document indirect_branch function attribute.
6554 * doc/invoke.texi: Document -mindirect-branch= option.
6556 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
6559 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
6561 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
6563 * ipa-inline.c (want_inline_small_function_p): Return false if
6564 inlining has already failed with CIF_FINAL_ERROR.
6565 (update_caller_keys): Call want_inline_small_function_p before
6567 (update_callee_keys): Likewise.
6569 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6571 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
6573 (rs6000_quadword_masked_address_p): Likewise.
6574 (quad_aligned_load_p): Likewise.
6575 (quad_aligned_store_p): Likewise.
6576 (const_load_sequence_p): Add comment to describe the outer-most loop.
6577 (mimic_memory_attributes_and_flags): New function.
6578 (rs6000_gen_stvx): Likewise.
6579 (replace_swapped_aligned_store): Likewise.
6580 (rs6000_gen_lvx): Likewise.
6581 (replace_swapped_aligned_load): Likewise.
6582 (replace_swapped_load_constant): Capitalize argument name in
6583 comment describing this function.
6584 (rs6000_analyze_swaps): Add a third pass to search for vector loads
6585 and stores that access quad-word aligned addresses and replace
6586 with stvx or lvx instructions when appropriate.
6587 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
6588 New function prototype.
6589 (rs6000_quadword_masked_address_p): Likewise.
6590 (rs6000_gen_lvx): Likewise.
6591 (rs6000_gen_stvx): Likewise.
6592 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
6593 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
6594 when memory address is aligned.
6595 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
6596 this split to select lvx instruction when memory address is aligned.
6597 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
6598 instruction when memory address is aligned.
6599 (*vsx_le_perm_load_v16qi): Likewise.
6600 (four unnamed splitters): Modify to select the stvx instruction
6601 when memory is aligned.
6603 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
6605 * predict.c (determine_unlikely_bbs): Handle correctly BBs
6606 which appears in the queue multiple times.
6608 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6609 Alan Hayward <alan.hayward@arm.com>
6610 David Sherwood <david.sherwood@arm.com>
6612 * tree-vectorizer.h (vec_lower_bound): New structure.
6613 (_loop_vec_info): Add check_nonzero and lower_bounds.
6614 (LOOP_VINFO_CHECK_NONZERO): New macro.
6615 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
6616 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
6617 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
6618 fields. Make seg_len the distance travelled, not including the
6620 (dr_direction_indicator): Declare.
6621 (dr_zero_step_indicator): Likewise.
6622 (dr_known_forward_stride_p): Likewise.
6623 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
6625 (runtime_alias_check_p): Allow runtime alias checks with
6627 (operator ==): Compare access_size and align.
6628 (prune_runtime_alias_test_list): Rework for new distinction between
6629 the access_size and seg_len.
6630 (create_intersect_range_checks_index): Likewise. Cope with polynomial
6632 (get_segment_min_max): New function.
6633 (create_intersect_range_checks): Use it.
6634 (dr_step_indicator): New function.
6635 (dr_direction_indicator): Likewise.
6636 (dr_zero_step_indicator): Likewise.
6637 (dr_known_forward_stride_p): Likewise.
6638 * tree-loop-distribution.c (data_ref_segment_size): Return
6639 DR_STEP * (niters - 1).
6640 (compute_alias_check_pairs): Update call to the dr_with_seg_len
6642 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
6643 (vect_preserves_scalar_order_p): New function, split out from...
6644 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
6645 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
6646 (vect_vfa_access_size): New function.
6647 (vect_vfa_align): Likewise.
6648 (vect_compile_time_alias): Take access_size_a and access_b arguments.
6649 (dump_lower_bound): New function.
6650 (vect_check_lower_bound): Likewise.
6651 (vect_small_gap_p): Likewise.
6652 (vectorizable_with_step_bound_p): Likewise.
6653 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
6654 depencies if the vectorization factor is 1. Convert the checks
6655 for nonzero steps into checks on the bounds of DR_STEP. Try using
6656 a bunds check for variable steps if the minimum required step is
6657 relatively small. Update calls to the dr_with_seg_len
6658 constructor and to vect_compile_time_alias.
6659 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
6661 (vect_loop_versioning): Call it.
6662 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
6664 (vect_estimate_min_profitable_iters): Account for any bounds checks.
6666 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6667 Alan Hayward <alan.hayward@arm.com>
6668 David Sherwood <david.sherwood@arm.com>
6670 * doc/sourcebuild.texi (vect_scatter_store): Document.
6671 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
6673 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
6675 * genopinit.c (main): Add supports_vec_scatter_store and
6676 supports_vec_scatter_store_cached to target_optabs.
6677 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
6678 IFN_MASK_SCATTER_STORE.
6679 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
6681 * internal-fn.h (internal_store_fn_p): Declare.
6682 (internal_fn_stored_value_index): Likewise.
6683 * internal-fn.c (scatter_store_direct): New macro.
6684 (expand_scatter_store_optab_fn): New function.
6685 (direct_scatter_store_optab_supported_p): New macro.
6686 (internal_store_fn_p): New function.
6687 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
6688 IFN_MASK_SCATTER_STORE.
6689 (internal_fn_mask_index): Likewise.
6690 (internal_fn_stored_value_index): New function.
6691 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
6693 * optabs-query.h (supports_vec_scatter_store_p): Declare.
6694 * optabs-query.c (supports_vec_scatter_store_p): New function.
6695 * tree-vectorizer.h (vect_get_store_rhs): Declare.
6696 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
6697 true for scatter stores.
6698 (vect_gather_scatter_fn_p): Handle scatter stores too.
6699 (vect_check_gather_scatter): Consider using scatter stores if
6700 supports_vec_scatter_store_p.
6701 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
6703 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6704 internal_fn_stored_value_index.
6705 (check_load_store_masking): Handle scatter stores too.
6706 (vect_get_store_rhs): Make public.
6707 (vectorizable_call): Use internal_store_fn_p.
6708 (vectorizable_store): Handle scatter store internal functions.
6709 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
6710 when deciding whether the end of the group has been reached.
6711 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
6712 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
6713 (mask_scatter_store<mode>): New insns.
6715 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6716 Alan Hayward <alan.hayward@arm.com>
6717 David Sherwood <david.sherwood@arm.com>
6719 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
6720 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
6721 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
6723 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
6724 Use vect_truncate_gather_scatter_offset if we can't treat the
6725 operation as a normal gather load or scatter store.
6726 (get_group_load_store_type): Take the gather_scatter_info
6727 as argument. Try using a gather load or scatter store for
6728 single-element groups.
6729 (get_load_store_type): Update calls to get_group_load_store_type
6730 and vect_use_strided_gather_scatters_p.
6732 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6733 Alan Hayward <alan.hayward@arm.com>
6734 David Sherwood <david.sherwood@arm.com>
6736 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
6737 optional tree argument.
6738 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
6740 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
6741 but continue to use the current value as a fallback.
6742 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
6743 to compare the updates.
6744 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
6745 (get_load_store_type): Use it when handling a strided access.
6746 (vect_get_strided_load_store_ops): New function.
6747 (vect_get_data_ptr_increment): Likewise.
6748 (vectorizable_load): Handle strided gather loads. Always pass
6749 a step to vect_create_data_ref_ptr and bump_vector_ptr.
6751 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6752 Alan Hayward <alan.hayward@arm.com>
6753 David Sherwood <david.sherwood@arm.com>
6755 * doc/md.texi (gather_load@var{m}): Document.
6756 (mask_gather_load@var{m}): Likewise.
6757 * genopinit.c (main): Add supports_vec_gather_load and
6758 supports_vec_gather_load_cached to target_optabs.
6759 * optabs-tree.c (init_tree_optimization_optabs): Use
6760 ggc_cleared_alloc to allocate target_optabs.
6761 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
6762 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
6764 * internal-fn.h (internal_load_fn_p): Declare.
6765 (internal_gather_scatter_fn_p): Likewise.
6766 (internal_fn_mask_index): Likewise.
6767 (internal_gather_scatter_fn_supported_p): Likewise.
6768 * internal-fn.c (gather_load_direct): New macro.
6769 (expand_gather_load_optab_fn): New function.
6770 (direct_gather_load_optab_supported_p): New macro.
6771 (direct_internal_fn_optab): New function.
6772 (internal_load_fn_p): Likewise.
6773 (internal_gather_scatter_fn_p): Likewise.
6774 (internal_fn_mask_index): Likewise.
6775 (internal_gather_scatter_fn_supported_p): Likewise.
6776 * optabs-query.c (supports_at_least_one_mode_p): New function.
6777 (supports_vec_gather_load_p): Likewise.
6778 * optabs-query.h (supports_vec_gather_load_p): Declare.
6779 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
6780 and memory_type field.
6781 (NUM_PATTERNS): Bump to 15.
6782 * tree-vect-data-refs.c: Include internal-fn.h.
6783 (vect_gather_scatter_fn_p): New function.
6784 (vect_describe_gather_scatter_call): Likewise.
6785 (vect_check_gather_scatter): Try using internal functions for
6786 gather loads. Recognize existing calls to a gather load function.
6787 (vect_analyze_data_refs): Consider using gather loads if
6788 supports_vec_gather_load_p.
6789 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
6790 (vect_get_gather_scatter_offset_type): Likewise.
6791 (vect_convert_mask_for_vectype): Likewise.
6792 (vect_add_conversion_to_patterm): Likewise.
6793 (vect_try_gather_scatter_pattern): Likewise.
6794 (vect_recog_gather_scatter_pattern): New pattern recognizer.
6795 (vect_vect_recog_func_ptrs): Add it.
6796 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6797 internal_fn_mask_index and internal_gather_scatter_fn_p.
6798 (check_load_store_masking): Take the gather_scatter_info as an
6799 argument and handle gather loads.
6800 (vect_get_gather_scatter_ops): New function.
6801 (vectorizable_call): Check internal_load_fn_p.
6802 (vectorizable_load): Likewise. Handle gather load internal
6804 (vectorizable_store): Update call to check_load_store_masking.
6805 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
6806 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
6807 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
6808 (aarch64_gather_scale_operand_d): New predicates.
6809 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
6810 (mask_gather_load<mode>): New insns.
6812 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6813 Alan Hayward <alan.hayward@arm.com>
6814 David Sherwood <david.sherwood@arm.com>
6816 * optabs.def (fold_left_plus_optab): New optab.
6817 * doc/md.texi (fold_left_plus_@var{m}): Document.
6818 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
6819 * internal-fn.c (fold_left_direct): Define.
6820 (expand_fold_left_optab_fn): Likewise.
6821 (direct_fold_left_optab_supported_p): Likewise.
6822 * fold-const-call.c (fold_const_fold_left): New function.
6823 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
6824 * tree-parloops.c (valid_reduction_p): New function.
6825 (gather_scalar_reductions): Use it.
6826 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
6827 (vect_finish_replace_stmt): Declare.
6828 * tree-vect-loop.c (fold_left_reduction_fn): New function.
6829 (needs_fold_left_reduction_p): New function, split out from...
6830 (vect_is_simple_reduction): ...here. Accept reductions that
6831 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
6832 (vect_force_simple_reduction): Also store the reduction type in
6833 the assignment's STMT_VINFO_REDUC_TYPE.
6834 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
6835 (merge_with_identity): New function.
6836 (vect_expand_fold_left): Likewise.
6837 (vectorize_fold_left_reduction): Likewise.
6838 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
6839 scalar phi in place for it. Check for target support and reject
6840 cases that would reassociate the operation. Defer the transform
6841 phase to vectorize_fold_left_reduction.
6842 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
6843 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
6844 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
6846 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6848 * tree-if-conv.c (predicate_mem_writes): Remove redundant
6849 call to ifc_temp_var.
6851 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6852 Alan Hayward <alan.hayward@arm.com>
6853 David Sherwood <david.sherwood@arm.com>
6855 * target.def (legitimize_address_displacement): Take the original
6856 offset as a poly_int.
6857 * targhooks.h (default_legitimize_address_displacement): Update
6859 * targhooks.c (default_legitimize_address_displacement): Likewise.
6860 * doc/tm.texi: Regenerate.
6861 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
6862 as an argument, moving assert of ad->disp == ad->disp_term to...
6863 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
6864 Try calling targetm.legitimize_address_displacement before expanding
6865 the address rather than afterwards, and adjust for the new interface.
6866 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
6867 Match the new hook interface. Handle SVE addresses.
6868 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
6871 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6873 * Makefile.in (OBJS): Add early-remat.o.
6874 * target.def (select_early_remat_modes): New hook.
6875 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
6876 * doc/tm.texi: Regenerate.
6877 * targhooks.h (default_select_early_remat_modes): Declare.
6878 * targhooks.c (default_select_early_remat_modes): New function.
6879 * timevar.def (TV_EARLY_REMAT): New timevar.
6880 * passes.def (pass_early_remat): New pass.
6881 * tree-pass.h (make_pass_early_remat): Declare.
6882 * early-remat.c: New file.
6883 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
6885 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
6887 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6888 Alan Hayward <alan.hayward@arm.com>
6889 David Sherwood <david.sherwood@arm.com>
6891 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
6892 vfm1 with a bound_epilog parameter.
6893 (vect_do_peeling): Update calls accordingly, and move the prologue
6894 call earlier in the function. Treat the base bound_epilog as 0 for
6895 fully-masked loops and retain vf - 1 for other loops. Add 1 to
6896 this base when peeling for gaps.
6897 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
6898 with fully-masked loops.
6899 (vect_estimate_min_profitable_iters): Handle the single peeled
6900 iteration in that case.
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 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
6907 single-element interleaving even if the size is not a power of 2.
6908 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
6909 accesses for single-element interleaving if the group size is
6912 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6913 Alan Hayward <alan.hayward@arm.com>
6914 David Sherwood <david.sherwood@arm.com>
6916 * doc/md.texi (fold_extract_last_@var{m}): Document.
6917 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
6918 * optabs.def (fold_extract_last_optab): New optab.
6919 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
6920 * internal-fn.c (fold_extract_direct): New macro.
6921 (expand_fold_extract_optab_fn): Likewise.
6922 (direct_fold_extract_optab_supported_p): Likewise.
6923 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
6924 * tree-vect-loop.c (vect_model_reduction_cost): Handle
6925 EXTRACT_LAST_REDUCTION.
6926 (get_initial_def_for_reduction): Do not create an initial vector
6927 for EXTRACT_LAST_REDUCTION reductions.
6928 (vectorizable_reduction): Leave the scalar phi in place for
6929 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
6930 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
6931 epilogue code for EXTRACT_LAST_REDUCTION and defer the
6932 transform phase to vectorizable_condition.
6933 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
6935 (vect_finish_stmt_generation): ...here.
6936 (vect_finish_replace_stmt): New function.
6937 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
6938 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
6940 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
6942 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6943 Alan Hayward <alan.hayward@arm.com>
6944 David Sherwood <david.sherwood@arm.com>
6946 * doc/md.texi (extract_last_@var{m}): Document.
6947 * optabs.def (extract_last_optab): New optab.
6948 * internal-fn.def (EXTRACT_LAST): New internal function.
6949 * internal-fn.c (cond_unary_direct): New macro.
6950 (expand_cond_unary_optab_fn): Likewise.
6951 (direct_cond_unary_optab_supported_p): Likewise.
6952 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
6953 loops using EXTRACT_LAST.
6954 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
6955 (extract_last_<mode>): ...this optab.
6956 (vec_extract<mode><Vel>): Update accordingly.
6958 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6959 Alan Hayward <alan.hayward@arm.com>
6960 David Sherwood <david.sherwood@arm.com>
6962 * target.def (empty_mask_is_expensive): New hook.
6963 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
6964 * doc/tm.texi: Regenerate.
6965 * targhooks.h (default_empty_mask_is_expensive): Declare.
6966 * targhooks.c (default_empty_mask_is_expensive): New function.
6967 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
6968 if the target says that empty masks are expensive.
6969 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
6971 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
6973 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6974 Alan Hayward <alan.hayward@arm.com>
6975 David Sherwood <david.sherwood@arm.com>
6977 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
6978 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
6979 (vect_use_loop_mask_for_alignment_p): New function.
6980 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
6981 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
6982 niters_skip argument. Make sure that the first niters_skip elements
6983 of the first iteration are inactive.
6984 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
6985 Update call to vect_set_loop_masks_directly.
6986 (get_misalign_in_elems): New function, split out from...
6987 (vect_gen_prolog_loop_niters): ...here.
6988 (vect_update_init_of_dr): Take a code argument that specifies whether
6989 the adjustment should be added or subtracted.
6990 (vect_update_init_of_drs): Likewise.
6991 (vect_prepare_for_masked_peels): New function.
6992 (vect_do_peeling): Skip prologue peeling if we're using a mask
6993 instead. Update call to vect_update_inits_of_drs.
6994 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6996 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
6997 alignment. Do not include the number of peeled iterations in
6998 the minimum threshold in that case.
6999 (vectorizable_induction): Adjust the start value down by
7000 LOOP_VINFO_MASK_SKIP_NITERS iterations.
7001 (vect_transform_loop): Call vect_prepare_for_masked_peels.
7002 Take the number of skipped iterations into account when calculating
7004 * tree-vect-stmts.c (vect_gen_while_not): New function.
7006 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7007 Alan Hayward <alan.hayward@arm.com>
7008 David Sherwood <david.sherwood@arm.com>
7010 * doc/sourcebuild.texi (vect_fully_masked): Document.
7011 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
7013 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
7015 (vect_analyze_loop_2): ...here. Don't check the vectorization
7016 factor against the number of loop iterations if the loop is
7019 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7020 Alan Hayward <alan.hayward@arm.com>
7021 David Sherwood <david.sherwood@arm.com>
7023 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
7024 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
7025 (dump_groups): Update accordingly.
7026 (iv_use::mem_type): New member variable.
7027 (address_p): New function.
7028 (record_use): Add a mem_type argument and initialize the new
7030 (record_group_use): Add a mem_type argument. Use address_p.
7031 Remove obsolete null checks of base_object. Update call to record_use.
7032 (find_interesting_uses_op): Update call to record_group_use.
7033 (find_interesting_uses_cond): Likewise.
7034 (find_interesting_uses_address): Likewise.
7035 (get_mem_type_for_internal_fn): New function.
7036 (find_address_like_use): Likewise.
7037 (find_interesting_uses_stmt): Try find_address_like_use before
7038 calling find_interesting_uses_op.
7039 (addr_offset_valid_p): Use the iv mem_type field as the type
7040 of the addressed memory.
7041 (add_autoinc_candidates): Likewise.
7042 (get_address_cost): Likewise.
7043 (split_small_address_groups_p): Use address_p.
7044 (split_address_groups): Likewise.
7045 (add_iv_candidate_for_use): Likewise.
7046 (autoinc_possible_for_pair): Likewise.
7047 (rewrite_groups): Likewise.
7048 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
7049 (determine_group_iv_cost): Update after split of USE_ADDRESS.
7050 (get_alias_ptr_type_for_ptr_address): New function.
7051 (rewrite_use_address): Rewrite address uses in calls that were
7052 identified by find_address_like_use.
7054 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7055 Alan Hayward <alan.hayward@arm.com>
7056 David Sherwood <david.sherwood@arm.com>
7058 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
7060 * gimple-expr.h (is_gimple_addressable: Likewise.
7061 * gimple-expr.c (is_gimple_address): Likewise.
7062 * internal-fn.c (expand_call_mem_ref): New function.
7063 (expand_mask_load_optab_fn): Use it.
7064 (expand_mask_store_optab_fn): Likewise.
7066 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7067 Alan Hayward <alan.hayward@arm.com>
7068 David Sherwood <david.sherwood@arm.com>
7070 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
7071 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
7072 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
7073 (cond_umax@var{mode}): Document.
7074 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
7075 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
7076 (cond_umin_optab, cond_umax_optab): New optabs.
7077 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
7078 (COND_IOR, COND_XOR): New internal functions.
7079 * internal-fn.h (get_conditional_internal_fn): Declare.
7080 * internal-fn.c (cond_binary_direct): New macro.
7081 (expand_cond_binary_optab_fn): Likewise.
7082 (direct_cond_binary_optab_supported_p): Likewise.
7083 (get_conditional_internal_fn): New function.
7084 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
7085 Cope with reduction statements that are vectorized as calls rather
7087 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
7088 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
7089 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
7090 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
7091 (UNSPEC_COND_EOR): New unspecs.
7092 (optab): Add mappings for them.
7093 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
7094 (sve_int_op, sve_fp_op): New int attributes.
7096 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7097 Alan Hayward <alan.hayward@arm.com>
7098 David Sherwood <david.sherwood@arm.com>
7100 * optabs.def (while_ult_optab): New optab.
7101 * doc/md.texi (while_ult@var{m}@var{n}): Document.
7102 * internal-fn.def (WHILE_ULT): New internal function.
7103 * internal-fn.h (direct_internal_fn_supported_p): New override
7104 that takes two types as argument.
7105 * internal-fn.c (while_direct): New macro.
7106 (expand_while_optab_fn): New function.
7107 (convert_optab_supported_p): Likewise.
7108 (direct_while_optab_supported_p): New macro.
7109 * wide-int.h (wi::udiv_ceil): New function.
7110 * tree-vectorizer.h (rgroup_masks): New structure.
7111 (vec_loop_masks): New typedef.
7112 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
7114 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
7115 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
7116 (vect_max_vf): New function.
7117 (slpeel_make_loop_iterate_ntimes): Delete.
7118 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
7119 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
7120 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
7121 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
7122 internal-fn.h, stor-layout.h and optabs-query.h.
7123 (vect_set_loop_mask): New function.
7124 (add_preheader_seq): Likewise.
7125 (add_header_seq): Likewise.
7126 (interleave_supported_p): Likewise.
7127 (vect_maybe_permute_loop_masks): Likewise.
7128 (vect_set_loop_masks_directly): Likewise.
7129 (vect_set_loop_condition_masked): Likewise.
7130 (vect_set_loop_condition_unmasked): New function, split out from
7131 slpeel_make_loop_iterate_ntimes.
7132 (slpeel_make_loop_iterate_ntimes): Rename to..
7133 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
7134 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
7135 (vect_do_peeling): Update call accordingly.
7136 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
7138 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7139 mask_compare_type, can_fully_mask_p and fully_masked_p.
7140 (release_vec_loop_masks): New function.
7141 (_loop_vec_info): Use it to free the loop masks.
7142 (can_produce_all_loop_masks_p): New function.
7143 (vect_get_max_nscalars_per_iter): Likewise.
7144 (vect_verify_full_masking): Likewise.
7145 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
7146 retries, and free the mask rgroups before retrying. Check loop-wide
7147 reasons for disallowing fully-masked loops. Make the final decision
7148 about whether use a fully-masked loop or not.
7149 (vect_estimate_min_profitable_iters): Do not assume that peeling
7150 for the number of iterations will be needed for fully-masked loops.
7151 (vectorizable_reduction): Disable fully-masked loops.
7152 (vectorizable_live_operation): Likewise.
7153 (vect_halve_mask_nunits): New function.
7154 (vect_double_mask_nunits): Likewise.
7155 (vect_record_loop_mask): Likewise.
7156 (vect_get_loop_mask): Likewise.
7157 (vect_transform_loop): Handle the case in which the final loop
7158 iteration might handle a partial vector. Call vect_set_loop_condition
7159 instead of slpeel_make_loop_iterate_ntimes.
7160 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
7161 (check_load_store_masking): New function.
7162 (prepare_load_store_mask): Likewise.
7163 (vectorizable_store): Handle fully-masked loops.
7164 (vectorizable_load): Likewise.
7165 (supportable_widening_operation): Use vect_halve_mask_nunits for
7167 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
7168 (vect_gen_while): New function.
7169 * config/aarch64/aarch64.md (umax<mode>3): New expander.
7170 (aarch64_uqdec<mode>): New insn.
7172 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7173 Alan Hayward <alan.hayward@arm.com>
7174 David Sherwood <david.sherwood@arm.com>
7176 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
7177 (reduc_xor_scal_optab): New optabs.
7178 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
7179 (reduc_xor_scal_@var{m}): Document.
7180 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
7181 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
7183 * fold-const-call.c (fold_const_call): Handle them.
7184 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
7185 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
7186 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
7187 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
7188 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
7189 (UNSPEC_XORV): New unspecs.
7190 (optab): Add entries for them.
7191 (BITWISEV): New int iterator.
7192 (bit_reduc_op): New int attributes.
7194 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7195 Alan Hayward <alan.hayward@arm.com>
7196 David Sherwood <david.sherwood@arm.com>
7198 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
7199 * internal-fn.def (VEC_SHL_INSERT): New internal function.
7200 * optabs.def (vec_shl_insert_optab): New optab.
7201 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
7202 (duplicate_and_interleave): Likewise.
7203 * tree-vect-loop.c: Include internal-fn.h.
7204 (neutral_op_for_slp_reduction): New function, split out from
7205 get_initial_defs_for_reduction.
7206 (get_initial_def_for_reduction): Handle option 2 for variable-length
7207 vectors by loading the neutral value into a vector and then shifting
7208 the initial value into element 0.
7209 (get_initial_defs_for_reduction): Replace the code argument with
7210 the neutral value calculated by neutral_op_for_slp_reduction.
7211 Use gimple_build_vector for constant-length vectors.
7212 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
7213 but the first group_size elements have a neutral value.
7214 Use duplicate_and_interleave otherwise.
7215 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
7216 Update call to get_initial_defs_for_reduction. Handle SLP
7217 reductions for variable-length vectors by creating one vector
7218 result for each scalar result, with the elements associated
7219 with other scalar results stubbed out with the neutral value.
7220 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
7221 Require IFN_VEC_SHL_INSERT for double reductions on
7222 variable-length vectors, or SLP reductions that have
7223 a neutral value. Require can_duplicate_and_interleave_p
7224 support for variable-length unchained SLP reductions if there
7225 is no neutral value, such as for MIN/MAX reductions. Also require
7226 the number of vector elements to be a multiple of the number of
7227 SLP statements when doing variable-length unchained SLP reductions.
7228 Update call to vect_create_epilog_for_reduction.
7229 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
7230 and remove initial values.
7231 (duplicate_and_interleave): Make public.
7232 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
7233 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
7235 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7236 Alan Hayward <alan.hayward@arm.com>
7237 David Sherwood <david.sherwood@arm.com>
7239 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
7240 (can_duplicate_and_interleave_p): New function.
7241 (vect_get_and_check_slp_defs): Take the vector of statements
7242 rather than just the current one. Remove excess parentheses.
7243 Restriction rejectinon of vect_constant_def and vect_external_def
7244 for variable-length vectors to boolean types, or types for which
7245 can_duplicate_and_interleave_p is false.
7246 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
7247 (duplicate_and_interleave): New function.
7248 (vect_get_constant_vectors): Use gimple_build_vector for
7249 constant-length vectors and suitable variable-length constant
7250 vectors. Use duplicate_and_interleave for other variable-length
7251 vectors. Don't defer the update when inserting new statements.
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 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
7258 min_profitable_iters doesn't go negative.
7260 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7261 Alan Hayward <alan.hayward@arm.com>
7262 David Sherwood <david.sherwood@arm.com>
7264 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
7265 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
7266 * optabs.def (vec_mask_load_lanes_optab): New optab.
7267 (vec_mask_store_lanes_optab): Likewise.
7268 * internal-fn.def (MASK_LOAD_LANES): New internal function.
7269 (MASK_STORE_LANES): Likewise.
7270 * internal-fn.c (mask_load_lanes_direct): New macro.
7271 (mask_store_lanes_direct): Likewise.
7272 (expand_mask_load_optab_fn): Handle masked operations.
7273 (expand_mask_load_lanes_optab_fn): New macro.
7274 (expand_mask_store_optab_fn): Handle masked operations.
7275 (expand_mask_store_lanes_optab_fn): New macro.
7276 (direct_mask_load_lanes_optab_supported_p): Likewise.
7277 (direct_mask_store_lanes_optab_supported_p): Likewise.
7278 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
7280 (vect_load_lanes_supported): Likewise.
7281 * tree-vect-data-refs.c (strip_conversion): New function.
7282 (can_group_stmts_p): Likewise.
7283 (vect_analyze_data_ref_accesses): Use it instead of checking
7284 for a pair of assignments.
7285 (vect_store_lanes_supported): Take a masked_p parameter.
7286 (vect_load_lanes_supported): Likewise.
7287 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
7288 vect_store_lanes_supported and vect_load_lanes_supported.
7289 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7290 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
7291 parameter. Don't allow gaps for masked accesses.
7292 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
7293 and vect_load_lanes_supported.
7294 (get_load_store_type): Take a masked_p parameter and update
7295 call to get_group_load_store_type.
7296 (vectorizable_store): Update call to get_load_store_type.
7297 Handle IFN_MASK_STORE_LANES.
7298 (vectorizable_load): Update call to get_load_store_type.
7299 Handle IFN_MASK_LOAD_LANES.
7301 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7302 Alan Hayward <alan.hayward@arm.com>
7303 David Sherwood <david.sherwood@arm.com>
7305 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
7307 * config/aarch64/aarch64-protos.h
7308 (aarch64_sve_struct_memory_operand_p): Declare.
7309 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
7310 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
7311 (VPRED, vpred): Handle SVE structure modes.
7312 * config/aarch64/constraints.md (Utx): New constraint.
7313 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
7314 (aarch64_sve_struct_nonimmediate_operand): New predicates.
7315 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
7316 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
7317 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
7318 structure modes. Split into pieces after RA.
7319 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
7320 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
7322 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
7323 SVE structure modes.
7324 (aarch64_classify_address): Likewise.
7325 (sizetochar): Move earlier in file.
7326 (aarch64_print_operand): Handle SVE register lists.
7327 (aarch64_array_mode): New function.
7328 (aarch64_sve_struct_memory_operand_p): Likewise.
7329 (TARGET_ARRAY_MODE): Redefine.
7331 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7332 Alan Hayward <alan.hayward@arm.com>
7333 David Sherwood <david.sherwood@arm.com>
7335 * target.def (array_mode): New target hook.
7336 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
7337 * doc/tm.texi: Regenerate.
7338 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
7339 * hooks.c (hook_optmode_mode_uhwi_none): New function.
7340 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
7342 * stor-layout.c (mode_for_array): Likewise. Support polynomial
7345 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7346 Alan Hayward <alan.hayward@arm.com>
7347 David Sherwood <david.sherwood@arm.com>
7349 * fold-const.c (fold_binary_loc): Check the argument types
7350 rather than the result type when testing for a vector operation.
7352 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7354 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
7355 * doc/tm.texi: Regenerate.
7357 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7358 Alan Hayward <alan.hayward@arm.com>
7359 David Sherwood <david.sherwood@arm.com>
7361 * doc/invoke.texi (-msve-vector-bits=): Document new option.
7362 (sve): Document new AArch64 extension.
7363 * doc/md.texi (w): Extend the description of the AArch64
7364 constraint to include SVE vectors.
7365 (Upl, Upa): Document new AArch64 predicate constraints.
7366 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
7368 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
7369 (msve-vector-bits=): New option.
7370 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
7371 SVE when these are disabled.
7372 (sve): New extension.
7373 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
7374 modes. Adjust their number of units based on aarch64_sve_vg.
7375 (MAX_BITSIZE_MODE_ANY_MODE): Define.
7376 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
7377 aarch64_addr_query_type.
7378 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
7379 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
7380 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
7381 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
7382 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
7383 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
7384 (aarch64_simd_imm_zero_p): Delete.
7385 (aarch64_check_zero_based_sve_index_immediate): Declare.
7386 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7387 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7388 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7389 (aarch64_sve_float_mul_immediate_p): Likewise.
7390 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7392 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
7393 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
7394 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
7395 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
7396 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
7397 (aarch64_regmode_natural_size): Likewise.
7398 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
7399 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
7401 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
7402 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
7403 for VG and the SVE predicate registers.
7404 (V_ALIASES): Add a "z"-prefixed alias.
7405 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
7406 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
7407 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
7408 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
7409 (REG_CLASS_NAMES): Add entries for them.
7410 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
7411 and the predicate registers.
7412 (aarch64_sve_vg): Declare.
7413 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
7414 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
7415 (REGMODE_NATURAL_SIZE): Define.
7416 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
7418 * config/aarch64/aarch64.c: Include cfgrtl.h.
7419 (simd_immediate_info): Add a constructor for series vectors,
7420 and an associated step field.
7421 (aarch64_sve_vg): New variable.
7422 (aarch64_dbx_register_number): Handle VG and the predicate registers.
7423 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
7424 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
7425 (VEC_ANY_DATA, VEC_STRUCT): New constants.
7426 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
7427 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
7428 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
7429 (aarch64_get_mask_mode): New functions.
7430 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
7431 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
7432 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
7433 predicate modes and predicate registers. Explicitly restrict
7434 GPRs to modes of 16 bytes or smaller. Only allow FP registers
7435 to store a vector mode if it is recognized by
7436 aarch64_classify_vector_mode.
7437 (aarch64_regmode_natural_size): New function.
7438 (aarch64_hard_regno_caller_save_mode): Return the original mode
7440 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
7441 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
7442 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
7443 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
7445 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
7446 does not overlap dest if the function is frame-related. Handle
7448 (aarch64_split_add_offset): New function.
7449 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
7450 them aarch64_add_offset.
7451 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
7452 and update call to aarch64_sub_sp.
7453 (aarch64_add_cfa_expression): New function.
7454 (aarch64_expand_prologue): Pass extra temporary registers to the
7455 functions above. Handle the case in which we need to emit new
7456 DW_CFA_expressions for registers that were originally saved
7457 relative to the stack pointer, but now have to be expressed
7458 relative to the frame pointer.
7459 (aarch64_output_mi_thunk): Pass extra temporary registers to the
7461 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
7462 IP0 and IP1 values for SVE frames.
7463 (aarch64_expand_vec_series): New function.
7464 (aarch64_expand_sve_widened_duplicate): Likewise.
7465 (aarch64_expand_sve_const_vector): Likewise.
7466 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
7467 Handle SVE constants. Use emit_move_insn to move a force_const_mem
7468 into the register, rather than emitting a SET directly.
7469 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
7470 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
7471 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
7472 (offset_9bit_signed_scaled_p): New functions.
7473 (aarch64_replicate_bitmask_imm): New function.
7474 (aarch64_bitmask_imm): Use it.
7475 (aarch64_cannot_force_const_mem): Reject expressions involving
7476 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
7477 (aarch64_classify_index): Handle SVE indices, by requiring
7478 a plain register index with a scale that matches the element size.
7479 (aarch64_classify_address): Handle SVE addresses. Assert that
7480 the mode of the address is VOIDmode or an integer mode.
7481 Update call to aarch64_classify_symbol.
7482 (aarch64_classify_symbolic_expression): Update call to
7483 aarch64_classify_symbol.
7484 (aarch64_const_vec_all_in_range_p): New function.
7485 (aarch64_print_vector_float_operand): Likewise.
7486 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
7487 "vN" for FP registers with SVE modes. Handle (const ...) vectors
7488 and the FP immediates 1.0 and 0.5.
7489 (aarch64_print_address_internal): Handle SVE addresses.
7490 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
7491 (aarch64_regno_regclass): Handle predicate registers.
7492 (aarch64_secondary_reload): Handle big-endian reloads of SVE
7494 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
7495 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
7496 (aarch64_convert_sve_vector_bits): New function.
7497 (aarch64_override_options): Use it to handle -msve-vector-bits=.
7498 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7500 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
7501 Handle SVE vector and predicate modes. Accept VL-based constants
7502 that need only one temporary register, and VL offsets that require
7503 no temporary registers.
7504 (aarch64_conditional_register_usage): Mark the predicate registers
7505 as fixed if SVE isn't available.
7506 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
7507 Return true for SVE vector and predicate modes.
7508 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
7509 rather than an unsigned int. Handle SVE modes.
7510 (aarch64_preferred_simd_mode): Update call accordingly. Handle
7512 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
7514 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7515 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7516 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7517 (aarch64_sve_float_mul_immediate_p): New functions.
7518 (aarch64_sve_valid_immediate): New function.
7519 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
7520 Explicitly reject structure modes. Check for INDEX constants.
7521 Handle PTRUE and PFALSE constants.
7522 (aarch64_check_zero_based_sve_index_immediate): New function.
7523 (aarch64_simd_imm_zero_p): Delete.
7524 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
7525 vector modes. Accept constants in the range of CNT[BHWD].
7526 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
7527 ask for an Advanced SIMD mode.
7528 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
7529 (aarch64_simd_vector_alignment): Handle SVE predicates.
7530 (aarch64_vectorize_preferred_vector_alignment): New function.
7531 (aarch64_simd_vector_alignment_reachable): Use it instead of
7533 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
7534 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
7536 (MAX_VECT_LEN): Delete.
7537 (expand_vec_perm_d): Add a vec_flags field.
7538 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
7539 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
7540 (aarch64_evpc_ext): Don't apply a big-endian lane correction
7542 (aarch64_evpc_rev): Rename to...
7543 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
7544 (aarch64_evpc_rev_global): New function.
7545 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
7546 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
7548 (aarch64_evpc_sve_tbl): New function.
7549 (aarch64_expand_vec_perm_const_1): Update after rename of
7550 aarch64_evpc_rev. Handle SVE permutes too, trying
7551 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
7552 than aarch64_evpc_tbl.
7553 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
7554 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
7555 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
7556 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
7557 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
7558 (aarch64_expand_sve_vcond): New functions.
7559 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
7560 of aarch64_vector_mode_p.
7561 (aarch64_dwarf_poly_indeterminate_value): New function.
7562 (aarch64_compute_pressure_classes): Likewise.
7563 (aarch64_can_change_mode_class): Likewise.
7564 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
7565 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
7566 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
7567 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
7568 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
7569 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
7570 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
7571 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
7573 (Dn, Dl, Dr): Accept const as well as const_vector.
7574 (Dz): Likewise. Compare against CONST0_RTX.
7575 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
7576 of "vector" where appropriate.
7577 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
7578 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
7579 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
7580 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
7581 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
7582 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
7583 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
7584 (v_int_equiv): Extend to SVE modes.
7585 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
7587 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
7588 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
7589 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
7590 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
7591 (SVE_COND_FP_CMP): New int iterators.
7592 (perm_hilo): Handle the new unpack unspecs.
7593 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
7595 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
7596 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
7597 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
7598 (aarch64_equality_operator, aarch64_constant_vector_operand)
7599 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
7600 (aarch64_sve_nonimmediate_operand): Likewise.
7601 (aarch64_sve_general_operand): Likewise.
7602 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
7603 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
7604 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
7605 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
7606 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
7607 (aarch64_sve_float_arith_immediate): Likewise.
7608 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
7609 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
7610 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
7611 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
7612 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
7613 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
7614 (aarch64_sve_float_arith_operand): Likewise.
7615 (aarch64_sve_float_arith_with_sub_operand): Likewise.
7616 (aarch64_sve_float_mul_operand): Likewise.
7617 (aarch64_sve_vec_perm_operand): Likewise.
7618 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
7619 (aarch64_mov_operand): Accept const_poly_int and const_vector.
7620 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
7621 as well as const_vector.
7622 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
7623 in file. Use CONST0_RTX and CONSTM1_RTX.
7624 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
7625 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
7626 Use aarch64_simd_imm_zero.
7627 * config/aarch64/aarch64-sve.md: New file.
7628 * config/aarch64/aarch64.md: Include it.
7629 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
7630 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
7631 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
7632 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
7633 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
7634 (sve): New attribute.
7635 (enabled): Disable instructions with the sve attribute unless
7637 (movqi, movhi): Pass CONST_POLY_INT operaneds through
7638 aarch64_expand_mov_immediate.
7639 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
7640 CNT[BHSD] immediates.
7641 (movti): Split CONST_POLY_INT moves into two halves.
7642 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
7643 Split additions that need a temporary here if the destination
7644 is the stack pointer.
7645 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
7646 (*add<mode>3_poly_1): New instruction.
7647 (set_clobber_cc): New expander.
7649 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7651 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
7652 parameter and use it instead of GET_MODE_SIZE (innermode). Use
7653 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
7654 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
7655 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
7656 Change innermode from fixed_mode_size to machine_mode.
7657 (simplify_subreg): Update call accordingly. Handle a constant-sized
7658 subreg of a variable-length CONST_VECTOR.
7660 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7661 Alan Hayward <alan.hayward@arm.com>
7662 David Sherwood <david.sherwood@arm.com>
7664 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
7665 (add_offset_to_base): New function, split out from...
7666 (create_mem_ref): ...here. When handling a scale other than 1,
7667 check first whether the address is valid without the offset.
7668 Add it into the base if so, leaving the index and scale as-is.
7670 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7673 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
7674 fold_for_warn before checking if arg2 is INTEGER_CST.
7676 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
7678 * config/rs6000/predicates.md (load_multiple_operation): Delete.
7679 (store_multiple_operation): Delete.
7680 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
7681 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
7682 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
7683 guarded by TARGET_STRING.
7684 (rs6000_output_load_multiple): Delete.
7685 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
7686 OPTION_MASK_STRING / TARGET_STRING handling.
7687 (print_operand) <'N', 'O'>: Add comment that these are unused now.
7688 (const rs6000_opt_masks) <"string">: Change mask to 0.
7689 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
7690 (MASK_STRING): Delete.
7691 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
7693 (load_multiple): Delete.
7700 (store_multiple): Delete.
7707 (movmemsi_8reg): Delete.
7708 (corresponding unnamed define_insn): Delete.
7709 (movmemsi_6reg): Delete.
7710 (corresponding unnamed define_insn): Delete.
7711 (movmemsi_4reg): Delete.
7712 (corresponding unnamed define_insn): Delete.
7713 (movmemsi_2reg): Delete.
7714 (corresponding unnamed define_insn): Delete.
7715 (movmemsi_1reg): Delete.
7716 (corresponding unnamed define_insn): Delete.
7717 * config/rs6000/rs6000.opt (mno-string): New.
7718 (mstring): Replace by deprecation warning stub.
7719 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
7721 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7723 * regrename.c (regrename_do_replace): If replacing the same
7724 reg multiple times, try to reuse last created gen_raw_REG.
7727 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
7728 main to workaround a bug in GDB.
7730 2018-01-12 Tom de Vries <tom@codesourcery.com>
7733 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
7735 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
7737 PR rtl-optimization/80481
7738 * ira-color.c (get_cap_member): New function.
7739 (allocnos_conflict_by_live_ranges_p): Use it.
7740 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
7741 (setup_slot_coalesced_allocno_live_ranges): Ditto.
7743 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
7746 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
7747 (*saddl_se_1): Ditto.
7749 (*ssubl_se_1): Ditto.
7751 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7753 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
7754 rather than wi::to_widest for DR_INITs.
7755 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
7756 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
7757 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
7759 (vect_analyze_group_access_1): Note that here.
7761 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7763 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
7764 polynomial type sizes.
7766 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7768 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
7769 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
7770 (gimple_add_tmp_var): Likewise.
7772 2018-01-12 Martin Liska <mliska@suse.cz>
7774 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
7775 (gimple_alloc_sizes): Likewise.
7776 (dump_gimple_statistics): Use PRIu64 in printf format.
7777 * gimple.h: Change uint64_t to int.
7779 2018-01-12 Martin Liska <mliska@suse.cz>
7781 * tree-core.h: Use uint64_t instead of int.
7782 * tree.c (tree_node_counts): Likewise.
7783 (tree_node_sizes): Likewise.
7784 (dump_tree_statistics): Use PRIu64 in printf format.
7786 2018-01-12 Martin Liska <mliska@suse.cz>
7788 * Makefile.in: As qsort_chk is implemented in vec.c, add
7789 vec.o to linkage of gencfn-macros.
7790 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
7791 passing the info to record_node_allocation_statistics.
7792 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
7794 * ggc-common.c (struct ggc_usage): Add operator== and use
7795 it in operator< and compare function.
7796 * mem-stats.h (struct mem_usage): Likewise.
7797 * vec.c (struct vec_usage): Remove operator< and compare
7798 function. Can be simply inherited.
7800 2018-01-12 Martin Jambor <mjambor@suse.cz>
7803 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
7804 * tree-ssa-math-opts.c: Include domwalk.h.
7805 (convert_mult_to_fma_1): New function.
7806 (fma_transformation_info): New type.
7807 (fma_deferring_state): Likewise.
7808 (cancel_fma_deferring): New function.
7809 (result_of_phi): Likewise.
7810 (last_fma_candidate_feeds_initial_phi): Likewise.
7811 (convert_mult_to_fma): Added deferring logic, split actual
7812 transformation to convert_mult_to_fma_1.
7813 (math_opts_dom_walker): New type.
7814 (math_opts_dom_walker::after_dom_children): New method, body moved
7815 here from pass_optimize_widening_mul::execute, added deferring logic
7817 (pass_optimize_widening_mul::execute): Moved most of code to
7818 math_opts_dom_walker::after_dom_children.
7819 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
7820 * config/i386/i386.c (ix86_option_override_internal): Added
7821 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
7823 2018-01-12 Richard Biener <rguenther@suse.de>
7826 * dwarf2out.c (gen_variable_die): Do not reset old_die for
7827 inline instance vars.
7829 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
7832 * config/rx/rx.c (rx_is_restricted_memory_address):
7835 2018-01-12 Richard Biener <rguenther@suse.de>
7837 PR tree-optimization/80846
7838 * target.def (split_reduction): New target hook.
7839 * targhooks.c (default_split_reduction): New function.
7840 * targhooks.h (default_split_reduction): Declare.
7841 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
7842 target requests first reduce vectors by combining low and high
7844 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
7845 (get_vectype_for_scalar_type_and_size): Export.
7846 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
7847 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
7848 * doc/tm.texi: Regenerate.
7849 * config/i386/i386.c (ix86_split_reduction): Implement
7850 TARGET_VECTORIZE_SPLIT_REDUCTION.
7852 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7855 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
7856 in PIC mode except for TARGET_VXWORKS_RTP.
7857 * config/sparc/sparc.c: Include cfgrtl.h.
7858 (TARGET_INIT_PIC_REG): Define.
7859 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
7860 (sparc_pic_register_p): New predicate.
7861 (sparc_legitimate_address_p): Use it.
7862 (sparc_legitimize_pic_address): Likewise.
7863 (sparc_delegitimize_address): Likewise.
7864 (sparc_mode_dependent_address_p): Likewise.
7865 (gen_load_pcrel_sym): Remove 4th parameter.
7866 (load_got_register): Adjust call to above. Remove obsolete stuff.
7867 (sparc_expand_prologue): Do not call load_got_register here.
7868 (sparc_flat_expand_prologue): Likewise.
7869 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
7870 (sparc_use_pseudo_pic_reg): New function.
7871 (sparc_init_pic_reg): Likewise.
7872 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7873 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
7875 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
7877 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
7878 Add item for branch_cost.
7880 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7882 PR rtl-optimization/83565
7883 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
7884 not extend the result to a larger mode for rotate operations.
7885 (num_sign_bit_copies1): Likewise.
7887 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7890 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
7892 Use values-Xc.o for -pedantic.
7893 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
7895 2018-01-12 Martin Liska <mliska@suse.cz>
7898 * ipa-devirt.c (final_warning_record::grow_type_warnings):
7900 (possible_polymorphic_call_targets): Use it.
7901 (ipa_devirt): Likewise.
7903 2018-01-12 Martin Liska <mliska@suse.cz>
7905 * profile-count.h (enum profile_quality): Use 0 as invalid
7906 enum value of profile_quality.
7908 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
7910 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
7911 -mext-string options.
7913 2018-01-12 Richard Biener <rguenther@suse.de>
7915 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
7916 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
7917 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
7919 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
7921 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
7923 * configure.ac (--with-long-double-format): Add support for the
7924 configuration option to change the default long double format on
7926 * config.gcc (powerpc*-linux*-*): Likewise.
7927 * configure: Regenerate.
7928 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
7929 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
7930 used without modification.
7932 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7934 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
7935 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
7936 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
7937 MISC_BUILTIN_SPEC_BARRIER.
7938 (rs6000_init_builtins): Likewise.
7939 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
7941 (speculation_barrier): New define_insn.
7942 * doc/extend.texi: Document __builtin_speculation_barrier.
7944 2018-01-11 Jakub Jelinek <jakub@redhat.com>
7947 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
7948 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
7949 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
7951 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
7952 integral modes instead of "ss" and "sd".
7953 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
7954 vectors with 32-bit and 64-bit elements.
7955 (vecdupssescalarmodesuffix): New mode attribute.
7956 (vec_dup<mode>): Use it.
7958 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
7961 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
7962 frame if argument is passed on stack.
7964 2018-01-11 Jakub Jelinek <jakub@redhat.com>
7967 * ree.c (combine_reaching_defs): Optimize also
7968 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
7969 reg2=any_extend(exp); reg1=reg2;, formatting fix.
7971 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
7974 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
7976 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
7979 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
7980 after they are computed.
7982 2018-01-11 Bin Cheng <bin.cheng@arm.com>
7984 PR tree-optimization/83695
7985 * gimple-loop-linterchange.cc
7986 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
7987 reset cached scev information after interchange.
7988 (pass_linterchange::execute): Remove call to scev_reset_htab.
7990 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7992 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
7993 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
7994 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
7995 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
7996 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
7997 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
7998 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
7999 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
8000 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
8001 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
8002 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
8003 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
8004 (V_lane_reg): Likewise.
8005 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
8007 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
8008 (vfmal_lane_low<mode>_intrinsic,
8009 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
8010 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
8011 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
8012 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
8013 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
8014 vfmsl_lane_high<mode>_intrinsic): New define_insns.
8016 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8018 * config/arm/arm-cpus.in (fp16fml): New feature.
8019 (ALL_SIMD): Add fp16fml.
8020 (armv8.2-a): Add fp16fml as an option.
8021 (armv8.3-a): Likewise.
8022 (armv8.4-a): Add fp16fml as part of fp16.
8023 * config/arm/arm.h (TARGET_FP16FML): Define.
8024 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
8026 * config/arm/arm-modes.def (V2HF): Define.
8027 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
8028 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
8029 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
8030 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
8031 vfmsl_low, vfmsl_high): New set of builtins.
8032 * config/arm/iterators.md (PLUSMINUS): New code iterator.
8033 (vfml_op): New code attribute.
8034 (VFMLHALVES): New int iterator.
8035 (VFML, VFMLSEL): New mode attributes.
8036 (V_reg): Define mapping for V2HF.
8037 (V_hi, V_lo): New mode attributes.
8038 (VF_constraint): Likewise.
8039 (vfml_half, vfml_half_selector): New int attributes.
8040 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
8042 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
8043 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
8045 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
8046 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
8047 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
8048 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
8050 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
8051 Document new effective target and option set.
8053 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8055 * config/arm/arm-cpus.in (armv8_4): New feature.
8056 (ARMv8_4a): New fgroup.
8057 (armv8.4-a): New arch.
8058 * config/arm/arm-tables.opt: Regenerate.
8059 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
8060 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
8061 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
8062 Add matching rules for -march=armv8.4-a and extensions.
8063 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
8065 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
8068 * config/rx/rx.md (BW): New mode attribute.
8069 (sync_lock_test_and_setsi): Add mode suffix to insn output.
8071 2018-01-11 Richard Biener <rguenther@suse.de>
8073 PR tree-optimization/83435
8074 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
8075 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
8076 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
8078 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8079 Alan Hayward <alan.hayward@arm.com>
8080 David Sherwood <david.sherwood@arm.com>
8082 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
8084 (aarch64_classify_address): Initialize it. Track polynomial offsets.
8085 (aarch64_print_address_internal): Use it to check for a zero offset.
8087 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8088 Alan Hayward <alan.hayward@arm.com>
8089 David Sherwood <david.sherwood@arm.com>
8091 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
8092 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
8093 Return a poly_int64 rather than a HOST_WIDE_INT.
8094 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
8095 rather than a HOST_WIDE_INT.
8096 * config/aarch64/aarch64.h (aarch64_frame): Protect with
8097 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
8098 hard_fp_offset, frame_size, initial_adjust, callee_offset and
8099 final_offset from HOST_WIDE_INT to poly_int64.
8100 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
8101 to_constant when getting the number of units in an Advanced SIMD
8103 (aarch64_builtin_vectorized_function): Check for a constant number
8105 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
8107 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
8108 attribute instead of GET_MODE_NUNITS.
8109 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
8110 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
8111 GET_MODE_SIZE for fixed-size registers.
8112 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
8113 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
8114 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
8115 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
8116 (aarch64_print_operand, aarch64_print_address_internal)
8117 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
8118 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
8119 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
8120 Handle polynomial GET_MODE_SIZE.
8121 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
8122 wider than SImode without modification.
8123 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
8124 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
8125 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
8126 passing and returning SVE modes.
8127 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
8128 rather than GEN_INT.
8129 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
8130 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
8131 (aarch64_allocate_and_probe_stack_space): Likewise.
8132 (aarch64_layout_frame): Cope with polynomial offsets.
8133 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
8134 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
8136 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
8137 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
8138 poly_int64 rather than a HOST_WIDE_INT.
8139 (aarch64_get_separate_components, aarch64_process_components)
8140 (aarch64_expand_prologue, aarch64_expand_epilogue)
8141 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
8142 (aarch64_anchor_offset): New function, split out from...
8143 (aarch64_legitimize_address): ...here.
8144 (aarch64_builtin_vectorization_cost): Handle polynomial
8145 TYPE_VECTOR_SUBPARTS.
8146 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
8148 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
8149 number of elements from the PARALLEL rather than the mode.
8150 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
8151 rather than GET_MODE_BITSIZE.
8152 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
8153 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
8154 (aarch64_expand_vec_perm_const_1): Handle polynomial
8155 d->perm.length () and d->perm elements.
8156 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
8157 Apply to_constant to d->perm elements.
8158 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
8159 polynomial CONST_VECTOR_NUNITS.
8160 (aarch64_move_pointer): Take amount as a poly_int64 rather
8162 (aarch64_progress_pointer): Avoid temporary variable.
8163 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
8164 the mode attribute instead of GET_MODE.
8166 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8167 Alan Hayward <alan.hayward@arm.com>
8168 David Sherwood <david.sherwood@arm.com>
8170 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
8171 x exists before using it.
8172 (aarch64_add_constant_internal): Rename to...
8173 (aarch64_add_offset_1): ...this. Replace regnum with separate
8174 src and dest rtxes. Handle the case in which they're different,
8175 including when the offset is zero. Replace scratchreg with an rtx.
8176 Use 2 additions if there is no spare register into which we can
8177 move a 16-bit constant.
8178 (aarch64_add_constant): Delete.
8179 (aarch64_add_offset): Replace reg with separate src and dest
8180 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
8181 Use aarch64_add_offset_1.
8182 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
8183 an rtx rather than an int. Take the delta as a poly_int64
8184 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
8185 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
8186 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
8187 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
8188 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
8190 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
8191 aarch64_add_constant.
8193 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8195 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
8196 Use scalar_float_mode.
8198 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8200 * config/aarch64/aarch64-simd.md
8201 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
8202 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
8203 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
8204 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
8205 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
8206 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
8207 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
8208 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
8209 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
8210 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
8212 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8215 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
8216 targ_options->x_arm_arch_string is non NULL.
8218 2018-01-11 Tamar Christina <tamar.christina@arm.com>
8220 * config/aarch64/aarch64.h
8221 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
8223 2018-01-11 Sudakshina Das <sudi.das@arm.com>
8226 * expmed.c (emit_store_flag_force): Swap if const op0
8227 and change VOIDmode to mode of op0.
8229 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8231 PR rtl-optimization/83761
8232 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
8233 than bytes to mode_for_size.
8235 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8238 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
8239 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
8242 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8245 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
8246 when in layout mode.
8247 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
8248 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
8251 2018-01-10 Michael Collison <michael.collison@arm.com>
8253 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
8254 * config/aarch64/aarch64-option-extension.def: Add
8255 AARCH64_OPT_EXTENSION of 'fp16fml'.
8256 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8257 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
8258 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
8259 * config/aarch64/constraints.md (Ui7): New constraint.
8260 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
8261 (VFMLA_SEL_W): Ditto.
8264 (VFMLA16_LOW): New int iterator.
8265 (VFMLA16_HIGH): Ditto.
8266 (UNSPEC_FMLAL): New unspec.
8267 (UNSPEC_FMLSL): Ditto.
8268 (UNSPEC_FMLAL2): Ditto.
8269 (UNSPEC_FMLSL2): Ditto.
8270 (f16mac): New code attribute.
8271 * config/aarch64/aarch64-simd-builtins.def
8272 (aarch64_fmlal_lowv2sf): Ditto.
8273 (aarch64_fmlsl_lowv2sf): Ditto.
8274 (aarch64_fmlalq_lowv4sf): Ditto.
8275 (aarch64_fmlslq_lowv4sf): Ditto.
8276 (aarch64_fmlal_highv2sf): Ditto.
8277 (aarch64_fmlsl_highv2sf): Ditto.
8278 (aarch64_fmlalq_highv4sf): Ditto.
8279 (aarch64_fmlslq_highv4sf): Ditto.
8280 (aarch64_fmlal_lane_lowv2sf): Ditto.
8281 (aarch64_fmlsl_lane_lowv2sf): Ditto.
8282 (aarch64_fmlal_laneq_lowv2sf): Ditto.
8283 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
8284 (aarch64_fmlalq_lane_lowv4sf): Ditto.
8285 (aarch64_fmlsl_lane_lowv4sf): Ditto.
8286 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
8287 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
8288 (aarch64_fmlal_lane_highv2sf): Ditto.
8289 (aarch64_fmlsl_lane_highv2sf): Ditto.
8290 (aarch64_fmlal_laneq_highv2sf): Ditto.
8291 (aarch64_fmlsl_laneq_highv2sf): Ditto.
8292 (aarch64_fmlalq_lane_highv4sf): Ditto.
8293 (aarch64_fmlsl_lane_highv4sf): Ditto.
8294 (aarch64_fmlalq_laneq_highv4sf): Ditto.
8295 (aarch64_fmlsl_laneq_highv4sf): Ditto.
8296 * config/aarch64/aarch64-simd.md:
8297 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
8298 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8299 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
8300 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8301 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
8302 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
8303 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
8304 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
8305 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
8306 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
8307 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
8308 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
8309 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
8310 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
8311 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
8312 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
8313 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
8314 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
8315 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
8316 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
8317 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
8318 (vfmlsl_low_u32): Ditto.
8319 (vfmlalq_low_u32): Ditto.
8320 (vfmlslq_low_u32): Ditto.
8321 (vfmlal_high_u32): Ditto.
8322 (vfmlsl_high_u32): Ditto.
8323 (vfmlalq_high_u32): Ditto.
8324 (vfmlslq_high_u32): Ditto.
8325 (vfmlal_lane_low_u32): Ditto.
8326 (vfmlsl_lane_low_u32): Ditto.
8327 (vfmlal_laneq_low_u32): Ditto.
8328 (vfmlsl_laneq_low_u32): Ditto.
8329 (vfmlalq_lane_low_u32): Ditto.
8330 (vfmlslq_lane_low_u32): Ditto.
8331 (vfmlalq_laneq_low_u32): Ditto.
8332 (vfmlslq_laneq_low_u32): Ditto.
8333 (vfmlal_lane_high_u32): Ditto.
8334 (vfmlsl_lane_high_u32): Ditto.
8335 (vfmlal_laneq_high_u32): Ditto.
8336 (vfmlsl_laneq_high_u32): Ditto.
8337 (vfmlalq_lane_high_u32): Ditto.
8338 (vfmlslq_lane_high_u32): Ditto.
8339 (vfmlalq_laneq_high_u32): Ditto.
8340 (vfmlslq_laneq_high_u32): Ditto.
8341 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
8342 (AARCH64_FL_FOR_ARCH8_4): New.
8343 (AARCH64_ISA_F16FML): New ISA flag.
8344 (TARGET_F16FML): New feature flag for fp16fml.
8345 (doc/invoke.texi): Document new fp16fml option.
8347 2018-01-10 Michael Collison <michael.collison@arm.com>
8349 * config/aarch64/aarch64-builtins.c:
8350 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
8351 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8352 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
8353 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
8354 (AARCH64_ISA_SHA3): New ISA flag.
8355 (TARGET_SHA3): New feature flag for sha3.
8356 * config/aarch64/iterators.md (sha512_op): New int attribute.
8357 (CRYPTO_SHA512): New int iterator.
8358 (UNSPEC_SHA512H): New unspec.
8359 (UNSPEC_SHA512H2): Ditto.
8360 (UNSPEC_SHA512SU0): Ditto.
8361 (UNSPEC_SHA512SU1): Ditto.
8362 * config/aarch64/aarch64-simd-builtins.def
8363 (aarch64_crypto_sha512hqv2di): New builtin.
8364 (aarch64_crypto_sha512h2qv2di): Ditto.
8365 (aarch64_crypto_sha512su0qv2di): Ditto.
8366 (aarch64_crypto_sha512su1qv2di): Ditto.
8367 (aarch64_eor3qv8hi): Ditto.
8368 (aarch64_rax1qv2di): Ditto.
8369 (aarch64_xarqv2di): Ditto.
8370 (aarch64_bcaxqv8hi): Ditto.
8371 * config/aarch64/aarch64-simd.md:
8372 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
8373 (aarch64_crypto_sha512su0qv2di): Ditto.
8374 (aarch64_crypto_sha512su1qv2di): Ditto.
8375 (aarch64_eor3qv8hi): Ditto.
8376 (aarch64_rax1qv2di): Ditto.
8377 (aarch64_xarqv2di): Ditto.
8378 (aarch64_bcaxqv8hi): Ditto.
8379 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
8380 (vsha512h2q_u64): Ditto.
8381 (vsha512su0q_u64): Ditto.
8382 (vsha512su1q_u64): Ditto.
8383 (veor3q_u16): Ditto.
8384 (vrax1q_u64): Ditto.
8386 (vbcaxq_u16): Ditto.
8387 * config/arm/types.md (crypto_sha512): New type attribute.
8388 (crypto_sha3): Ditto.
8389 (doc/invoke.texi): Document new sha3 option.
8391 2018-01-10 Michael Collison <michael.collison@arm.com>
8393 * config/aarch64/aarch64-builtins.c:
8394 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
8395 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8396 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
8397 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
8398 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
8399 (AARCH64_ISA_SM4): New ISA flag.
8400 (TARGET_SM4): New feature flag for sm4.
8401 * config/aarch64/aarch64-simd-builtins.def
8402 (aarch64_sm3ss1qv4si): Ditto.
8403 (aarch64_sm3tt1aq4si): Ditto.
8404 (aarch64_sm3tt1bq4si): Ditto.
8405 (aarch64_sm3tt2aq4si): Ditto.
8406 (aarch64_sm3tt2bq4si): Ditto.
8407 (aarch64_sm3partw1qv4si): Ditto.
8408 (aarch64_sm3partw2qv4si): Ditto.
8409 (aarch64_sm4eqv4si): Ditto.
8410 (aarch64_sm4ekeyqv4si): Ditto.
8411 * config/aarch64/aarch64-simd.md:
8412 (aarch64_sm3ss1qv4si): Ditto.
8413 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
8414 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
8415 (aarch64_sm4eqv4si): Ditto.
8416 (aarch64_sm4ekeyqv4si): Ditto.
8417 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
8418 (sm3part_op): Ditto.
8419 (CRYPTO_SM3TT): Ditto.
8420 (CRYPTO_SM3PART): Ditto.
8421 (UNSPEC_SM3SS1): New unspec.
8422 (UNSPEC_SM3TT1A): Ditto.
8423 (UNSPEC_SM3TT1B): Ditto.
8424 (UNSPEC_SM3TT2A): Ditto.
8425 (UNSPEC_SM3TT2B): Ditto.
8426 (UNSPEC_SM3PARTW1): Ditto.
8427 (UNSPEC_SM3PARTW2): Ditto.
8428 (UNSPEC_SM4E): Ditto.
8429 (UNSPEC_SM4EKEY): Ditto.
8430 * config/aarch64/constraints.md (Ui2): New constraint.
8431 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
8432 * config/arm/types.md (crypto_sm3): New type attribute.
8433 (crypto_sm4): Ditto.
8434 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
8435 (vsm3tt1aq_u32): Ditto.
8436 (vsm3tt1bq_u32): Ditto.
8437 (vsm3tt2aq_u32): Ditto.
8438 (vsm3tt2bq_u32): Ditto.
8439 (vsm3partw1q_u32): Ditto.
8440 (vsm3partw2q_u32): Ditto.
8441 (vsm4eq_u32): Ditto.
8442 (vsm4ekeyq_u32): Ditto.
8443 (doc/invoke.texi): Document new sm4 option.
8445 2018-01-10 Michael Collison <michael.collison@arm.com>
8447 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
8448 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
8449 (AARCH64_FL_FOR_ARCH8_4): New.
8450 (AARCH64_FL_V8_4): New flag.
8451 (doc/invoke.texi): Document new armv8.4-a option.
8453 2018-01-10 Michael Collison <michael.collison@arm.com>
8455 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8456 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
8457 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
8458 * config/aarch64/aarch64-option-extension.def: Add
8459 AARCH64_OPT_EXTENSION of 'sha2'.
8460 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
8461 (crypto): Disable sha2 and aes if crypto disabled.
8462 (crypto): Enable aes and sha2 if enabled.
8463 (simd): Disable sha2 and aes if simd disabled.
8464 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
8466 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
8467 (TARGET_SHA2): New feature flag for sha2.
8468 (TARGET_AES): New feature flag for aes.
8469 * config/aarch64/aarch64-simd.md:
8470 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
8471 conditional on TARGET_AES.
8472 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
8473 (aarch64_crypto_sha1hsi): Make pattern conditional
8475 (aarch64_crypto_sha1hv4si): Ditto.
8476 (aarch64_be_crypto_sha1hv4si): Ditto.
8477 (aarch64_crypto_sha1su1v4si): Ditto.
8478 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
8479 (aarch64_crypto_sha1su0v4si): Ditto.
8480 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
8481 (aarch64_crypto_sha256su0v4si): Ditto.
8482 (aarch64_crypto_sha256su1v4si): Ditto.
8483 (doc/invoke.texi): Document new aes and sha2 options.
8485 2018-01-10 Martin Sebor <msebor@redhat.com>
8487 PR tree-optimization/83781
8488 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
8491 2018-01-11 Martin Sebor <msebor@gmail.com>
8492 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8494 PR tree-optimization/83501
8495 PR tree-optimization/81703
8497 * tree-ssa-strlen.c (get_string_cst): Rename...
8498 (get_string_len): ...to this. Handle global constants.
8499 (handle_char_store): Adjust.
8501 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
8502 Jim Wilson <jimw@sifive.com>
8504 * config/riscv/riscv-protos.h (riscv_output_return): New.
8505 * config/riscv/riscv.c (struct machine_function): New naked_p field.
8506 (riscv_attribute_table, riscv_output_return),
8507 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
8508 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
8509 (riscv_compute_frame_info): Only compute frame->mask if not a naked
8511 (riscv_expand_prologue): Add early return for naked function.
8512 (riscv_expand_epilogue): Likewise.
8513 (riscv_function_ok_for_sibcall): Return false for naked function.
8514 (riscv_set_current_function): New.
8515 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
8516 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
8517 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
8518 * doc/extend.texi (RISC-V Function Attributes): New.
8520 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
8522 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
8523 check for 128-bit long double before checking TCmode.
8524 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
8525 128-bit long doubles before checking TFmode or TCmode.
8526 (FLOAT128_IBM_P): Likewise.
8528 2018-01-10 Martin Sebor <msebor@redhat.com>
8530 PR tree-optimization/83671
8531 * builtins.c (c_strlen): Unconditionally return zero for the empty
8533 Use -Warray-bounds for warnings.
8534 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
8535 for non-constant array indices with COMPONENT_REF, arrays of
8536 arrays, and pointers to arrays.
8537 (gimple_fold_builtin_strlen): Determine and set length range for
8538 non-constant character arrays.
8540 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
8543 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
8546 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
8548 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
8550 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8553 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
8554 VECTOR_MEM_ALTIVEC_OR_VSX_P.
8555 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
8556 indexed_or_indirect_operand predicate.
8557 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
8558 (*vsx_le_perm_load_v8hi): Likewise.
8559 (*vsx_le_perm_load_v16qi): Likewise.
8560 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
8561 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
8562 (*vsx_le_perm_store_v8hi): Likewise.
8563 (*vsx_le_perm_store_v16qi): Likewise.
8564 (eight unnamed splitters): Likewise.
8566 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8568 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
8569 * config/rs6000/emmintrin.h: Likewise.
8570 * config/rs6000/mmintrin.h: Likewise.
8571 * config/rs6000/xmmintrin.h: Likewise.
8573 2018-01-10 David Malcolm <dmalcolm@redhat.com>
8576 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
8578 * tree.c (tree_nop_conversion): Return true for location wrapper
8580 (maybe_wrap_with_location): New function.
8581 (selftest::check_strip_nops): New function.
8582 (selftest::test_location_wrappers): New function.
8583 (selftest::tree_c_tests): Call it.
8584 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
8585 (maybe_wrap_with_location): New decl.
8586 (EXPR_LOCATION_WRAPPER_P): New macro.
8587 (location_wrapper_p): New inline function.
8588 (tree_strip_any_location_wrapper): New inline function.
8590 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
8593 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
8594 stack_realign_offset for the largest alignment of stack slot
8596 (ix86_find_max_used_stack_alignment): New function.
8597 (ix86_finalize_stack_frame_flags): Use it. Set
8598 max_used_stack_alignment if we don't realign stack.
8599 * config/i386/i386.h (machine_function): Add
8600 max_used_stack_alignment.
8602 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
8604 * config/arm/arm.opt (-mbranch-cost): New option.
8605 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
8608 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
8611 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
8612 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
8614 2018-01-10 Richard Biener <rguenther@suse.de>
8617 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
8618 early out so it also covers the case where we have a non-NULL
8621 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8623 PR tree-optimization/83753
8624 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
8625 for non-strided grouped accesses if the number of elements is 1.
8627 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8630 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
8631 * i386.h (TARGET_USE_GATHER): Define.
8632 * x86-tune.def (X86_TUNE_USE_GATHER): New.
8634 2018-01-10 Martin Liska <mliska@suse.cz>
8637 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
8638 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
8640 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
8641 CLEANUP_NO_PARTITIONING is not set.
8643 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8645 * doc/rtl.texi: Remove documentation of (const ...) wrappers
8646 for vectors, as a partial revert of r254296.
8647 * rtl.h (const_vec_p): Delete.
8648 (const_vec_duplicate_p): Don't test for vector CONSTs.
8649 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
8650 * expmed.c (make_tree): Likewise.
8653 * common.md (E, F): Use CONSTANT_P instead of checking for
8655 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
8656 checking for CONST_VECTOR.
8658 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8661 * predict.c (force_edge_cold): Handle in more sane way edges
8664 2018-01-09 Carl Love <cel@us.ibm.com>
8666 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
8668 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
8669 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
8670 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
8671 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
8672 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
8673 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
8674 * config/rs6000/rs6000-protos.h: Add extern defition for
8675 rs6000_generate_float2_double_code.
8676 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
8678 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
8679 (float2_v2df): Add define_expand.
8681 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
8684 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
8685 op_mode in the force_to_mode call.
8687 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8689 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
8690 instead of checking each element individually.
8691 (aarch64_evpc_uzp): Likewise.
8692 (aarch64_evpc_zip): Likewise.
8693 (aarch64_evpc_ext): Likewise.
8694 (aarch64_evpc_rev): Likewise.
8695 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
8696 instead of checking each element individually. Return true without
8698 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
8699 whether all selected elements come from the same input, instead of
8700 checking each element individually. Remove calls to gen_rtx_REG,
8701 start_sequence and end_sequence and instead assert that no rtl is
8704 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8706 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
8707 order of HIGH and CONST checks.
8709 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8711 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
8712 if the destination isn't an SSA_NAME.
8714 2018-01-09 Richard Biener <rguenther@suse.de>
8716 PR tree-optimization/83668
8717 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
8719 (canonicalize_loop_form): ... here, renamed from ...
8720 (canonicalize_loop_closed_ssa_form): ... this and amended to
8721 swap successor edges for loop exit blocks to make us use
8722 the RPO order we need for initial schedule generation.
8724 2018-01-09 Joseph Myers <joseph@codesourcery.com>
8726 PR tree-optimization/64811
8727 * match.pd: When optimizing comparisons with Inf, avoid
8728 introducing or losing exceptions from comparisons with NaN.
8730 2018-01-09 Martin Liska <mliska@suse.cz>
8733 * asan.c (shadow_mem_size): Add gcc_assert.
8735 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
8737 Don't save registers in main().
8740 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
8741 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
8742 * config/avr/avr.c (avr_set_current_function): Don't error if
8743 naked, OS_task or OS_main are specified at the same time.
8744 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
8746 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
8748 * common/config/avr/avr-common.c (avr_option_optimization_table):
8749 Switch on -mmain-is-OS_task for optimizing compilations.
8751 2018-01-09 Richard Biener <rguenther@suse.de>
8753 PR tree-optimization/83572
8754 * graphite.c: Include cfganal.h.
8755 (graphite_transform_loops): Connect infinite loops to exit
8756 and remove fake edges at the end.
8758 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8760 * ipa-inline.c (edge_badness): Revert accidental checkin.
8762 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8765 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
8766 symbols; not inline clones.
8768 2018-01-09 Jakub Jelinek <jakub@redhat.com>
8771 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
8772 hard registers. Formatting fixes.
8774 PR preprocessor/83722
8775 * gcc.c (try_generate_repro): Pass
8776 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
8777 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
8780 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
8781 Kito Cheng <kito.cheng@gmail.com>
8783 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
8784 (riscv_leaf_function_p): Delete.
8785 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
8787 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8789 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
8791 (do_ifelse): New function.
8792 (do_isel): New function.
8793 (do_sub3): New function.
8794 (do_add3): New function.
8795 (do_load_mask_compare): New function.
8796 (do_overlap_load_compare): New function.
8797 (expand_compare_loop): New function.
8798 (expand_block_compare): Call expand_compare_loop() when appropriate.
8799 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
8801 (-mblock-compare-inline-loop-limit): New option.
8803 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8806 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
8807 Reverse order of second and third operands in first alternative.
8808 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
8809 of first and second elements in UNSPEC_VPERMR vector.
8810 (altivec_expand_vec_perm_le): Likewise.
8812 2018-01-08 Jeff Law <law@redhat.com>
8814 PR rtl-optimizatin/81308
8815 * tree-switch-conversion.c (cfg_altered): New file scoped static.
8816 (process_switch): If group_case_labels makes a change, then set
8818 (pass_convert_switch::execute): If a switch is converted, then
8819 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
8821 PR rtl-optimization/81308
8822 * recog.c (split_all_insns): Conditionally cleanup the CFG after
8825 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
8827 PR target/83663 - Revert r255946
8828 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
8829 generation for cases where splatting a value is not useful.
8830 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
8831 across a vec_duplicate and a paradoxical subreg forming a vector
8832 mode to a vec_concat.
8834 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8836 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
8837 -march=armv8.3-a variants.
8838 * config/arm/t-multilib: Likewise.
8839 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
8841 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8843 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
8845 (cceq_ior_compare_complement): Give it a name so I can use it, and
8846 change boolean_or_operator predicate to boolean_operator so it can
8847 be used to generate a crand.
8848 (eqne): New code iterator.
8849 (bd/bd_neg): New code_attrs.
8850 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
8851 a single define_insn.
8852 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
8853 decrement (bdnzt/bdnzf/bdzt/bdzf).
8854 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
8855 with the new names of the branch decrement patterns, and added the
8856 names of the branch decrement conditional patterns.
8858 2018-01-08 Richard Biener <rguenther@suse.de>
8860 PR tree-optimization/83563
8861 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
8864 2018-01-08 Richard Biener <rguenther@suse.de>
8867 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
8869 2018-01-08 Richard Biener <rguenther@suse.de>
8871 PR tree-optimization/83685
8872 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
8873 references to abnormals.
8875 2018-01-08 Richard Biener <rguenther@suse.de>
8878 * dwarf2out.c (output_indirect_strings): Handle empty
8879 skeleton_debug_str_hash.
8880 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
8882 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
8884 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
8885 (emit_store_direct): Likewise.
8886 (arc_trampoline_adjust_address): Likewise.
8887 (arc_asm_trampoline_template): New function.
8888 (arc_initialize_trampoline): Use asm_trampoline_template.
8889 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
8890 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
8891 * config/arc/arc.md (flush_icache): Delete pattern.
8893 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
8895 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
8896 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
8899 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
8902 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
8903 by not USED_FOR_TARGET.
8904 (make_pass_resolve_sw_modes): Likewise.
8906 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
8908 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
8911 2018-01-08 Richard Biener <rguenther@suse.de>
8914 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
8916 2018-01-08 Richard Biener <rguenther@suse.de>
8919 * match.pd ((t * 2) / 2) -> t): Add missing :c.
8921 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
8924 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
8925 basic blocks with a small number of successors.
8926 (convert_control_dep_chain_into_preds): Improve handling of
8928 (dump_predicates): Split apart into...
8929 (dump_pred_chain): ...here...
8930 (dump_pred_info): ...and here.
8931 (can_one_predicate_be_invalidated_p): Add debugging printfs.
8932 (can_chain_union_be_invalidated_p): Improve check for invalidation
8934 (uninit_uses_cannot_happen): Avoid unnecessary if
8935 convert_control_dep_chain_into_preds yielded nothing.
8937 2018-01-06 Martin Sebor <msebor@redhat.com>
8939 PR tree-optimization/83640
8940 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
8941 subtracting negative offset from size.
8942 (builtin_access::overlap): Adjust offset bounds of the access to fall
8943 within the size of the object if possible.
8945 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
8947 PR rtl-optimization/83699
8948 * expmed.c (extract_bit_field_1): Restrict the vector usage of
8949 extract_bit_field_as_subreg to cases in which the extracted
8950 value is also a vector.
8952 * lra-constraints.c (process_alt_operands): Test for the equivalence
8953 substitutions when detecting a possible reload cycle.
8955 2018-01-06 Jakub Jelinek <jakub@redhat.com>
8958 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
8959 by default if flag_selective_schedling{,2}. Formatting fixes.
8961 PR rtl-optimization/83682
8962 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
8963 if it has non-VECTOR_MODE element mode.
8964 (vec_duplicate_p): Likewise.
8967 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
8968 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
8970 2018-01-05 Jakub Jelinek <jakub@redhat.com>
8973 * config/i386/i386-builtin.def
8974 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
8975 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
8976 Require also OPTION_MASK_ISA_AVX512F in addition to
8977 OPTION_MASK_ISA_GFNI.
8978 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
8979 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
8980 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
8981 to OPTION_MASK_ISA_GFNI.
8982 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
8983 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
8984 OPTION_MASK_ISA_AVX512BW.
8985 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
8986 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
8987 addition to OPTION_MASK_ISA_GFNI.
8988 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
8989 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
8990 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
8991 to OPTION_MASK_ISA_GFNI.
8992 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
8993 a requirement for all ISAs rather than any of them with a few
8995 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
8997 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
8998 bitmasks to be enabled with 3 exceptions, instead of requiring any
8999 enabled ISA with lots of exceptions.
9000 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
9001 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
9002 Change avx512bw in isa attribute to avx512f.
9003 * config/i386/sgxintrin.h: Add license boilerplate.
9004 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
9005 to __AVX512F__ and __AVX512VL to __AVX512VL__.
9006 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
9007 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
9009 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
9010 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
9011 temporarily sse2 rather than sse if not enabled already.
9014 * config/i386/sse.md (VI248_VLBW): Rename to ...
9015 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
9016 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
9017 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
9018 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
9019 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
9020 mode iterator instead of VI248_VLBW.
9022 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
9024 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
9025 (record_modified): Skip clobbers; add debug output.
9026 (param_change_prob): Use sreal frequencies.
9028 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9030 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
9031 punt for user-aligned variables.
9033 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9035 * tree-chrec.c (chrec_contains_symbols): Return true for
9038 2018-01-05 Sudakshina Das <sudi.das@arm.com>
9041 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
9042 of (x|y) == x for BICS pattern.
9044 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9046 PR tree-optimization/83605
9047 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
9048 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
9051 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
9053 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
9054 * config/epiphany/rtems.h: New file.
9056 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9057 Uros Bizjak <ubizjak@gmail.com>
9060 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
9061 QIreg_operand instead of register_operand predicate.
9062 * config/i386/i386.c (ix86_rop_should_change_byte_p,
9063 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
9064 comments instead of -fmitigate[-_]rop.
9066 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9069 * cgraphunit.c (symbol_table::compile): Switch to text_section
9070 before calling assembly_start debug hook.
9071 * run-rtl-passes.c (run_rtl_passes): Likewise.
9074 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9076 * tree-vrp.c (extract_range_from_binary_expr_1): Check
9077 range_int_cst_p rather than !symbolic_range_p before calling
9078 extract_range_from_multiplicative_op_1.
9080 2018-01-04 Jeff Law <law@redhat.com>
9082 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
9083 redundant test in assertion.
9085 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9087 * doc/rtl.texi: Document machine_mode wrapper classes.
9089 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9091 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
9094 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9096 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
9097 the VEC_PERM_EXPR fold to fail.
9099 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9102 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
9103 to switched_sections.
9105 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9108 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
9111 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
9114 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
9115 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
9117 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9120 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
9121 is BLKmode and bitpos not zero or mode change is needed.
9123 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9126 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
9129 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
9132 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
9133 instead of MULT rtx. Update all corresponding splitters.
9135 (*ssub<modesuffix>): Ditto.
9137 (*cmp_sadd_di): Update split patterns.
9138 (*cmp_sadd_si): Ditto.
9139 (*cmp_sadd_sidi): Ditto.
9140 (*cmp_ssub_di): Ditto.
9141 (*cmp_ssub_si): Ditto.
9142 (*cmp_ssub_sidi): Ditto.
9143 * config/alpha/predicates.md (const23_operand): New predicate.
9144 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
9145 Look for ASHIFT, not MULT inner operand.
9146 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
9148 2018-01-04 Martin Liska <mliska@suse.cz>
9150 PR gcov-profile/83669
9151 * gcov.c (output_intermediate_file): Add version to intermediate
9153 * doc/gcov.texi: Document new field 'version' in intermediate
9154 file format. Fix location of '-k' option of gcov command.
9156 2018-01-04 Martin Liska <mliska@suse.cz>
9159 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
9161 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9163 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
9165 2018-01-03 Martin Sebor <msebor@redhat.com>
9167 PR tree-optimization/83655
9168 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
9169 checking calls with invalid arguments.
9171 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9173 * tree-vect-stmts.c (vect_get_store_rhs): New function.
9174 (vectorizable_mask_load_store): Delete.
9175 (vectorizable_call): Return false for masked loads and stores.
9176 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
9177 instead of gimple_assign_rhs1.
9178 (vectorizable_load): Handle IFN_MASK_LOAD.
9179 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
9181 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9183 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
9185 (vectorizable_mask_load_store): ...here.
9186 (vectorizable_load): ...and here.
9188 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9190 * tree-vect-stmts.c (vect_build_all_ones_mask)
9191 (vect_build_zero_merge_argument): New functions, split out from...
9192 (vectorizable_load): ...here.
9194 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9196 * tree-vect-stmts.c (vect_check_store_rhs): New function,
9198 (vectorizable_mask_load_store): ...here.
9199 (vectorizable_store): ...and here.
9201 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9203 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
9205 (vectorizable_mask_load_store): ...here.
9207 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9209 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
9210 (vect_model_store_cost): Take a vec_load_store_type instead of a
9212 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
9213 (vect_model_store_cost): Take a vec_load_store_type instead of a
9215 (vectorizable_mask_load_store): Update accordingly.
9216 (vectorizable_store): Likewise.
9217 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
9219 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9221 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
9222 IFN_MASK_LOAD calls here rather than...
9223 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
9225 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9226 Alan Hayward <alan.hayward@arm.com>
9227 David Sherwood <david.sherwood@arm.com>
9229 * expmed.c (extract_bit_field_1): For vector extracts,
9230 fall back to extract_bit_field_as_subreg if vec_extract
9233 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9234 Alan Hayward <alan.hayward@arm.com>
9235 David Sherwood <david.sherwood@arm.com>
9237 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
9238 they are variable or constant sized.
9239 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
9240 slots for constant-sized data.
9242 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9243 Alan Hayward <alan.hayward@arm.com>
9244 David Sherwood <david.sherwood@arm.com>
9246 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
9247 handling COND_EXPRs with boolean comparisons, try to find a better
9248 basis for the mask type than the boolean itself.
9250 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9252 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
9253 is calculated and how it can be overridden.
9254 * genmodes.c (max_bitsize_mode_any_mode): New variable.
9255 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
9257 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
9260 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9261 Alan Hayward <alan.hayward@arm.com>
9262 David Sherwood <david.sherwood@arm.com>
9264 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
9265 Remove the mode argument.
9266 (aarch64_simd_valid_immediate): Remove the mode and inverse
9268 * config/aarch64/iterators.md (bitsize): New iterator.
9269 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
9270 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
9271 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
9272 aarch64_simd_valid_immediate.
9273 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
9274 (aarch64_reg_or_bic_imm): Likewise.
9275 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
9276 with an insn_type enum and msl with a modifier_type enum.
9277 Replace element_width with a scalar_mode. Change the shift
9278 to unsigned int. Add constructors for scalar_float_mode and
9279 scalar_int_mode elements.
9280 (aarch64_vect_float_const_representable_p): Delete.
9281 (aarch64_can_const_movi_rtx_p)
9282 (aarch64_simd_scalar_immediate_valid_for_move)
9283 (aarch64_simd_make_constant): Update call to
9284 aarch64_simd_valid_immediate.
9285 (aarch64_advsimd_valid_immediate_hs): New function.
9286 (aarch64_advsimd_valid_immediate): Likewise.
9287 (aarch64_simd_valid_immediate): Remove mode and inverse
9288 arguments. Rewrite to use the above. Use const_vec_duplicate_p
9289 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
9290 and aarch64_float_const_representable_p on the result.
9291 (aarch64_output_simd_mov_immediate): Remove mode argument.
9292 Update call to aarch64_simd_valid_immediate and use of
9293 simd_immediate_info.
9294 (aarch64_output_scalar_simd_mov_immediate): Update call
9297 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9298 Alan Hayward <alan.hayward@arm.com>
9299 David Sherwood <david.sherwood@arm.com>
9301 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
9302 (mode_nunits): Likewise CONST_MODE_NUNITS.
9303 * machmode.def (ADJUST_NUNITS): Document.
9304 * genmodes.c (mode_data::need_nunits_adj): New field.
9305 (blank_mode): Update accordingly.
9306 (adj_nunits): New variable.
9307 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
9309 (emit_mode_size_inline): Set need_bytesize_adj for all modes
9310 listed in adj_nunits.
9311 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
9312 listed in adj_nunits. Don't emit case statements for such modes.
9313 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
9314 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
9315 nothing if adj_nunits is nonnull.
9316 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
9317 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
9318 (emit_mode_fbit): Update use of print_maybe_const_decl.
9319 (emit_move_size): Likewise. Treat the array as non-const
9321 (emit_mode_adjustments): Handle adj_nunits.
9323 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9325 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
9326 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
9327 (VECTOR_MODES): Use it.
9328 (make_vector_modes): Take the prefix as an argument.
9330 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9331 Alan Hayward <alan.hayward@arm.com>
9332 David Sherwood <david.sherwood@arm.com>
9334 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
9335 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
9336 for MODE_VECTOR_BOOL.
9337 * machmode.def (VECTOR_BOOL_MODE): Document.
9338 * genmodes.c (VECTOR_BOOL_MODE): New macro.
9339 (make_vector_bool_mode): New function.
9340 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
9342 * lto-streamer-in.c (lto_input_mode_table): Likewise.
9343 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
9345 * stor-layout.c (int_mode_for_mode): Likewise.
9346 * tree.c (build_vector_type_for_mode): Likewise.
9347 * varasm.c (output_constant_pool_2): Likewise.
9348 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
9349 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
9350 for MODE_VECTOR_BOOL.
9351 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
9352 of mode class checks.
9353 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
9354 instead of a list of mode class checks.
9355 (expand_vector_scalar_condition): Likewise.
9356 (type_for_widest_vector_mode): Handle BImode as an inner mode.
9358 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9359 Alan Hayward <alan.hayward@arm.com>
9360 David Sherwood <david.sherwood@arm.com>
9362 * machmode.h (mode_size): Change from unsigned short to
9364 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
9365 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9366 or if measurement_type is not polynomial.
9367 (fixed_size_mode::includes_p): Check for constant-sized modes.
9368 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
9369 return a poly_uint16 rather than an unsigned short.
9370 (emit_mode_size): Change the type of mode_size from unsigned short
9371 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
9372 (emit_mode_adjustments): Cope with polynomial vector sizes.
9373 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9375 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9377 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
9378 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
9379 * caller-save.c (setup_save_areas): Likewise.
9380 (replace_reg_with_saved_mem): Likewise.
9381 * calls.c (emit_library_call_value_1): Likewise.
9382 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
9383 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
9384 (gen_lowpart_for_combine): Likewise.
9385 * convert.c (convert_to_integer_1): Likewise.
9386 * cse.c (equiv_constant, cse_insn): Likewise.
9387 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
9388 (cselib_subst_to_values): Likewise.
9389 * dce.c (word_dce_process_block): Likewise.
9390 * df-problems.c (df_word_lr_mark_ref): Likewise.
9391 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
9392 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
9393 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
9394 (rtl_for_decl_location): Likewise.
9395 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
9396 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
9397 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
9398 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
9399 (expand_expr_real_1): Likewise.
9400 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
9401 (pad_below): Likewise.
9402 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9403 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
9404 * ira.c (get_subreg_tracking_sizes): Likewise.
9405 * ira-build.c (ira_create_allocno_objects): Likewise.
9406 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
9407 (ira_sort_regnos_for_alter_reg): Likewise.
9408 * ira-costs.c (record_operand_costs): Likewise.
9409 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
9410 (resolve_simple_move): Likewise.
9411 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
9412 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
9413 (lra_constraints): Likewise.
9414 (CONST_POOL_OK_P): Reject variable-sized modes.
9415 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
9416 (add_pseudo_to_slot, lra_spill): Likewise.
9417 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9418 * optabs-query.c (get_best_extraction_insn): Likewise.
9419 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9420 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
9421 (expand_mult_highpart, valid_multiword_target_p): Likewise.
9422 * recog.c (offsettable_address_addr_space_p): Likewise.
9423 * regcprop.c (maybe_mode_change): Likewise.
9424 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
9425 * regrename.c (build_def_use): Likewise.
9426 * regstat.c (dump_reg_info): Likewise.
9427 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
9428 (find_reloads, find_reloads_subreg_address): Likewise.
9429 * reload1.c (eliminate_regs_1): Likewise.
9430 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
9431 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
9432 (simplify_binary_operation_1, simplify_subreg): Likewise.
9433 * targhooks.c (default_function_arg_padding): Likewise.
9434 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
9435 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
9436 (verify_gimple_assign_ternary): Likewise.
9437 * tree-inline.c (estimate_move_cost): Likewise.
9438 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9439 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
9440 (get_address_cost_ainc): Likewise.
9441 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
9442 (vect_supportable_dr_alignment): Likewise.
9443 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9444 (vectorizable_reduction): Likewise.
9445 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
9446 (vectorizable_operation, vectorizable_load): Likewise.
9447 * tree.c (build_same_sized_truth_vector_type): Likewise.
9448 * valtrack.c (cleanup_auto_inc_dec): Likewise.
9449 * var-tracking.c (emit_note_insn_var_location): Likewise.
9450 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
9451 (ADDR_VEC_ALIGN): Likewise.
9453 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9454 Alan Hayward <alan.hayward@arm.com>
9455 David Sherwood <david.sherwood@arm.com>
9457 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
9459 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9460 or if measurement_type is polynomial.
9461 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
9462 * combine.c (make_extraction): Likewise.
9463 * dse.c (find_shift_sequence): Likewise.
9464 * dwarf2out.c (mem_loc_descriptor): Likewise.
9465 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
9466 (extract_bit_field, extract_low_bits): Likewise.
9467 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
9468 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
9469 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
9470 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
9471 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9472 * reload.c (find_reloads): Likewise.
9473 * reload1.c (alter_reg): Likewise.
9474 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
9475 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
9476 * tree-if-conv.c (predicate_mem_writes): Likewise.
9477 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
9478 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
9479 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
9480 * valtrack.c (dead_debug_insert_temp): Likewise.
9481 * varasm.c (mergeable_constant_section): Likewise.
9482 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
9484 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9485 Alan Hayward <alan.hayward@arm.com>
9486 David Sherwood <david.sherwood@arm.com>
9488 * expr.c (expand_assignment): Cope with polynomial mode sizes
9489 when assigning to a CONCAT.
9491 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9492 Alan Hayward <alan.hayward@arm.com>
9493 David Sherwood <david.sherwood@arm.com>
9495 * machmode.h (mode_precision): Change from unsigned short to
9497 (mode_to_precision): Return a poly_uint16 rather than an unsigned
9499 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
9500 or if measurement_type is not polynomial.
9501 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
9502 in which the mode is already known to be a scalar_int_mode.
9503 * genmodes.c (emit_mode_precision): Change the type of mode_precision
9504 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
9506 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9507 for GET_MODE_PRECISION.
9508 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9509 for GET_MODE_PRECISION.
9510 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
9512 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
9513 (expand_field_assignment, make_extraction): Likewise.
9514 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
9515 (get_last_value): Likewise.
9516 * convert.c (convert_to_integer_1): Likewise.
9517 * cse.c (cse_insn): Likewise.
9518 * expr.c (expand_expr_real_1): Likewise.
9519 * lra-constraints.c (simplify_operand_subreg): Likewise.
9520 * optabs-query.c (can_atomic_load_p): Likewise.
9521 * optabs.c (expand_atomic_load): Likewise.
9522 (expand_atomic_store): Likewise.
9523 * ree.c (combine_reaching_defs): Likewise.
9524 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
9525 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
9526 * tree.h (type_has_mode_precision_p): Likewise.
9527 * ubsan.c (instrument_si_overflow): Likewise.
9529 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9530 Alan Hayward <alan.hayward@arm.com>
9531 David Sherwood <david.sherwood@arm.com>
9533 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
9534 polynomial numbers of units.
9535 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
9536 (valid_vector_subparts_p): New function.
9537 (build_vector_type): Remove temporary shim and take the number
9538 of units as a poly_uint64 rather than an int.
9539 (build_opaque_vector_type): Take the number of units as a
9540 poly_uint64 rather than an int.
9541 * tree.c (build_vector_from_ctor): Handle polynomial
9542 TYPE_VECTOR_SUBPARTS.
9543 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
9544 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
9545 (build_vector_from_val): If the number of units is variable,
9546 use build_vec_duplicate_cst for constant operands and
9547 VEC_DUPLICATE_EXPR otherwise.
9548 (make_vector_type): Remove temporary is_constant ().
9549 (build_vector_type, build_opaque_vector_type): Take the number of
9550 units as a poly_uint64 rather than an int.
9551 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
9553 * cfgexpand.c (expand_debug_expr): Likewise.
9554 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
9555 (store_constructor, expand_expr_real_1): Likewise.
9556 (const_scalar_mask_from_tree): Likewise.
9557 * fold-const-call.c (fold_const_reduction): Likewise.
9558 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
9559 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
9560 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
9561 (fold_relational_const): Likewise.
9562 (native_interpret_vector): Likewise. Change the size from an
9563 int to an unsigned int.
9564 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
9565 TYPE_VECTOR_SUBPARTS.
9566 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
9567 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
9568 duplicating a non-constant operand into a variable-length vector.
9569 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
9570 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
9571 * ipa-icf.c (sem_variable::equals): Likewise.
9572 * match.pd: Likewise.
9573 * omp-simd-clone.c (simd_clone_subparts): Likewise.
9574 * print-tree.c (print_node): Likewise.
9575 * stor-layout.c (layout_type): Likewise.
9576 * targhooks.c (default_builtin_vectorization_cost): Likewise.
9577 * tree-cfg.c (verify_gimple_comparison): Likewise.
9578 (verify_gimple_assign_binary): Likewise.
9579 (verify_gimple_assign_ternary): Likewise.
9580 (verify_gimple_assign_single): Likewise.
9581 * tree-pretty-print.c (dump_generic_node): Likewise.
9582 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9583 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
9584 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
9585 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
9586 (vect_shift_permute_load_chain): Likewise.
9587 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
9588 (expand_vector_condition, optimize_vector_constructor): Likewise.
9589 (lower_vec_perm, get_compute_type): Likewise.
9590 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9591 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
9592 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
9593 (vect_recog_mask_conversion_pattern): Likewise.
9594 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
9595 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
9596 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9597 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
9598 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
9599 (vectorizable_shift, vectorizable_operation, vectorizable_store)
9600 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
9601 (supportable_widening_operation): Likewise.
9602 (supportable_narrowing_operation): Likewise.
9603 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
9605 * varasm.c (output_constant): Likewise.
9607 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9608 Alan Hayward <alan.hayward@arm.com>
9609 David Sherwood <david.sherwood@arm.com>
9611 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
9612 so that both the length == 3 and length != 3 cases set up their
9613 own permute vectors. Add comments explaining why we know the
9614 number of elements is constant.
9615 (vect_permute_load_chain): Likewise.
9617 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9618 Alan Hayward <alan.hayward@arm.com>
9619 David Sherwood <david.sherwood@arm.com>
9621 * machmode.h (mode_nunits): Change from unsigned char to
9623 (ONLY_FIXED_SIZE_MODES): New macro.
9624 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
9625 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
9626 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
9628 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
9629 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
9630 or if measurement_type is not polynomial.
9631 * genmodes.c (ZERO_COEFFS): New macro.
9632 (emit_mode_nunits_inline): Make mode_nunits_inline return a
9634 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
9635 Use ZERO_COEFFS when emitting initializers.
9636 * data-streamer.h (bp_pack_poly_value): New function.
9637 (bp_unpack_poly_value): Likewise.
9638 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9639 for GET_MODE_NUNITS.
9640 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9641 for GET_MODE_NUNITS.
9642 * tree.c (make_vector_type): Remove temporary shim and make
9643 the real function take the number of units as a poly_uint64
9645 (build_vector_type_for_mode): Handle polynomial nunits.
9646 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
9647 * emit-rtl.c (const_vec_series_p_1): Likewise.
9648 (gen_rtx_CONST_VECTOR): Likewise.
9649 * fold-const.c (test_vec_duplicate_folding): Likewise.
9650 * genrecog.c (validate_pattern): Likewise.
9651 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
9652 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9653 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
9654 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
9655 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
9656 * rtlanal.c (subreg_get_info): Likewise.
9657 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9658 (vect_grouped_load_supported): Likewise.
9659 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
9660 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
9661 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
9662 (simplify_const_unary_operation, simplify_binary_operation_1)
9663 (simplify_const_binary_operation, simplify_ternary_operation)
9664 (test_vector_ops_duplicate, test_vector_ops): Likewise.
9665 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
9666 instead of CONST_VECTOR_NUNITS.
9667 * varasm.c (output_constant_pool_2): Likewise.
9668 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
9669 explicit-encoded elements in the XVEC for variable-length vectors.
9671 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9673 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
9675 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9676 Alan Hayward <alan.hayward@arm.com>
9677 David Sherwood <david.sherwood@arm.com>
9679 * coretypes.h (fixed_size_mode): Declare.
9680 (fixed_size_mode_pod): New typedef.
9681 * builtins.h (target_builtins::x_apply_args_mode)
9682 (target_builtins::x_apply_result_mode): Change type to
9683 fixed_size_mode_pod.
9684 * builtins.c (apply_args_size, apply_result_size, result_vector)
9685 (expand_builtin_apply_args_1, expand_builtin_apply)
9686 (expand_builtin_return): Update accordingly.
9688 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9690 * cse.c (hash_rtx_cb): Hash only the encoded elements.
9691 * cselib.c (cselib_hash_rtx): Likewise.
9692 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
9693 CONST_VECTOR encoding.
9695 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9696 Jeff Law <law@redhat.com>
9699 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
9700 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
9701 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
9702 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
9705 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
9706 explicitly probe *sp in a noreturn function if there were any callee
9707 register saves or frame pointer is needed.
9709 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9712 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
9713 BLKmode for ternary, binary or unary expressions.
9716 * var-tracking.c (delete_vta_debug_insn): New inline function.
9717 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
9718 insns from get_insns () to NULL instead of each bb separately.
9719 Use delete_vta_debug_insn. No longer static.
9720 (vt_debug_insns_local, variable_tracking_main_1): Adjust
9721 delete_vta_debug_insns callers.
9722 * rtl.h (delete_vta_debug_insns): Declare.
9723 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
9724 instead of variable_tracking_main.
9726 2018-01-03 Martin Sebor <msebor@redhat.com>
9728 PR tree-optimization/83603
9729 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
9730 arguments past the endof the argument list in functions declared
9731 without a prototype.
9732 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
9733 Avoid checking when arguments are null.
9735 2018-01-03 Martin Sebor <msebor@redhat.com>
9738 * doc/extend.texi (attribute const): Fix a typo.
9739 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
9740 issuing -Wsuggest-attribute for void functions.
9742 2018-01-03 Martin Sebor <msebor@redhat.com>
9744 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
9745 offset_int::from instead of wide_int::to_shwi.
9746 (maybe_diag_overlap): Remove assertion.
9747 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
9748 * gimple-ssa-sprintf.c (format_directive): Same.
9749 (parse_directive): Same.
9750 (sprintf_dom_walker::compute_format_length): Same.
9751 (try_substitute_return_value): Same.
9753 2018-01-03 Jeff Law <law@redhat.com>
9756 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
9757 non-constant residual for zero at runtime and avoid probing in
9758 that case. Reorganize code for trailing problem to mirror handling
9761 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9763 PR tree-optimization/83501
9764 * tree-ssa-strlen.c (get_string_cst): New.
9765 (handle_char_store): Call get_string_cst.
9767 2018-01-03 Martin Liska <mliska@suse.cz>
9769 PR tree-optimization/83593
9770 * tree-ssa-strlen.c: Include tree-cfg.h.
9771 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
9772 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
9773 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
9775 (strlen_dom_walker::before_dom_children): Call
9776 gimple_purge_dead_eh_edges. Dump tranformation with details
9778 (strlen_dom_walker::before_dom_children): Update call by adding
9779 new argument cleanup_eh.
9780 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
9782 2018-01-03 Martin Liska <mliska@suse.cz>
9785 * cif-code.def (VARIADIC_THUNK): New enum value.
9786 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
9789 2018-01-03 Jan Beulich <jbeulich@suse.com>
9791 * sse.md (mov<mode>_internal): Tighten condition for when to use
9792 vmovdqu<ssescalarsize> for TI and OI modes.
9794 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9796 Update copyright years.
9798 2018-01-03 Martin Liska <mliska@suse.cz>
9801 * ipa-visibility.c (function_and_variable_visibility): Skip
9802 functions with noipa attribure.
9804 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9806 * gcc.c (process_command): Update copyright notice dates.
9807 * gcov-dump.c (print_version): Ditto.
9808 * gcov.c (print_version): Ditto.
9809 * gcov-tool.c (print_version): Ditto.
9810 * gengtype.c (create_file): Ditto.
9811 * doc/cpp.texi: Bump @copying's copyright year.
9812 * doc/cppinternals.texi: Ditto.
9813 * doc/gcc.texi: Ditto.
9814 * doc/gccint.texi: Ditto.
9815 * doc/gcov.texi: Ditto.
9816 * doc/install.texi: Ditto.
9817 * doc/invoke.texi: Ditto.
9819 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9821 * vector-builder.h (vector_builder::m_full_nelts): Change from
9822 unsigned int to poly_uint64.
9823 (vector_builder::full_nelts): Update prototype accordingly.
9824 (vector_builder::new_vector): Likewise.
9825 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
9826 (vector_builder::operator ==): Likewise.
9827 (vector_builder::finalize): Likewise.
9828 * int-vector-builder.h (int_vector_builder::int_vector_builder):
9829 Take the number of elements as a poly_uint64 rather than an
9831 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
9832 from unsigned int to poly_uint64.
9833 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
9834 (vec_perm_indices::new_vector): Likewise.
9835 (vec_perm_indices::length): Likewise.
9836 (vec_perm_indices::nelts_per_input): Likewise.
9837 (vec_perm_indices::input_nelts): Likewise.
9838 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
9839 number of elements per input as a poly_uint64 rather than an
9840 unsigned int. Use the original encoding for variable-length
9841 vectors, rather than clamping each individual element.
9842 For the second and subsequent elements in each pattern,
9843 clamp the step and base before clamping their sum.
9844 (vec_perm_indices::series_p): Handle polynomial element counts.
9845 (vec_perm_indices::all_in_range_p): Likewise.
9846 (vec_perm_indices_to_tree): Likewise.
9847 (vec_perm_indices_to_rtx): Likewise.
9848 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
9849 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
9850 (tree_vector_builder::new_binary_operation): Handle polynomial
9851 element counts. Return false if we need to know the number
9852 of elements at compile time.
9853 * fold-const.c (fold_vec_perm): Punt if the number of elements
9854 isn't known at compile time.
9856 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9858 * vec-perm-indices.h (vec_perm_builder): Change element type
9859 from HOST_WIDE_INT to poly_int64.
9860 (vec_perm_indices::element_type): Update accordingly.
9861 (vec_perm_indices::clamp): Handle polynomial element_types.
9862 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9863 (vec_perm_indices::all_in_range_p): Likewise.
9864 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
9866 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
9867 polynomial vec_perm_indices element types.
9868 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
9869 * fold-const.c (fold_vec_perm): Likewise.
9870 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
9871 * tree-vect-generic.c (lower_vec_perm): Likewise.
9872 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9873 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
9874 element type to HOST_WIDE_INT.
9876 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9877 Alan Hayward <alan.hayward@arm.com>
9878 David Sherwood <david.sherwood@arm.com>
9880 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
9881 rather than an int. Use plus_constant.
9882 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
9883 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
9885 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9886 Alan Hayward <alan.hayward@arm.com>
9887 David Sherwood <david.sherwood@arm.com>
9889 * calls.c (emit_call_1, expand_call): Change struct_value_size from
9890 a HOST_WIDE_INT to a poly_int64.
9892 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9893 Alan Hayward <alan.hayward@arm.com>
9894 David Sherwood <david.sherwood@arm.com>
9896 * calls.c (load_register_parameters): Cope with polynomial
9897 mode sizes. Require a constant size for BLKmode parameters
9898 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
9899 forces a parameter to be padded at the lsb end in order to
9900 fill a complete number of words, require the parameter size
9901 to be ordered wrt UNITS_PER_WORD.
9903 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9904 Alan Hayward <alan.hayward@arm.com>
9905 David Sherwood <david.sherwood@arm.com>
9907 * reload1.c (spill_stack_slot_width): Change element type
9908 from unsigned int to poly_uint64_pod.
9909 (alter_reg): Treat mode sizes as polynomial.
9911 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9912 Alan Hayward <alan.hayward@arm.com>
9913 David Sherwood <david.sherwood@arm.com>
9915 * reload.c (complex_word_subreg_p): New function.
9916 (reload_inner_reg_of_subreg, push_reload): Use it.
9918 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9919 Alan Hayward <alan.hayward@arm.com>
9920 David Sherwood <david.sherwood@arm.com>
9922 * lra-constraints.c (process_alt_operands): Reject matched
9923 operands whose sizes aren't ordered.
9924 (match_reload): Refer to this check here.
9926 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9927 Alan Hayward <alan.hayward@arm.com>
9928 David Sherwood <david.sherwood@arm.com>
9930 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
9931 that the mode size is in the set {1, 2, 4, 8, 16}.
9933 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9934 Alan Hayward <alan.hayward@arm.com>
9935 David Sherwood <david.sherwood@arm.com>
9937 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
9938 Use plus_constant instead of gen_rtx_PLUS.
9940 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9941 Alan Hayward <alan.hayward@arm.com>
9942 David Sherwood <david.sherwood@arm.com>
9944 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
9945 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
9946 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
9947 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
9948 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
9949 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
9950 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
9951 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
9952 * config/i386/i386.c (ix86_push_rounding): ...this new function.
9953 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
9955 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
9956 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
9957 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
9958 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
9959 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
9960 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
9961 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
9962 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
9963 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
9964 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
9966 * expr.c (emit_move_resolve_push): Treat the input and result
9967 of PUSH_ROUNDING as a poly_int64.
9968 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
9969 (emit_push_insn): Likewise.
9970 * lra-eliminations.c (mark_not_eliminable): Likewise.
9971 * recog.c (push_operand): Likewise.
9972 * reload1.c (elimination_effects): Likewise.
9973 * rtlanal.c (nonzero_bits1): Likewise.
9974 * calls.c (store_one_arg): Likewise. Require the padding to be
9975 known at compile time.
9977 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9978 Alan Hayward <alan.hayward@arm.com>
9979 David Sherwood <david.sherwood@arm.com>
9981 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
9982 Use plus_constant instead of gen_rtx_PLUS.
9984 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9985 Alan Hayward <alan.hayward@arm.com>
9986 David Sherwood <david.sherwood@arm.com>
9988 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
9991 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9992 Alan Hayward <alan.hayward@arm.com>
9993 David Sherwood <david.sherwood@arm.com>
9995 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
9996 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
9997 via stack temporaries. Treat the mode size as polynomial too.
9999 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10000 Alan Hayward <alan.hayward@arm.com>
10001 David Sherwood <david.sherwood@arm.com>
10003 * expr.c (expand_expr_real_2): When handling conversions involving
10004 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
10005 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
10006 as a poly_uint64 too.
10008 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10009 Alan Hayward <alan.hayward@arm.com>
10010 David Sherwood <david.sherwood@arm.com>
10012 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
10014 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10015 Alan Hayward <alan.hayward@arm.com>
10016 David Sherwood <david.sherwood@arm.com>
10018 * combine.c (can_change_dest_mode): Handle polynomial
10019 REGMODE_NATURAL_SIZE.
10020 * expmed.c (store_bit_field_1): Likewise.
10021 * expr.c (store_constructor): Likewise.
10022 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
10023 and polynomial REGMODE_NATURAL_SIZE.
10024 (gen_lowpart_common): Likewise.
10025 * reginfo.c (record_subregs_of_mode): Likewise.
10026 * rtlanal.c (read_modify_subreg_p): Likewise.
10028 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10029 Alan Hayward <alan.hayward@arm.com>
10030 David Sherwood <david.sherwood@arm.com>
10032 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
10033 numbers of elements.
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 * match.pd: Cope with polynomial numbers of vector elements.
10041 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10042 Alan Hayward <alan.hayward@arm.com>
10043 David Sherwood <david.sherwood@arm.com>
10045 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
10046 in a POINTER_PLUS_EXPR.
10048 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10049 Alan Hayward <alan.hayward@arm.com>
10050 David Sherwood <david.sherwood@arm.com>
10052 * omp-simd-clone.c (simd_clone_subparts): New function.
10053 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
10054 (ipa_simd_modify_function_body): Likewise.
10056 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10057 Alan Hayward <alan.hayward@arm.com>
10058 David Sherwood <david.sherwood@arm.com>
10060 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
10061 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
10062 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
10063 (expand_vector_condition, vector_element): Likewise.
10064 (subparts_gt): New function.
10065 (get_compute_type): Use subparts_gt.
10066 (count_type_subparts): Delete.
10067 (expand_vector_operations_1): Use subparts_gt instead of
10068 count_type_subparts.
10070 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10071 Alan Hayward <alan.hayward@arm.com>
10072 David Sherwood <david.sherwood@arm.com>
10074 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
10075 (vect_compile_time_alias): ...this new function. Do the calculation
10076 on poly_ints rather than trees.
10077 (vect_prune_runtime_alias_test_list): Update call accordingly.
10079 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10080 Alan Hayward <alan.hayward@arm.com>
10081 David Sherwood <david.sherwood@arm.com>
10083 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
10085 (vect_schedule_slp_instance): Likewise.
10087 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10088 Alan Hayward <alan.hayward@arm.com>
10089 David Sherwood <david.sherwood@arm.com>
10091 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
10092 constant and extern definitions for variable-length vectors.
10093 (vect_get_constant_vectors): Note that the number of units
10094 is known to be constant.
10096 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10097 Alan Hayward <alan.hayward@arm.com>
10098 David Sherwood <david.sherwood@arm.com>
10100 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
10101 of units as polynomial. Choose between WIDE and NARROW based
10104 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10105 Alan Hayward <alan.hayward@arm.com>
10106 David Sherwood <david.sherwood@arm.com>
10108 * tree-vect-stmts.c (simd_clone_subparts): New function.
10109 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
10111 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10112 Alan Hayward <alan.hayward@arm.com>
10113 David Sherwood <david.sherwood@arm.com>
10115 * tree-vect-stmts.c (vectorizable_call): Treat the number of
10116 vectors as polynomial. Use build_index_vector for
10117 IFN_GOMP_SIMD_LANE.
10119 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10120 Alan Hayward <alan.hayward@arm.com>
10121 David Sherwood <david.sherwood@arm.com>
10123 * tree-vect-stmts.c (get_load_store_type): Treat the number of
10124 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
10125 for variable-length vectors.
10126 (vectorizable_mask_load_store): Treat the number of units as
10127 polynomial, asserting that it is constant if the condition has
10128 already been enforced.
10129 (vectorizable_store, vectorizable_load): Likewise.
10131 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10132 Alan Hayward <alan.hayward@arm.com>
10133 David Sherwood <david.sherwood@arm.com>
10135 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
10136 of units as polynomial. Punt if we can't tell at compile time
10137 which vector contains the final result.
10139 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10140 Alan Hayward <alan.hayward@arm.com>
10141 David Sherwood <david.sherwood@arm.com>
10143 * tree-vect-loop.c (vectorizable_induction): Treat the number
10144 of units as polynomial. Punt on SLP inductions. Use an integer
10145 VEC_SERIES_EXPR for variable-length integer reductions. Use a
10146 cast of such a series for variable-length floating-point
10149 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10150 Alan Hayward <alan.hayward@arm.com>
10151 David Sherwood <david.sherwood@arm.com>
10153 * tree.h (build_index_vector): Declare.
10154 * tree.c (build_index_vector): New function.
10155 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
10156 of units as polynomial, forcibly converting it to a constant if
10157 vectorizable_reduction has already enforced the condition.
10158 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
10159 to create a {1,2,3,...} vector.
10160 (vectorizable_reduction): Treat the number of units as polynomial.
10161 Choose vectype_in based on the largest scalar element size rather
10162 than the smallest number of units. Enforce the restrictions
10165 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10166 Alan Hayward <alan.hayward@arm.com>
10167 David Sherwood <david.sherwood@arm.com>
10169 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
10170 number of units as polynomial.
10172 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10173 Alan Hayward <alan.hayward@arm.com>
10174 David Sherwood <david.sherwood@arm.com>
10176 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
10177 * target.def (autovectorize_vector_sizes): Return the vector sizes
10178 by pointer, using vector_sizes rather than a bitmask.
10179 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
10180 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
10181 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
10183 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
10184 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
10185 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
10186 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
10187 * omp-general.c (omp_max_vf): Likewise.
10188 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10189 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
10190 * tree-vect-loop.c (vect_analyze_loop): Likewise.
10191 * tree-vect-slp.c (vect_slp_bb): Likewise.
10192 * doc/tm.texi: Regenerate.
10193 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
10195 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
10196 the vector size as a poly_uint64 rather than an unsigned int.
10197 (current_vector_size): Change from an unsigned int to a poly_uint64.
10198 (get_vectype_for_scalar_type): Update accordingly.
10199 * tree.h (build_truth_vector_type): Take the size and number of
10200 units as a poly_uint64 rather than an unsigned int.
10201 (build_vector_type): Add a temporary overload that takes
10202 the number of units as a poly_uint64 rather than an unsigned int.
10203 * tree.c (make_vector_type): Likewise.
10204 (build_truth_vector_type): Take the number of units as a poly_uint64
10205 rather than an unsigned int.
10207 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10208 Alan Hayward <alan.hayward@arm.com>
10209 David Sherwood <david.sherwood@arm.com>
10211 * target.def (get_mask_mode): Take the number of units and length
10212 as poly_uint64s rather than unsigned ints.
10213 * targhooks.h (default_get_mask_mode): Update accordingly.
10214 * targhooks.c (default_get_mask_mode): Likewise.
10215 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
10216 * doc/tm.texi: Regenerate.
10218 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10219 Alan Hayward <alan.hayward@arm.com>
10220 David Sherwood <david.sherwood@arm.com>
10222 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
10223 * omp-general.c (omp_max_vf): Likewise.
10224 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
10225 (expand_omp_simd): Handle polynomial safelen.
10226 * omp-low.c (omplow_simd_context): Add a default constructor.
10227 (omplow_simd_context::max_vf): Change from int to poly_uint64.
10228 (lower_rec_simd_input_clauses): Update accordingly.
10229 (lower_rec_input_clauses): Likewise.
10231 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10232 Alan Hayward <alan.hayward@arm.com>
10233 David Sherwood <david.sherwood@arm.com>
10235 * tree-vectorizer.h (vect_nunits_for_cost): New function.
10236 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
10237 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
10238 (vect_analyze_slp_cost): Likewise.
10239 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
10240 (vect_model_load_cost): Likewise.
10242 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10243 Alan Hayward <alan.hayward@arm.com>
10244 David Sherwood <david.sherwood@arm.com>
10246 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
10247 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
10248 from an unsigned int * to a poly_uint64_pod *.
10249 (calculate_unrolling_factor): New function.
10250 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
10252 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10253 Alan Hayward <alan.hayward@arm.com>
10254 David Sherwood <david.sherwood@arm.com>
10256 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
10257 from an unsigned int to a poly_uint64.
10258 (_loop_vec_info::slp_unrolling_factor): Likewise.
10259 (_loop_vec_info::vectorization_factor): Change from an int
10261 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
10262 (vect_get_num_vectors): New function.
10263 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
10264 (vect_get_num_copies): Use vect_get_num_vectors.
10265 (vect_analyze_data_ref_dependences): Change max_vf from an int *
10266 to an unsigned int *.
10267 (vect_analyze_data_refs): Change min_vf from an int * to a
10269 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10270 than an unsigned HOST_WIDE_INT.
10271 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
10272 (vect_analyze_data_ref_dependence): Change max_vf from an int *
10273 to an unsigned int *.
10274 (vect_analyze_data_ref_dependences): Likewise.
10275 (vect_compute_data_ref_alignment): Handle polynomial vf.
10276 (vect_enhance_data_refs_alignment): Likewise.
10277 (vect_prune_runtime_alias_test_list): Likewise.
10278 (vect_shift_permute_load_chain): Likewise.
10279 (vect_supportable_dr_alignment): Likewise.
10280 (dependence_distance_ge_vf): Take the vectorization factor as a
10281 poly_uint64 rather than an unsigned HOST_WIDE_INT.
10282 (vect_analyze_data_refs): Change min_vf from an int * to a
10284 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
10285 vfm1 as a poly_uint64 rather than an int. Make the same change
10286 for the returned bound_scalar.
10287 (vect_gen_vector_loop_niters): Handle polynomial vf.
10288 (vect_do_peeling): Likewise. Update call to
10289 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
10290 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
10292 * tree-vect-loop.c (vect_determine_vectorization_factor)
10293 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
10294 (vect_get_known_peeling_cost): Likewise.
10295 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
10296 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
10297 (vect_transform_loop): Likewise. Use the lowest possible VF when
10298 updating the upper bounds of the loop.
10299 (vect_min_worthwhile_factor): Make static. Return an unsigned int
10300 rather than an int.
10301 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
10302 polynomial unroll factors.
10303 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
10304 (vect_make_slp_decision): Likewise.
10305 (vect_supported_load_permutation_p): Likewise, and polynomial
10307 (vect_analyze_slp_cost): Handle polynomial vf.
10308 (vect_slp_analyze_node_operations): Likewise.
10309 (vect_slp_analyze_bb_1): Likewise.
10310 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10311 than an unsigned HOST_WIDE_INT.
10312 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
10313 (vectorizable_load): Handle polynomial vf.
10314 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
10316 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
10318 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10319 Alan Hayward <alan.hayward@arm.com>
10320 David Sherwood <david.sherwood@arm.com>
10322 * match.pd: Handle bit operations involving three constants
10323 and try to fold one pair.
10325 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10327 * tree-vect-loop-manip.c: Include gimple-fold.h.
10328 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
10329 niters_maybe_zero parameters. Handle other cases besides a step of 1.
10330 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
10331 Add a path that uses a step of VF instead of 1, but disable it
10333 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
10334 and niters_no_overflow parameters. Update calls to
10335 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
10336 Create a new SSA name if the latter choses to use a ste other
10337 than zero, and return it via niters_vector_mult_vf_var.
10338 * tree-vect-loop.c (vect_transform_loop): Update calls to
10339 vect_do_peeling, vect_gen_vector_loop_niters and
10340 slpeel_make_loop_iterate_ntimes.
10341 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
10342 (vect_gen_vector_loop_niters): Update declarations after above changes.
10344 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
10346 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
10347 128-bit round to integer instructions.
10348 (ceil<mode>2): Likewise.
10349 (btrunc<mode>2): Likewise.
10350 (round<mode>2): Likewise.
10352 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10354 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
10355 unaligned VSX load/store on P8/P9.
10356 (expand_block_clear): Allow the use of unaligned VSX
10357 load/store on P8/P9.
10359 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10361 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
10363 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
10364 swap associated with both a load and a store.
10366 2018-01-02 Andrew Waterman <andrew@sifive.com>
10368 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
10369 * config/riscv/riscv.md (clear_cache): Use it.
10371 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
10373 * web.c: Remove out-of-date comment.
10375 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10377 * expr.c (fixup_args_size_notes): Check that any existing
10378 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
10379 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
10380 (emit_single_push_insn): ...here.
10382 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10384 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
10385 (const_vector_encoded_nelts): New function.
10386 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
10387 (const_vector_int_elt, const_vector_elt): Declare.
10388 * emit-rtl.c (const_vector_int_elt_1): New function.
10389 (const_vector_elt): Likewise.
10390 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
10391 of CONST_VECTOR_ELT.
10393 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10395 * expr.c: Include rtx-vector-builder.h.
10396 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
10397 directly on the tree encoding.
10398 (const_vector_from_tree): Likewise.
10399 * optabs.c: Include rtx-vector-builder.h.
10400 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
10401 sequence of "u" values.
10402 * vec-perm-indices.c: Include rtx-vector-builder.h.
10403 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
10404 directly on the vec_perm_indices encoding.
10406 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10408 * doc/rtl.texi (const_vector): Describe new encoding scheme.
10409 * Makefile.in (OBJS): Add rtx-vector-builder.o.
10410 * rtx-vector-builder.h: New file.
10411 * rtx-vector-builder.c: Likewise.
10412 * rtl.h (rtx_def::u2): Add a const_vector field.
10413 (CONST_VECTOR_NPATTERNS): New macro.
10414 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
10415 (CONST_VECTOR_DUPLICATE_P): Likewise.
10416 (CONST_VECTOR_STEPPED_P): Likewise.
10417 (CONST_VECTOR_ENCODED_ELT): Likewise.
10418 (const_vec_duplicate_p): Check for a duplicated vector encoding.
10419 (unwrap_const_vec_duplicate): Likewise.
10420 (const_vec_series_p): Check for a non-duplicated vector encoding.
10421 Say that the function only returns true for integer vectors.
10422 * emit-rtl.c: Include rtx-vector-builder.h.
10423 (gen_const_vec_duplicate_1): Delete.
10424 (gen_const_vector): Call gen_const_vec_duplicate instead of
10425 gen_const_vec_duplicate_1.
10426 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
10427 (gen_const_vec_duplicate): Use rtx_vector_builder.
10428 (gen_const_vec_series): Likewise.
10429 (gen_rtx_CONST_VECTOR): Likewise.
10430 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
10431 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10432 Build a new vector rather than modifying a CONST_VECTOR in-place.
10433 (handle_special_swappables): Update call accordingly.
10434 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
10435 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10436 Build a new vector rather than modifying a CONST_VECTOR in-place.
10437 (handle_special_swappables): Update call accordingly.
10439 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10441 * simplify-rtx.c (simplify_const_binary_operation): Use
10442 CONST_VECTOR_ELT instead of XVECEXP.
10444 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10446 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
10447 the selector elements to be different from the data elements
10448 if the selector is a VECTOR_CST.
10449 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
10450 ssizetype for the selector.
10452 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10454 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
10455 before testing each element individually.
10456 * tree-vect-generic.c (lower_vec_perm): Likewise.
10458 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10460 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
10461 * selftest-run-tests.c (selftest::run_tests): Call it.
10462 * vector-builder.h (vector_builder::operator ==): New function.
10463 (vector_builder::operator !=): Likewise.
10464 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
10465 (vec_perm_indices::all_from_input_p): New function.
10466 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10467 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
10468 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
10469 instead of reading the VECTOR_CST directly. Detect whether both
10470 vector inputs are the same before constructing the vec_perm_indices,
10471 and update the number of inputs argument accordingly. Use the
10472 utility functions added above. Only construct sel2 if we need to.
10474 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10476 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
10477 the broadcast of the low byte.
10478 (expand_mult_highpart): Use an explicit encoding for the permutes.
10479 * optabs-query.c (can_mult_highpart_p): Likewise.
10480 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
10481 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10482 (vectorizable_bswap): Likewise.
10483 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
10484 explicit encoding for the power-of-2 permutes.
10485 (vect_permute_store_chain): Likewise.
10486 (vect_grouped_load_supported): Likewise.
10487 (vect_permute_load_chain): Likewise.
10489 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10491 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
10492 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
10493 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
10494 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10495 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
10496 (vect_gen_perm_mask_any): Likewise.
10498 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10500 * int-vector-builder.h: New file.
10501 * vec-perm-indices.h: Include int-vector-builder.h.
10502 (vec_perm_indices): Redefine as an int_vector_builder.
10503 (auto_vec_perm_indices): Delete.
10504 (vec_perm_builder): Redefine as a stand-alone class.
10505 (vec_perm_indices::vec_perm_indices): New function.
10506 (vec_perm_indices::clamp): Likewise.
10507 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
10508 (vec_perm_indices::new_vector): New function.
10509 (vec_perm_indices::new_expanded_vector): Update for new
10510 vec_perm_indices class.
10511 (vec_perm_indices::rotate_inputs): New function.
10512 (vec_perm_indices::all_in_range_p): Operate directly on the
10513 encoded form, without computing elided elements.
10514 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
10515 encoding. Update for new vec_perm_indices class.
10516 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
10517 the given vec_perm_builder.
10518 (expand_vec_perm_var): Update vec_perm_builder constructor.
10519 (expand_mult_highpart): Use vec_perm_builder instead of
10520 auto_vec_perm_indices.
10521 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
10522 vec_perm_indices instead of auto_vec_perm_indices. Use a single
10523 or double series encoding as appropriate.
10524 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
10525 vec_perm_indices instead of auto_vec_perm_indices.
10526 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10527 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10528 (vect_permute_store_chain): Likewise.
10529 (vect_grouped_load_supported): Likewise.
10530 (vect_permute_load_chain): Likewise.
10531 (vect_shift_permute_load_chain): Likewise.
10532 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10533 (vect_transform_slp_perm_load): Likewise.
10534 (vect_schedule_slp_instance): Likewise.
10535 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10536 (vectorizable_mask_load_store): Likewise.
10537 (vectorizable_bswap): Likewise.
10538 (vectorizable_store): Likewise.
10539 (vectorizable_load): Likewise.
10540 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
10541 vec_perm_indices instead of auto_vec_perm_indices. Use
10542 tree_to_vec_perm_builder to read the vector from a tree.
10543 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
10544 vec_perm_builder instead of a vec_perm_indices.
10545 (have_whole_vector_shift): Use vec_perm_builder and
10546 vec_perm_indices instead of auto_vec_perm_indices. Leave the
10547 truncation to calc_vec_perm_mask_for_shift.
10548 (vect_create_epilog_for_reduction): Likewise.
10549 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
10550 from auto_vec_perm_indices to vec_perm_indices.
10551 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10552 instead of changing individual elements.
10553 (aarch64_vectorize_vec_perm_const): Use new_vector to install
10554 the vector in d.perm.
10555 * config/arm/arm.c (expand_vec_perm_d::perm): Change
10556 from auto_vec_perm_indices to vec_perm_indices.
10557 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10558 instead of changing individual elements.
10559 (arm_vectorize_vec_perm_const): Use new_vector to install
10560 the vector in d.perm.
10561 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
10562 Update vec_perm_builder constructor.
10563 (rs6000_expand_interleave): Likewise.
10564 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
10565 (rs6000_expand_interleave): Likewise.
10567 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10569 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
10570 to qimode could truncate the indices.
10571 * optabs.c (expand_vec_perm_var): Likewise.
10573 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10575 * Makefile.in (OBJS): Add vec-perm-indices.o.
10576 * vec-perm-indices.h: New file.
10577 * vec-perm-indices.c: Likewise.
10578 * target.h (vec_perm_indices): Replace with a forward class
10580 (auto_vec_perm_indices): Move to vec-perm-indices.h.
10581 * optabs.h: Include vec-perm-indices.h.
10582 (expand_vec_perm): Delete.
10583 (selector_fits_mode_p, expand_vec_perm_var): Declare.
10584 (expand_vec_perm_const): Declare.
10585 * target.def (vec_perm_const_ok): Replace with...
10586 (vec_perm_const): ...this new hook.
10587 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
10588 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
10589 * doc/tm.texi: Regenerate.
10590 * optabs.def (vec_perm_const): Delete.
10591 * doc/md.texi (vec_perm_const): Likewise.
10592 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
10593 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
10594 expand_vec_perm for constant permutation vectors. Assert that
10595 the mode of variable permutation vectors is the integer equivalent
10596 of the mode that is being permuted.
10597 * optabs-query.h (selector_fits_mode_p): Declare.
10598 * optabs-query.c: Include vec-perm-indices.h.
10599 (selector_fits_mode_p): New function.
10600 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
10601 is defined, instead of checking whether the vec_perm_const_optab
10602 exists. Use targetm.vectorize.vec_perm_const instead of
10603 targetm.vectorize.vec_perm_const_ok. Check whether the indices
10604 fit in the vector mode before using a variable permute.
10605 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
10606 vec_perm_indices instead of an rtx.
10607 (expand_vec_perm): Replace with...
10608 (expand_vec_perm_const): ...this new function. Take the selector
10609 as a vec_perm_indices rather than an rtx. Also take the mode of
10610 the selector. Update call to shift_amt_for_vec_perm_mask.
10611 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
10612 Use vec_perm_indices::new_expanded_vector to expand the original
10613 selector into bytes. Check whether the indices fit in the vector
10614 mode before using a variable permute.
10615 (expand_vec_perm_var): Make global.
10616 (expand_mult_highpart): Use expand_vec_perm_const.
10617 * fold-const.c: Includes vec-perm-indices.h.
10618 * tree-ssa-forwprop.c: Likewise.
10619 * tree-vect-data-refs.c: Likewise.
10620 * tree-vect-generic.c: Likewise.
10621 * tree-vect-loop.c: Likewise.
10622 * tree-vect-slp.c: Likewise.
10623 * tree-vect-stmts.c: Likewise.
10624 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
10626 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
10627 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
10628 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
10629 (aarch64_vectorize_vec_perm_const): ...this new function.
10630 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10631 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10632 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
10633 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
10634 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10635 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10636 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
10638 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
10639 check for NEON modes.
10640 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
10641 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
10642 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
10643 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
10645 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
10646 the old VEC_PERM_CONST conditions.
10647 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
10648 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
10649 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
10650 (ia64_vectorize_vec_perm_const_ok): Merge into...
10651 (ia64_vectorize_vec_perm_const): ...this new function.
10652 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
10653 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
10654 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
10655 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
10656 * config/mips/mips.c (mips_expand_vec_perm_const)
10657 (mips_vectorize_vec_perm_const_ok): Merge into...
10658 (mips_vectorize_vec_perm_const): ...this new function.
10659 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
10660 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
10661 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
10662 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
10663 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
10664 (rs6000_expand_vec_perm_const): Delete.
10665 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
10667 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10668 (altivec_expand_vec_perm_const_le): Take each operand individually.
10669 Operate on constant selectors rather than rtxes.
10670 (altivec_expand_vec_perm_const): Likewise. Update call to
10671 altivec_expand_vec_perm_const_le.
10672 (rs6000_expand_vec_perm_const): Delete.
10673 (rs6000_vectorize_vec_perm_const_ok): Delete.
10674 (rs6000_vectorize_vec_perm_const): New function.
10675 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10676 an element count and rtx array.
10677 (rs6000_expand_extract_even): Update call accordingly.
10678 (rs6000_expand_interleave): Likewise.
10679 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
10680 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
10681 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
10682 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
10683 (rs6000_expand_vec_perm_const): Delete.
10684 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10685 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10686 (altivec_expand_vec_perm_const_le): Take each operand individually.
10687 Operate on constant selectors rather than rtxes.
10688 (altivec_expand_vec_perm_const): Likewise. Update call to
10689 altivec_expand_vec_perm_const_le.
10690 (rs6000_expand_vec_perm_const): Delete.
10691 (rs6000_vectorize_vec_perm_const_ok): Delete.
10692 (rs6000_vectorize_vec_perm_const): New function. Remove stray
10693 reference to the SPE evmerge intructions.
10694 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10695 an element count and rtx array.
10696 (rs6000_expand_extract_even): Update call accordingly.
10697 (rs6000_expand_interleave): Likewise.
10698 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
10699 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
10701 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10703 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10705 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
10706 vector mode and that that mode matches the mode of the data
10708 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
10709 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
10710 directly using expand_vec_perm_1 when forcing selectors into
10712 (expand_vec_perm_var): New function, split out from expand_vec_perm.
10714 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10716 * optabs-query.h (can_vec_perm_p): Delete.
10717 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
10718 * optabs-query.c (can_vec_perm_p): Split into...
10719 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
10720 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
10721 particular selector is valid.
10722 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10723 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10724 (vect_grouped_load_supported): Likewise.
10725 (vect_shift_permute_load_chain): Likewise.
10726 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10727 (vect_transform_slp_perm_load): Likewise.
10728 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10729 (vectorizable_bswap): Likewise.
10730 (vect_gen_perm_mask_checked): Likewise.
10731 * fold-const.c (fold_ternary_loc): Likewise. Don't take
10732 implementations of variable permutation vectors into account
10733 when deciding which selector to use.
10734 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
10735 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
10736 with a false third argument.
10737 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
10738 to test whether the constant selector is valid and can_vec_perm_var_p
10739 to test whether a variable selector is valid.
10741 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10743 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
10744 * optabs-query.c (can_vec_perm_p): Likewise.
10745 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
10746 instead of vec_perm_indices.
10747 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
10748 (vect_gen_perm_mask_checked): Likewise,
10749 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
10750 (vect_gen_perm_mask_checked): Likewise,
10752 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10754 * optabs-query.h (qimode_for_vec_perm): Declare.
10755 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
10756 (qimode_for_vec_perm): ...this new function.
10757 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
10759 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10761 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
10762 does not have a conditional at the top.
10764 2018-01-02 Richard Biener <rguenther@suse.de>
10766 * ipa-inline.c (big_speedup_p): Fix expression.
10768 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10771 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
10774 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10778 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
10779 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
10780 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
10781 cond_taken_branch_cost 3->4.
10783 2018-01-01 Jakub Jelinek <jakub@redhat.com>
10785 PR tree-optimization/83581
10786 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
10787 TODO_cleanup_cfg if any changes have been made.
10789 PR middle-end/83608
10790 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
10791 convert_modes if target mode has the right side, but different mode
10794 PR middle-end/83609
10795 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
10796 last argument when extracting from CONCAT. If either from_real or
10797 from_imag is NULL, use expansion through memory. If result is not
10798 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
10799 the parts directly to inner mode, if even that fails, use expansion
10802 PR middle-end/83623
10803 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
10804 check for bswap in mode rather than HImode and use that in expand_unop
10807 Copyright (C) 2018 Free Software Foundation, Inc.
10809 Copying and distribution of this file, with or without modification,
10810 are permitted in any medium without royalty provided the copyright
10811 notice and this notice are preserved.