1 2018-04-10 Martin Liska <mliska@suse.cz>
3 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4 cgraph_edge and ipa_ref.
6 2018-04-10 Jakub Jelinek <jakub@redhat.com>
11 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
12 computation of the VEC_MERGE selector from mask.
13 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
14 Fix decoding of the VEC_MERGE selector into mask.
16 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
18 PR tree-optimization/85286
19 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
21 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
23 * final.c (final_1): Set insn_last_address as well as
26 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29 * explow.c (emit_stack_probe): Call validize_mem on memory location
30 before passing it to gen_probe_stack. Create address operand and
31 legitimize it for the probe_stack_address case.
33 2018-04-09 Jan Hubicka <jh@suse.cz>
36 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
37 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
38 * tree.c (free_lang_data_in_type): Fix handling of binfos;
40 (free_lang_data): Rebuild type inheritance graph.
42 2018-04-09 Martin Sebor <msebor@redhat.com>
44 * invoke.texi (-finline-small-functions): Mention other optimization
46 (-findirect-inlining, -fpartial-inlining): Same.
47 (-finline-functions-called-once): Same.
48 (-freorder-blocks-and-partition): Same.
50 2018-04-09 Jan Hubicka <jh@suse.cz>
53 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
54 jumps; choose last target that matches the criteria (i.e.
55 no partition changes for non-crossing jumps).
56 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
57 support for redirecting crossing jumps to non-crossing.
59 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
61 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
62 also for naked functions.
64 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
66 * config/arc/arc.md (add_shift): New pattern.
67 (add_shift2): Likewise.
68 (sub_shift): Likewise.
69 (sub_shift_cmp0_noout): Likewise.
70 (compare_si_ashiftsi): Likewise.
71 (xbfu_cmp0_noout): New combine pattern.
72 (xbfu_cmp0"): Likewise.
73 (movsi_set_cc_insn): Place the predicable variant first.
74 (commutative_binary_cmp0_noout): Remove clobber.
75 (commutative_binary_cmp0): New pattern.
76 (noncommutative_binary_cmp0): Likewise.
77 (noncommutative_binary_cmp0_noout): Likewise.
78 (noncommutative_binary_comparison_result_used): Removed.
79 (rsub_cmp0): New pattern.
80 (rsub_cmp0_noout): Likewise.
81 (extzvsi): Changed, keep only meaningful variants.
82 (SQH, SEZ): New iterators.
83 (SQH_postfix): New mode attribute.
84 (SEZ_prefix): New code attribute.
85 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
86 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
87 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
89 (noncommutative_operator): Check the availability of barrel
92 2018-04-09 Richard Biener <rguenther@suse.de>
94 PR tree-optimization/85284
95 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
96 Only use the niter constraining form of simple_iv when the exit
99 2018-04-09 Tom de Vries <tom@codesourcery.com>
102 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
103 (define_expand "*memory_barrier"): New define_expand.
104 (define_insn "memory_barrier"): New insn.
106 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
108 PR rtl-optimization/80463
109 PR rtl-optimization/83972
110 PR rtl-optimization/83480
112 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
113 correct producer for the insn.
114 (tidy_control_flow): Fixup seqnos in case of debug insns.
116 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
118 PR rtl-optimization/83913
120 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
121 different sched-times when merging exprs.
123 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
125 PR rtl-optimization/83962
127 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
128 tidy_fallthru_edge and tidy_control_flow.
130 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
132 PR rtl-optimization/83530
134 * sel-sched.c (force_next_insn): New global variable.
135 (remove_insn_for_debug): When force_next_insn is true, also leave only
136 next insn in the ready list.
137 (sel_sched_region): When the region wasn't scheduled, make another pass
138 over it with force_next_insn set to 1.
140 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
142 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
144 * config/nds32/constants.md (unspec_volatile_element): Add enum values
145 for interrupt control.
146 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
147 functions for interrupt control.
148 * config/nds32/nds32-intrinsic.md: Likewise.
149 * config/nds32/nds32_intrinsic.h: Likewise.
150 * config/nds32/nds32.h (nds32_builtins): Likewise.
152 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
154 * config/nds32/nds32.c (nds32_init_machine_status,
155 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
156 strict_aligned_p field.
157 (nds32_expand_to_rtl_hook): New function.
158 (TARGET_EXPAND_TO_RTL_HOOK): Define.
159 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
161 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
162 Chung-Ju Wu <jasonwucj@gmail.com>
164 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
165 * config/nds32/nds32-n7.md: New file.
166 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
167 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
169 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
170 * config/nds32/nds32.md (pipeline_model): Add n7.
171 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
172 * config/nds32/pipelines.md: Include n7 settings.
174 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
175 Chung-Ju Wu <jasonwucj@gmail.com>
177 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
178 * config/nds32/nds32-e8.md: New file.
179 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
180 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
182 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
183 * config/nds32/nds32.md (pipeline_model): Add e8.
184 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
185 * config/nds32/pipelines.md: Include e8 settings.
187 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
188 Chung-Ju Wu <jasonwucj@gmail.com>
190 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
191 * config/nds32/nds32-n8.md: New file.
192 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
193 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
195 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
196 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
197 * config/nds32/nds32.md (pipeline_model): Add n8.
198 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
199 * config/nds32/pipelines.md: Include n8 settings.
201 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
202 Chung-Ju Wu <jasonwucj@gmail.com>
204 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
205 * config/nds32/nds32-n9-2r1w.md: New file.
206 * config/nds32/nds32-n9-3r2w.md: New file.
207 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
208 nds32_register_ports): New or modify for cpu n9.
209 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
211 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
212 * config/nds32/nds32-utils.c: New file.
213 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
214 TARGET_MUL_SLOW): Define.
215 * config/nds32/nds32.md (pipeline_model): New attribute.
216 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
217 New options that support cpu n9.
218 * config/nds32/pipelines.md: Include n9 settings.
219 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
221 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
223 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
224 information if necessary.
225 (output_cond_branch_compare_zero): Likewise.
226 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
227 (nds32_target_alignment): Refine for alignment.
228 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
229 (FUNCTION_BOUNDARY): Modify.
230 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
232 * config/nds32/nds32.opt (malways-align, malign-functions): New.
234 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
236 * config/nds32/constants.md (unspec_volatile_element): Add values for
237 TLB operation and data prefetch.
238 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
239 functions for TLB operation and data prefetch.
240 * config/nds32/nds32-intrinsic.md: Likewise.
241 * config/nds32/nds32_intrinsic.h: Likewise.
242 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
243 (nds32_print_operand): Likewise.
244 * config/nds32/nds32.h (nds32_builtins): Likewise.
246 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
247 Andrew Pinski <pinsika@gcc.gnu.org>
250 * match.pd: Use constant_boolean_node of correct type instead of
251 boolean_true_node or boolean_false_node for simplifying
252 pointer comparisons to zero.
254 2018-04-07 Jakub Jelinek <jakub@redhat.com>
256 PR tree-optimization/80021
257 * tree.c (verify_type_variant): Make error call in verify_variant_match
258 translatable and remove final full stop.
260 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
262 * config/nds32/constants.md (unspec_volatile_element): Add
263 UNSPEC_VOLATILE_EH_RETURN.
264 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
265 nds32_output_stack_pop): Support dwarf exception handling process.
266 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
267 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
268 exception handling process.
269 (nds32_compute_stack_frame): Likewise.
270 (nds32_return_addr_rtx): Likewise.
271 (nds32_initial_elimination_offset): Likewise.
272 (nds32_expand_prologue): Likewise.
273 (nds32_expand_epilogue): Likewise.
274 (nds32_dynamic_chain_address): New function.
275 * config/nds32/nds32.h (machine_function): Add fields for dwarf
277 (DYNAMIC_CHAIN_ADDRESS): Define.
278 (EH_RETURN_DATA_REGNO): Define.
279 (EH_RETURN_STACKADJ_RTX): Define.
280 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
281 patterns for dwarf exception handling.
283 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
285 * config/nds32/nds32.h: Clean up obsolete macros.
287 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
289 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
290 Add enum values for particular instructions.
291 * config/nds32/nds32-intrinsic.c: Implementation of expanding
292 particular intrinsic functions.
293 * config/nds32/nds32-intrinsic.md: Likewise.
294 * config/nds32/nds32_intrinsic.h: Likewise.
295 * config/nds32/nds32.h (nds32_builtins): Likewise.
296 * config/nds32/nds32.md (type): Add pbsad and pbsada.
297 (btst, ave): New patterns for particular instructions.
299 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
301 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
302 Add enum values for atomic load/store and memory sync.
303 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
305 * config/nds32/nds32-intrinsic.md: Likewise.
306 * config/nds32/nds32_intrinsic.h: Likewise.
307 * config/nds32/nds32.h (nds32_builtins): Likewise.
309 2018-04-07 Jakub Jelinek <jakub@redhat.com>
311 PR tree-optimization/85257
312 * fold-const.c (native_encode_vector): If not all elts could fit
313 and off is -1, return 0 rather than offset.
314 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
315 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
316 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
317 adjust buffer in native_interpret_expr call.
319 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
321 * config/nds32/constants.md (unspec_volatile_element): Add cache
323 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
324 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
325 * config/nds32/nds32.c (nds32_cctl_names): New.
326 (nds32_print_operand): Handle cache control register names.
327 * config/nds32/nds32.h (nds32_builtins): New enum values.
328 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
330 * config/nds32/nds32.md (type): Add mmu.
331 * config/nds32/pipelines.md (simple_insn): Add mmu.
333 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
335 * config/nds32/nds32.md (type): Remove call.
336 * config/nds32/pipelines.md (simple_insn): Likewise.
338 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
340 * config/nds32/constants.md (unspec_volatile_element): Add
341 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
342 UNSPEC_VOLATILE_FMFCFG.
343 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
344 description for fmfcfg and fmfcsr.
345 (bdesc_1arg): Add fmtcsr.
346 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
347 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
348 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
349 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
350 unspec_fmfcfg): New patterns.
351 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
352 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
353 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
354 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
355 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
356 __nds32__fmfcfg): Define.
358 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
360 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
361 intrinsic register names.
362 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
363 intrinsic register enum values and macros.
365 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
367 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
368 for load/store addressing form.
369 (nds32_print_operand_address): Likewise.
371 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
374 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
375 based on LABEL_REF. Remove useless assertion.
376 (pic_address_needs_scratch): Fix formatting.
377 (sparc_legitimize_pic_address): Minor tweaks.
378 (sparc_delegitimize_address): Adjust assertion accordingly.
379 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
380 into symbolic_operand.
381 (movsi_high_pic_label_ref): Likewise.
382 (movsi_lo_sum_pic_label_ref): Likewise.
383 (movdi_pic_label_ref): Likewise.
384 (movdi_high_pic_label_ref): Likewise.
385 (movdi_lo_sum_pic_label_ref): Likewise.
387 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
389 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
390 custom LIB_SPEC setup.
392 2018-04-06 Ruslan Bukin <br@bsdpad.com>
393 Kito Cheng <kito.cheng@gmail.com>
395 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
396 * config/riscv/freebsd.h: New.
398 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
400 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
401 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
404 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
405 Kito Cheng <kito.cheng@gmail.com>
407 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
408 nds32_output_call, nds32_symbol_binds_local_p): New functions.
409 * config/nds32/nds32-protos.h (nds32_output_call,
410 nds32_output_return): Declare.
411 * config/nds32/nds32.md: Refine all the call and return patterns.
413 2018-04-06 Jakub Jelinek <jakub@redhat.com>
416 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
417 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
419 PR rtl-optimization/84872
420 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
421 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
424 2018-04-06 Tamar Christina <tamar.christina@arm.com>
426 * expr.c (copy_blkmode_to_reg): Revert 254862.
427 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
429 2018-04-06 Richard Biener <rguenther@suse.de>
432 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
433 after seeing a component reference with an adjacent field. Treat
434 refs to arrays at struct end of external decls similar to
435 refs to unconstrained commons.
437 2018-04-06 Jakub Jelinek <jakub@redhat.com>
440 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
441 look through SAVE_EXPRs with non-side-effects argument. Adjust
443 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
446 2018-04-06 Richard Biener <rguenther@suse.de>
449 * alias.c (find_base_term): New wrapper around find_base_term
450 unwinding CSELIB_VAL_PTR changes.
451 (find_base_term): Do not restore CSELIB_VAL_PTR during the
454 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
456 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
458 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
459 constant definitions.
460 ("nop"): lr 0,0 -> nopr r0
461 ("nop_lr0", "nop_lr1"): New insn definitions.
463 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
465 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
466 NDS32_V3PUSH_AVAILABLE_P macro.
468 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
469 Chung-Ju Wu <jasonwucj@gmail.com>
471 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
472 (nds32*-*-*): Add float and fpu_config into supported_defaults.
473 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
474 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
475 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
476 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
477 * config/nds32/constraints.md: New constraints and checking for hard
479 * config/nds32/iterators.md: New mode iterator and attribute for hard
481 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
483 * config/nds32/nds32-fpu.md: New file.
484 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
485 deal with hard float code generation.
486 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
488 (abi_type, float_reg_number): New enum type.
489 * config/nds32/nds32-predicates.c: New predicates for hard float.
490 * config/nds32/nds32-protos.h: Declare functions for hard float.
491 * config/nds32/nds32.c: Implementation for hard float configuration.
492 * config/nds32/nds32.h: Definitions for hard float configuration.
493 * config/nds32/nds32.md: Include hard float machine description and
494 modify patterns for hard float configuration.
495 * config/nds32/nds32.opt: New options for hard float configuration.
496 * config/nds32/predicates.md: New predicates for hard float
499 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
501 * common/config/nds32/nds32-common.c
502 (nds32_option_optimization_table): Enable -mreleax-hint by default.
504 2018-04-05 Jakub Jelinek <jakub@redhat.com>
507 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
508 CONSTRUCTOR_ELT (ctor, ...)->value.
510 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
513 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
515 2018-04-05 Tom de Vries <tom@codesourcery.com>
518 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
521 2018-04-05 Shiva Chen <shiva0217@gmail.com>
522 Kito Cheng <kito.cheng@gmail.com>
524 * config/nds32/constraints.md (U33): Fine-tune checking condition.
525 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
526 * config/nds32/nds32.h (nds32_16bit_address_type): Add
527 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
529 2018-04-05 Shiva Chen <shiva0217@gmail.com>
530 Kito Cheng <kito.cheng@gmail.com>
532 * config/nds32/constraints.md (Ufe): New memory constraint.
533 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
534 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
535 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
537 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
538 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
540 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
542 * config/nds32/nds32.md: Use optimize_size in the condition for
543 alu-shift instructions.
545 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
547 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
549 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
551 * config/nds32/nds32.md (negsi2): Refine pattern.
553 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
554 Chung-Ju Wu <jasonwucj@gmail.com>
556 * config/nds32/iterators.md (shift_rotate): New code iterator.
557 (shift): New code attribute.
558 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
559 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
560 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
561 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
563 (andsi3, *andsi3): Ditto.
564 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
565 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
566 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
567 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
568 nds32_ior_operand, nds32_xor_operand): New predicates.
570 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
572 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
573 (addsi3, subsi3): ... this.
575 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
577 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
579 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
581 * config/nds32/nds32.md: Adjust indention.
583 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
585 * config/nds32/nds32.md (feature): New attribute.
587 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
589 * config/nds32/nds32.md (subtype): New attribute.
591 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
594 * config/arm/arm-builtins.c (arm_expand_builtin): Change
595 expansion to perform a bitwise AND of the argument followed by a
596 boolean negation of the result.
598 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
600 PR rtl-optimization/84878
601 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
602 the basic block. Assert the use reference is not artificial and that
603 it has an associated insn.
605 2018-04-04 Michael Matz <matz@suse.de>
607 * builtins.c (compute_objsize): Pass correct operand
608 to array_at_struct_end_p.
610 2018-04-04 Richard Biener <rguenther@suse.de>
613 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
614 from contexts for DINFO_LEVEL_TERSE and below.
616 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
618 * config/nds32/nds32-doubleword.md (move_<mode>): Require
619 resiter_operand condition.
620 * config/nds32/nds32.md (*move<mode>): Ditto.
622 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
623 Monk Chiang <sh.chiang04@gmail.com>
625 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
627 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
629 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
631 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
632 Kito Cheng <kito.cheng@gmail.com>
634 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
635 nds32_cond_code_str, output_cond_branch,
636 output_cond_branch_compare_zero, nds32_expand_cbranch,
637 nds32_expand_cstore, nds32_expand_movcc,
638 nds32_output_cbranchsi4_equality_zero,
639 nds32_output_cbranchsi4_equality_reg,
640 nds32_output_cbranchsi4_equality_reg_or_const_int,
641 nds32_output_cbranchsi4_greater_less_zero: New functions.
642 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
643 nds32_expand_cstore, nds32_expand_movcc,
644 nds32_output_cbranchsi4_equality_zero,
645 nds32_output_cbranchsi4_equality_reg,
646 nds32_output_cbranchsi4_equality_reg_or_const_int,
647 nds32_output_cbranchsi4_greater_less_zero): Declare.
648 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
649 nds32_rimm11s_operand): New predicates.
650 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
651 * config/nds32/nds32.md: Rewrite all the branch and conditional move
654 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
656 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
657 * config/nds32/nds32.md: Ditto.
658 * config/nds32/pipelines.md: Ditto.
660 2018-04-04 Richard Biener <rguenther@suse.de>
662 PR tree-optimization/85168
663 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
664 propagating abnormals.
666 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
668 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
670 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
671 Kito Cheng <kito.cheng@gmail.com>
673 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
674 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
675 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
676 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
677 * config/nds32/nds32.md (sibcall_internal): New.
678 (sibcall_register): Remove.
679 (sibcall_immediate): Remove.
680 (sibcall_value_internal): New.
681 (sibcall_value_register): Remove.
682 (sibcall_value_immediate): Remove.
683 * config/nds32/predicates.md (nds32_general_register_operand): New.
684 (nds32_call_address_operand): New.
686 2018-04-03 Jakub Jelinek <jakub@redhat.com>
688 PR rtl-optimization/85167
689 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
690 bb_defs if *split_p, instead preinitialize it to NULL.
692 PR tree-optimization/85156
693 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
694 evaluating the argument multiple times.
696 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
698 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
700 (_mm_cvtpd_ps): Likewise.
701 (_mm_cvttpd_epi32): Likewise.
702 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
703 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
704 vector, pixel, and bool following altivec.h include.
706 2018-04-03 Martin Sebor <msebor@redhat.com>
708 * doc/extend.texi (Common Function Attributes): Clarify.
709 (const attribute): Likewise.
710 (pure attribute): Likewise.
712 2018-04-03 Jakub Jelinek <jakub@redhat.com>
715 * config/i386/i386.c (ix86_expand_vector_set): Use
716 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
718 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
720 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
721 instructions when changing rounding bits to preserve precision bits
722 in the x87 control word.
724 2018-04-03 Martin Liska <mliska@suse.cz>
726 PR tree-optimization/82491
727 * rtl.h (strip_offset_and_add): Replace += suboffset with
728 poly_uint64 () + suboffset.
730 2018-03-29 Martin Liska <mliska@suse.cz>
731 Martin Jambor <mjambor@suse.cz>
734 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
735 param_type is not an integral or pointer type.
737 2018-04-03 Richard Biener <rguenther@suse.de>
739 * sese.h (recompute_all_dominators): Remove.
741 2018-04-02 Martin Sebor <msebor@redhat.com>
743 * doc/invoke.texi (-Wrestrict): Fix typos.
745 2018-04-02 Jim Wilson <jimw@sifive.com>
747 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
748 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
749 (<optab>di3, <optab>si3_extend): Likewise.
750 (<optab>si3_mask, <optab>si3_mask_1): New.
751 (<optab>di3_mask, <optab>di3_mask_1): New.
752 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
753 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
754 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
756 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
758 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
761 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
763 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
764 (nds32_canonicalize_comparison): New function.
766 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
767 Kito Cheng <kito.cheng@gmail.com>
768 Kuan-Lin Chen <kuanlinchentw@gmail.com>
770 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
771 * config/nds32/constants.md (unspec_volatile_element): Add
772 UNSPEC_VOLATILE_RELAX_GROUP.
773 * config/nds32/nds32-relax-opt.c: New file.
774 * config/nds32/nds32-predicates.c
775 (nds32_symbol_load_store_p): New function.
776 * config/nds32/nds32-protos.h
777 (nds32_symbol_load_store_p): Declare function.
778 (make_pass_nds32_relax_opt): Declare new rtl pass function.
779 * config/nds32/nds32.c
780 (nds32_register_pass): New function to register pass.
781 (nds32_register_passes): New function to register passes.
782 * config/nds32/nds32.md (relax_group): New pattern.
783 * config/nds32/nds32.opt (mrelax-hint): New option.
784 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
786 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
788 * config/nds32/t-nds32: Modify files dependency.
790 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
792 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
793 (PROFILE_HOOK): Define its implementation.
795 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
797 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
798 type and 32-bit size.
800 2018-04-01 Jakub Jelinek <jakub@redhat.com>
803 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
804 (V_128_256): New mode iterator.
805 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
806 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
807 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
809 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
810 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
812 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
815 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
816 NaN inputs correctly.
818 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
821 * config/rs6000/vsx.md (??r): New mode attribute.
822 (*vsx_mov<mode>_64bit): Use it.
823 (*vsx_mov<mode>_32bit): Likewise.
825 2018-03-30 Martin Sebor <msebor@redhat.com>
827 PR tree-optimization/84818
828 * builtins.c (check_access): Use warning_n.
830 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
833 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
835 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
838 2018-03-30 Julia Koval <julia.koval@intel.com>
841 * x86-tune.def (movx, partial_reg_dependency): Enable for
844 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
847 * lra-constraints.c (process_alt_operands): Move setting
848 this_alternative_matches below.
850 2018-03-29 Martin Liska <mliska@suse.cz>
853 * doc/invoke.texi: Document how LTO works with debug info.
854 Describe auto-load support of binutils. Mention 'x86-64'
855 as valid option value of -march option.
857 2018-03-29 Jakub Jelinek <jakub@redhat.com>
859 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
862 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
863 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
864 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
867 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
870 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
871 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
872 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
873 for RS6000_BTM_POWERPC64.
874 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
875 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
876 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
881 2018-03-28 Carl Love <cel@us.ibm.com>
884 2017-09-27 Carl Love <cel@us.ibm.com>
886 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
887 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
888 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
891 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
893 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
894 instead of __vector bool.
895 (_mm_max_pu8): Likewise.
896 (_mm_min_pi16): Likewise.
898 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
901 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
905 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
907 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
908 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
909 (div_extend): Likewise.
910 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
912 (__builtin_divweuo): Likewise.
913 (__builtin_divdeo): Likewise.
914 (__builtin_divdeuo): Likewise.
916 2018-03-28 Jakub Jelinek <jakub@redhat.com>
919 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
920 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
922 PR tree-optimization/82004
923 * gimple-match-head.c (optimize_pow_to_exp): New function.
924 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
925 Don't fold to exp if optimize_pow_to_exp is false.
927 2018-03-28 Martin Liska <mliska@suse.cz>
930 * calls.c (initialize_argument_information): Fix trailing space.
931 * common.opt: Fix typo and provide better explanation for
932 -fsanitize-coverage option.
933 * config/i386/i386.opt: Fix typo.
935 2018-03-28 Jakub Jelinek <jakub@redhat.com>
936 Martin Liska <mliska@suse.cz>
939 * gimplify.c (asan_poison_variable): Don't do the check for
940 gimplify_omp_ctxp here.
941 (gimplify_decl_expr): Do it here.
942 (gimplify_target_expr): Likewise.
944 2018-03-28 Martin Liska <mliska@suse.cz>
947 * config/i386/i386.c (ix86_function_arg_advance): Do not call
948 chkp_type_bounds_count if MPX is not enabled.
950 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
952 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
954 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
957 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
958 function to create the function decl for complex long double
959 multiply and divide for -mabi=ieeelongdouble.
960 (init_float128_ieee): Call it.
962 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
965 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
966 -fcf-protection=branch -mibt.
967 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
969 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
972 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
974 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
977 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
978 extern array declarations.
980 2018-03-27 Richard Biener <rguenther@suse.de>
983 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
984 explicit single_use checks.
986 2018-03-27 Richard Biener <rguenther@suse.de>
988 PR tree-optimization/85082
989 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
992 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
994 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
995 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
996 Turn on fasynchronous-unwind-tables and funwind-tables.
998 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
1001 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
1002 (*bmi_blsr_<mode>_ccz): Ditto.
1004 2018-03-26 Tom de Vries <tom@codesourcery.com>
1006 PR tree-optimization/85063
1007 * omp-general.c (offloading_function_p): New function. Factor out
1009 * omp-offload.c (pass_omp_target_link::gate): ... here.
1010 * omp-general.h (offloading_function_p): Declare.
1011 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
1012 with attribute omp declare target for offloading functions.
1014 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
1016 PR tree-optimization/84005
1017 * tree-data-ref.h (get_base_for_alignment): Declare.
1018 * tree-data-ref.c (get_base_for_alignment_1): New function.
1019 (get_base_for_alignment): Likewise.
1020 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
1021 get_base_for_alignment to find a suitable base object, instead
1022 of always using drb->base_address.
1024 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1027 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
1028 known size by default.
1030 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
1033 * lra-constraints.c (process_alt_operands): Don't match BLKmode
1034 and non BLKmode operands.
1036 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1039 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
1040 Clean up attributes.
1042 2018-03-23 Richard Biener <rguenther@suse.de>
1045 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
1046 we are going to emit early debug for LTO.
1048 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1051 * function.c (match_asm_constraints_1): Don't optimize if input
1052 doesn't satisfy general_operand predicate for output's mode.
1055 * alias.c (write_dependence_p): Don't require for x_canonicalized
1056 non-VOIDmode if x has VOIDmode.
1059 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
1060 just don't try to optimize it rather than assert it never happens.
1062 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1064 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
1065 macro expansions for definition of ST_INTERNAL_<mode> and
1066 LD_INTERNAL_<mode> builtins.
1067 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
1069 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
1071 (altivec_expand_st_builtin): Likewise.
1072 (altivec_expand_builtin): Remove calls to deleted functions.
1073 (rs6000_address_for_altivec): Delete this function.
1074 * config/rs6000/vector.md: Remove expands for
1075 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
1077 2018-03-22 Sudakshina Das <sudi.das@arm.com>
1080 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
1081 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
1082 re-computing once computed.
1083 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
1084 (arm_init_machine_status): Initialize
1085 machine->static_chain_stack_bytes.
1087 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1090 * doc/extend.texi: Add four new prototypes for vec_ld.
1091 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
1092 definitions for more logical presentation.
1093 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
1094 entries for V1TI variants of __builtin_altivec_ld builtin.
1095 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
1096 handling of V1TI variant of LVX icode pattern.
1097 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
1098 (rs6000_gimple_fold_builtin): Likewise.
1099 (altivec_init_builtins): Add code to define
1100 __builtin_altivec_lvx_v1ti function.
1102 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1105 * function.c (match_asm_constraints_1): Don't do the optimization
1106 if input isn't a REG, SUBREG, MEM or constant.
1108 2018-03-22 Tom de Vries <tom@codesourcery.com>
1110 PR tree-optimization/84956
1111 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
1112 bb_has_abnormal_pred.
1114 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1117 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
1118 DECL_INITIAL (decl) to decl at the end.
1119 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
1122 2018-03-21 Joseph Myers <joseph@codesourcery.com>
1124 * doc/extend.texi (__builtin_tgmath): Document when complex
1125 integer types are treated as _Complex _Float64.
1127 2018-03-21 Tom de Vries <tom@codesourcery.com>
1129 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
1131 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1133 PR tree-optimization/84960
1134 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
1135 if it is ENTRY block, move them into single succ of ENTRY in that case.
1137 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
1139 PR tree-optimization/84811
1140 * poly-int.h (poly_span_traits): Remove the T3 parameter and
1141 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
1142 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
1143 (known_subrange_p): Update accordingly. Cast each value involved
1144 in the size comparison, rather than casting the result of the
1147 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1149 PR tree-optimization/84982
1150 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
1151 by flipping the least significant bit rather than all bits from
1152 bitpos to bitpos + bitsize - 1.
1154 2018-03-21 Nathan Sidwell <nathan@acm.org>
1156 * doc/extend.texi (Deprecated Features): Remove mention of
1157 long-deleted deprecations.
1159 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1162 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
1163 * configure: Regenerate.
1165 2018-03-21 Tom de Vries <tom@codesourcery.com>
1167 PR tree-optimization/83126
1168 * tree-parloops.c (num_phis): New function.
1169 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
1171 2018-03-21 Nathan Sidwell <nathan@acm.org>
1173 * doc/extend.texi (Deprecated Features): Update deprecated flags,
1174 mention anon-struct/union members and trailing attributes.
1176 2018-03-21 Bin Cheng <bin.cheng@arm.com>
1178 PR tree-optimization/84969
1179 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
1180 builtin memset partitions if they set different rhs values.
1182 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1184 PR rtl-optimization/84989
1185 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
1186 VEC_DUPLICATE with scalar result mode.
1188 2018-03-21 Martin Liska <mliska@suse.cz>
1191 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
1192 not intended return statement.
1194 2018-03-21 Martin Liska <mliska@suse.cz>
1197 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
1198 (chkp_find_bound_slots_1): Limit number of iterations.
1200 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
1203 * Minor grammar fixes for x86 options.
1205 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1208 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
1209 holding REG_CFA_RESTORE notes, instead turn them into a USE.
1211 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
1214 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
1215 (altivec_lvx_<mode>_1op): Likewise.
1216 (altivec_stvx_<mode>_2op): Likewise.
1217 (altivec_stvx_<mode>_1op): Likewise.
1218 (altivec_lvx_<VM2:mode>): New define_expand.
1219 (altivec_stvx_<VM2:mode>): Likewise.
1220 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
1221 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1222 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
1223 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1224 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
1225 (rs6000_gen_lvx): Likewise.
1226 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
1227 (altivec_expand_stv_builtin): Likewise.
1228 (altivec_expand_builtin): Likewise.
1229 * config/rs6000/vector.md: Likewise.
1231 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1234 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
1237 2018-03-20 Richard Biener <rguenther@suse.de>
1240 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
1241 sign-conversions as zero, fall back to standard scalar_stmt
1244 2018-03-20 Martin Liska <mliska@suse.cz>
1247 * predict.c (rebuild_frequencies): Handle case when we have
1248 PROFILE_ABSENT, but flag_guess_branch_prob is false.
1250 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1253 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
1254 flag_section_anchors.
1255 * varasm.c (use_blocks_for_decl_p): Remove hack for
1256 dw2_force_const_mem.
1259 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
1261 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
1262 be created, use lowpart_subreg of operands[0] rather than operands[0]
1264 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
1265 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
1266 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
1267 and n constraint instead of aarch64_shift_imm_di and Usd.
1268 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
1269 (*aarch64_<optab>_reg_minus<mode>3): ... this.
1271 2018-03-20 Sudakshina Das <sudi.das@arm.com>
1274 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
1275 to favor GPR over NEON registers.
1276 (<shift>di3_neon): Likewise.
1278 2018-03-20 Tom de Vries <tom@codesourcery.com>
1281 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
1282 (nvptx_process_pars): Emit bar.sync asap and alap.
1284 2018-03-20 Tom de Vries <tom@codesourcery.com>
1287 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
1288 seen_label if seen_label is already set.
1290 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1293 * config/i386/i386.c (fold_builtin_cpu): For features above 31
1294 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
1295 Use 1U instead of 1. Formatting fixes.
1298 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
1299 instead of TREE_TYPE (s1) for the return value.
1301 2018-03-19 Jakub Jelinek <jakub@redhat.com>
1303 PR tree-optimization/84946
1304 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
1305 bitsize + bitsize in poly_uint64 rather than poly_int64.
1308 * dwarf2asm.c: Include fold-const.c.
1309 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
1310 of decl rather than decl itself.
1312 PR rtl-optimization/84643
1313 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
1315 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
1318 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
1319 calling assemble_variable.
1321 2018-03-19 Sudakshina Das <sudi.das@arm.com>
1324 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
1325 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
1327 2018-03-19 Jim Wilson <jimw@sifive.com>
1330 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
1331 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
1332 (riscv_first_stack_step): Likewise.
1333 (riscv_option_override): Use STACK_BOUNDARY instead of
1335 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
1337 (BIGGEST_ALIGNMENT): Set to 128.
1338 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
1339 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
1342 2018-03-19 Richard Biener <rguenther@suse.de>
1344 PR tree-optimization/84933
1345 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
1346 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
1348 2018-03-19 Richard Biener <rguenther@suse.de>
1350 PR tree-optimization/84859
1351 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
1352 (cond_if_else_store_replacement): Perform sinking operation on
1353 single-store BBs regardless of MAX_STORES_TO_SINK setting.
1354 Generalize what a BB with a single eligible store is.
1356 2018-03-19 Richard Biener <rguenther@suse.de>
1358 PR tree-optimization/84929
1359 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
1360 chrec_is_positive against non-chrec arg.
1362 2018-03-19 Tamar Christina <tamar.christina@arm.com>
1365 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
1367 2018-03-18 Martin Liska <mliska@suse.cz>
1369 PR rtl-optimization/84635
1370 * regrename.c (build_def_use): Use matches_mode only when
1373 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
1375 PR tree-optimization/84913
1376 * tree-vect-loop.c (vectorizable_reduction): Don't try to
1377 vectorize chains of COND_EXPRs.
1379 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1381 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
1383 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1385 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
1387 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1389 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
1391 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1392 Kito Cheng <kito.cheng@gmail.com>
1394 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
1395 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
1396 (nds32_adjust_reg_alloc_order): New function.
1397 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
1399 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
1401 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
1402 nds32_print_operand, nds32_print_operand_address): Use
1403 HOST_WIDE_INT_PRINT_DEC instead.
1405 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1407 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
1409 2018-03-17 Jakub Jelinek <jakub@redhat.com>
1412 * config/i386/i386.c (initial_ix86_tune_features,
1413 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
1415 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
1416 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
1417 rather than 1u << ix86_tune. Formatting fix.
1418 (ix86_option_override_internal): Change ix86_arch_mask from
1419 unsigned int to unsigned HOST_WIDE_INT, initialize to
1420 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
1421 (ix86_function_specific_restore): Likewise.
1423 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1426 * postreload.c (reload_combine_recognize_pattern): Perform
1427 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
1428 truncate_int_for_mode the result for the destination's mode.
1431 * hsa-gen.c (mem_type_for_type): Fix comment typo.
1432 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
1434 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1437 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
1440 * lra-assigns.c (lra_split_hard_reg_for): Don't use
1441 regno_allocno_class_array and sorted_pseudos.
1442 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
1443 insns where regno is used.
1445 2018-03-16 Martin Liska <mliska@suse.cz>
1448 * multiple_target.c (create_dispatcher_calls): Redirect
1449 reference in the symbol table.
1451 2018-03-16 Martin Liska <mliska@suse.cz>
1454 * multiple_target.c (create_dispatcher_calls): Redirect also
1457 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1461 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
1462 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
1463 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
1465 2018-03-16 Julia Koval <julia.koval@intel.com>
1467 * doc/invoke.texi (Skylake Server): Add CLWB.
1468 Cannonlake): Remove CLWB.
1470 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1472 PR tree-optimization/84841
1473 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
1475 (FLOAT_ONE_CONST_TYPE): Define.
1476 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
1477 (sort_by_operand_rank): Put entries with higher constant_type last
1478 rather than first to match comments.
1480 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
1482 * config/nios2/nios2.md (movsi_internal): Fix thinko in
1485 2018-03-15 Jakub Jelinek <jakub@redhat.com>
1488 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
1489 check and use address of target always.
1491 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
1494 * config/i386/i386.c (indirect_thunk_needed): Update comments.
1495 (indirect_thunk_bnd_needed): Likewise.
1496 (indirect_thunks_used): Likewise.
1497 (indirect_thunks_bnd_used): Likewise.
1498 (indirect_return_needed): New.
1499 (indirect_return_bnd_needed): Likewise.
1500 (output_indirect_thunk_function): Add a bool argument for
1502 (output_indirect_thunk_function): Don't generate alias for
1503 function return thunk.
1504 (ix86_code_end): Call output_indirect_thunk_function to generate
1505 function return thunks.
1506 (ix86_output_function_return): Set indirect_return_bnd_needed
1507 and indirect_return_needed instead of indirect_thunk_bnd_needed
1508 and indirect_thunk_needed.
1510 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
1512 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
1513 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
1514 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
1516 2018-03-15 David Malcolm <dmalcolm@redhat.com>
1517 Paul Hua <paul.hua.gm@gmail.com>
1520 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
1522 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
1524 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
1525 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
1528 2018-03-15 Tamar Christina <tamar.christina@arm.com>
1531 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
1532 instead of GET_MODE_SIZE when comparing Units.
1534 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1537 * varasm.c (hash_section): Return an unchangeble hash value
1538 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
1539 Return !aarch64_can_use_per_function_literal_pools_p ().
1541 2018-03-15 Jakub Jelinek <jakub@redhat.com>
1544 * optabs.c (emit_conditional_move): Pass address of cmode's copy
1545 rather than address of cmode as last argument to prepare_cmp_insn.
1547 2018-03-15 Julia Koval <julia.koval@intel.com>
1549 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
1550 F_AVX512VNNI, F_AVX512BITALG): New.
1552 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
1555 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
1556 insn for floating-point loads and stores.
1558 2018-03-14 Carl Love <cel@us.ibm.com>
1560 * config/rs6000/rs6000-c.c: Add macro definitions for
1561 ALTIVEC_BUILTIN_VEC_PERMXOR.
1562 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
1563 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
1564 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
1565 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
1567 * config/doc/extend.texi: Add prototypes for vec_permxor.
1569 2018-03-14 David Malcolm <dmalcolm@redhat.com>
1572 * diagnostic-show-locus.c (class layout_point): Convert m_line
1573 from int to linenum_type.
1574 (line_span::comparator): Use linenum "compare" function when
1575 comparing line numbers.
1576 (test_line_span): New function.
1577 (layout_range::contains_point): Convert param "row" from int to
1579 (layout_range::intersects_line_p): Likewise.
1580 (layout::will_show_line_p): Likewise.
1581 (layout::print_source_line): Likewise.
1582 (layout::should_print_annotation_line_p): Likewise.
1583 (layout::print_annotation_line): Likewise.
1584 (layout::print_leading_fixits): Likewise.
1585 (layout::annotation_line_showed_range_p): Likewise.
1586 (struct line_corrections): Likewise for field m_row.
1587 (line_corrections::line_corrections): Likewise for param "row".
1588 (layout::print_trailing_fixits): Likewise.
1589 (layout::get_state_at_point): Likewise.
1590 (layout::get_x_bound_for_row): Likewise.
1591 (layout::print_line): Likewise.
1592 (diagnostic_show_locus): Likewise for locals "last_line" and
1594 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
1595 * input.c (selftest::test_linenum_comparisons): New function.
1596 (selftest::input_c_tests): Call it.
1597 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
1598 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
1599 * selftest.h (ASSERT_GT): New macro.
1600 (ASSERT_GT_AT): New macro.
1601 (ASSERT_LT): New macro.
1602 (ASSERT_LT_AT): New macro.
1604 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
1606 PR rtl-optimization/84780
1607 * combine.c (distribute_links): Don't make a link based on pc_rtx.
1609 2018-03-14 Martin Liska <mliska@suse.cz>
1611 * tree.c (record_node_allocation_statistics): Use
1612 get_stats_node_kind.
1613 (get_stats_node_kind): New function extracted from
1614 record_node_allocation_statistics.
1615 (free_node): Use get_stats_node_kind.
1617 2018-03-14 Richard Biener <rguenther@suse.de>
1619 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
1620 that the value-set of ANTIC_IN doesn't grow.
1623 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1625 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1626 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1628 2018-03-14 Julia Koval <julia.koval@intel.com>
1630 * config.gcc (icelake-client, icelake-server): New.
1632 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
1633 (initial_ix86_arch_features): Ditto.
1634 (PTA_SKYLAKE): Add SGX.
1635 (PTA_ICELAKE): Remove.
1636 (PTA_ICELAKE_CLIENT): New.
1637 (PTA_ICELAKE_SERVER): New.
1638 (ix86_option_override_internal): Split up icelake on icelake client and
1640 (get_builtin_code_for_version): Ditto.
1641 (fold_builtin_cpu): Ditto.
1642 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
1643 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
1644 * config/i386/i386.h (processor_type): Ditto.
1645 * doc/invoke.texi: Ditto.
1647 2018-03-14 Jakub Jelinek <jakub@redhat.com>
1650 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
1651 INTEGER_CST offset, add it together with bitpos / 8 and
1652 sign extend based on POINTER_SIZE.
1656 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
1659 * config/i386/constraints.md (Yc): New register constraint.
1660 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
1661 Use Yc constraint for alternative 2 of operand 0. Remove
1662 preferred_for_speed attribute.
1664 2018-03-14 Richard Biener <rguenther@suse.de>
1666 PR tree-optimization/84830
1667 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
1668 with the old one to avoid oscillations.
1670 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
1673 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
1675 (assign_by_spills): Return a flag of reload assignment failure.
1676 Do not process the reload assignment failures. Do not spill other
1677 reload pseudos if they has the same reg class. Update n if
1679 (lra_assign): Add a return arg. Set up from the result of
1680 assign_by_spills call.
1681 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1682 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1683 usage_insns if it is not NULL.
1684 (spill_hard_reg_in_range): New function.
1685 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1686 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1687 function prototypes.
1688 (lra_assign): Change prototype.
1689 * lra.c (lra): Add code to deal with fails by splitting hard reg
1692 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
1694 * config/riscv/riscv.opt (mrelax): New option.
1695 * config/riscv/riscv.c (riscv_file_start): Emit ".option
1696 "norelax" when riscv_mrelax is disabled.
1697 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
1699 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1702 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
1703 reassociation for int modes.
1705 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1707 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
1708 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
1710 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
1711 * config/aarch64/aarch64-sve.md
1712 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
1713 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
1714 (*extend<mode><Vwide>2): Rename to...
1715 (aarch64_sve_extend<mode><Vwide>2): ...this.
1716 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
1717 renaming the old pattern to...
1718 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
1720 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
1721 define_expand, renaming the old pattern to...
1722 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
1723 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
1724 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
1725 account when deciding which SVE instruction the optab should use.
1726 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
1728 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1730 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
1731 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
1732 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
1733 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
1734 (tlsdesc_small_<mode>): Turn a define_expand and use
1735 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
1736 (tlsdesc_small_advsimd_<mode>): ...this.
1737 (tlsdesc_small_sve_<mode>): New pattern.
1739 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1741 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
1742 (UNSPEC_UMUL_HIGHPART): New constants.
1743 (MUL_HIGHPART): New int iteraor.
1744 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
1745 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
1747 (*<su>mul<mode>3_highpart): New define_insn.
1749 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1752 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1755 2018-03-13 Martin Liska <mliska@suse.cz>
1758 * (sem_item_optimizer::sem_item_optimizer): Initialize new
1760 (sem_item_optimizer::~sem_item_optimizer): Release it.
1761 (sem_item_optimizer::merge_classes): Register variable aliases.
1762 (sem_item_optimizer::fixup_pt_set): New function.
1763 (sem_item_optimizer::fixup_points_to_sets): Likewise.
1764 * ipa-icf.h: Declare new variables and functions.
1766 2018-03-13 Jakub Jelinek <jakub@redhat.com>
1769 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
1770 integer_pow2p@2 and test integer_pow2p in condition.
1771 (A < 0 ? C : 0): Similarly for @1.
1774 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
1775 characters starting at p contain '\0' character, don't look beyond
1779 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
1780 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
1783 * reg-stack.c (change_stack): Change update_end var from int to
1784 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
1785 also call set_block_for_insn on the newly added insns and rescan.
1788 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
1789 on the last operand.
1792 * tree.c (stabilize_reference_1): Return save_expr (e) for
1793 STATEMENT_LIST even if it doesn't have side-effects.
1795 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
1797 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
1799 2018-03-12 Renlin Li <renlin.li@arm.com>
1801 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
1802 aarch64_output_scalar_simd_mov_immediate.
1804 2018-03-12 Martin Sebor <msebor@redhat.com>
1806 PR tree-optimization/83456
1807 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
1808 for perfectly overlapping calls to memcpy.
1809 (gimple_fold_builtin_memory_chk): Same.
1810 (gimple_fold_builtin_strcpy): Handle no-warning.
1811 (gimple_fold_builtin_stxcpy_chk): Same.
1812 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
1814 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
1816 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
1817 parameter. Use it for SFmode.
1818 (rs6000_function_arg_advance_1): Adjust.
1819 (rs6000_function_arg): Adjust.
1820 (rs6000_gimplify_va_arg): Pass false for that new parameter.
1822 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
1824 PR rtl-optimization/84169
1825 PR rtl-optimization/84780
1826 * combine.c (can_combine_p): Check for a 2-insn combination whether
1827 the destination register is used between the two insns, too.
1829 2018-03-12 Richard Biener <rguenther@suse.de>
1831 PR tree-optimization/84803
1832 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
1833 for refs DR analysis didn't process.
1835 2018-03-12 Richard Biener <rguenther@suse.de>
1837 PR tree-optimization/84777
1838 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
1839 force-vectorize loops ignore whether we are optimizing for size.
1841 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
1843 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
1844 (TARGET_MD_ASM_ADJUST): Define.
1846 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
1847 Kito Cheng <kito.cheng@gmail.com>
1848 Chung-Ju Wu <jasonwucj@gmail.com>
1850 * config/nds32/nds32.c (nds32_compute_stack_frame,
1851 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
1852 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
1853 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
1854 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
1855 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
1856 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
1857 * config/nds32/nds32.md (prologue, epilogue): Use macro
1858 NDS32_V3PUSH_AVAILABLE_P to do checking.
1860 2018-03-11 Jakub Jelinek <jakub@redhat.com>
1863 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
1864 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
1865 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
1866 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
1867 addition of most attributes on !orig_type_die or the attribute not
1868 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
1870 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1871 Chung-Ju Wu <jasonwucj@gmail.com>
1873 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
1875 * config/nds32/nds32.opt (mvh): New option.
1877 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1878 Chung-Ju Wu <jasonwucj@gmail.com>
1880 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
1882 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
1883 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
1886 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1887 Chung-Ju Wu <jasonwucj@gmail.com>
1889 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
1891 * config/nds32/nds32-multiple.md (strlensi): New pattern.
1892 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
1894 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
1895 Kito Cheng <kito.cheng@gmail.com>
1896 Chung-Ju Wu <jasonwucj@gmail.com>
1898 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
1899 UNSPEC_FFMISM and UNSPEC_FLMISM.
1900 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
1901 for ffb, ffmism and flmism.
1902 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
1903 (unspec_ffmism): Ditto.
1904 (unspec_flmism): Ditto.
1905 (nds32_expand_builtin_impl): Check if string extension is available.
1906 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
1907 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
1909 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
1912 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1915 * lra-assigns.c (assign_by_spills): Return a flag of reload
1916 assignment failure. Do not process the reload assignment
1917 failures. Do not spill other reload pseudos if they has the same
1919 (lra_assign): Add a return arg. Set up from the result of
1920 assign_by_spills call.
1921 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1922 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1923 usage_insns if it is not NULL.
1924 (spill_hard_reg_in_range): New function.
1925 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1926 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1927 function prototypes.
1928 (lra_assign): Change prototype.
1929 * lra.c (lra): Add code to deal with fails by splitting hard reg
1932 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
1935 * config/i386/i386.opt: Replace Enforcment with Enforcement.
1937 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
1940 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
1941 (dw_val_node): Add val_symbolic_view.
1942 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
1943 (symview_upper_bound): New.
1944 (new_line_info_table): Initialize symviews_since_reset.
1945 (dwarf2out_source_line): Count symviews_since_reset and set
1946 symview_upper_bound.
1947 (dw_val_equal_p): Handle symview.
1948 (add_AT_symview): New.
1949 (print_dw_val): Handle symview.
1950 (attr_checksum, attr_checksum_ordered): Likewise.
1951 (same_dw_val_p, size_of_die): Likewise.
1952 (value_format, output_die): Likewise.
1953 (add_high_low_attributes): Use add_AT_symview for entry_view.
1954 (dwarf2out_finish): Reset symview_upper_bound, clear
1957 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
1960 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
1961 Add strict argument and use it.
1962 (rs6000_split_multireg_move): Update for new strict argument.
1963 (mem_operand_gpr): Disallow all non-offsettable addresses.
1964 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
1966 2018-03-09 Jakub Jelinek <jakub@redhat.com>
1969 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
1970 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
1971 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
1974 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
1975 decl, use remap_type if we want to use the type.
1977 2018-03-09 Martin Sebor <msebor@redhat.com>
1979 PR tree-optimization/84526
1980 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1982 (builtin_access::generic_overlap): Be prepared to handle non-array
1985 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
1987 PR rtl-optimization/84682
1988 * lra-constraints.c (process_address_1): Check is_address flag
1989 for address constraints.
1990 (process_alt_operands): Likewise.
1991 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
1992 preprocess_constraints.
1993 * recog.h (preprocess_constraints): Add oploc parameter.
1995 * recog.c (preprocess_constraints): Test address_operand for
1996 CT_ADDRESS constraints.
1998 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2001 * lra-assigns.c (assign_by_spills): Return a flag of reload
2002 assignment failure. Do not process the reload assignment
2003 failures. Do not spill other reload pseudos if they has the same
2005 (lra_assign): Add a return arg. Set up from the result of
2006 assign_by_spills call.
2007 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2008 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2009 usage_insns if it is not NULL.
2010 (spill_hard_reg_in_range): New function.
2011 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2012 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2013 function prototypes.
2014 (lra_assign): Change prototype.
2015 * lra.c (lra): Add code to deal with fails by splitting hard reg
2018 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2021 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
2022 Accept complain bool parameter. Only emit errors if it is true.
2023 (arm_parse_cpu_option_name): Likewise.
2024 (arm_target_thumb_only): Adjust callers of the above.
2025 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
2026 prototype to take a default true bool parameter.
2027 (arm_parse_arch_option_name): Likewise.
2029 2018-03-09 David Malcolm <dmalcolm@redhat.com>
2030 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
2034 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
2035 * configure: Regenerate.
2036 * configure.ac ("linker --version-script option"): New.
2037 ("linker soname option"): New.
2039 2018-03-09 Richard Biener <rguenther@suse.de>
2041 PR tree-optimization/84775
2042 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
2043 immediate uses of predicate stmts and mark them modified.
2046 PR tree-optimization/84178
2047 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2049 (version_loop_for_if_conversion): Delay update_ssa call.
2050 (tree_if_conversion): Delay update_ssa until after predicate
2053 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
2056 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
2057 when the function accesses prior frames.
2059 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2062 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
2063 gen_llsym, otherwise call maybe_gen_llsym.
2066 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
2067 has ',' character inside of it.
2069 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2072 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
2073 as clobbering CC_REGNUM.
2075 2018-03-08 Richard Biener <rguenther@suse.de>
2078 * tree-scalar-evolution.c: Include tree-into-ssa.h.
2079 (follow_copies_to_constant): Do not follow SSA names registered
2082 2018-03-08 Richard Biener <rguenther@suse.de>
2084 PR tree-optimization/84178
2085 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2087 (version_loop_for_if_conversion): Delay update_ssa call.
2088 (tree_if_conversion): Delay update_ssa until after predicate
2091 2018-03-08 David Malcolm <dmalcolm@redhat.com>
2093 PR tree-optimization/84178
2094 * tree-if-conv.c (release_bb_predicate): Remove the
2095 the assertion that the stmts have NULL use_ops.
2096 Discard the statements, asserting that they haven't
2097 yet been added to a BB.
2099 2018-03-08 Richard Biener <rguenther@suse.de>
2101 PR tree-optimization/84746
2102 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
2103 (phi_translate): Pass in destination ANTIC_OUT set.
2104 (phi_translate_1): Likewise. For a simplified result lookup
2105 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
2106 (phi_translate_set): Adjust.
2107 (do_pre_regular_insertion): Likewise.
2108 (do_pre_partial_partial_insertion): Likewise.
2110 2018-03-08 Martin Liska <mliska@suse.cz>
2112 PR gcov-profile/84735
2113 * doc/gcov.texi: Document usage of profile files.
2114 * gcov-io.h: Document changes in the format.
2116 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
2120 * dwarf2out.c (struct dw_line_info_table): Update comments for
2122 (FORCE_RESET_NEXT_VIEW): New.
2123 (FORCE_RESETTING_VIEW_P): New.
2124 (RESETTING_VIEW_P): Check for -1 too.
2125 (ZERO_VIEW_P): Likewise.
2126 (new_line_info_table): Force-reset next view.
2127 (dwarf2out_begin_function): Likewise.
2128 (dwarf2out_source_line): Simplify zero_view_p initialization.
2129 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
2130 view directly. Omit view when omitting .loc at line 0.
2132 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2134 PR tree-optimization/84740
2135 * tree-switch-conversion.c (process_switch): Call build_constructors
2136 only if info.phi_count is non-zero.
2138 PR tree-optimization/84739
2139 * tree-tailcall.c (find_tail_calls): Check call arguments against
2140 DECL_ARGUMENTS (current_function_decl) rather than
2141 DECL_ARGUMENTS (func) when checking for tail recursion.
2143 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2145 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
2146 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
2147 Volker Reichelt's entry and add entries for people that perform
2148 GCC fuzzy testing and report numerous bugs.
2150 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
2153 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
2154 readonly data in sdata, if that is disabled.
2155 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
2156 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
2157 -mreadonly-in-sdata option.
2159 2018-03-07 Martin Sebor <msebor@redhat.com>
2161 PR tree-optimization/84468
2162 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
2163 basic block when looking for nul assignment.
2165 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
2168 * except.h (output_function_exception_table): Adjust prototype.
2169 * except.c (output_function_exception_table): Remove FNNAME parameter
2170 and add SECTION parameter. Ouput one part of the table at a time.
2171 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
2172 the first part of the exception table and emit unwind directives.
2173 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
2174 (i386_pe_seh_cold_init): Likewise.
2175 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
2176 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
2177 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
2178 (ix86_output_call_insn): Emit a nop in one more case for SEH.
2179 * config/i386/winnt.c: Include except.h.
2180 (struct seh_frame_state): Add reg_offset, after_prologue and
2181 in_cold_section fields.
2182 (i386_pe_seh_end_prologue): Set seh->after_prologue.
2183 (i386_pe_seh_cold_init): New function.
2184 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
2185 to seh->in_cold_section.
2186 (seh_emit_push): Record the offset of the push.
2187 (seh_emit_save): Record the offet of the save.
2188 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
2189 Test seh->after_prologue to disregard the epilogue.
2190 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
2191 (i386_pe_end_cold_function): New function.
2193 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2196 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
2197 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
2200 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
2202 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
2203 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
2206 * multiple_target.c: Include tree-inline.h and intl.h.
2207 (expand_target_clones): Diagnose and fail if node->definition and
2208 !tree_versionable_function_p (node->decl).
2210 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
2212 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
2214 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
2215 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
2216 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
2218 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2221 * combine.c (try_combine): Use reg_or_subregno instead of handling
2222 just paradoxical SUBREGs and REGs.
2224 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
2226 * config/arc/arc.c (arc_finalize_pic): Remove function.
2227 (arc_must_save_register): We use single base PIC register, remove
2228 checks to save/restore the PIC register.
2229 (arc_expand_prologue): Likewise.
2230 * config/arc/arc-protos.h (arc_set_default_type_attributes):
2232 (arc_verify_short): Likewise.
2233 (arc_attr_type): Likewise.
2234 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
2235 (walk_stores): Likewise.
2236 (arc_address_cost): Make it static.
2237 (arc_verify_short): Likewise.
2238 (branch_dest): Likewise.
2239 (arc_attr_type): Likewise.
2240 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
2241 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
2242 (arc_final_prescan_insn): Remove inserting the nops due to
2243 hardware hazards. It is done in reorg step.
2244 (insn_length_variant_t): Remove.
2245 (insn_length_parameters_t): Likewise.
2246 (arc_insn_length_parameters): Likewise.
2247 (arc_get_insn_variants): Likewise.
2248 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
2250 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2253 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
2256 PR tree-optimization/84687
2257 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
2259 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
2261 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2263 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
2264 Rename to ppc_speculation_barrier.
2265 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
2266 __builtin_ppc_speculation_barrier.
2268 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2271 * combine.c (combine_simplify_rtx): Don't try to simplify if
2272 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
2275 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
2277 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
2278 to 32 bytes when compiling for POWER9.
2280 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2283 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
2284 regparm >= 3 with no arg reg available also for calls with
2285 flag_force_indirect_call. Pass decl to ix86_function_regparm.
2288 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
2290 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
2292 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
2295 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
2297 2018-03-05 Richard Biener <rguenther@suse.de>
2299 PR tree-optimization/84486
2300 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
2301 When inserting a __builtin_assume_aligned call set the LHS
2302 SSA name alignment info accordingly.
2304 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
2306 PR tree-optimization/84114
2307 * config/aarch64/aarch64.c (aarch64_reassociation_width)
2308 Avoid reassociation of FLOAT_MODE addition.
2310 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
2312 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
2313 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
2314 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
2315 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
2316 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
2317 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
2318 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
2320 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
2321 __builtin_ia32_wbinvd): New builtins.
2322 (SPECIAL_ARGS2): New.
2323 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
2324 (SPECIAL_ARGS2): New.
2325 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
2326 (ix86_valid_target_attribute_inner_p): Ditto.
2327 (ix86_init_mmx_sse_builtins): Add special_args2.
2328 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
2329 TARGET_WBNOINVD_P): New.
2330 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
2331 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
2332 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
2333 * config/i386/immintrin.h (_wbinvd): New intrinsic.
2334 * config/i386/pconfigintrin.h: New file.
2335 * config/i386/wbnoinvdintrin.h: Ditto.
2336 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
2337 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
2339 2018-03-05 Richard Biener <rguenther@suse.de>
2341 PR tree-optimization/84670
2342 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2344 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2345 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2346 (compute_antic_aux): Only assert the number of values in ANTIC_IN
2347 doesn't grow if all successors (recursively) were visited at least
2350 2018-03-05 Richard Biener <rguenther@suse.de>
2352 PR tree-optimization/84650
2353 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
2354 if executed in the loop pipeline.
2356 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
2358 * doc/configfiles.texi (Configuration Files): Move info about
2359 conditionalizing $target-protos.h to...
2360 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
2361 differs from $target-protos.h.
2363 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
2364 Chung-Ju Wu <jasonwucj@gmail.com>
2366 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
2367 * config/nds32/nds32-multiple.md (setmemsi): Define.
2368 * config/nds32/nds32-memory-manipulation.c
2369 (nds32_gen_dup_4_byte_to_word_value): New.
2370 (emit_setmem_word_loop): New.
2371 (emit_setmem_byte_loop): New.
2372 (nds32_expand_setmem_loop): New.
2373 (nds32_expand_setmem_loop_v3m): New.
2374 (nds32_expand_setmem_unroll): New.
2375 (nds32_expand_setmem): New.
2377 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2378 Chung-Ju Wu <jasonwucj@gmail.com>
2380 * config/nds32/nds32-memory-manipulation.c
2381 (nds32_emit_load_store): New.
2382 (nds32_emit_post_inc_load_store): New.
2383 (nds32_emit_mem_move): New.
2384 (nds32_emit_mem_move_block): New.
2385 (nds32_expand_movmemsi_loop_unknown_size): New.
2386 (nds32_expand_movmemsi_loop_known_size): New.
2387 (nds32_expand_movmemsi_loop): New.
2388 (nds32_expand_movmemsi_unroll): New.
2389 (nds32_expand_movmemqi): Rename ...
2390 (nds32_expand_movmemsi): ... to this.
2391 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
2392 (movmemsi): ... to this.
2393 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
2394 (nds32_expand_movmemsi): ... to this.
2396 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2397 Monk Chiang <sh.chiang04@gmail.com>
2398 Chung-Ju Wu <jasonwucj@gmail.com>
2400 * config/nds32/nds32-protos.h
2401 (nds32_expand_load_multiple): New arguments.
2402 (nds32_expand_store_multiple): Ditto.
2403 (nds32_valid_multiple_load_store): Rename ...
2404 (nds32_valid_multiple_load_store_p): ... to this.
2405 * config/nds32/nds32-memory-manipulation.c
2406 (nds32_expand_load_multiple): Refine implementation.
2407 (nds32_expand_store_multiple): Ditto.
2408 * config/nds32/nds32-multiple.md
2409 (load_multiple): Update nds32_expand_load_multiple interface.
2410 (store_multiple): Update nds32_expand_store_multiple interface.
2411 * config/nds32/nds32-predicates.c
2412 (nds32_valid_multiple_load_store): Rename ...
2413 (nds32_valid_multiple_load_store_p): ... to this and refine
2415 * config/nds32/predicates.md
2416 (nds32_load_multiple_and_update_address_operation): New predicate.
2417 (nds32_store_multiple_and_update_address_operation): New predicate.
2419 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2420 Chung-Ju Wu <jasonwucj@gmail.com>
2422 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
2423 (combo): New attribute.
2424 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
2426 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
2428 * config/nds32/nds32.opt: Change -mcmodel= default value.
2430 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
2431 Monk Chiang <sh.chiang04@gmail.com>
2432 Chung-Ju Wu <jasonwucj@gmail.com>
2434 * config/nds32/constants.md (unspec_element): New enum.
2435 * config/nds32/constraints.md (Umw): New constraint.
2436 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
2437 * config/nds32/nds32-intrinsic.md: Likewise.
2438 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
2439 (nds32_valid_smw_lwm_base_p): New.
2440 (nds32_output_smw_single_word): New.
2441 (nds32_output_lmw_single_word): New.
2442 (nds32_expand_unaligned_load): New.
2443 (nds32_expand_unaligned_store): New.
2444 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
2445 (nds32_output_smw_single_word): Declare.
2446 (nds32_output_lmw_single_word): Declare.
2447 (nds32_expand_unaligned_load): Declare.
2448 (nds32_expand_unaligned_store): Declare.
2449 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
2450 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
2451 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
2452 NDS32_BUILTIN_UASTORE_DW.
2453 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
2456 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
2457 Kito Cheng <kito.cheng@gmail.com>
2458 Chung-Ju Wu <jasonwucj@gmail.com>
2460 * config/nds32/nds32-intrinsic.c
2461 (nds32_expand_builtin_null_ftype_reg): Delete.
2462 (nds32_expand_builtin_reg_ftype_imm): Ditto.
2463 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
2464 (nds32_read_argument): New.
2465 (nds32_legitimize_target): Ditto.
2466 (nds32_legitimize_argument): Ditto.
2467 (nds32_check_constant_argument): Ditto.
2468 (nds32_expand_unop_builtin): Ditto.
2469 (nds32_expand_unopimm_builtin): Ditto.
2470 (nds32_expand_binop_builtin): Ditto.
2471 (nds32_builtin_decl_impl): Ditto.
2472 (builtin_description): Ditto.
2473 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
2474 (nds32_init_builtins_impl): Ditto.
2475 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
2476 (nds32_builtin_decl): New.
2477 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
2478 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
2480 2018-03-02 Jeff Law <law@redhat.com>
2482 * reorg.c (stop_search_p): Handle DEBUG_INSN.
2483 (redundant_insn, fill_simple_delay_slots): Likewise.
2484 (fill_slots_from_thread): Likewise.
2485 * resource.c (mark_referenced_resources): Likewise.
2486 (mark_set_resources, find_dead_or_set_registers): Likewise.
2488 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2490 * substring-locations.h (format_warning_va): Formatting fix for
2492 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
2494 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
2495 * substring-locations.c: Include intl.h.
2496 (format_warning_va): Turned into small wrapper around
2497 format_warning_n_va, renamed to ...
2498 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
2499 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
2501 (format_warning_at_substring_n): New function.
2502 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
2503 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
2504 format_warning_at_substring with just a shorter name instead of
2505 const function pointer.
2506 (fmtwarn_n): New function.
2507 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
2508 appropriate, get rid of all the fmtstr temporaries, move conditionals
2509 with G_() wrapped string literals directly into fmtwarn arguments,
2510 cast dir.len to (int), formatting fixes.
2512 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
2514 * doc/invoke.texi: Remove "Cilk Plus" references.
2516 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2517 Richard Biener <rguenther@suse.de>
2520 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
2521 for error or warning attributes if CALL_FROM_THUNK_P is set.
2524 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2527 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
2528 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
2531 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
2532 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
2534 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
2535 instead of -1U in last predictors element's probability member.
2537 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2540 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2541 arguments if they are comparable.
2543 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
2545 PR tree-optimization/84634
2546 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
2547 masks and masked_loop_p with a single loop_masks, making sure it's
2548 null for bb vectorization.
2550 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
2552 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
2553 (vect_analyze_data_ref_access): Use loop->safe_len rather than
2554 loop->force_vectorize to check whether there is no alias.
2556 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2559 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
2561 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
2563 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
2564 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
2565 instead of a loop around prev_real_insn.
2566 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
2570 * config/i386/i386.c (ix86_print_operand): Use conditional
2571 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
2574 2018-03-02 Richard Biener <rguenther@suse.de>
2576 PR tree-optimization/84427
2577 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
2578 (bitmap_set_subtract_values): Rewrite to handle multiple
2581 (prune_clobbered_mems): Likewise.
2582 (phi_translate): Take edge instead of pred/phiblock.
2583 (phi_translate_1): Likewise.
2584 (phi_translate_set): Likewise. Insert all translated
2585 exprs for a value into the set, keeping possibly multiple
2586 expressions per value.
2587 (compute_antic_aux): Adjust for phi_translate changes.
2588 When intersecting union the expressions and prune those
2589 not in the final value set, keeping possibly multiple
2590 expressions per value. Do not use value-insertion
2591 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
2592 all expressions. Add verification that the value-sets
2593 only shrink during iteration.
2594 (compute_partial_antic_aux): Adjust for the phi_translate changes.
2595 (do_pre_regular_insertion): Likewise.
2596 (do_pre_partial_partial_insertion): Likewise.
2598 2018-03-02 Richard Biener <rguenther@suse.de>
2601 * config/darwin.c (saved_debug_info_level): New static global.
2602 (darwin_asm_lto_start): Disable debug info generation for LTO out.
2603 (darwin_asm_lto_end): Restore debug info generation settings.
2605 2018-03-01 Martin Liska <mliska@suse.cz>
2608 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
2611 2018-03-01 Richard Biener <rguenther@suse.de>
2614 * dwarf2out.c (gen_variable_die): Properly handle late VLA
2615 type annotation with LTO when debug was disabled at compile-time.
2617 2018-03-01 Matthew Fortune <mfortune@gmail.com>
2619 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
2621 (mips_final_postscan_insn): Likewise.
2623 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
2625 PR rtl-optimization/84528
2626 * alias.c (init_alias_target): Add commentary.
2627 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
2628 a unique base value if the frame pointer is not eliminated
2629 to the stack pointer.
2631 2018-03-01 Tom de Vries <tom@codesourcery.com>
2633 PR rtl-optimization/83327
2634 * lra-int.h (hard_regs_spilled_into): Declare.
2635 * lra.c (hard_regs_spilled_into): Define.
2636 (init_reg_info): Init hard_regs_spilled_into.
2637 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
2638 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
2639 (process_bb_lives): Handle hard_regs_spilled_into.
2640 (lra_create_live_ranges_1): Before doing liveness propagation, clear
2641 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
2643 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
2645 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
2646 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
2647 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
2648 * config/rs6000/aix72.h: New file.
2650 2018-02-28 Jakub Jelinek <jakub@redhat.com>
2652 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
2653 instead of warning_at with conditional singular and plural messages
2657 * stor-layout.c (update_alignment_for_field): For
2658 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
2659 && !DECL_PACKED (field), do the alignment update, just use
2660 only desired_align instead of MAX (type_align, desired_align)
2662 (place_field): Don't do known_align < desired_align handling
2663 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
2664 is non-NULL, instead do it after rli->prev_field handling and
2665 only if not within a bitfield word. For DECL_PACKED (field)
2666 use type_align of BITS_PER_UNIT.
2668 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
2670 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
2671 superfluous parentheses and trailing spaces.
2673 2018-02-28 Richard Biener <rguenther@suse.de>
2675 PR tree-optimization/84584
2676 * graphite-scop-detection.c (scop_detection::add_scop): Discard
2677 SCoPs with fake exit edge.
2679 2018-02-28 Martin Liska <mliska@suse.cz>
2682 * timevar.c (timer::print): Fix format to properly print 100%
2685 2018-02-28 Richard Biener <rguenther@suse.de>
2688 * genmatch.c (capture_info::walk_match): Do not mark
2689 captured expressions without operands as expr_p given
2690 they act more like predicates and should be subject to
2691 "lost tail" side-effect preserving.
2693 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
2695 PR rtl-optimization/81611
2696 * auto-inc-dec.c (attempt_change): Move dead note from
2697 mem_insn if it's the next use of regno
2698 (find_address): Take address use of reg holding
2699 non-incremented value. Add parm to limit search to the named
2701 (merge_in_block): Attempt to use a mem insn that is the next
2702 use of the original regno.
2704 2018-02-27 Martin Sebor <msebor@redhat.com>
2707 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
2708 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
2710 2018-02-27 Martin Sebor <msebor@redhat.com>
2712 PR translation/84207
2713 * diagnostic-core.h (warning_n, error_n, inform_n): Change
2714 n argument to unsigned HOST_WIDE_INT.
2715 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
2716 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
2717 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
2718 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
2720 2018-02-27 Richard Biener <rguenther@suse.de>
2722 PR tree-optimization/84512
2723 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
2724 Do not use the estimate returned from record_stmt_cost for
2725 the scalar iteration cost but sum properly using add_stmt_cost.
2727 2018-02-27 Richard Biener <rguenther@suse.de>
2729 PR tree-optimization/84466
2730 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2731 Adjust last change to less strictly validate use operands.
2733 2018-02-27 Martin Liska <mliska@suse.cz>
2735 PR gcov-profile/84548
2736 * gcov.c (process_file): Allow partial overlap and consider it
2737 also as group functions.
2738 (output_lines): Properly calculate range of lines for a group.
2740 2018-02-27 Martin Liska <mliska@suse.cz>
2742 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
2743 'ggc' suffixes. Change first column width.
2744 (timer::print): Fix formatting of the column.
2746 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
2748 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
2749 preserve inline entry blocks for the sake of debug inline
2750 entry point markers alone.
2751 (remove_unused_locals): Suggest in comments a better place to
2752 force the preservation of inline entry blocks that are
2753 otherwise unused, but do not preserve them.
2755 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2757 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
2759 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2762 * config/i386/constraints.md (Bs): Replace
2763 ix86_indirect_branch_register with
2764 TARGET_INDIRECT_BRANCH_REGISTER.
2766 * config/i386/i386.md (indirect_jump): Likewise.
2767 (tablejump): Likewise.
2768 (*sibcall_memory): Likewise.
2769 (*sibcall_value_memory): Likewise.
2770 Peepholes of indirect call and jump via memory: Likewise.
2771 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
2772 (*sibcall_value_GOT_32): Likewise.
2773 * config/i386/predicates.md (indirect_branch_operand): Likewise.
2774 (GOT_memory_operand): Likewise.
2775 (call_insn_operand): Likewise.
2776 (sibcall_insn_operand): Likewise.
2777 (GOT32_symbol_operand): Likewise.
2778 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
2780 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2782 PR rtl-optimization/83496
2783 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
2784 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
2785 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
2786 redundant insn, if any.
2787 (relax_delay_slots): Likewise.
2788 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
2790 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
2792 PR tree-optimization/83965
2793 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
2794 that grouped statements are part of a reduction chain. Return
2795 true if the statement is not marked as a reduction itself but
2797 (vect_recog_dot_prod_pattern): Don't check whether the statement
2798 is part of a group here.
2799 (vect_recog_sad_pattern): Likewise.
2800 (vect_recog_widen_sum_pattern): Likewise.
2802 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2805 * final.c (rest_of_clean_state): Also look for calls inside sequences.
2807 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2810 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
2812 (ix86_output_indirect_function_return): New prototype.
2813 (ix86_split_simple_return_pop_internal): Likewise.
2814 * config/i386/i386.c (indirect_return_via_cx): New.
2815 (indirect_return_via_cx_bnd): Likewise.
2816 (indirect_thunk_name): Handle return va CX_REG.
2817 (output_indirect_thunk_function): Create alias for
2818 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
2819 (ix86_output_indirect_jmp): Remove the bool argument.
2820 (ix86_output_indirect_function_return): New function.
2821 (ix86_split_simple_return_pop_internal): Likewise.
2822 * config/i386/i386.md (*indirect_jump): Don't pass false
2823 to ix86_output_indirect_jmp.
2824 (*tablejump_1): Likewise.
2825 (simple_return_pop_internal): Change it to define_insn_and_split.
2826 Call ix86_split_simple_return_pop_internal to split it for
2828 (simple_return_indirect_internal): Call
2829 ix86_output_indirect_function_return instead of
2830 ix86_output_indirect_jmp.
2832 2018-02-26 Jakub Jelinek <jakub@redhat.com>
2835 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
2836 memset and value initialization afterwards.
2838 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
2840 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
2842 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2845 * common/config/aarch64/aarch64-common.c
2846 (aarch_option_optimization_table[]): Switch
2847 off fomit-frame-pointer
2849 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
2850 Chung-Ju Wu <jasonwucj@gmail.com>
2852 * config/nds32/nds32-multiple.md (load_multiple): Disallow
2854 (store_multiple): Ditto.
2856 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
2858 * config.gcc: Add --with-cpu support for nds32 target.
2859 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
2860 * config/nds32/nds32.opt: Add -mcpu= option.
2862 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
2864 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
2865 isel=yes): Warn for these deprecated options.
2867 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
2869 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
2870 ISA_2_5_MASKS_EMBEDDED.
2872 2018-02-23 Jakub Jelinek <jakub@redhat.com>
2874 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
2875 p->max as pointers rather than using iterative_hash_expr.
2877 2018-02-23 Carl Love <cel@us.ibm.com>
2879 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
2880 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
2882 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
2883 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
2884 P8V_BUILTIN_VEC_VUNSIGNED2.
2886 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
2889 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
2890 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
2892 * lra-constraints.c (curr_insn_transform): Set up
2893 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
2895 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
2896 * lra-eliminations.c (spill_pseudos): Ditto.
2897 (process_insn_for_elimination): Ditto.
2898 * lra-lives.c (reg_early_clobber_p): Use the new macros.
2899 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
2900 LRA_NON_CLOBBERED_ALT.
2902 2018-02-22 Martin Sebor <msebor@redhat.com>
2904 PR tree-optimization/84480
2905 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
2906 to maybe_diag_stxncpy_trunc. Call it.
2907 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
2908 from gimple_fold_builtin_strcpy. Print inlining stack.
2909 (handle_builtin_stxncpy): Print inlining stack.
2910 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
2912 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
2915 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
2916 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
2917 and -fcheck-pointer-bounds are used together.
2918 (indirect_thunk_prefix): New enum.
2919 (indirect_thunk_need_prefix): New function.
2920 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
2921 "_nt" instead of "_bnd" for NOTRACK prefix.
2922 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
2923 (output_indirect_thunk_function): Likewise.
2925 (ix86_code_end): Update output_indirect_thunk_function calls.
2926 (ix86_output_indirect_branch_via_reg): Replace
2927 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
2928 (ix86_output_indirect_branch_via_push): Likewise.
2929 (ix86_output_function_return): Likewise.
2930 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
2931 incompatible with -fcf-protection=branch and
2932 -fcheck-pointer-bounds.
2934 2018-02-22 Steve Ellcey <sellcey@cavium.com>
2937 * config/aarch64/aarch64.c (aarch64_print_address_internal):
2938 Change gcc_assert call to output_operand_lossage.
2940 2018-02-22 Steve Ellcey <sellcey@cavium.com>
2942 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
2944 2018-02-22 DJ Delorie <dj@redhat.com>
2945 Sebastian Perta <sebastian.perta@renesas.com>
2946 Oleg Endo <olegendo@gcc.gnu.org>
2948 * config/rx/rx.c (rx_rtx_costs): New function.
2949 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
2951 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
2953 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
2955 2018-02-22 Martin Liska <mliska@suse.cz>
2958 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
2959 Add "native" as a possible value.
2961 2018-02-22 Martin Liska <mliska@suse.cz>
2964 * config/i386/i386.c (ix86_option_override_internal):
2965 Add "native" as a possible value for -march and -mtune.
2967 2018-02-22 Jakub Jelinek <jakub@redhat.com>
2970 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
2971 to all type variants.
2973 PR tree-optimization/84503
2974 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
2975 width as info->bitpos + info->bitsize - start.
2976 (merged_store_group::merge_overlapping): Simplify width computation.
2977 (check_no_overlap): New function.
2978 (imm_store_chain_info::try_coalesce_bswap): Compute expected
2979 start + width and last_order of the group, fail if check_no_overlap
2981 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
2982 to group if check_no_overlap fails.
2984 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2986 * config/rs6000/altivec.md: Delete contraint arguments to
2987 define_expand, define_split, and define_peephole2, and in
2988 define_insn_and_split if always unused.
2989 * config/rs6000/darwin.md: Ditto.
2990 * config/rs6000/dfp.md: Ditto.
2991 * config/rs6000/rs6000.md: Ditto.
2992 * config/rs6000/sync.md: Ditto.
2993 * config/rs6000/vector.md: Ditto.
2994 * config/rs6000/vsx.md: Ditto.
2996 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2998 * config/rs6000/altivec.md: Write output control strings as braced
2999 blocks instead of double-quoted strings.
3000 * config/rs6000/darwin.md: Ditto.
3001 * config/rs6000/rs6000.md: Ditto.
3002 * config/rs6000/vector.md: Ditto.
3003 * config/rs6000/vsx.md: Ditto.
3005 2018-02-21 Jason Merrill <jason@redhat.com>
3007 PR c++/84314 - ICE with templates and fastcall attribute.
3008 * attribs.c (build_type_attribute_qual_variant): Remove assert.
3010 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3012 * ipa-cp.c (determine_versionability): Fix comment typos.
3014 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3017 * ipa-cp.c (determine_versionability): Do not version functions caling
3020 2018-02-21 Martin Liska <mliska@suse.cz>
3023 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
3024 Add "native" as a possible value.
3025 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
3026 the macro when native cpu detection is available.
3028 2018-02-21 Martin Liska <mliska@suse.cz>
3031 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
3032 Add "native" as a possible value.
3033 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
3034 when native cpu detection is available.
3036 2018-02-21 Jakub Jelinek <jakub@redhat.com>
3037 Martin Sebor <msebor@redhat.com>
3039 PR tree-optimization/84478
3040 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
3042 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
3043 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
3044 support which is conservatively correct, for 2 only stay conservative
3045 for maxlen. Formatting and comment capitalization fixes. Add STRICT
3046 argument to the 2 argument get_range_strlen, adjust 6 arg
3047 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
3049 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
3050 (gimple_fold_builtin_strlen): Pass true as last argument to
3053 2018-02-20 Martin Sebor <msebor@redhat.com>
3056 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
3057 (builtin_memref::set_base_and_offset): Same. Handle inner references.
3058 (builtin_memref::builtin_memref): Factor out parts into
3059 set_base_and_offset and call it.
3061 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3064 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
3065 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
3066 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
3067 search at the associated MODE_INT.
3069 2018-02-20 Jeff Law <law@redhat.com>
3072 PR tree-optimization/81592
3074 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
3075 for range data rather than using global data.
3076 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
3077 range data rather than using global data.
3078 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
3079 pass it to children as needed.
3080 (struct directive::fmtresult): Similarly.
3081 (struct directive::set_width): Similarly.
3082 (struct directive::set_precision): Similarly.
3083 (format_integer, format_directive, parse_directive): Similarly.
3084 (format_none): Accept unnamed vr_values parameter.
3085 (format_percent, format_floating, format_character): Similarly.
3086 (format_string, format_plain): Similarly.
3087 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
3088 the EVRP range analyzer for range data rather than using global data.
3089 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
3090 gimple-ssa-evrp-analyze.h
3091 (class sprintf_dom_walker): Add after_dom_children member function.
3092 Add evrp_range_analyzer member.
3093 (sprintf_dom_walker::before_dom_children): Call into the EVRP
3094 range analyzer as needed.
3095 (sprintf_dom_walker::after_dom_children): New member function.
3096 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
3098 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
3099 (evrp_range_analyzer::pop_to_marker): Likewise.
3101 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3103 PR tree-optimization/84419
3104 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
3105 with the required type if its current type is compatible but
3108 2018-02-20 Jakub Jelinek <jakub@redhat.com>
3111 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
3112 after vectorization.
3114 2018-02-20 Martin Liska <mliska@suse.cz>
3117 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
3118 possible values if we don't have a hint.
3120 2018-02-20 Martin Liska <mliska@suse.cz>
3124 * final.c (shorten_branches): Build align_tab array with one
3126 * opts.c (finish_options): Add alignment option limit check.
3127 (MAX_CODE_ALIGN): Likewise.
3128 (MAX_CODE_ALIGN_VALUE): Likewise.
3129 * doc/invoke.texi: Document maximum allowed option value for
3130 all -falign-* options.
3132 2018-02-19 Jakub Jelinek <jakub@redhat.com>
3135 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
3136 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
3137 * var-tracking.c (emit_note_insn_var_location): Remove all references
3138 to NOTE_INSN_CALL_ARG_LOCATION.
3139 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
3140 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
3141 Use copy_rtx_if_shared.
3142 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
3143 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
3144 (dwarf2out_var_location): Remove handling of
3145 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
3147 * final.c (final_scan_insn): Remove all references to
3148 NOTE_INSN_CALL_ARG_LOCATION.
3149 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
3150 before dumping final insns.
3151 * except.c (emit_note_eh_region_end): Remove all references to
3152 NOTE_INSN_CALL_ARG_LOCATION.
3153 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
3154 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
3155 * config/arc/arc.c (hwloop_optimize): Likewise.
3156 * config/arm/arm.c (create_fix_barrier): Likewise.
3157 * config/s390/s390.c (s390_chunkify_start): Likewise.
3158 * config/sh/sh.c (find_barrier): Likewise.
3159 * config/i386/i386.c (rest_of_insert_endbranch,
3160 ix86_seh_fixup_eh_fallthru): Likewise.
3161 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
3162 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
3163 * config/frv/frv.c (frv_function_prologue): Likewise.
3164 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
3166 (note_outside_basic_block_p): Remove all references to
3167 NOTE_INSN_CALL_ARG_LOCATION.
3168 * gengtype.c (adjust_field_rtx_def): Likewise.
3169 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
3171 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
3172 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
3175 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
3178 PR tree-optimization/84452
3179 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
3180 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
3183 2018-02-19 Martin Liska <mliska@suse.cz>
3186 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
3188 2018-02-19 Martin Liska <mliska@suse.cz>
3189 Richard Sandiford <richard.sandiford@linaro.org>
3191 PR tree-optimization/82491
3192 * gimple-fold.c (get_base_constructor): Make earlier bail out
3195 2018-02-19 Carl Love <cel@us.ibm.com>
3197 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
3198 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
3200 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
3201 P8V_BUILTIN_VEC_NEG.
3203 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
3205 * config/rl78/rl78.md (movdf): New define expand.
3207 2018-02-19 Martin Liska <mliska@suse.cz>
3210 * doc/invoke.texi: Fix typo.
3211 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
3213 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
3215 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
3216 handle rs6000_single_float and rs6000_double_float specially for
3219 2018-02-16 Jeff Law <law@redhat.com>
3221 * config/rx/rx.c (add_pop_cfi_notes): New function.;
3224 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3227 * ipa-inline.c (inline_small_functions): Fix a typo.
3229 2018-02-16 Nathan Sidwell <nathan@acm.org>
3231 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
3233 2018-02-16 Carl Love <cel@us.ibm.com>
3235 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
3236 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
3237 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
3238 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
3239 expansion to P8V_BUILTIN_VEC_FLOAT2.
3241 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
3243 PR rtl-optimization/70023
3244 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
3245 src_regno into account.
3247 2018-02-16 Carl Love <cel@us.ibm.com>
3249 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
3250 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
3251 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
3252 * config/rs6000/rs6000.c: Remove case statements for
3253 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
3254 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
3255 and P9V_BUILTIN_VEC_VINSERT4B.
3256 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
3257 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
3258 * config/rs6000/vsx.md:
3259 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
3262 2018-02-16 Carl Love <cel@us.ibm.com>
3264 * config/rs6000/altivec.h: Add builtin names vec_extract4b
3266 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
3268 * config/rs6000/rs6000-c.c: Add the definitions for
3269 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
3270 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
3271 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
3272 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
3273 definition for insert4b and define insn *insert3b_internal.
3274 * doc/extend.texi: Add documentation for vec_extract4b.
3276 2018-02-16 Nathan Sidwell <nathan@acm.org>
3278 * doc/extend.texi (Backwards Compatibility): Mention friend
3279 injection. Note for-scope is deprecated.
3280 * doc/invoke.texi (-ffriend-injection): Deprecate.
3282 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
3284 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
3285 that moved to I2, also allow destinations that are a paradoxical
3286 subreg (instead of a normal reg).
3288 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
3291 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
3294 2018-02-16 Richard Biener <rguenther@suse.de>
3296 PR tree-optimization/84037
3297 PR tree-optimization/84016
3299 * config/i386/i386.c (ix86_builtin_vectorization_cost):
3300 Adjust vec_construct for the fact we need additional higher latency
3301 128bit inserts for AVX256 and AVX512 vector builds.
3302 (ix86_add_stmt_cost): Scale vector construction cost for
3305 2018-02-16 Richard Biener <rguenther@suse.de>
3307 PR tree-optimization/84417
3308 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
3309 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
3310 (non_rewritable_lvalue_p): Likewise, use poly-ints.
3312 2018-02-16 Martin Liska <mliska@suse.cz>
3315 * internal-fn.def (ASAN_CHECK): Set proper flags.
3316 (ASAN_MARK): Likewise.
3318 2018-02-16 Julia Koval <julia.koval@intel.com>
3320 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
3321 from PTA_CANNONLAKE.
3323 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3326 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
3327 Use ++iter rather than iter++ for std::list iterators.
3328 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
3329 defer deleting them until all nodes in the forest are processed. Do
3330 free even leaf nodes. Change to_process into auto_vec.
3333 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
3334 * vec.h (vec_default_construct): Use memset instead of placement new
3335 if BROKEN_VALUE_INITIALIZATION is defined.
3336 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
3337 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
3340 PR rtl-optimization/83723
3341 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
3342 * lra.c (lra_substitute_pseudo): Likewise. If true, use
3343 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
3345 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
3347 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
3349 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
3351 PR rtl-optimization/81443
3352 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
3353 from inner REGs to paradoxical SUBREGs.
3355 2018-02-16 Richard Biener <rguenther@suse.de>
3357 PR tree-optimization/84399
3358 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3359 For operands we can analyze at their definition make sure we can
3360 analyze them at each use as well.
3362 2018-02-16 Richard Biener <rguenther@suse.de>
3364 PR tree-optimization/84190
3365 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
3366 volatile accesses if the decl isn't volatile.
3368 2018-02-15 Jason Merrill <jason@redhat.com>
3370 PR c++/84314 - ICE with templates and fastcall attribute.
3371 * attribs.c (build_type_attribute_qual_variant): Don't clobber
3372 TYPE_CANONICAL on an existing type.
3374 2018-02-15 Jakub Jelinek <jakub@redhat.com>
3376 PR tree-optimization/84383
3377 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
3378 dstoff nor call operand_equal_p if dstbase is NULL.
3380 PR tree-optimization/84334
3381 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
3382 also a CONSTANT_CLASS_P, punt.
3384 2018-02-14 Jim Wilson <jimw@sifive.com>
3386 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
3387 first SMALL_OPERAND check. New local min_second_step. Move assert
3388 to where locals are set. Add TARGET_RVC support.
3389 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
3391 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
3393 * doc/invoke.texi: Correct -Wformat-overflow code sample.
3395 2018-02-14 Martin Sebor <msebor@redhat.com>
3397 PR tree-optimization/83698
3398 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
3399 arrays constrain the offset range to their bounds.
3400 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
3401 (builtin_access::overlap): Avoid setting the size of overlap if it's
3403 (maybe_diag_overlap): Also consider arrays when deciding what values
3404 of offsets to include in diagnostics.
3406 2018-02-14 Martin Sebor <msebor@redhat.com>
3409 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
3410 that correspond to the kind of a declaration.
3412 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
3415 * config/pa/pa.md: Load address of PIC label using the linkage table
3416 if the label is nonlocal.
3418 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
3420 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
3421 warning message if user requests -maltivec=be.
3422 * doc/invoke.texi: Document deprecation of -maltivec=be.
3424 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
3427 * config/rs6000/rs6000-c.c: Update definitions for
3428 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
3429 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
3431 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3434 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
3435 add _get_ssp intrinsics. Remove argument from
3436 __builtin_ia32_rdssp[d|q].
3437 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
3438 * config/i386/i386-builtin.def: Remove argument from
3439 __builtin_ia32_rdssp[d|q].
3440 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
3441 ix86_expand_special_args_builtin for _rdssp[d|q].
3442 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
3443 Clear register before usage.
3444 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
3445 Add documentation for new _get_ssp and _inc_ssp intrinsics.
3447 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
3449 PR tree-optimization/84357
3450 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
3451 operand 1 of an ARRAY_REF too.
3453 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
3456 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
3457 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3459 (set_of_reg): New struct.
3460 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
3461 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
3462 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3464 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
3465 Split into bitclr, bitset, bitinvert patterns if appropriate.
3466 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
3467 use rx_fuse_in_memory_bitop.
3468 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
3469 to named insn, correct maximum insn length.
3471 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
3474 * machmode.def: Define a complex mode for PARTIAL_INT.
3475 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
3477 * doc/rtl.texi: Document CSPImode.
3478 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
3480 (msp430_hard_regno_nregs_with_padding): Likewise.
3482 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
3485 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
3487 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
3489 PR rtl-optimization/84169
3490 * combine.c (try_combine): New variable split_i2i3. Set it to true if
3491 we generated a parallel as new i3 and we split that to new i2 and i3
3492 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
3493 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
3494 those to i2, not i1. Partially rewrite this scan code.
3496 2018-02-13 Jakub Jelinek <jakub@redhat.com>
3499 * stor-layout.c (place_field): For variable length fields, adjust
3500 offset_align afterwards not just based on the field's alignment,
3501 but also on the size.
3504 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
3505 of exps and logs in the use_exp2 case.
3507 2018-02-13 Jeff Law <law@redhat.com>
3509 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
3512 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
3515 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
3519 * common.opt (gas-loc-support, gas-locview-support): New.
3520 (ginline-points, ginternal-reset-location-views): New.
3521 * doc/invoke.texi: Document them. Use @itemx where intended.
3522 (gvariable-location-views): Adjust.
3523 * target.def (reset_location_view): New.
3524 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
3525 (TARGET_RESET_LOCATION_VIEW): New.
3526 * doc/tm.texi: Rebuilt.
3527 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
3528 (dwarf2out_default_as_locview_support): New.
3529 (output_asm_line_debug_info): Use option variables.
3530 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
3531 (output_loc_list): Likewise.
3532 (add_high_low_attributes): Check option variables.
3533 Don't output entry view attribute in strict mode.
3534 (gen_inlined_subroutine_die): Check option variables.
3535 (dwarf2out_inline_entry): Likewise.
3536 (init_sections_and_labels): Likewise.
3537 (dwarf2out_early_finish): Likewise.
3538 (maybe_reset_location_view): New, from...
3539 (dwarf2out_var_location): ... here. Call it.
3540 * debug.h (dwarf2out_default_as_loc_support): Declare.
3541 (dwarf2out_default_as_locview_support): Declare.
3542 * hooks.c (hook_int_rtx_insn_0): New.
3543 * hooks.h (hook_int_rtx_insn_0): Declare.
3544 * toplev.c (process_options): Take -gas-loc-support and
3545 -gas-locview-support from dwarf2out. Enable
3546 -gvariable-location-views by default only with locview
3547 assembler support. Enable -ginternal-reset-location-views by
3548 default only if the target defines the corresponding hook.
3549 Enable -ginline-points by default if location views are
3550 enabled; force it disabled if statement frontiers are
3552 * tree-inline.c (expand_call_inline): Check option variables.
3553 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
3555 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
3557 PR tree-optimization/84321
3558 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
3559 handling. Also check whether the anti-range contains any values
3560 that satisfy the mask; switch to a VR_RANGE if not.
3562 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
3565 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
3567 2018-02-13 Martin Jambor <mjambor@suse.cz>
3570 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
3571 of call statements, also set location of a load to a temporary.
3573 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
3575 * config/rl78/rl78.c (add_vector_labels): New function.
3576 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
3577 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
3578 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
3579 which checks that no arguments are passed.
3580 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
3581 * doc/extend.texi: Documentation for the new attribute.
3583 2018-02-13 Andreas Schwab <schwab@suse.de>
3585 * config/riscv/linux.h (CPP_SPEC): Define.
3587 2018-02-13 Jakub Jelinek <jakub@redhat.com>
3590 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
3591 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
3592 OPTION_MASK_ISA_AES as first argument to def_builtin_const
3593 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
3594 instead of OPTION_MASK_ISA_PCLMUL as first argument to
3595 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
3596 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
3597 temporarily for AES and PCLMUL builtins.
3599 PR tree-optimization/84339
3600 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
3601 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
3605 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
3606 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
3607 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
3609 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
3611 * omp-simd-clone.h: New file.
3612 * omp-simd-clone.c: Include omp-simd-clone.h.
3613 (expand_simd_clones): No longer static.
3614 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
3615 cgraph.h and omp-simd-clone.h.
3616 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
3617 (vect_recog_widen_shift_pattern): Formatting fix.
3618 (vect_pattern_recog_1): Don't check optab for calls.
3621 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
3622 operands[2] into a REG before using gen_lowpart on it.
3624 2018-02-12 Jeff Law <law@redhat.com>
3627 * config/sh/sh.c (find_barrier): Consider a sibling call
3630 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
3631 successfully back substituting a reg.
3633 2018-02-12 Richard Biener <rguenther@suse.de>
3635 PR tree-optimization/84037
3636 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
3637 parameter, move visited init to caller.
3638 (vect_slp_analyze_operations): Separate cost from validity
3639 check, initialize visited once for all instances.
3640 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
3642 * tree-vect-stmts.c (vect_model_simple_cost): Make early
3644 (vect_model_promotion_demotion_cost): Likewise.
3645 (vectorizable_bswap): Guard cost modeling with !slp_node
3646 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
3648 (vectorizable_call): Likewise.
3649 (vectorizable_conversion): Likewise.
3650 (vectorizable_assignment): Likewise.
3651 (vectorizable_shift): Likewise.
3652 (vectorizable_operation): Likewise.
3653 (vectorizable_store): Likewise.
3654 (vectorizable_load): Likewise.
3655 (vectorizable_condition): Likewise.
3656 (vectorizable_comparison): Likewise.
3658 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
3661 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
3662 (ASAN_MARK): Fix fnspec to account for return value, change pointer
3663 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
3665 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3668 * params.def (inline-min-speedup): Increase from 8 to 15.
3669 (max-inline-insns-auto): Decrease from 40 to 30.
3670 * ipa-split.c (consider_split): Add some buffer for function to
3671 be considered inlining candidate.
3672 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
3675 2018-02-12 Richard Biener <rguenther@suse.de>
3677 PR tree-optimization/84037
3678 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
3679 matched stmts if we cannot swap the non-matched ones.
3681 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3683 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
3684 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
3685 _mm_maskz_scalef_round_ss): New intrinsics.
3686 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
3687 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
3688 __builtin_ia32_scalefss_round): Remove.
3689 (__builtin_ia32_scalefsd_mask_round,
3690 __builtin_ia32_scalefss_mask_round): New intrinsics.
3691 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
3692 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
3693 ((match_operand:VF_128 2 "<round_nimm_predicate>"
3694 "<round_constraint>")): Changed to ...
3695 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
3696 "<round_scalar_constraint>")): ... this.
3697 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
3698 %0, %1, %2<round_op3>}"): Changed to ...
3699 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
3700 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
3701 %2<round_scalar_mask_op3>}"): ... this.
3702 * config/i386/subst.md (round_scalar_nimm_predicate): New.
3704 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3706 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
3707 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
3708 (_mm_maskz_sqrt_round_ss): New intrinsics.
3709 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
3710 (__builtin_ia32_sqrtsd_mask_round)
3711 (__builtin_ia32_sqrtss_mask_round): New builtins.
3712 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
3713 (__builtin_ia32_sqrtss_round): Remove.
3714 (__builtin_ia32_sqrtsd_mask_round)
3715 (__builtin_ia32_sqrtss_mask_round): New builtins.
3716 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
3717 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
3718 ((match_operand:VF_128 1 "vector_operand"
3719 "xBm,<round_constraint>")): Changed to ...
3720 ((match_operand:VF_128 1 "vector_operand"
3721 "xBm,<round_scalar_constraint>")): ... this.
3722 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
3723 %0, %2, %<iptr>1<round_op3>}): Changed to ...
3724 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
3725 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
3726 %<iptr>1<round_scalar_mask_op3>}): ... this.
3727 ((set_attr "prefix" "<round_prefix>")): Changed to ...
3728 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
3730 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
3733 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
3734 Cast vec_cmpeq result to correct type.
3735 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
3736 Cast vec_cmpgt result to correct type.
3738 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
3740 * final.c (final_scan_insn_1): Renamed from...
3741 (final_scan_insn): ... this. New wrapper, to recover
3742 seen from the outermost call in recursive ones.
3743 * config/sparc/sparc.c (output_return): Drop seen from call.
3744 (output_sibcall): Likewise.
3745 * config/visium/visium.c (output_branch): Likewise.
3747 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
3749 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
3752 2018-02-10 Alan Modra <amodra@gmail.com>
3755 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
3756 Specify LR as an input.
3758 2018-02-10 Jakub Jelinek <jakub@redhat.com>
3761 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
3762 remove_member_access_dummy_vars): New functions.
3763 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
3764 lower_omp_1, execute_lower_omp): Use them.
3766 PR rtl-optimization/84308
3767 * shrink-wrap.c (spread_components): Release todo vector.
3769 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
3771 PR rtl-optimization/57193
3772 * ira-color.c (struct allocno_color_data): Add member
3773 conflict_allocno_hard_prefs.
3774 (update_conflict_allocno_hard_prefs): New.
3775 (bucket_allocno_compare_func): Add a preference based on
3776 conflict_allocno_hard_prefs.
3777 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
3778 (color_allocnos): Remove a dead code. Initiate
3779 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
3781 2018-02-09 Jakub Jelinek <jakub@redhat.com>
3784 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
3785 constraint from =wa to wa. Avoid a subreg on the output operand,
3786 instead use a pseudo and subreg it in a move.
3787 (p9_xxbrd_<mode>): Changed to ...
3788 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
3789 (p9_xxbrd_v2df): New expander.
3790 (p9_xxbrw_<mode>): Changed to ...
3791 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
3792 (p9_xxbrw_v4sf): New expander.
3794 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
3796 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
3798 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
3801 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
3802 multiply in 32-bit mode.
3803 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
3804 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
3807 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
3809 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
3810 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
3811 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
3812 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
3814 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
3817 * dwarf2out.c (is_trivial_indirect_ref): New function.
3818 (dwarf2out_late_global_decl): Do not generate a location
3819 attribute for variables that have a non-trivial DECL_VALUE_EXPR
3820 and that are not defined in the current unit.
3822 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
3824 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
3825 instead of a libcall for UNORDERED.
3827 2018-02-09 Tamar Christina <tamar.christina@arm.com>
3830 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
3831 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
3833 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3836 * config/s390/s390.c (s390_set_current_function): Invoke
3837 s390_indirect_branch_settings also if fndecl didn't change.
3839 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
3841 * config/rs6000/rs6000.md (blockage): Set length to zero.
3843 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
3845 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
3847 2018-02-09 Jakub Jelinek <jakub@redhat.com>
3850 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
3851 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
3855 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
3856 PARALLEL incoming that failed vt_get_decl_and_offset check.
3859 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
3860 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
3862 (get_variable_section): For decls in named .bss* sections pass true as
3863 second argument to bss_initializer_p.
3865 2018-02-09 Marek Polacek <polacek@redhat.com>
3866 Jakub Jelinek <jakub@redhat.com>
3869 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
3870 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
3871 Sync some changes from cxx_fold_indirect_ref.
3873 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
3875 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
3877 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
3878 (BLOCK_INLINE_ENTRY_LABEL): New.
3879 (dwarf2out_var_location): Disregard inline entry markers.
3880 (inline_entry_data): New struct.
3881 (inline_entry_data_hasher): New hashtable type.
3882 (inline_entry_data_hasher::hash): New.
3883 (inline_entry_data_hasher::equal): New.
3884 (inline_entry_data_table): New variable.
3885 (add_high_low_attributes): Add DW_AT_entry_pc and
3886 DW_AT_GNU_entry_view attributes if a pending entry is found
3887 in inline_entry_data_table. Add old entry_pc attribute only
3888 if debug nonbinding markers are disabled.
3889 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
3890 markers are enabled.
3891 (block_within_block_p, dwarf2out_inline_entry): New.
3892 (dwarf2out_finish): Check that no entries remained in
3893 inline_entry_data_table.
3894 * final.c (reemit_insn_block_notes): Handle inline entry notes.
3895 (final_scan_insn, notice_source_line): Likewise.
3896 (rest_of_clean_state): Skip inline entry markers.
3897 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
3899 * gimple.c (gimple_build_debug_inline_entry): New.
3900 * gimple.h (enum gimple_debug_subcode): Add
3901 GIMPLE_DEBUG_INLINE_ENTRY.
3902 (gimple_build_debug_inline_entry): Declare.
3903 (gimple_debug_inline_entry_p): New.
3904 (gimple_debug_nonbind_marker_p): Adjust.
3905 * insn-notes.def (INLINE_ENTRY): New.
3906 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
3907 inline entry marker notes.
3908 (print_insn): Likewise.
3909 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
3910 (INSN_DEBUG_MARKER_KIND): Likewise.
3911 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
3912 * tree-inline.c (expand_call_inline): Build and insert
3913 debug_inline_entry stmt.
3914 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
3915 inline entry blocks early, if nonbind markers are enabled.
3916 (dump_scope_block): Dump fragment info.
3917 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
3918 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
3919 (gimple_build_debug_inline_entry): New.
3920 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
3921 Enable/disable inline entry points too.
3922 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
3923 (DEBUG_INSN): Describe inline entry markers.
3925 * common.opt (gvariable-location-views): New.
3926 (gvariable-location-views=incompat5): New.
3927 * config.in: Rebuilt.
3928 * configure: Rebuilt.
3929 * configure.ac: Test assembler for view support.
3930 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
3931 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
3932 * dwarf2out.c (var_loc_view): New typedef.
3933 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
3934 (dwarf2out_locviews_in_attribute): New.
3935 (dwarf2out_locviews_in_loclist): New.
3936 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
3937 (enum dw_line_info_opcode): Add LI_adv_address.
3938 (struct dw_line_info_table): Add view.
3939 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
3940 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
3941 (zero_view_p): New variable.
3942 (ZERO_VIEW_P): New macro.
3943 (output_asm_line_debug_info): New.
3944 (struct var_loc_node): Add view.
3945 (add_AT_view_list, AT_loc_list): New.
3946 (add_var_loc_to_decl): Add view param. Test it against last.
3947 (new_loc_list): Add view params. Record them.
3948 (AT_loc_list_ptr): Handle loc and view lists.
3949 (view_list_to_loc_list_val_node): New.
3950 (print_dw_val): Handle dw_val_class_view_list.
3951 (size_of_die): Likewise.
3952 (value_format): Likewise.
3953 (loc_list_has_views): New.
3954 (gen_llsym): Set vl_symbol too.
3955 (maybe_gen_llsym, skip_loc_list_entry): New.
3956 (dwarf2out_maybe_output_loclist_view_pair): New.
3957 (output_loc_list): Output view list or entries too.
3958 (output_view_list_offset): New.
3959 (output_die): Handle dw_val_class_view_list.
3960 (output_dwarf_version): New.
3961 (output_compilation_unit_header): Use it.
3962 (output_skeleton_debug_sections): Likewise.
3963 (output_rnglists, output_line_info): Likewise.
3964 (output_pubnames, output_aranges): Update version comments.
3965 (output_one_line_info_table): Output view numbers in asm comments.
3966 (dw_loc_list): Determine current endview, pass it to new_loc_list.
3967 Call maybe_gen_llsym.
3968 (loc_list_from_tree_1): Adjust.
3969 (add_AT_location_description): Create view list attribute if
3970 needed, check it's absent otherwise.
3971 (convert_cfa_to_fb_loc_list): Adjust.
3972 (maybe_emit_file): Call output_asm_line_debug_info for test.
3973 (dwarf2out_var_location): Reset views as needed. Precompute
3974 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
3975 attribute. Set view.
3976 (new_line_info_table): Reset next view.
3977 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
3978 (dwarf2out_source_line): Likewise. Output view resets and labels to
3979 the assembler, or select appropriate line info opcodes.
3980 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
3981 (optimize_string_length): Catch it. Adjust.
3982 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
3983 dw_val_class_view_list, and remove it if no longer needed.
3984 (hash_loc_list): Hash view numbers.
3985 (loc_list_hasher::equal): Compare them.
3986 (optimize_location_lists): Check whether a view list symbol is
3987 needed, and whether the locview attribute is present, and
3988 whether they match. Remove the locview attribute if no longer
3990 (index_location_lists): Call skip_loc_list_entry for test.
3991 (dwarf2out_finish): Call output_asm_line_debug_info for test.
3992 Use output_dwarf_version.
3993 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
3994 (struct dw_val_node): Add val_view_list.
3995 * final.c (SEEN_NEXT_VIEW): New.
3996 (set_next_view_needed): New.
3997 (clear_next_view_needed): New.
3998 (maybe_output_next_view): New.
3999 (final_start_function): Rename to...
4000 (final_start_function_1): ... this. Take pointer to FIRST,
4001 add SEEN parameter. Emit param bindings in the initial view.
4002 (final_start_function): Reintroduce SEEN-less interface.
4003 (final): Rename to...
4004 (final_1): ... this. Take SEEN parameter. Output final pending
4005 next view at the end.
4006 (final): Reintroduce seen-less interface.
4007 (final_scan_insn): Output pending next view before switching
4008 sections or ending a block. Mark the next view as needed when
4009 outputting variable locations. Notify debug backend of section
4010 changes, and of location view changes.
4011 (rest_of_handle_final): Adjust.
4012 * toplev.c (process_options): Autodetect value for debug variable
4013 location views option. Warn on incompat5 without -gdwarf-5.
4014 * doc/invoke.texi (gvariable-location-views): New.
4015 (gvariable-location-views=incompat5): New.
4016 (gno-variable-location-views): New.
4018 2018-02-08 David Malcolm <dmalcolm@redhat.com>
4020 PR tree-optimization/84136
4021 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
4022 that the result of find_edge is non-NULL.
4024 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4027 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
4028 storing integer register in SImode. Fix cost of 256 and 512
4029 byte aligned SSE register store.
4031 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4033 * config/i386/i386.c (ix86_multiplication_cost): Fix
4034 multiplication cost for TARGET_AVX512DQ.
4036 2018-02-08 Marek Polacek <polacek@redhat.com>
4038 PR tree-optimization/84238
4039 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
4042 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4044 PR tree-optimization/84265
4045 * tree-vect-stmts.c (vectorizable_store): Don't treat
4046 VMAT_CONTIGUOUS accesses as grouped.
4047 (vectorizable_load): Likewise.
4049 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4051 PR tree-optimization/81635
4052 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
4053 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
4054 (test_round_for_mask): New functions.
4055 (wide_int_cc_tests): Call test_round_for_mask.
4056 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
4057 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
4058 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
4059 range returned by get_range_info.
4061 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4064 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
4065 * symtab.c: Include builtins.h
4066 (symtab_node::output_to_lto_symbol_table_p): Move here
4067 from lto-streamer-out.c:output_symbol_p.
4068 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
4069 (output_symbol_p): Move all logic to symtab.c
4070 (produce_symtab): Update.
4072 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4074 * config/s390/s390-opts.h (enum indirect_branch): Define.
4075 * config/s390/s390-protos.h (s390_return_addr_from_memory)
4076 (s390_indirect_branch_via_thunk)
4077 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
4078 (enum s390_indirect_branch_type): Define.
4079 * config/s390/s390.c (struct s390_frame_layout, struct
4080 machine_function): Remove.
4081 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
4082 (indirect_branch_table_label_no, indirect_branch_table_name):
4084 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
4085 (enum s390_indirect_branch_option): Define.
4086 (s390_return_addr_from_memory): New function.
4087 (s390_handle_string_attribute): New function.
4088 (s390_attribute_table): Add new attribute handler.
4089 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
4090 (s390_indirect_branch_via_thunk): New function.
4091 (s390_indirect_branch_via_inline_thunk): New function.
4092 (s390_function_ok_for_sibcall): When jumping via thunk disallow
4093 sibling call optimization for non z10 compiles.
4094 (s390_emit_call): Force indirect branch target to be a single
4095 register. Add r1 clobber for non-z10 compiles.
4096 (s390_emit_epilogue): Emit return jump via return_use expander.
4097 (s390_reorg): Handle JUMP_INSNs as execute targets.
4098 (s390_option_override_internal): Perform validity checks for the
4099 new command line options.
4100 (s390_indirect_branch_attrvalue): New function.
4101 (s390_indirect_branch_settings): New function.
4102 (s390_set_current_function): Invoke s390_indirect_branch_settings.
4103 (s390_output_indirect_thunk_function): New function.
4104 (s390_code_end): Implement target hook.
4105 (s390_case_values_threshold): Implement target hook.
4106 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
4108 * config/s390/s390.h (struct s390_frame_layout)
4109 (struct machine_function): Move here from s390.c.
4110 (TARGET_INDIRECT_BRANCH_NOBP_RET)
4111 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
4112 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
4113 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
4114 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
4115 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
4116 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
4117 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
4118 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
4119 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
4120 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
4121 (mnemonic attribute): Add values which aren't recognized
4123 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
4124 pattern for branch conversion. Fix mnemonic attribute.
4125 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
4126 indirect branch via thunk if requested.
4127 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
4128 ("*indirect_jump"): Disable for branch conversion using out of
4130 ("indirect_jump_via_thunk<mode>_z10")
4131 ("indirect_jump_via_thunk<mode>")
4132 ("indirect_jump_via_inlinethunk<mode>_z10")
4133 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
4134 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
4135 ("casesi_jump_via_inlinethunk<mode>_z10")
4136 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
4137 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
4138 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
4139 ("*indirect2_jump"): Disable for branch conversion.
4140 ("casesi_jump"): Turn into expander and expand patterns for branch
4142 ("return_use"): New expander.
4143 ("*return"): Emit return via thunk and rename it to ...
4144 ("*return<mode>"): ... this one.
4145 * config/s390/s390.opt: Add new options and and enum for the
4148 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4150 * lra-constraints.c (match_reload): Unconditionally use
4151 gen_lowpart_SUBREG, rather than selecting between that
4152 and equivalent gen_rtx_SUBREG code.
4154 2018-02-08 Richard Biener <rguenther@suse.de>
4156 PR tree-optimization/84233
4157 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
4158 changed flag instead of boguously re-using phi_inserted.
4160 2018-02-08 Martin Jambor <mjambor@suse.cz>
4162 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
4163 static local variables.
4165 2018-02-08 Richard Biener <rguenther@suse.de>
4167 PR tree-optimization/84278
4168 * tree-vect-stmts.c (vectorizable_store): When looking for
4169 smaller vector types to perform grouped strided loads/stores
4170 make sure the mode is supported by the target.
4171 (vectorizable_load): Likewise.
4173 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4175 * config/aarch64/aarch64.c (aarch64_components_for_bb):
4176 Increase LDP/STP opportunities by adding adjacent callee-saves.
4178 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4180 PR rtl-optimization/84068
4181 PR rtl-optimization/83459
4182 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
4184 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
4186 PR tree-optimization/84224
4187 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
4188 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
4191 2018-02-07 Iain Sandoe <iain@codesourcery.com>
4194 * config/rs6000/altivec.md (*restore_world): Remove LR use.
4195 * config/rs6000/predicates.md (restore_world_operation): Adjust op
4196 count, remove one USE.
4198 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
4200 * doc/install.texi (Configuration): Document the
4201 --with-long-double-format={ibm,ieee} PowerPC configuration
4205 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
4206 Convert from define_expand to be define_insn_and_split. Rework
4207 float/double/_Float128 conversions to QI/HI/SImode to work with
4208 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
4209 conversions to QI/HImode types did a store and then a load to
4210 truncate the value. For conversions to VSX registers, don't split
4211 the insn, instead emit the code directly. Use the code iterator
4212 any_fix to combine signed and unsigned conversions.
4213 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
4214 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
4215 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4216 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
4217 (fix_<mode>di2_hw): Likewise.
4218 (fixuns_<mode>di2_hw): Likewise.
4219 (fix_<mode>si2_hw): Likewise.
4220 (fixuns_<mode>si2_hw): Likewise.
4221 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
4222 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4223 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
4224 fix<uns>_trunc<SFDF:mode>si2_p8.
4225 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
4227 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
4228 (fix<uns>_<mode>_mem): Likewise.
4229 (fctiw<u>z_<mode>_mem): Likewise.
4230 (fix<uns>_<mode>_mem): Likewise.
4231 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
4232 the register allocator from doing a direct move to the GPRs to do
4233 a store, and instead use the ISA 3.0 store byte/half-word from
4234 vector register instruction. For IEEE 128-bit floating point,
4235 also optimize stores of 32-bit ints.
4236 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
4238 2018-02-07 Alan Hayward <alan.hayward@arm.com>
4240 * genextract.c (push_pathstr_operand): New function to support
4242 (walk_rtx): Call push_pathstr_operand.
4243 (print_path): Support [a-zA-Z].
4245 2018-02-07 Richard Biener <rguenther@suse.de>
4247 PR tree-optimization/84037
4248 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
4249 (cse_and_gimplify_to_preheader): Declare.
4250 (vect_get_place_in_interleaving_chain): Likewise.
4251 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4253 (_loop_vec_info::~_loop_vec_info): Delete it.
4254 (cse_and_gimplify_to_preheader): New function.
4255 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
4256 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
4257 (vectorizable_load): Likewise. For grouped stores always base
4258 the IV on the first element.
4259 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
4260 condition before gimplifying.
4262 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4264 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
4265 *DIV_EXPR and *MOD_EXPR.
4267 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
4270 * config/i386/i386.c (ix86_option_override_internal): Mask out
4271 the CF_SET bit when checking -fcf-protection.
4273 2018-02-07 Tom de Vries <tom@codesourcery.com>
4276 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
4279 2018-02-07 Richard Biener <rguenther@suse.de>
4281 PR tree-optimization/84204
4282 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
4285 PR tree-optimization/84205
4286 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
4287 special-case isl_ast_op_zdiv_r.
4289 PR tree-optimization/84223
4290 * graphite-scop-detection.c (gather_bbs::before_dom_children):
4291 Only add conditions from within the region.
4292 (gather_bbs::after_dom_children): Adjust.
4294 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
4297 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
4298 * config/avr/avr.md: Only post-reload split REG-REG moves if
4299 either register is GENERAL_REG_P.
4301 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4303 PR tree-optimization/84235
4304 * tree-ssa-scopedtables.c
4305 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
4306 if the subtraction is performed in floating point type where NaNs are
4307 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
4308 build 1. Formatting fix.
4310 2018-02-06 Jakub Jelinek <jakub@redhat.com>
4313 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
4314 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
4315 and skip it regardless of bb boundaries. Use CALL_P macro,
4316 don't test INSN_P (insn) together with CALL_P or JUMP_P check
4317 unnecessarily, formatting fix.
4319 2018-02-06 Michael Collison <michael.collison@arm.com>
4321 * config/arm/thumb2.md:
4322 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
4323 (*thumb_mov_notscc): Ditto.
4325 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
4328 * config/rs6000/rs6000.md (su code attribute): Use "u" for
4329 unsigned_fix, not "s".
4331 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4333 * configure.ac (gcc_fn_eh_frame_ro): New function.
4334 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
4335 correct .eh_frame permissions.
4336 * configure: Regenerate.
4338 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
4340 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
4343 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4345 * config/rs6000/rs6000.c (rs6000_option_override_internal):
4346 Display warning message for -mno-speculate-indirect-jumps.
4348 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
4350 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
4352 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
4354 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
4356 PR tree-optimization/84225
4357 * tree-eh.c (find_trapping_overflow): Only call
4358 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
4360 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4363 * config/i386/i386.c: Reimplement the check of possible options
4364 -mibt/-mshstk conbination. Change error messages.
4365 * doc/invoke.texi: Fix a typo: remove extra '='.
4367 2018-02-06 Marek Polacek <polacek@redhat.com>
4369 PR tree-optimization/84228
4370 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
4372 2018-02-06 Tamar Christina <tamar.christina@arm.com>
4375 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
4376 emitted arch directives.
4377 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
4378 __ARM_FEATURE_COPROC before changing architectures.
4380 2018-02-06 Richard Biener <rguenther@suse.de>
4382 * config/i386/i386.c (print_reg): Fix typo.
4383 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
4385 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
4387 * configure: Regenerate.
4389 2018-02-05 Martin Sebor <msebor@redhat.com>
4391 PR tree-optimization/83369
4392 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
4395 2018-02-05 Martin Liska <mliska@suse.cz>
4397 * doc/invoke.texi: Cherry-pick upstream r323995.
4399 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
4401 * ira.c (ira_init_register_move_cost): Adjust comment.
4403 2018-02-05 Martin Liska <mliska@suse.cz>
4405 PR gcov-profile/84137
4406 * doc/gcov.texi: Fix typo in documentation.
4408 2018-02-05 Martin Liska <mliska@suse.cz>
4410 PR gcov-profile/83879
4411 * doc/gcov.texi: Document necessity of --dynamic-list-data when
4412 using dlopen functionality.
4414 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
4416 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
4417 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
4418 _mm_maskz_range_ss, _mm_mask_range_round_ss,
4419 _mm_maskz_range_round_ss): New intrinsics.
4420 (__builtin_ia32_rangesd128_round)
4421 (__builtin_ia32_rangess128_round): Remove.
4422 (__builtin_ia32_rangesd128_mask_round,
4423 __builtin_ia32_rangess128_mask_round): New builtins.
4424 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
4425 __builtin_ia32_rangess128_round): Remove.
4426 (__builtin_ia32_rangesd128_mask_round,
4427 __builtin_ia32_rangess128_mask_round): New builtins.
4428 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
4429 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
4430 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
4431 "<round_saeonly_constraint>")): Changed to ...
4432 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
4433 "<round_saeonly_scalar_constraint>")): ... this.
4434 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
4435 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
4436 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
4437 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
4438 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
4440 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
4442 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
4444 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
4445 Remove all values except native, 8540 and 8548.
4447 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
4449 * config/i386/i386.c (ix86_output_function_return): Pass
4450 INVALID_REGNUM, instead of -1, as invalid register number to
4451 indirect_thunk_name and output_indirect_thunk.
4453 2018-02-02 Julia Koval <julia.koval@intel.com>
4455 * config.gcc: Add -march=icelake.
4456 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
4457 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
4458 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
4459 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
4460 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
4461 (processor_target_table): Add icelake.
4462 (ix86_option_override_internal): Handle new PTAs.
4463 (get_builtin_code_for_version): Handle icelake.
4464 (M_INTEL_COREI7_ICELAKE): New.
4465 (fold_builtin_cpu): Handle icelake.
4466 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
4467 * doc/invoke.texi: Add -march=icelake.
4469 2018-02-02 Julia Koval <julia.koval@intel.com>
4471 * config/i386/i386.c (ix86_option_override_internal): Change flags type
4472 to wide_int_bitmask.
4473 * wide-int-bitmask.h: New.
4475 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4478 * config/i386/i386.md: Replace Pmode with word_mode in
4479 builtin_setjmp_setup and builtin_longjmp to support x32.
4481 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
4485 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
4487 (rs6000_supported_cpu_names): New static variable.
4488 (linux_cpu_translation_table): Likewise.
4489 (elf_platform) <cpu>: Define new static variable and use it.
4490 Translate kernel AT_PLATFORM name to canonical name if needed.
4491 Error if platform name is unknown.
4493 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
4496 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
4498 2018-02-01 Jeff Law <law@redhat.com>
4501 * config/i386/i386.c (release_scratch_register_on_entry): Add new
4502 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
4503 the scratch if RELEASE_VIA_POP is false.
4504 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
4505 If we have to save a temporary register, decrement SIZE appropriately.
4506 Pass new arguments to release_scratch_register_on_entry.
4507 (ix86_adjust_stack_and_probe): Likewise.
4508 (ix86_emit_probe_stack_range): Pass new arguments to
4509 release_scratch_register_on_entry.
4511 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
4513 PR rtl-optimization/84157
4514 * combine.c (change_zero_ext): Use REG_P predicate in
4515 front of HARD_REGISTER_P predicate.
4517 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
4519 * config/avr/avr.c (avr_option_override): Move disabling of
4520 -fdelete-null-pointer-checks to...
4521 * common/config/avr/avr-common.c (avr_option_optimization_table):
4524 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4526 PR tree-optimization/81635
4527 * tree-data-ref.c (split_constant_offset_1): For types that
4528 wrap on overflow, try to use range info to prove that wrapping
4531 2018-02-01 Renlin Li <renlin.li@arm.com>
4534 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
4536 (aarch64_register_move_cost): Likewise.
4537 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
4539 (REG_CLASS_NAMES): Likewise.
4540 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
4541 TAILCALL_ADDR_REGS. Remove IP registers.
4542 * config/aarch64/aarch64.md (Ucs): Update register constraint.
4544 2018-02-01 Richard Biener <rguenther@suse.de>
4546 * domwalk.h (dom_walker::dom_walker): Add additional constructor
4547 for specifying RPO order and allow NULL for that.
4548 * domwalk.c (dom_walker::dom_walker): Likewise.
4549 (dom_walker::walk): Handle NULL RPO order.
4550 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
4552 (rewrite_update_dom_walker): Likewise.
4553 (mark_def_dom_walker): Likewise.
4555 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4557 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
4558 (aarch64_maybe_expand_sve_subreg_move): Declare.
4559 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
4560 * config/aarch64/predicates.md (aarch64_any_register_operand): New
4562 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
4563 that are semantically a reverse operation.
4564 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
4565 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
4566 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
4568 (aarch64_can_change_mode_class): For big-endian, forbid changes
4569 between two SVE modes if they have different element sizes.
4571 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4573 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
4574 the TImode handling for big-endian targets.
4576 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4578 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
4579 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
4581 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
4582 Remove BSWAP handing for big-endian targets and use the form of
4583 LD1RQ appropariate for the mode.
4585 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4587 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
4588 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
4591 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4594 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
4595 check for operands that need to go through aarch64_sve_reload_be.
4597 2018-02-01 Jakub Jelinek <jakub@redhat.com>
4599 PR tree-optimization/81661
4600 PR tree-optimization/84117
4601 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
4602 * tree-eh.c: Include gimplify.h.
4603 (find_trapping_overflow, replace_trapping_overflow,
4604 rewrite_to_non_trapping_overflow): New functions.
4605 * tree-vect-loop.c: Include tree-eh.h.
4606 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
4607 * tree-data-ref.c: Include tree-eh.h.
4608 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
4610 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
4612 PR rtl-optimization/84123
4613 * combine.c (change_zero_ext): Check if hard register satisfies
4614 can_change_dest_mode before calling gen_lowpart_SUBREG.
4616 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
4619 * ira.c (ira_init_register_move_cost): Remove assert.
4621 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4623 PR rtl-optimization/84071
4624 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
4625 * doc/tm.texi: Regenerate.
4627 2018-01-31 Richard Biener <rguenther@suse.de>
4629 PR tree-optimization/84132
4630 * tree-data-ref.c (analyze_miv_subscript): Properly
4631 check whether evolution_function_is_affine_multivariate_p
4632 before calling gcd_of_steps_may_divide_p.
4634 2018-01-31 Julia Koval <julia.koval@intel.com>
4637 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
4638 * config/i386/i386.md (rdpid_rex64) New.
4639 (rdpid): Make 32bit only.
4641 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
4644 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
4645 an IDENTIFIER_NODE for FUNCTION_TYPE's.
4647 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4650 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4652 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
4654 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4656 PR rtl-optimization/84071
4657 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
4658 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
4660 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4662 * config/arc/arc.c (arc_handle_aux_attribute): New function.
4663 (arc_attribute_table): Add 'aux' attribute.
4664 (arc_in_small_data_p): Consider aux like variables.
4665 (arc_is_aux_reg_p): New function.
4666 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
4667 (arc_get_aux_arg): New function.
4668 (prepare_move_operands): Handle aux-register access.
4669 (arc_handle_aux_attribute): New function.
4670 * doc/extend.texi (ARC Variable attributes): Add subsection.
4672 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4674 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
4675 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
4676 (arc_attribute_table): Add 'uncached' attribute.
4677 (arc_print_operand): Print '.di' flag for uncached memory
4679 (arc_in_small_data_p): Do not consider for small data the uncached
4681 (arc_is_uncached_mem_p): New function.
4682 * config/arc/predicates.md (compact_store_memory_operand): Check
4683 for uncached memory accesses.
4684 (nonvol_nonimm_operand): Likewise.
4685 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
4687 2018-01-31 Jakub Jelinek <jakub@redhat.com>
4690 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
4691 falign-loops=): Add Optimization flag.
4693 2018-01-30 Jeff Law <law@redhat.com>
4696 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
4697 INT_REGISTERS_SAVED. Check it prior to calling
4698 get_scratch_register_on_entry.
4699 (ix86_adjust_stack_and_probe): Similarly.
4700 (ix86_emit_probe_stack_range): Similarly.
4701 (ix86_expand_prologue): Corresponding changes.
4703 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4706 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
4707 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
4709 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
4712 * lra-constraints.c (curr_insn_transform): Process AND in the
4715 2018-01-30 Jakub Jelinek <jakub@redhat.com>
4717 PR rtl-optimization/83986
4718 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
4719 dependence against last_pending_memory_flush in addition to
4722 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
4724 PR tree-optimization/81611
4725 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
4728 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4731 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
4734 2018-01-30 Richard Biener <rguenther@suse.de>
4735 Jakub Jelinek <jakub@redhat.com>
4737 PR tree-optimization/84111
4738 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
4739 inner loops added during recursion, as they don't have up-to-date
4742 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4745 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
4746 (can_inline_edge_by_limits_p): ... here.
4747 (can_early_inline_edge_p, check_callers,
4748 update_caller_keys, update_callee_keys, recursive_inlining,
4749 add_new_edges_to_heap, speculation_useful_p,
4750 inline_small_functions,
4751 inline_small_functions, flatten_function,
4752 inline_to_all_callers_1): Update.
4754 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4756 * profile-count.c (profile_count::combine_with_ipa_count): Handle
4759 2018-01-30 Richard Biener <rguenther@suse.de>
4761 PR tree-optimization/83008
4762 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
4763 invariant and constant vector uses in stmts when they need
4766 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4769 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
4770 * configure: Regenerate.
4772 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4774 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
4776 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
4777 Use gen_rtx_REG rather than gen_lowpart.
4779 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4781 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
4782 rather than 0 when creating partial subregs.
4784 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4786 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
4789 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
4792 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
4793 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
4794 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
4795 flags. This restores the settings used before the 2017-07-24.
4796 Turning off pre increment/decrement/modify allows IVOPTS to
4797 optimize DF/SF loops where the index is an int.
4799 2018-01-29 Richard Biener <rguenther@suse.de>
4800 Kelvin Nilsen <kelvin@gcc.gnu.org>
4803 * tree-vect-stmts.c (vectorizable_call): Don't call
4804 targetm.vectorize_builtin_md_vectorized_function if callee is
4807 2018-01-22 Carl Love <cel@us.ibm.com>
4809 * doc/extend.tex: Fix typo in second arg in
4810 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
4812 2018-01-29 Richard Biener <rguenther@suse.de>
4814 PR tree-optimization/84086
4815 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
4816 (flush_ssaname_freelist): When SSA names were released reset
4817 the SCEV hash table.
4819 2018-01-29 Richard Biener <rguenther@suse.de>
4821 PR tree-optimization/84057
4822 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
4823 removed paths when removing edges.
4825 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
4827 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
4828 -mfunction-return=@var{choice}.
4830 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
4833 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
4835 (layout::print_source_line): Likewise.
4836 (test_get_line_width_without_trailing_whitespace): Add test cases.
4838 2018-01-27 Jakub Jelinek <jakub@redhat.com>
4841 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
4844 2018-01-26 Jim Wilson <jimw@sifive.com>
4846 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
4848 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
4851 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4853 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
4854 and CMP + SUB-immediate -> SUBS.
4856 2018-01-26 Martin Sebor <msebor@redhat.com>
4858 PR tree-optimization/83896
4859 * tree-ssa-strlen.c (get_string_len): Rename...
4860 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
4861 Avoid assuming length is constant.
4862 (handle_char_store): Use HOST_WIDE_INT for string length.
4864 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
4867 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
4868 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
4870 2018-01-26 Richard Biener <rguenther@suse.de>
4872 PR rtl-optimization/84003
4873 * dse.c (record_store): Only record redundant stores when
4874 the earlier store aliases at least all accesses the later one does.
4876 2018-01-26 Jakub Jelinek <jakub@redhat.com>
4878 PR rtl-optimization/83985
4879 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
4880 REG_CFA_RESTORE insns.
4881 (delete_unmarked_insns): Don't ignore separate shrink wrapping
4882 REG_CFA_RESTORE insns here.
4885 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
4886 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
4888 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4890 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
4891 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
4892 (arc_init): Likewise.
4893 (arc_override_options): Likewise.
4894 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
4896 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
4898 * config/arc/arc.h (TARGET_DBNZ): Define.
4899 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
4900 properly set the tune attribute.
4901 (dbnz): Use TARGET_DBNZ guard.
4902 * config/arc/arc.opt (mtune): Add core3 option.
4904 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4906 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
4907 recognize new pic like addresses.
4908 (arc_delegitimize_address): Clean up.
4910 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4912 * config/arc/arc-arches.def: Option mrf16 valid for all
4914 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
4915 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
4916 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
4917 * config/arc/arc-tables.opt: Regenerate.
4918 * config/arc/arc.c (arc_conditional_register_usage): Handle
4919 reduced register file case.
4920 (arc_file_start): Set must have build attributes.
4921 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
4923 * config/arc/arc.opt (mrf16): Add new option.
4924 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
4925 * config/arc/genmultilib.awk: Handle new mrf16 option.
4926 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
4927 * config/arc/t-multilib: Regenerate.
4928 * doc/invoke.texi (ARC Options): Document mrf16 option.
4930 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4932 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
4933 * config/arc/arc.c (arc_handle_secure_attribute): New function.
4934 (arc_attribute_table): Add 'secure_call' attribute.
4935 (arc_print_operand): Print secure call operand.
4936 (arc_function_ok_for_sibcall): Don't optimize tail calls when
4938 (arc_is_secure_call_p): New function. * config/arc/arc.md
4939 (call_i): Add support for sjli instruction.
4940 (call_value_i): Likewise.
4941 * config/arc/constraints.md (Csc): New constraint.
4943 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4944 John Eric Martin <John.Martin@emmicro-us.com>
4946 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
4947 * config/arc/arc.c (_arc_jli_section): New struct.
4948 (arc_jli_section): New type.
4949 (rc_jli_sections): New static variable.
4950 (arc_handle_jli_attribute): New function.
4951 (arc_attribute_table): Add jli_always and jli_fixed attribute.
4952 (arc_file_end): New function.
4953 (TARGET_ASM_FILE_END): Define.
4954 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
4955 (arc_add_jli_section): New function.
4956 (jli_call_scan): Likewise.
4957 (arc_reorg): Call jli_call_scan.
4958 (arc_output_addsi): Remove 'S' from printing asm operand.
4959 (arc_is_jli_call_p): New function.
4960 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
4962 (movhi_insn): Likewise.
4963 (movsi_insn): Likewise.
4964 (movsi_set_cc_insn): Likewise.
4965 (loadqi_update): Likewise.
4966 (load_zeroextendqisi_update): Likewise.
4967 (load_signextendqisi_update): Likewise.
4968 (loadhi_update): Likewise.
4969 (load_zeroextendhisi_update): Likewise.
4970 (load_signextendhisi_update): Likewise.
4971 (loadsi_update): Likewise.
4972 (loadsf_update): Likewise.
4973 (movsicc_insn): Likewise.
4974 (bset_insn): Likewise.
4975 (bxor_insn): Likewise.
4976 (bclr_insn): Likewise.
4977 (bmsk_insn): Likewise.
4978 (bicsi3_insn): Likewise.
4979 (cmpsi_cc_c_insn): Likewise.
4980 (movsi_ne): Likewise.
4981 (movsi_cond_exec): Likewise.
4982 (clrsbsi2): Likewise.
4993 (call_i): Remove 'S' asm letter, add jli instruction.
4994 (call_value_i): Likewise.
4995 * config/arc/arc.op (mjli-always): New option.
4996 * config/arc/constraints.md (Cji): New constraint.
4997 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
4999 (subsf3_fpx): Likewise.
5000 (mulsf3_fpx): Likewise.
5001 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
5003 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
5005 * doc/invoke.texi (ARC): Document mjli-always option.
5007 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
5009 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
5010 avoid addition with 0 and use incw and decw where possible.
5012 2018-01-26 Richard Biener <rguenther@suse.de>
5014 PR tree-optimization/81082
5015 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
5016 association if it requires casting to unsigned.
5017 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
5018 from fold_plusminus_mult_expr to catch important cases late when
5019 range info is available.
5021 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5023 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
5024 * configure.ac (hidden_linkonce): New test.
5025 * configure: Regenerate.
5026 * config.in: Regenerate.
5028 2018-01-26 Julia Koval <julia.koval@intel.com>
5030 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
5031 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
5032 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
5033 _mm_mask_bitshuffle_epi64_mask): Fix type.
5034 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
5035 USI_FTYPE_V4DI_V4DI_USI): Remove.
5036 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
5037 __builtin_ia32_vpshufbitqmb256_mask,
5038 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
5039 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
5040 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
5042 2018-01-26 Alan Modra <amodra@gmail.com>
5045 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
5046 UNSPEC_VBPERMQ. Sort other unspecs.
5048 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
5050 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
5052 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
5055 * predict.c (drop_profile): Do not push/pop cfun; update also
5057 (handle_missing_profiles): Fix logic looking for zero profiles.
5059 2018-01-25 Jakub Jelinek <jakub@redhat.com>
5062 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
5063 on functions with #pragma omp declare simd or functions with simd
5065 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
5066 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
5067 Remove trailing \n from warning_at calls.
5069 2018-01-25 Tom de Vries <tom@codesourcery.com>
5072 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5073 for neutered workers.
5075 2018-01-24 Joseph Myers <joseph@codesourcery.com>
5078 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
5079 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
5081 2018-01-24 Jeff Law <law@redhat.com>
5084 * i386.c (get_probe_interval): Move to earlier point.
5085 (ix86_compute_frame_layout): If -fstack-clash-protection and
5086 the frame is larger than the probe interval, then use pushes
5087 to save registers rather than reg->mem moves.
5088 (ix86_expand_prologue): Remove conditional for int_registers_saved
5091 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
5094 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
5095 min/max for never referenced object.
5097 2018-01-24 Jakub Jelinek <jakub@redhat.com>
5100 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
5102 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
5103 attributes from DECL_ATTRIBUTES (decl) without affecting
5104 DECL_ATTRIBUTES (current_function_decl).
5105 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
5106 functions with non-NULL DECL_ABSTRACT_ORIGIN.
5108 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
5110 PR tree-optimization/83979
5111 * fold-const.c (fold_comparison): Use constant_boolean_node
5112 instead of boolean_{true,false}_node.
5114 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
5116 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
5119 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5121 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5122 Simplify the clause that sets the length attribute.
5123 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5124 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
5125 clause that sets the length attribute.
5126 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5128 2018-01-24 Tom de Vries <tom@codesourcery.com>
5131 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
5132 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
5133 Add strict parameter.
5134 (prevent_branch_around_nothing): Insert dummy insn between branch to
5135 label and label with no ptx insn inbetween.
5136 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
5138 2018-01-24 Tom de Vries <tom@codesourcery.com>
5141 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5142 for neutered threads in warp.
5143 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
5145 2018-01-24 Richard Biener <rguenther@suse.de>
5147 PR tree-optimization/83176
5148 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
5151 2018-01-24 Richard Biener <rguenther@suse.de>
5153 PR tree-optimization/82819
5154 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
5155 code generating pluses that are no-ops in the target precision.
5157 2018-01-24 Richard Biener <rguenther@suse.de>
5160 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
5162 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5164 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
5165 to merge probabilities.
5166 * predict.c (probably_never_executed): Also mark as cold functions
5167 with global 0 profile and guessed local profile.
5168 * profile-count.c (profile_probability::combine_with_count): New
5170 * profile-count.h (profile_probability::operator*,
5171 profile_probability::operator*=, profile_probability::operator/,
5172 profile_probability::operator/=): Reduce precision to adjusted
5173 and set value to guessed on contradictory divisions.
5174 (profile_probability::combine_with_freq): Remove.
5175 (profile_probability::combine_wiht_count): Declare.
5176 (profile_count::force_nonzero):: Set to adjusted.
5177 (profile_count::probability_in):: Set quality to adjusted.
5178 * tree-ssa-tail-merge.c (replace_block_by): Use
5181 2018-01-23 Andrew Waterman <andrew@sifive.com>
5182 Jim Wilson <jimw@sifive.com>
5184 * config/riscv/riscv.c (riscv_stack_boundary): New.
5185 (riscv_option_override): Set riscv_stack_boundary. Handle
5186 riscv_preferred_stack_boundary_arg.
5187 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
5188 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
5189 (STACK_BOUNDARY): Set to riscv_stack_boundary.
5190 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
5191 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
5192 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
5194 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
5197 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
5198 of struct ix86_frame.
5199 (ix86_expand_epilogue): Likewise. Add a local variable for
5200 the reg_save_offset field in struct ix86_frame.
5202 2018-01-23 Bin Cheng <bin.cheng@arm.com>
5204 PR tree-optimization/82604
5205 * tree-loop-distribution.c (enum partition_kind): New enum item
5206 PKIND_PARTIAL_MEMSET.
5207 (partition_builtin_p): Support above new enum item.
5208 (generate_code_for_partition): Ditto.
5209 (compute_access_range): Differentiate cases that equality can be
5210 proven at all loops, the innermost loops or no loops.
5211 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
5212 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
5213 (finalize_partitions, distribute_loop): Don't fuse partition of
5214 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
5215 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
5218 2018-01-23 Martin Liska <mliska@suse.cz>
5220 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
5221 order to ignore the predictor.
5222 (PRED_POLYMORPHIC_CALL): Likewise.
5223 (PRED_RECURSIVE_CALL): Likewise.
5225 2018-01-23 Martin Liska <mliska@suse.cz>
5227 * tree-profile.c (tree_profiling): Print function header to
5228 aware reader which function we are working on.
5229 * value-prof.c (gimple_find_values_to_profile): Do not print
5230 not interesting value histograms.
5232 2018-01-23 Martin Liska <mliska@suse.cz>
5234 * profile-count.h (enum profile_quality): Add
5235 profile_uninitialized as the first value. Do not number values
5236 as they are zero based.
5237 (profile_count::verify): Update sanity check.
5238 (profile_probability::verify): Likewise.
5240 2018-01-23 Nathan Sidwell <nathan@acm.org>
5242 * doc/invoke.texi (ffor-scope): Deprecate.
5244 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5246 PR tree-optimization/83510
5247 * domwalk.c (set_all_edges_as_executable): New function.
5248 (dom_walker::dom_walker): Convert bool param
5249 "skip_unreachable_blocks" to enum reachability. Move setup of
5250 edge flags to set_all_edges_as_executable and only do it when
5251 reachability is REACHABLE_BLOCKS.
5252 * domwalk.h (enum dom_walker::reachability): New enum.
5253 (dom_walker::dom_walker): Convert bool param
5254 "skip_unreachable_blocks" to enum reachability.
5255 (set_all_edges_as_executable): New decl.
5256 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
5257 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
5259 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
5260 but converting true to REACHABLE_BLOCKS.
5261 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
5263 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
5264 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
5265 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
5267 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
5268 if check_all_array_refs will be called.
5270 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5272 * tree.c (selftest::test_location_wrappers): Add more test
5275 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5277 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
5278 (selftest::test_bit_in_range): Likewise.
5280 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5283 * doc/sourcebuild.texi (vect_float): Say that the selector
5284 only describes the situation when -funsafe-math-optimizations is on.
5285 (vect_float_strict): Document.
5287 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5289 PR tree-optimization/83965
5290 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
5291 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
5292 instead of checking only for a reduction.
5293 (vect_recog_widen_sum_pattern): Likewise.
5295 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5297 * predict.c (probably_never_executed): Only use precise profile info.
5298 (compute_function_frequency): Skip after inlining hack since we now
5299 have quality checking.
5301 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5303 * profile-count.h (profile_probability::very_unlikely,
5304 profile_probability::unlikely, profile_probability::even): Set
5305 precision to guessed.
5307 2018-01-23 Richard Biener <rguenther@suse.de>
5309 PR tree-optimization/83963
5310 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
5311 Properly terminate dominator walk when crossing the exit edge not
5312 when visiting its source block.
5314 2018-01-23 Jakub Jelinek <jakub@redhat.com>
5317 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
5318 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
5320 2018-01-22 Jakub Jelinek <jakub@redhat.com>
5322 PR tree-optimization/83957
5323 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
5324 semicolon after for body surrounded by braces.
5326 PR tree-optimization/83081
5327 * profile-count.h (profile_probability::split): New method.
5328 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
5329 Use profile_probability::split.
5330 (do_compare_rtx_and_jump): Fix adjustment of probabilities
5331 when splitting a single conditional jump into 2.
5333 2018-01-22 David Malcolm <dmalcolm@redhat.com>
5335 PR tree-optimization/69452
5336 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
5339 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5341 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
5342 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
5343 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
5345 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5347 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
5348 * config/rl78/rl78.md (movdi): New define_expand.
5349 * config/rl78/rl78.c (rl78_split_movdi): New function.
5351 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
5354 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
5356 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
5357 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
5358 128-bit to produce an UNSPEC move to get the double word with the
5359 signbit and then a shift directly to do signbit.
5360 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
5361 implementation with a new version that just does either a direct
5362 move or a regular move. Move memory interface to separate insns.
5363 Move insns so they are next to the expander.
5364 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
5365 with signbit move. Split big and little endian case.
5366 (signbit<mode>2_dm_mem_le): Likewise.
5367 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
5368 (signbit<mode>2_dm2): Likewise.
5370 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5372 * config/rl78/rl78.md (anddi3): New define_expand.
5374 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5376 * config/rl78/rl78.md (umindi3): New define_expand.
5378 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5380 * config/rl78/rl78.md (smindi3): New define_expand.
5382 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5384 * config/rl78/rl78.md (smaxdi3): New define_expand.
5386 2018-01-22 Carl Love <cel@us.ibm.com>
5388 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
5389 LVX_V1TI): Add macro expansion.
5390 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
5391 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
5392 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
5393 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
5394 Change check to determine if the instruction is a byte reversing
5395 entry. Fix typo in comment.
5396 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
5397 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
5398 Add def_builtin calls for new builtins.
5399 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
5400 Add define_insn expansion.
5402 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5404 * config/rl78/rl78.md (umaxdi3): New define_expand.
5406 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5408 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
5409 for non-QImode registers.
5411 2018-01-22 Richard Biener <rguenther@suse.de>
5413 PR tree-optimization/83963
5414 * graphite-scop-detection.c (scop_detection::get_sese): Delay
5415 including the loop exit block.
5416 (scop_detection::merge_sese): Likewise.
5417 (scop_detection::add_scop): Do it here instead.
5419 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5421 * doc/sourcebuild.texi (arm_softfloat): Document.
5423 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
5426 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
5427 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
5428 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
5430 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5431 David Edelsohn <dje.gcc@gmail.com>
5434 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5435 Change "crset eq" to "crset 2".
5436 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5437 (*call_indirect_aix<mode>_nospec): Likewise.
5438 (*call_value_indirect_aix<mode>_nospec): Likewise.
5439 (*call_indirect_elfv2<mode>_nospec): Likewise.
5440 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
5441 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
5442 change assembly output from . to $.
5443 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5444 (indirect_jump<mode>_nospec): Change assembly output from . to $.
5445 (*tablejump<mode>_internal1_nospec): Likewise.
5447 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
5450 * config/sh/sh_optimize_sett_clrt.cc:
5451 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
5453 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
5455 PR tree-optimization/83940
5456 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
5457 offset_dt to vect_constant_def rather than vect_unknown_def_type.
5458 (vect_check_load_store_mask): Add a mask_dt_out parameter and
5459 use it to pass back the definition type.
5460 (vect_check_store_rhs): Likewise rhs_dt_out.
5461 (vect_build_gather_load_calls): Add a mask_dt argument and use
5462 it instead of a call to vect_is_simple_use.
5463 (vectorizable_store): Update calls to vect_check_load_store_mask
5464 and vect_check_store_rhs. Use the dt returned by the latter instead
5465 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
5466 instead of calls to vect_is_simple_use. Pass the scalar rather
5467 than the vector operand to vect_is_simple_use when handling
5468 second and subsequent copies of an rhs value.
5469 (vectorizable_load): Update calls to vect_check_load_store_mask
5470 and vect_build_gather_load_calls. Use the cached mask_dt and
5471 gs_info.offset_dt instead of calls to vect_is_simple_use.
5473 2018-01-20 Jakub Jelinek <jakub@redhat.com>
5476 * tree-emutls.c: Include gimplify.h.
5477 (lower_emutls_2): New function.
5478 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
5479 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
5480 it before further processing.
5483 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
5484 UINTVAL (trueop1) instead of INTVAL (op1).
5486 2018-01-19 Jakub Jelinek <jakub@redhat.com>
5490 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
5491 INCOMING_FRAME_SP_OFFSET if not defined.
5492 (scan_trace): Add ENTRY argument. If true and
5493 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
5494 emit a note to adjust the CFA offset.
5495 (create_cfi_notes): Adjust scan_trace callers.
5496 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
5497 INCOMING_FRAME_SP_OFFSET in the CIE.
5498 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
5499 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
5501 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
5502 * doc/tm.texi: Regenerated.
5504 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5506 PR rtl-optimization/83147
5507 * lra-constraints.c (remove_inheritance_pseudos): Use
5508 lra_substitute_pseudo_within_insn.
5510 2018-01-19 Tom de Vries <tom@codesourcery.com>
5511 Cesar Philippidis <cesar@codesourcery.com>
5514 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
5516 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
5519 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
5520 spaces for function labels.
5522 2018-01-19 Martin Liska <mliska@suse.cz>
5524 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
5525 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
5526 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
5527 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
5528 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
5529 (PRED_CONST_RETURN): Change from 69 to 65.
5530 (PRED_NULL_RETURN): Change from 91 to 71.
5531 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
5532 (PRED_LOOP_GUARD): Change from 66 to 73.
5534 2018-01-19 Martin Liska <mliska@suse.cz>
5536 * predict.c (predict_insn_def): Add new assert.
5537 (struct branch_predictor): Change type to signed integer.
5538 (test_prediction_value_range): Amend test to cover
5540 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
5541 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
5542 (PRED_LOOP_ITERATIONS_MAX): Likewise.
5543 (PRED_LOOP_IV_COMPARE): Likewise.
5544 * predict.h (PROB_UNINITIALIZED): Define new constant.
5546 2018-01-19 Martin Liska <mliska@suse.cz>
5548 * predict.c (dump_prediction): Add new format for
5549 analyze_brprob.py script which is enabled with -details
5551 * profile-count.h (precise_p): New function.
5553 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
5555 PR tree-optimization/83922
5556 * tree-vect-loop.c (vect_verify_full_masking): Return false if
5557 there are no statements that need masking.
5558 (vect_active_double_reduction_p): New function.
5559 (vect_analyze_loop_operations): Use it when handling phis that
5560 are not in the loop header.
5562 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
5564 PR tree-optimization/83914
5565 * tree-vect-loop.c (vectorizable_induction): Don't convert
5566 init_expr or apply the peeling adjustment for inductions
5567 that are nested within the vectorized loop.
5569 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5571 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
5574 2018-01-18 Jakub Jelinek <jakub@redhat.com>
5578 * function.h (gimplify_parameters): Add gimple_seq * argument.
5579 * function.c: Include gimple.h and options.h.
5580 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
5581 for the added local temporaries if needed.
5582 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
5583 if there are any parameter cleanups, wrap whole body into a
5584 try/finally with the cleanups.
5586 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
5589 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
5590 Use GET_MODE_CLASS for scalar floating point.
5592 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
5596 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
5597 Fix call of call_cgraph_insertion_hooks.
5599 2018-01-18 Martin Sebor <msebor@redhat.com>
5601 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
5603 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
5606 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
5609 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
5612 * common.opt: (-ffile-prefix-map): New option.
5613 * opts.c (common_handle_option): Defer it.
5614 * opts-global.c (handle_common_deferred_options): Handle it.
5615 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
5616 * file-prefix-map.h: New file.
5617 (remap_debug_filename, add_debug_prefix_map): ...here.
5618 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
5619 * final.c (debug_prefix_map, add_debug_prefix_map
5620 remap_debug_filename): Move to...
5621 * file-prefix-map.c: New file.
5622 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
5623 generalize, get rid of alloca(), use strrchr() instead of strchr().
5624 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
5625 Implement in terms of add_prefix_map().
5626 (remap_macro_filename, remap_debug_filename): Implement in term of
5628 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
5629 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
5630 * dbxout.c: Include file-prefix-map.h.
5631 * varasm.c: Likewise.
5632 * vmsdbgout.c: Likewise.
5633 * xcoffout.c: Likewise.
5634 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
5635 * doc/cppopts.texi (-fmacro-prefix-map): Document.
5636 * doc/invoke.texi (-ffile-prefix-map): Document.
5637 (-fdebug-prefix-map): Update description.
5639 2018-01-18 Martin Liska <mliska@suse.cz>
5641 * config/i386/i386.c (indirect_thunk_name): Document that also
5643 (output_indirect_thunk): Document why both instructions
5644 (pause and lfence) are generated.
5646 2018-01-18 Richard Biener <rguenther@suse.de>
5648 PR tree-optimization/83887
5649 * graphite-scop-detection.c
5650 (scop_detection::get_nearest_dom_with_single_entry): Remove.
5651 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
5652 (scop_detection::merge_sese): Re-implement with a flood-fill
5653 algorithm that properly finds a SESE region if it exists.
5655 2018-01-18 Jakub Jelinek <jakub@redhat.com>
5658 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
5659 pointer_diff optimizations use view_convert instead of convert.
5661 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5663 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5664 Generate different code for -mno-speculate-indirect-jumps.
5665 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5666 (*call_indirect_aix<mode>): Disable for
5667 -mno-speculate-indirect-jumps.
5668 (*call_indirect_aix<mode>_nospec): New define_insn.
5669 (*call_value_indirect_aix<mode>): Disable for
5670 -mno-speculate-indirect-jumps.
5671 (*call_value_indirect_aix<mode>_nospec): New define_insn.
5672 (*sibcall_nonlocal_sysv<mode>): Generate different code for
5673 -mno-speculate-indirect-jumps.
5674 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5676 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
5678 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
5679 long double type, set the flags for noting the default long double
5680 type, even if we don't pass or return a long double type.
5682 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
5685 * ipa-inline.c (flatten_function): Do not overwrite final inlining
5688 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
5690 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
5691 support for merge[hl].
5692 (fold_mergehl_helper): New helper function.
5693 (tree-vector-builder.h): New #include for tree_vector_builder usage.
5694 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
5695 (altivec_vmrglw_direct): Add xxmrglw insn.
5697 2018-01-17 Andrew Waterman <andrew@sifive.com>
5699 * config/riscv/riscv.c (riscv_conditional_register_usage): If
5700 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
5702 2018-01-17 David Malcolm <dmalcolm@redhat.com>
5705 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
5706 call the lto_location_cache before reading the
5707 DECL_SOURCE_LOCATION of the types.
5709 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
5710 Richard Sandiford <richard.sandiford@linaro.org>
5712 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
5713 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
5714 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
5715 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
5716 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
5718 * config/aarch64/constraints.md (aarch64_movti_operand):
5720 * config/aarch64/predicates.md (Uti): Add new constraint.
5722 2018-01-17 Carl Love <cel@us.ibm.com>
5724 * config/rs6000/vsx.md (define_expand xl_len_r,
5725 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
5726 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
5728 (define_expand, define_insn): Move the shift left from the
5729 define_insn to the define_expand for lxvl and stxvl instructions.
5730 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
5731 and XL_LEN_R definitions to PURE.
5733 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
5735 * config/i386/i386.c (indirect_thunk_name): Declare regno
5736 as unsigned int. Compare regno with INVALID_REGNUM.
5737 (output_indirect_thunk): Ditto.
5738 (output_indirect_thunk_function): Ditto.
5739 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
5740 in the call to output_indirect_thunk_function.
5742 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
5745 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
5746 rather than the size of inner_type to determine the stack slot size
5747 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
5749 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
5752 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
5755 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
5757 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
5758 endian Linux systems to optionally enable multilibs for selecting
5759 the long double type if the user configured an explicit type.
5760 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
5761 have no long double multilibs if not defined.
5762 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
5763 warn if the user used -mabi={ieee,ibm}longdouble and we built
5764 multilibs for long double.
5765 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
5766 appropriate multilib option.
5767 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
5769 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
5770 for building long double multilibs.
5771 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
5773 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
5775 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
5778 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
5780 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
5783 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
5786 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
5789 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5791 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
5792 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
5794 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
5796 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
5797 different rtl trees depending on TARGET_64BIT.
5798 (rs6000_gen_lvx): Likewise.
5800 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5802 * config/visium/visium.md (nop): Tweak comment.
5803 (hazard_nop): Likewise.
5805 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5807 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
5808 -mspeculate-indirect-jumps.
5809 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
5810 for -mno-speculate-indirect-jumps.
5811 (*call_indirect_elfv2<mode>_nospec): New define_insn.
5812 (*call_value_indirect_elfv2<mode>): Disable for
5813 -mno-speculate-indirect-jumps.
5814 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
5815 (indirect_jump): Emit different RTL for
5816 -mno-speculate-indirect-jumps.
5817 (*indirect_jump<mode>): Disable for
5818 -mno-speculate-indirect-jumps.
5819 (*indirect_jump<mode>_nospec): New define_insn.
5820 (tablejump): Emit different RTL for
5821 -mno-speculate-indirect-jumps.
5822 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
5823 (tablejumpsi_nospec): New define_expand.
5824 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
5825 (tablejumpdi_nospec): New define_expand.
5826 (*tablejump<mode>_internal1): Disable for
5827 -mno-speculate-indirect-jumps.
5828 (*tablejump<mode>_internal1_nospec): New define_insn.
5829 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
5832 2018-01-16 Artyom Skrobov tyomitch@gmail.com
5834 * caller-save.c (insert_save): Drop unnecessary parameter. All
5837 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5838 Richard Biener <rguenth@suse.de>
5841 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
5842 return early, inline manually is_gimple_sizepos. Make sure if we
5843 call gimplify_expr we don't end up with a gimple constant.
5844 * tree.c (variably_modified_type_p): Don't return true for
5845 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
5846 * gimplify.h (is_gimple_sizepos): Remove.
5848 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5850 PR tree-optimization/83857
5851 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
5852 vectorizable_live_operation for pure SLP statements.
5853 (vectorizable_live_operation): Handle PHIs.
5855 2018-01-16 Richard Biener <rguenther@suse.de>
5857 PR tree-optimization/83867
5858 * tree-vect-stmts.c (vect_transform_stmt): Precompute
5859 nested_in_vect_loop_p since the scalar stmt may get invalidated.
5861 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5864 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
5865 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
5866 If off is not INTEGER_CST, issue a may not be aligned warning
5867 rather than isn't aligned. Use isn%'t rather than isn't.
5868 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
5870 <case MULT_EXPR>: Improve the case when bottom and one of the
5871 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
5872 operand, in that case check if the other operand is multiple of
5873 bottom divided by the INTEGER_CST operand.
5875 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5878 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
5879 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
5880 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
5881 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
5882 * config/pa/pa.c (pa_function_arg_advance): Likewise.
5883 (pa_function_arg, pa_arg_partial_bytes): Likewise.
5884 (pa_function_arg_size): New function.
5886 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5888 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
5889 in a separate statement.
5891 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5893 PR tree-optimization/83847
5894 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
5895 group gathers and scatters.
5897 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5899 PR rtl-optimization/86620
5900 * params.def (max-sched-ready-insns): Bump minimum value to 1.
5902 PR rtl-optimization/83213
5903 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
5904 to last if both are JUMP_INSNs.
5906 PR tree-optimization/83843
5907 * gimple-ssa-store-merging.c
5908 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
5909 store_immediate_info for bswap/nop orig_stores.
5911 2018-01-15 Andrew Waterman <andrew@sifive.com>
5913 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
5915 <UDIV>: Increase cost if !TARGET_DIV.
5917 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
5919 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
5920 (define_attr "cr_logical_3op"): New.
5921 (cceq_ior_compare): Adjust.
5922 (cceq_ior_compare_complement): Adjust.
5923 (*cceq_rev_compare): Adjust.
5924 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
5925 (is_cracked_insn): Adjust.
5926 (insn_must_be_first_in_group): Adjust.
5927 * config/rs6000/40x.md: Adjust.
5928 * config/rs6000/440.md: Adjust.
5929 * config/rs6000/476.md: Adjust.
5930 * config/rs6000/601.md: Adjust.
5931 * config/rs6000/603.md: Adjust.
5932 * config/rs6000/6xx.md: Adjust.
5933 * config/rs6000/7450.md: Adjust.
5934 * config/rs6000/7xx.md: Adjust.
5935 * config/rs6000/8540.md: Adjust.
5936 * config/rs6000/cell.md: Adjust.
5937 * config/rs6000/e300c2c3.md: Adjust.
5938 * config/rs6000/e500mc.md: Adjust.
5939 * config/rs6000/e500mc64.md: Adjust.
5940 * config/rs6000/e5500.md: Adjust.
5941 * config/rs6000/e6500.md: Adjust.
5942 * config/rs6000/mpc.md: Adjust.
5943 * config/rs6000/power4.md: Adjust.
5944 * config/rs6000/power5.md: Adjust.
5945 * config/rs6000/power6.md: Adjust.
5946 * config/rs6000/power7.md: Adjust.
5947 * config/rs6000/power8.md: Adjust.
5948 * config/rs6000/power9.md: Adjust.
5949 * config/rs6000/rs64.md: Adjust.
5950 * config/rs6000/titan.md: Adjust.
5952 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5954 * config/i386/predicates.md (indirect_branch_operand): Rewrite
5955 ix86_indirect_branch_register logic.
5957 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5959 * config/i386/constraints.md (Bs): Update
5960 ix86_indirect_branch_register check. Don't check
5961 ix86_indirect_branch_register with GOT_memory_operand.
5963 * config/i386/predicates.md (GOT_memory_operand): Don't check
5964 ix86_indirect_branch_register here.
5965 (GOT32_symbol_operand): Likewise.
5967 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5969 * config/i386/predicates.md (constant_call_address_operand):
5970 Rewrite ix86_indirect_branch_register logic.
5971 (sibcall_insn_operand): Likewise.
5973 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5975 * config/i386/constraints.md (Bs): Replace
5976 ix86_indirect_branch_thunk_register with
5977 ix86_indirect_branch_register.
5979 * config/i386/i386.md (indirect_jump): Likewise.
5980 (tablejump): Likewise.
5981 (*sibcall_memory): Likewise.
5982 (*sibcall_value_memory): Likewise.
5983 Peepholes of indirect call and jump via memory: Likewise.
5984 * config/i386/i386.opt: Likewise.
5985 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5986 (GOT_memory_operand): Likewise.
5987 (call_insn_operand): Likewise.
5988 (sibcall_insn_operand): Likewise.
5989 (GOT32_symbol_operand): Likewise.
5991 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5994 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
5995 type rather than type addr's type points to.
5996 (expand_omp_atomic_mutex): Likewise.
5997 (expand_omp_atomic): Likewise.
5999 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6002 * config/i386/i386.c (output_indirect_thunk_function): Use
6003 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
6004 for __x86_return_thunk.
6006 2018-01-15 Richard Biener <rguenther@suse.de>
6009 * expmed.c (extract_bit_field_1): Fix typo.
6011 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6014 * config/arm/iterators.md (VF): New mode iterator.
6015 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
6016 Remove integer-related logic from pattern.
6017 (neon_vabd<mode>_3): Likewise.
6019 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6022 * common.opt (fstrict-overflow): No longer an alias.
6023 (fwrapv-pointer): New option.
6024 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
6025 also for pointer types based on flag_wrapv_pointer.
6026 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
6027 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
6028 opts->x_flag_wrapv got set.
6029 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
6030 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
6031 POINTER_TYPE_OVERFLOW_UNDEFINED.
6032 * match.pd: Likewise in address comparison pattern.
6033 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
6035 2018-01-15 Richard Biener <rguenther@suse.de>
6038 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
6039 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
6040 Reset type names to their identifier if their TYPE_DECL doesn't
6041 have linkage (and thus is used for ODR and devirt).
6042 (save_debug_info_for_decl): Remove.
6043 (save_debug_info_for_type): Likewise.
6044 (add_tree_to_fld_list): Adjust.
6045 * tree-pretty-print.c (dump_generic_node): Make dumping of
6046 type names more robust.
6048 2018-01-15 Richard Biener <rguenther@suse.de>
6050 * BASE-VER: Bump to 8.0.1.
6052 2018-01-14 Martin Sebor <msebor@redhat.com>
6055 * builtins.c (check_access): Avoid warning when the no-warning bit
6058 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
6060 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
6061 * ira-color (allocno_hard_regs_compare): Likewise.
6063 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
6066 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
6067 Use .pushsection/.popsection.
6069 2018-01-14 Martin Sebor <msebor@redhat.com>
6072 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
6074 2018-01-14 Jakub Jelinek <jakub@redhat.com>
6076 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
6077 entry from extra_headers.
6078 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
6079 extra_headers, make the list bitwise identical to the i?86-*-* one.
6081 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6083 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
6084 -mcmodel=large with -mindirect-branch=thunk,
6085 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
6086 -mfunction-return=thunk-extern.
6087 * doc/invoke.texi: Document -mcmodel=large is incompatible with
6088 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
6089 -mfunction-return=thunk and -mfunction-return=thunk-extern.
6091 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6093 * config/i386/i386.c (print_reg): Print the name of the full
6094 integer register without '%'.
6095 (ix86_print_operand): Handle 'V'.
6096 * doc/extend.texi: Document 'V' modifier.
6098 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6100 * config/i386/constraints.md (Bs): Disallow memory operand for
6101 -mindirect-branch-register.
6103 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6104 (GOT_memory_operand): Likewise.
6105 (call_insn_operand): Likewise.
6106 (sibcall_insn_operand): Likewise.
6107 (GOT32_symbol_operand): Likewise.
6108 * config/i386/i386.md (indirect_jump): Call convert_memory_address
6109 for -mindirect-branch-register.
6110 (tablejump): Likewise.
6111 (*sibcall_memory): Likewise.
6112 (*sibcall_value_memory): Likewise.
6113 Disallow peepholes of indirect call and jump via memory for
6114 -mindirect-branch-register.
6115 (*call_pop): Replace m with Bw.
6116 (*call_value_pop): Likewise.
6117 (*sibcall_pop_memory): Replace m with Bs.
6118 * config/i386/i386.opt (mindirect-branch-register): New option.
6119 * doc/invoke.texi: Document -mindirect-branch-register option.
6121 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6123 * config/i386/i386-protos.h (ix86_output_function_return): New.
6124 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
6125 set function_return_type.
6126 (indirect_thunk_name): Add ret_p to indicate thunk for function
6128 (output_indirect_thunk_function): Pass false to
6129 indirect_thunk_name.
6130 (ix86_output_indirect_branch_via_reg): Likewise.
6131 (ix86_output_indirect_branch_via_push): Likewise.
6132 (output_indirect_thunk_function): Create alias for function
6133 return thunk if regno < 0.
6134 (ix86_output_function_return): New function.
6135 (ix86_handle_fndecl_attribute): Handle function_return.
6136 (ix86_attribute_table): Add function_return.
6137 * config/i386/i386.h (machine_function): Add
6138 function_return_type.
6139 * config/i386/i386.md (simple_return_internal): Use
6140 ix86_output_function_return.
6141 (simple_return_internal_long): Likewise.
6142 * config/i386/i386.opt (mfunction-return=): New option.
6143 (indirect_branch): Mention -mfunction-return=.
6144 * doc/extend.texi: Document function_return function attribute.
6145 * doc/invoke.texi: Document -mfunction-return= option.
6147 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6149 * config/i386/i386-opts.h (indirect_branch): New.
6150 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
6151 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
6152 with local indirect jump when converting indirect call and jump.
6153 (ix86_set_indirect_branch_type): New.
6154 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
6155 (indirectlabelno): New.
6156 (indirect_thunk_needed): Likewise.
6157 (indirect_thunk_bnd_needed): Likewise.
6158 (indirect_thunks_used): Likewise.
6159 (indirect_thunks_bnd_used): Likewise.
6160 (INDIRECT_LABEL): Likewise.
6161 (indirect_thunk_name): Likewise.
6162 (output_indirect_thunk): Likewise.
6163 (output_indirect_thunk_function): Likewise.
6164 (ix86_output_indirect_branch_via_reg): Likewise.
6165 (ix86_output_indirect_branch_via_push): Likewise.
6166 (ix86_output_indirect_branch): Likewise.
6167 (ix86_output_indirect_jmp): Likewise.
6168 (ix86_code_end): Call output_indirect_thunk_function if needed.
6169 (ix86_output_call_insn): Call ix86_output_indirect_branch if
6171 (ix86_handle_fndecl_attribute): Handle indirect_branch.
6172 (ix86_attribute_table): Add indirect_branch.
6173 * config/i386/i386.h (machine_function): Add indirect_branch_type
6174 and has_local_indirect_jump.
6175 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
6177 (tablejump): Likewise.
6178 (*indirect_jump): Use ix86_output_indirect_jmp.
6179 (*tablejump_1): Likewise.
6180 (simple_return_indirect_internal): Likewise.
6181 * config/i386/i386.opt (mindirect-branch=): New option.
6182 (indirect_branch): New.
6185 (thunk-inline): Likewise.
6186 (thunk-extern): Likewise.
6187 * doc/extend.texi: Document indirect_branch function attribute.
6188 * doc/invoke.texi: Document -mindirect-branch= option.
6190 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
6193 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
6195 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
6197 * ipa-inline.c (want_inline_small_function_p): Return false if
6198 inlining has already failed with CIF_FINAL_ERROR.
6199 (update_caller_keys): Call want_inline_small_function_p before
6201 (update_callee_keys): Likewise.
6203 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6205 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
6207 (rs6000_quadword_masked_address_p): Likewise.
6208 (quad_aligned_load_p): Likewise.
6209 (quad_aligned_store_p): Likewise.
6210 (const_load_sequence_p): Add comment to describe the outer-most loop.
6211 (mimic_memory_attributes_and_flags): New function.
6212 (rs6000_gen_stvx): Likewise.
6213 (replace_swapped_aligned_store): Likewise.
6214 (rs6000_gen_lvx): Likewise.
6215 (replace_swapped_aligned_load): Likewise.
6216 (replace_swapped_load_constant): Capitalize argument name in
6217 comment describing this function.
6218 (rs6000_analyze_swaps): Add a third pass to search for vector loads
6219 and stores that access quad-word aligned addresses and replace
6220 with stvx or lvx instructions when appropriate.
6221 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
6222 New function prototype.
6223 (rs6000_quadword_masked_address_p): Likewise.
6224 (rs6000_gen_lvx): Likewise.
6225 (rs6000_gen_stvx): Likewise.
6226 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
6227 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
6228 when memory address is aligned.
6229 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
6230 this split to select lvx instruction when memory address is aligned.
6231 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
6232 instruction when memory address is aligned.
6233 (*vsx_le_perm_load_v16qi): Likewise.
6234 (four unnamed splitters): Modify to select the stvx instruction
6235 when memory is aligned.
6237 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
6239 * predict.c (determine_unlikely_bbs): Handle correctly BBs
6240 which appears in the queue multiple times.
6242 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6243 Alan Hayward <alan.hayward@arm.com>
6244 David Sherwood <david.sherwood@arm.com>
6246 * tree-vectorizer.h (vec_lower_bound): New structure.
6247 (_loop_vec_info): Add check_nonzero and lower_bounds.
6248 (LOOP_VINFO_CHECK_NONZERO): New macro.
6249 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
6250 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
6251 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
6252 fields. Make seg_len the distance travelled, not including the
6254 (dr_direction_indicator): Declare.
6255 (dr_zero_step_indicator): Likewise.
6256 (dr_known_forward_stride_p): Likewise.
6257 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
6259 (runtime_alias_check_p): Allow runtime alias checks with
6261 (operator ==): Compare access_size and align.
6262 (prune_runtime_alias_test_list): Rework for new distinction between
6263 the access_size and seg_len.
6264 (create_intersect_range_checks_index): Likewise. Cope with polynomial
6266 (get_segment_min_max): New function.
6267 (create_intersect_range_checks): Use it.
6268 (dr_step_indicator): New function.
6269 (dr_direction_indicator): Likewise.
6270 (dr_zero_step_indicator): Likewise.
6271 (dr_known_forward_stride_p): Likewise.
6272 * tree-loop-distribution.c (data_ref_segment_size): Return
6273 DR_STEP * (niters - 1).
6274 (compute_alias_check_pairs): Update call to the dr_with_seg_len
6276 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
6277 (vect_preserves_scalar_order_p): New function, split out from...
6278 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
6279 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
6280 (vect_vfa_access_size): New function.
6281 (vect_vfa_align): Likewise.
6282 (vect_compile_time_alias): Take access_size_a and access_b arguments.
6283 (dump_lower_bound): New function.
6284 (vect_check_lower_bound): Likewise.
6285 (vect_small_gap_p): Likewise.
6286 (vectorizable_with_step_bound_p): Likewise.
6287 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
6288 depencies if the vectorization factor is 1. Convert the checks
6289 for nonzero steps into checks on the bounds of DR_STEP. Try using
6290 a bunds check for variable steps if the minimum required step is
6291 relatively small. Update calls to the dr_with_seg_len
6292 constructor and to vect_compile_time_alias.
6293 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
6295 (vect_loop_versioning): Call it.
6296 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
6298 (vect_estimate_min_profitable_iters): Account for any bounds checks.
6300 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6301 Alan Hayward <alan.hayward@arm.com>
6302 David Sherwood <david.sherwood@arm.com>
6304 * doc/sourcebuild.texi (vect_scatter_store): Document.
6305 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
6307 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
6309 * genopinit.c (main): Add supports_vec_scatter_store and
6310 supports_vec_scatter_store_cached to target_optabs.
6311 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
6312 IFN_MASK_SCATTER_STORE.
6313 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
6315 * internal-fn.h (internal_store_fn_p): Declare.
6316 (internal_fn_stored_value_index): Likewise.
6317 * internal-fn.c (scatter_store_direct): New macro.
6318 (expand_scatter_store_optab_fn): New function.
6319 (direct_scatter_store_optab_supported_p): New macro.
6320 (internal_store_fn_p): New function.
6321 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
6322 IFN_MASK_SCATTER_STORE.
6323 (internal_fn_mask_index): Likewise.
6324 (internal_fn_stored_value_index): New function.
6325 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
6327 * optabs-query.h (supports_vec_scatter_store_p): Declare.
6328 * optabs-query.c (supports_vec_scatter_store_p): New function.
6329 * tree-vectorizer.h (vect_get_store_rhs): Declare.
6330 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
6331 true for scatter stores.
6332 (vect_gather_scatter_fn_p): Handle scatter stores too.
6333 (vect_check_gather_scatter): Consider using scatter stores if
6334 supports_vec_scatter_store_p.
6335 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
6337 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6338 internal_fn_stored_value_index.
6339 (check_load_store_masking): Handle scatter stores too.
6340 (vect_get_store_rhs): Make public.
6341 (vectorizable_call): Use internal_store_fn_p.
6342 (vectorizable_store): Handle scatter store internal functions.
6343 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
6344 when deciding whether the end of the group has been reached.
6345 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
6346 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
6347 (mask_scatter_store<mode>): New insns.
6349 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6350 Alan Hayward <alan.hayward@arm.com>
6351 David Sherwood <david.sherwood@arm.com>
6353 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
6354 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
6355 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
6357 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
6358 Use vect_truncate_gather_scatter_offset if we can't treat the
6359 operation as a normal gather load or scatter store.
6360 (get_group_load_store_type): Take the gather_scatter_info
6361 as argument. Try using a gather load or scatter store for
6362 single-element groups.
6363 (get_load_store_type): Update calls to get_group_load_store_type
6364 and vect_use_strided_gather_scatters_p.
6366 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6367 Alan Hayward <alan.hayward@arm.com>
6368 David Sherwood <david.sherwood@arm.com>
6370 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
6371 optional tree argument.
6372 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
6374 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
6375 but continue to use the current value as a fallback.
6376 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
6377 to compare the updates.
6378 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
6379 (get_load_store_type): Use it when handling a strided access.
6380 (vect_get_strided_load_store_ops): New function.
6381 (vect_get_data_ptr_increment): Likewise.
6382 (vectorizable_load): Handle strided gather loads. Always pass
6383 a step to vect_create_data_ref_ptr and bump_vector_ptr.
6385 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6386 Alan Hayward <alan.hayward@arm.com>
6387 David Sherwood <david.sherwood@arm.com>
6389 * doc/md.texi (gather_load@var{m}): Document.
6390 (mask_gather_load@var{m}): Likewise.
6391 * genopinit.c (main): Add supports_vec_gather_load and
6392 supports_vec_gather_load_cached to target_optabs.
6393 * optabs-tree.c (init_tree_optimization_optabs): Use
6394 ggc_cleared_alloc to allocate target_optabs.
6395 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
6396 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
6398 * internal-fn.h (internal_load_fn_p): Declare.
6399 (internal_gather_scatter_fn_p): Likewise.
6400 (internal_fn_mask_index): Likewise.
6401 (internal_gather_scatter_fn_supported_p): Likewise.
6402 * internal-fn.c (gather_load_direct): New macro.
6403 (expand_gather_load_optab_fn): New function.
6404 (direct_gather_load_optab_supported_p): New macro.
6405 (direct_internal_fn_optab): New function.
6406 (internal_load_fn_p): Likewise.
6407 (internal_gather_scatter_fn_p): Likewise.
6408 (internal_fn_mask_index): Likewise.
6409 (internal_gather_scatter_fn_supported_p): Likewise.
6410 * optabs-query.c (supports_at_least_one_mode_p): New function.
6411 (supports_vec_gather_load_p): Likewise.
6412 * optabs-query.h (supports_vec_gather_load_p): Declare.
6413 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
6414 and memory_type field.
6415 (NUM_PATTERNS): Bump to 15.
6416 * tree-vect-data-refs.c: Include internal-fn.h.
6417 (vect_gather_scatter_fn_p): New function.
6418 (vect_describe_gather_scatter_call): Likewise.
6419 (vect_check_gather_scatter): Try using internal functions for
6420 gather loads. Recognize existing calls to a gather load function.
6421 (vect_analyze_data_refs): Consider using gather loads if
6422 supports_vec_gather_load_p.
6423 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
6424 (vect_get_gather_scatter_offset_type): Likewise.
6425 (vect_convert_mask_for_vectype): Likewise.
6426 (vect_add_conversion_to_patterm): Likewise.
6427 (vect_try_gather_scatter_pattern): Likewise.
6428 (vect_recog_gather_scatter_pattern): New pattern recognizer.
6429 (vect_vect_recog_func_ptrs): Add it.
6430 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6431 internal_fn_mask_index and internal_gather_scatter_fn_p.
6432 (check_load_store_masking): Take the gather_scatter_info as an
6433 argument and handle gather loads.
6434 (vect_get_gather_scatter_ops): New function.
6435 (vectorizable_call): Check internal_load_fn_p.
6436 (vectorizable_load): Likewise. Handle gather load internal
6438 (vectorizable_store): Update call to check_load_store_masking.
6439 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
6440 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
6441 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
6442 (aarch64_gather_scale_operand_d): New predicates.
6443 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
6444 (mask_gather_load<mode>): New insns.
6446 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6447 Alan Hayward <alan.hayward@arm.com>
6448 David Sherwood <david.sherwood@arm.com>
6450 * optabs.def (fold_left_plus_optab): New optab.
6451 * doc/md.texi (fold_left_plus_@var{m}): Document.
6452 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
6453 * internal-fn.c (fold_left_direct): Define.
6454 (expand_fold_left_optab_fn): Likewise.
6455 (direct_fold_left_optab_supported_p): Likewise.
6456 * fold-const-call.c (fold_const_fold_left): New function.
6457 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
6458 * tree-parloops.c (valid_reduction_p): New function.
6459 (gather_scalar_reductions): Use it.
6460 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
6461 (vect_finish_replace_stmt): Declare.
6462 * tree-vect-loop.c (fold_left_reduction_fn): New function.
6463 (needs_fold_left_reduction_p): New function, split out from...
6464 (vect_is_simple_reduction): ...here. Accept reductions that
6465 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
6466 (vect_force_simple_reduction): Also store the reduction type in
6467 the assignment's STMT_VINFO_REDUC_TYPE.
6468 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
6469 (merge_with_identity): New function.
6470 (vect_expand_fold_left): Likewise.
6471 (vectorize_fold_left_reduction): Likewise.
6472 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
6473 scalar phi in place for it. Check for target support and reject
6474 cases that would reassociate the operation. Defer the transform
6475 phase to vectorize_fold_left_reduction.
6476 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
6477 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
6478 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
6480 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6482 * tree-if-conv.c (predicate_mem_writes): Remove redundant
6483 call to ifc_temp_var.
6485 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6486 Alan Hayward <alan.hayward@arm.com>
6487 David Sherwood <david.sherwood@arm.com>
6489 * target.def (legitimize_address_displacement): Take the original
6490 offset as a poly_int.
6491 * targhooks.h (default_legitimize_address_displacement): Update
6493 * targhooks.c (default_legitimize_address_displacement): Likewise.
6494 * doc/tm.texi: Regenerate.
6495 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
6496 as an argument, moving assert of ad->disp == ad->disp_term to...
6497 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
6498 Try calling targetm.legitimize_address_displacement before expanding
6499 the address rather than afterwards, and adjust for the new interface.
6500 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
6501 Match the new hook interface. Handle SVE addresses.
6502 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
6505 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6507 * Makefile.in (OBJS): Add early-remat.o.
6508 * target.def (select_early_remat_modes): New hook.
6509 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
6510 * doc/tm.texi: Regenerate.
6511 * targhooks.h (default_select_early_remat_modes): Declare.
6512 * targhooks.c (default_select_early_remat_modes): New function.
6513 * timevar.def (TV_EARLY_REMAT): New timevar.
6514 * passes.def (pass_early_remat): New pass.
6515 * tree-pass.h (make_pass_early_remat): Declare.
6516 * early-remat.c: New file.
6517 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
6519 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
6521 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6522 Alan Hayward <alan.hayward@arm.com>
6523 David Sherwood <david.sherwood@arm.com>
6525 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
6526 vfm1 with a bound_epilog parameter.
6527 (vect_do_peeling): Update calls accordingly, and move the prologue
6528 call earlier in the function. Treat the base bound_epilog as 0 for
6529 fully-masked loops and retain vf - 1 for other loops. Add 1 to
6530 this base when peeling for gaps.
6531 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
6532 with fully-masked loops.
6533 (vect_estimate_min_profitable_iters): Handle the single peeled
6534 iteration in that case.
6536 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6537 Alan Hayward <alan.hayward@arm.com>
6538 David Sherwood <david.sherwood@arm.com>
6540 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
6541 single-element interleaving even if the size is not a power of 2.
6542 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
6543 accesses for single-element interleaving if the group size is
6546 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6547 Alan Hayward <alan.hayward@arm.com>
6548 David Sherwood <david.sherwood@arm.com>
6550 * doc/md.texi (fold_extract_last_@var{m}): Document.
6551 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
6552 * optabs.def (fold_extract_last_optab): New optab.
6553 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
6554 * internal-fn.c (fold_extract_direct): New macro.
6555 (expand_fold_extract_optab_fn): Likewise.
6556 (direct_fold_extract_optab_supported_p): Likewise.
6557 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
6558 * tree-vect-loop.c (vect_model_reduction_cost): Handle
6559 EXTRACT_LAST_REDUCTION.
6560 (get_initial_def_for_reduction): Do not create an initial vector
6561 for EXTRACT_LAST_REDUCTION reductions.
6562 (vectorizable_reduction): Leave the scalar phi in place for
6563 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
6564 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
6565 epilogue code for EXTRACT_LAST_REDUCTION and defer the
6566 transform phase to vectorizable_condition.
6567 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
6569 (vect_finish_stmt_generation): ...here.
6570 (vect_finish_replace_stmt): New function.
6571 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
6572 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
6574 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
6576 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6577 Alan Hayward <alan.hayward@arm.com>
6578 David Sherwood <david.sherwood@arm.com>
6580 * doc/md.texi (extract_last_@var{m}): Document.
6581 * optabs.def (extract_last_optab): New optab.
6582 * internal-fn.def (EXTRACT_LAST): New internal function.
6583 * internal-fn.c (cond_unary_direct): New macro.
6584 (expand_cond_unary_optab_fn): Likewise.
6585 (direct_cond_unary_optab_supported_p): Likewise.
6586 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
6587 loops using EXTRACT_LAST.
6588 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
6589 (extract_last_<mode>): ...this optab.
6590 (vec_extract<mode><Vel>): Update accordingly.
6592 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6593 Alan Hayward <alan.hayward@arm.com>
6594 David Sherwood <david.sherwood@arm.com>
6596 * target.def (empty_mask_is_expensive): New hook.
6597 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
6598 * doc/tm.texi: Regenerate.
6599 * targhooks.h (default_empty_mask_is_expensive): Declare.
6600 * targhooks.c (default_empty_mask_is_expensive): New function.
6601 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
6602 if the target says that empty masks are expensive.
6603 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
6605 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
6607 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6608 Alan Hayward <alan.hayward@arm.com>
6609 David Sherwood <david.sherwood@arm.com>
6611 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
6612 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
6613 (vect_use_loop_mask_for_alignment_p): New function.
6614 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
6615 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
6616 niters_skip argument. Make sure that the first niters_skip elements
6617 of the first iteration are inactive.
6618 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
6619 Update call to vect_set_loop_masks_directly.
6620 (get_misalign_in_elems): New function, split out from...
6621 (vect_gen_prolog_loop_niters): ...here.
6622 (vect_update_init_of_dr): Take a code argument that specifies whether
6623 the adjustment should be added or subtracted.
6624 (vect_update_init_of_drs): Likewise.
6625 (vect_prepare_for_masked_peels): New function.
6626 (vect_do_peeling): Skip prologue peeling if we're using a mask
6627 instead. Update call to vect_update_inits_of_drs.
6628 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6630 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
6631 alignment. Do not include the number of peeled iterations in
6632 the minimum threshold in that case.
6633 (vectorizable_induction): Adjust the start value down by
6634 LOOP_VINFO_MASK_SKIP_NITERS iterations.
6635 (vect_transform_loop): Call vect_prepare_for_masked_peels.
6636 Take the number of skipped iterations into account when calculating
6638 * tree-vect-stmts.c (vect_gen_while_not): New function.
6640 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6641 Alan Hayward <alan.hayward@arm.com>
6642 David Sherwood <david.sherwood@arm.com>
6644 * doc/sourcebuild.texi (vect_fully_masked): Document.
6645 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
6647 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
6649 (vect_analyze_loop_2): ...here. Don't check the vectorization
6650 factor against the number of loop iterations if the loop is
6653 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6654 Alan Hayward <alan.hayward@arm.com>
6655 David Sherwood <david.sherwood@arm.com>
6657 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
6658 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
6659 (dump_groups): Update accordingly.
6660 (iv_use::mem_type): New member variable.
6661 (address_p): New function.
6662 (record_use): Add a mem_type argument and initialize the new
6664 (record_group_use): Add a mem_type argument. Use address_p.
6665 Remove obsolete null checks of base_object. Update call to record_use.
6666 (find_interesting_uses_op): Update call to record_group_use.
6667 (find_interesting_uses_cond): Likewise.
6668 (find_interesting_uses_address): Likewise.
6669 (get_mem_type_for_internal_fn): New function.
6670 (find_address_like_use): Likewise.
6671 (find_interesting_uses_stmt): Try find_address_like_use before
6672 calling find_interesting_uses_op.
6673 (addr_offset_valid_p): Use the iv mem_type field as the type
6674 of the addressed memory.
6675 (add_autoinc_candidates): Likewise.
6676 (get_address_cost): Likewise.
6677 (split_small_address_groups_p): Use address_p.
6678 (split_address_groups): Likewise.
6679 (add_iv_candidate_for_use): Likewise.
6680 (autoinc_possible_for_pair): Likewise.
6681 (rewrite_groups): Likewise.
6682 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
6683 (determine_group_iv_cost): Update after split of USE_ADDRESS.
6684 (get_alias_ptr_type_for_ptr_address): New function.
6685 (rewrite_use_address): Rewrite address uses in calls that were
6686 identified by find_address_like_use.
6688 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6689 Alan Hayward <alan.hayward@arm.com>
6690 David Sherwood <david.sherwood@arm.com>
6692 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
6694 * gimple-expr.h (is_gimple_addressable: Likewise.
6695 * gimple-expr.c (is_gimple_address): Likewise.
6696 * internal-fn.c (expand_call_mem_ref): New function.
6697 (expand_mask_load_optab_fn): Use it.
6698 (expand_mask_store_optab_fn): Likewise.
6700 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6701 Alan Hayward <alan.hayward@arm.com>
6702 David Sherwood <david.sherwood@arm.com>
6704 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
6705 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
6706 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
6707 (cond_umax@var{mode}): Document.
6708 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
6709 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
6710 (cond_umin_optab, cond_umax_optab): New optabs.
6711 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
6712 (COND_IOR, COND_XOR): New internal functions.
6713 * internal-fn.h (get_conditional_internal_fn): Declare.
6714 * internal-fn.c (cond_binary_direct): New macro.
6715 (expand_cond_binary_optab_fn): Likewise.
6716 (direct_cond_binary_optab_supported_p): Likewise.
6717 (get_conditional_internal_fn): New function.
6718 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
6719 Cope with reduction statements that are vectorized as calls rather
6721 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
6722 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
6723 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
6724 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
6725 (UNSPEC_COND_EOR): New unspecs.
6726 (optab): Add mappings for them.
6727 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
6728 (sve_int_op, sve_fp_op): New int attributes.
6730 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6731 Alan Hayward <alan.hayward@arm.com>
6732 David Sherwood <david.sherwood@arm.com>
6734 * optabs.def (while_ult_optab): New optab.
6735 * doc/md.texi (while_ult@var{m}@var{n}): Document.
6736 * internal-fn.def (WHILE_ULT): New internal function.
6737 * internal-fn.h (direct_internal_fn_supported_p): New override
6738 that takes two types as argument.
6739 * internal-fn.c (while_direct): New macro.
6740 (expand_while_optab_fn): New function.
6741 (convert_optab_supported_p): Likewise.
6742 (direct_while_optab_supported_p): New macro.
6743 * wide-int.h (wi::udiv_ceil): New function.
6744 * tree-vectorizer.h (rgroup_masks): New structure.
6745 (vec_loop_masks): New typedef.
6746 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
6748 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
6749 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
6750 (vect_max_vf): New function.
6751 (slpeel_make_loop_iterate_ntimes): Delete.
6752 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
6753 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
6754 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
6755 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
6756 internal-fn.h, stor-layout.h and optabs-query.h.
6757 (vect_set_loop_mask): New function.
6758 (add_preheader_seq): Likewise.
6759 (add_header_seq): Likewise.
6760 (interleave_supported_p): Likewise.
6761 (vect_maybe_permute_loop_masks): Likewise.
6762 (vect_set_loop_masks_directly): Likewise.
6763 (vect_set_loop_condition_masked): Likewise.
6764 (vect_set_loop_condition_unmasked): New function, split out from
6765 slpeel_make_loop_iterate_ntimes.
6766 (slpeel_make_loop_iterate_ntimes): Rename to..
6767 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
6768 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
6769 (vect_do_peeling): Update call accordingly.
6770 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
6772 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6773 mask_compare_type, can_fully_mask_p and fully_masked_p.
6774 (release_vec_loop_masks): New function.
6775 (_loop_vec_info): Use it to free the loop masks.
6776 (can_produce_all_loop_masks_p): New function.
6777 (vect_get_max_nscalars_per_iter): Likewise.
6778 (vect_verify_full_masking): Likewise.
6779 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
6780 retries, and free the mask rgroups before retrying. Check loop-wide
6781 reasons for disallowing fully-masked loops. Make the final decision
6782 about whether use a fully-masked loop or not.
6783 (vect_estimate_min_profitable_iters): Do not assume that peeling
6784 for the number of iterations will be needed for fully-masked loops.
6785 (vectorizable_reduction): Disable fully-masked loops.
6786 (vectorizable_live_operation): Likewise.
6787 (vect_halve_mask_nunits): New function.
6788 (vect_double_mask_nunits): Likewise.
6789 (vect_record_loop_mask): Likewise.
6790 (vect_get_loop_mask): Likewise.
6791 (vect_transform_loop): Handle the case in which the final loop
6792 iteration might handle a partial vector. Call vect_set_loop_condition
6793 instead of slpeel_make_loop_iterate_ntimes.
6794 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
6795 (check_load_store_masking): New function.
6796 (prepare_load_store_mask): Likewise.
6797 (vectorizable_store): Handle fully-masked loops.
6798 (vectorizable_load): Likewise.
6799 (supportable_widening_operation): Use vect_halve_mask_nunits for
6801 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
6802 (vect_gen_while): New function.
6803 * config/aarch64/aarch64.md (umax<mode>3): New expander.
6804 (aarch64_uqdec<mode>): New insn.
6806 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6807 Alan Hayward <alan.hayward@arm.com>
6808 David Sherwood <david.sherwood@arm.com>
6810 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
6811 (reduc_xor_scal_optab): New optabs.
6812 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
6813 (reduc_xor_scal_@var{m}): Document.
6814 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
6815 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
6817 * fold-const-call.c (fold_const_call): Handle them.
6818 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
6819 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
6820 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
6821 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
6822 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
6823 (UNSPEC_XORV): New unspecs.
6824 (optab): Add entries for them.
6825 (BITWISEV): New int iterator.
6826 (bit_reduc_op): New int attributes.
6828 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6829 Alan Hayward <alan.hayward@arm.com>
6830 David Sherwood <david.sherwood@arm.com>
6832 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
6833 * internal-fn.def (VEC_SHL_INSERT): New internal function.
6834 * optabs.def (vec_shl_insert_optab): New optab.
6835 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
6836 (duplicate_and_interleave): Likewise.
6837 * tree-vect-loop.c: Include internal-fn.h.
6838 (neutral_op_for_slp_reduction): New function, split out from
6839 get_initial_defs_for_reduction.
6840 (get_initial_def_for_reduction): Handle option 2 for variable-length
6841 vectors by loading the neutral value into a vector and then shifting
6842 the initial value into element 0.
6843 (get_initial_defs_for_reduction): Replace the code argument with
6844 the neutral value calculated by neutral_op_for_slp_reduction.
6845 Use gimple_build_vector for constant-length vectors.
6846 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
6847 but the first group_size elements have a neutral value.
6848 Use duplicate_and_interleave otherwise.
6849 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
6850 Update call to get_initial_defs_for_reduction. Handle SLP
6851 reductions for variable-length vectors by creating one vector
6852 result for each scalar result, with the elements associated
6853 with other scalar results stubbed out with the neutral value.
6854 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
6855 Require IFN_VEC_SHL_INSERT for double reductions on
6856 variable-length vectors, or SLP reductions that have
6857 a neutral value. Require can_duplicate_and_interleave_p
6858 support for variable-length unchained SLP reductions if there
6859 is no neutral value, such as for MIN/MAX reductions. Also require
6860 the number of vector elements to be a multiple of the number of
6861 SLP statements when doing variable-length unchained SLP reductions.
6862 Update call to vect_create_epilog_for_reduction.
6863 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
6864 and remove initial values.
6865 (duplicate_and_interleave): Make public.
6866 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
6867 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
6869 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6870 Alan Hayward <alan.hayward@arm.com>
6871 David Sherwood <david.sherwood@arm.com>
6873 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
6874 (can_duplicate_and_interleave_p): New function.
6875 (vect_get_and_check_slp_defs): Take the vector of statements
6876 rather than just the current one. Remove excess parentheses.
6877 Restriction rejectinon of vect_constant_def and vect_external_def
6878 for variable-length vectors to boolean types, or types for which
6879 can_duplicate_and_interleave_p is false.
6880 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
6881 (duplicate_and_interleave): New function.
6882 (vect_get_constant_vectors): Use gimple_build_vector for
6883 constant-length vectors and suitable variable-length constant
6884 vectors. Use duplicate_and_interleave for other variable-length
6885 vectors. Don't defer the update when inserting new statements.
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.c (vect_estimate_min_profitable_iters): Make sure
6892 min_profitable_iters doesn't go negative.
6894 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6895 Alan Hayward <alan.hayward@arm.com>
6896 David Sherwood <david.sherwood@arm.com>
6898 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
6899 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
6900 * optabs.def (vec_mask_load_lanes_optab): New optab.
6901 (vec_mask_store_lanes_optab): Likewise.
6902 * internal-fn.def (MASK_LOAD_LANES): New internal function.
6903 (MASK_STORE_LANES): Likewise.
6904 * internal-fn.c (mask_load_lanes_direct): New macro.
6905 (mask_store_lanes_direct): Likewise.
6906 (expand_mask_load_optab_fn): Handle masked operations.
6907 (expand_mask_load_lanes_optab_fn): New macro.
6908 (expand_mask_store_optab_fn): Handle masked operations.
6909 (expand_mask_store_lanes_optab_fn): New macro.
6910 (direct_mask_load_lanes_optab_supported_p): Likewise.
6911 (direct_mask_store_lanes_optab_supported_p): Likewise.
6912 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
6914 (vect_load_lanes_supported): Likewise.
6915 * tree-vect-data-refs.c (strip_conversion): New function.
6916 (can_group_stmts_p): Likewise.
6917 (vect_analyze_data_ref_accesses): Use it instead of checking
6918 for a pair of assignments.
6919 (vect_store_lanes_supported): Take a masked_p parameter.
6920 (vect_load_lanes_supported): Likewise.
6921 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
6922 vect_store_lanes_supported and vect_load_lanes_supported.
6923 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6924 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
6925 parameter. Don't allow gaps for masked accesses.
6926 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
6927 and vect_load_lanes_supported.
6928 (get_load_store_type): Take a masked_p parameter and update
6929 call to get_group_load_store_type.
6930 (vectorizable_store): Update call to get_load_store_type.
6931 Handle IFN_MASK_STORE_LANES.
6932 (vectorizable_load): Update call to get_load_store_type.
6933 Handle IFN_MASK_LOAD_LANES.
6935 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6936 Alan Hayward <alan.hayward@arm.com>
6937 David Sherwood <david.sherwood@arm.com>
6939 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
6941 * config/aarch64/aarch64-protos.h
6942 (aarch64_sve_struct_memory_operand_p): Declare.
6943 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
6944 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
6945 (VPRED, vpred): Handle SVE structure modes.
6946 * config/aarch64/constraints.md (Utx): New constraint.
6947 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
6948 (aarch64_sve_struct_nonimmediate_operand): New predicates.
6949 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
6950 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
6951 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
6952 structure modes. Split into pieces after RA.
6953 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
6954 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
6956 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
6957 SVE structure modes.
6958 (aarch64_classify_address): Likewise.
6959 (sizetochar): Move earlier in file.
6960 (aarch64_print_operand): Handle SVE register lists.
6961 (aarch64_array_mode): New function.
6962 (aarch64_sve_struct_memory_operand_p): Likewise.
6963 (TARGET_ARRAY_MODE): Redefine.
6965 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6966 Alan Hayward <alan.hayward@arm.com>
6967 David Sherwood <david.sherwood@arm.com>
6969 * target.def (array_mode): New target hook.
6970 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
6971 * doc/tm.texi: Regenerate.
6972 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
6973 * hooks.c (hook_optmode_mode_uhwi_none): New function.
6974 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
6976 * stor-layout.c (mode_for_array): Likewise. Support polynomial
6979 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6980 Alan Hayward <alan.hayward@arm.com>
6981 David Sherwood <david.sherwood@arm.com>
6983 * fold-const.c (fold_binary_loc): Check the argument types
6984 rather than the result type when testing for a vector operation.
6986 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6988 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
6989 * doc/tm.texi: Regenerate.
6991 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6992 Alan Hayward <alan.hayward@arm.com>
6993 David Sherwood <david.sherwood@arm.com>
6995 * doc/invoke.texi (-msve-vector-bits=): Document new option.
6996 (sve): Document new AArch64 extension.
6997 * doc/md.texi (w): Extend the description of the AArch64
6998 constraint to include SVE vectors.
6999 (Upl, Upa): Document new AArch64 predicate constraints.
7000 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
7002 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
7003 (msve-vector-bits=): New option.
7004 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
7005 SVE when these are disabled.
7006 (sve): New extension.
7007 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
7008 modes. Adjust their number of units based on aarch64_sve_vg.
7009 (MAX_BITSIZE_MODE_ANY_MODE): Define.
7010 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
7011 aarch64_addr_query_type.
7012 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
7013 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
7014 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
7015 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
7016 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
7017 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
7018 (aarch64_simd_imm_zero_p): Delete.
7019 (aarch64_check_zero_based_sve_index_immediate): Declare.
7020 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7021 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7022 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7023 (aarch64_sve_float_mul_immediate_p): Likewise.
7024 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7026 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
7027 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
7028 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
7029 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
7030 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
7031 (aarch64_regmode_natural_size): Likewise.
7032 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
7033 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
7035 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
7036 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
7037 for VG and the SVE predicate registers.
7038 (V_ALIASES): Add a "z"-prefixed alias.
7039 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
7040 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
7041 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
7042 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
7043 (REG_CLASS_NAMES): Add entries for them.
7044 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
7045 and the predicate registers.
7046 (aarch64_sve_vg): Declare.
7047 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
7048 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
7049 (REGMODE_NATURAL_SIZE): Define.
7050 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
7052 * config/aarch64/aarch64.c: Include cfgrtl.h.
7053 (simd_immediate_info): Add a constructor for series vectors,
7054 and an associated step field.
7055 (aarch64_sve_vg): New variable.
7056 (aarch64_dbx_register_number): Handle VG and the predicate registers.
7057 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
7058 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
7059 (VEC_ANY_DATA, VEC_STRUCT): New constants.
7060 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
7061 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
7062 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
7063 (aarch64_get_mask_mode): New functions.
7064 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
7065 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
7066 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
7067 predicate modes and predicate registers. Explicitly restrict
7068 GPRs to modes of 16 bytes or smaller. Only allow FP registers
7069 to store a vector mode if it is recognized by
7070 aarch64_classify_vector_mode.
7071 (aarch64_regmode_natural_size): New function.
7072 (aarch64_hard_regno_caller_save_mode): Return the original mode
7074 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
7075 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
7076 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
7077 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
7079 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
7080 does not overlap dest if the function is frame-related. Handle
7082 (aarch64_split_add_offset): New function.
7083 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
7084 them aarch64_add_offset.
7085 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
7086 and update call to aarch64_sub_sp.
7087 (aarch64_add_cfa_expression): New function.
7088 (aarch64_expand_prologue): Pass extra temporary registers to the
7089 functions above. Handle the case in which we need to emit new
7090 DW_CFA_expressions for registers that were originally saved
7091 relative to the stack pointer, but now have to be expressed
7092 relative to the frame pointer.
7093 (aarch64_output_mi_thunk): Pass extra temporary registers to the
7095 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
7096 IP0 and IP1 values for SVE frames.
7097 (aarch64_expand_vec_series): New function.
7098 (aarch64_expand_sve_widened_duplicate): Likewise.
7099 (aarch64_expand_sve_const_vector): Likewise.
7100 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
7101 Handle SVE constants. Use emit_move_insn to move a force_const_mem
7102 into the register, rather than emitting a SET directly.
7103 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
7104 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
7105 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
7106 (offset_9bit_signed_scaled_p): New functions.
7107 (aarch64_replicate_bitmask_imm): New function.
7108 (aarch64_bitmask_imm): Use it.
7109 (aarch64_cannot_force_const_mem): Reject expressions involving
7110 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
7111 (aarch64_classify_index): Handle SVE indices, by requiring
7112 a plain register index with a scale that matches the element size.
7113 (aarch64_classify_address): Handle SVE addresses. Assert that
7114 the mode of the address is VOIDmode or an integer mode.
7115 Update call to aarch64_classify_symbol.
7116 (aarch64_classify_symbolic_expression): Update call to
7117 aarch64_classify_symbol.
7118 (aarch64_const_vec_all_in_range_p): New function.
7119 (aarch64_print_vector_float_operand): Likewise.
7120 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
7121 "vN" for FP registers with SVE modes. Handle (const ...) vectors
7122 and the FP immediates 1.0 and 0.5.
7123 (aarch64_print_address_internal): Handle SVE addresses.
7124 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
7125 (aarch64_regno_regclass): Handle predicate registers.
7126 (aarch64_secondary_reload): Handle big-endian reloads of SVE
7128 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
7129 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
7130 (aarch64_convert_sve_vector_bits): New function.
7131 (aarch64_override_options): Use it to handle -msve-vector-bits=.
7132 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7134 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
7135 Handle SVE vector and predicate modes. Accept VL-based constants
7136 that need only one temporary register, and VL offsets that require
7137 no temporary registers.
7138 (aarch64_conditional_register_usage): Mark the predicate registers
7139 as fixed if SVE isn't available.
7140 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
7141 Return true for SVE vector and predicate modes.
7142 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
7143 rather than an unsigned int. Handle SVE modes.
7144 (aarch64_preferred_simd_mode): Update call accordingly. Handle
7146 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
7148 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7149 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7150 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7151 (aarch64_sve_float_mul_immediate_p): New functions.
7152 (aarch64_sve_valid_immediate): New function.
7153 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
7154 Explicitly reject structure modes. Check for INDEX constants.
7155 Handle PTRUE and PFALSE constants.
7156 (aarch64_check_zero_based_sve_index_immediate): New function.
7157 (aarch64_simd_imm_zero_p): Delete.
7158 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
7159 vector modes. Accept constants in the range of CNT[BHWD].
7160 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
7161 ask for an Advanced SIMD mode.
7162 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
7163 (aarch64_simd_vector_alignment): Handle SVE predicates.
7164 (aarch64_vectorize_preferred_vector_alignment): New function.
7165 (aarch64_simd_vector_alignment_reachable): Use it instead of
7167 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
7168 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
7170 (MAX_VECT_LEN): Delete.
7171 (expand_vec_perm_d): Add a vec_flags field.
7172 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
7173 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
7174 (aarch64_evpc_ext): Don't apply a big-endian lane correction
7176 (aarch64_evpc_rev): Rename to...
7177 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
7178 (aarch64_evpc_rev_global): New function.
7179 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
7180 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
7182 (aarch64_evpc_sve_tbl): New function.
7183 (aarch64_expand_vec_perm_const_1): Update after rename of
7184 aarch64_evpc_rev. Handle SVE permutes too, trying
7185 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
7186 than aarch64_evpc_tbl.
7187 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
7188 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
7189 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
7190 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
7191 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
7192 (aarch64_expand_sve_vcond): New functions.
7193 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
7194 of aarch64_vector_mode_p.
7195 (aarch64_dwarf_poly_indeterminate_value): New function.
7196 (aarch64_compute_pressure_classes): Likewise.
7197 (aarch64_can_change_mode_class): Likewise.
7198 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
7199 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
7200 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
7201 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
7202 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
7203 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
7204 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
7205 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
7207 (Dn, Dl, Dr): Accept const as well as const_vector.
7208 (Dz): Likewise. Compare against CONST0_RTX.
7209 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
7210 of "vector" where appropriate.
7211 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
7212 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
7213 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
7214 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
7215 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
7216 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
7217 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
7218 (v_int_equiv): Extend to SVE modes.
7219 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
7221 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
7222 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
7223 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
7224 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
7225 (SVE_COND_FP_CMP): New int iterators.
7226 (perm_hilo): Handle the new unpack unspecs.
7227 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
7229 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
7230 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
7231 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
7232 (aarch64_equality_operator, aarch64_constant_vector_operand)
7233 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
7234 (aarch64_sve_nonimmediate_operand): Likewise.
7235 (aarch64_sve_general_operand): Likewise.
7236 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
7237 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
7238 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
7239 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
7240 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
7241 (aarch64_sve_float_arith_immediate): Likewise.
7242 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
7243 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
7244 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
7245 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
7246 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
7247 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
7248 (aarch64_sve_float_arith_operand): Likewise.
7249 (aarch64_sve_float_arith_with_sub_operand): Likewise.
7250 (aarch64_sve_float_mul_operand): Likewise.
7251 (aarch64_sve_vec_perm_operand): Likewise.
7252 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
7253 (aarch64_mov_operand): Accept const_poly_int and const_vector.
7254 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
7255 as well as const_vector.
7256 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
7257 in file. Use CONST0_RTX and CONSTM1_RTX.
7258 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
7259 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
7260 Use aarch64_simd_imm_zero.
7261 * config/aarch64/aarch64-sve.md: New file.
7262 * config/aarch64/aarch64.md: Include it.
7263 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
7264 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
7265 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
7266 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
7267 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
7268 (sve): New attribute.
7269 (enabled): Disable instructions with the sve attribute unless
7271 (movqi, movhi): Pass CONST_POLY_INT operaneds through
7272 aarch64_expand_mov_immediate.
7273 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
7274 CNT[BHSD] immediates.
7275 (movti): Split CONST_POLY_INT moves into two halves.
7276 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
7277 Split additions that need a temporary here if the destination
7278 is the stack pointer.
7279 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
7280 (*add<mode>3_poly_1): New instruction.
7281 (set_clobber_cc): New expander.
7283 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7285 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
7286 parameter and use it instead of GET_MODE_SIZE (innermode). Use
7287 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
7288 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
7289 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
7290 Change innermode from fixed_mode_size to machine_mode.
7291 (simplify_subreg): Update call accordingly. Handle a constant-sized
7292 subreg of a variable-length CONST_VECTOR.
7294 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7295 Alan Hayward <alan.hayward@arm.com>
7296 David Sherwood <david.sherwood@arm.com>
7298 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
7299 (add_offset_to_base): New function, split out from...
7300 (create_mem_ref): ...here. When handling a scale other than 1,
7301 check first whether the address is valid without the offset.
7302 Add it into the base if so, leaving the index and scale as-is.
7304 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7307 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
7308 fold_for_warn before checking if arg2 is INTEGER_CST.
7310 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
7312 * config/rs6000/predicates.md (load_multiple_operation): Delete.
7313 (store_multiple_operation): Delete.
7314 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
7315 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
7316 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
7317 guarded by TARGET_STRING.
7318 (rs6000_output_load_multiple): Delete.
7319 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
7320 OPTION_MASK_STRING / TARGET_STRING handling.
7321 (print_operand) <'N', 'O'>: Add comment that these are unused now.
7322 (const rs6000_opt_masks) <"string">: Change mask to 0.
7323 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
7324 (MASK_STRING): Delete.
7325 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
7327 (load_multiple): Delete.
7334 (store_multiple): Delete.
7341 (movmemsi_8reg): Delete.
7342 (corresponding unnamed define_insn): Delete.
7343 (movmemsi_6reg): Delete.
7344 (corresponding unnamed define_insn): Delete.
7345 (movmemsi_4reg): Delete.
7346 (corresponding unnamed define_insn): Delete.
7347 (movmemsi_2reg): Delete.
7348 (corresponding unnamed define_insn): Delete.
7349 (movmemsi_1reg): Delete.
7350 (corresponding unnamed define_insn): Delete.
7351 * config/rs6000/rs6000.opt (mno-string): New.
7352 (mstring): Replace by deprecation warning stub.
7353 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
7355 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7357 * regrename.c (regrename_do_replace): If replacing the same
7358 reg multiple times, try to reuse last created gen_raw_REG.
7361 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
7362 main to workaround a bug in GDB.
7364 2018-01-12 Tom de Vries <tom@codesourcery.com>
7367 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
7369 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
7371 PR rtl-optimization/80481
7372 * ira-color.c (get_cap_member): New function.
7373 (allocnos_conflict_by_live_ranges_p): Use it.
7374 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
7375 (setup_slot_coalesced_allocno_live_ranges): Ditto.
7377 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
7380 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
7381 (*saddl_se_1): Ditto.
7383 (*ssubl_se_1): Ditto.
7385 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7387 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
7388 rather than wi::to_widest for DR_INITs.
7389 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
7390 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
7391 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
7393 (vect_analyze_group_access_1): Note that here.
7395 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7397 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
7398 polynomial type sizes.
7400 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7402 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
7403 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
7404 (gimple_add_tmp_var): Likewise.
7406 2018-01-12 Martin Liska <mliska@suse.cz>
7408 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
7409 (gimple_alloc_sizes): Likewise.
7410 (dump_gimple_statistics): Use PRIu64 in printf format.
7411 * gimple.h: Change uint64_t to int.
7413 2018-01-12 Martin Liska <mliska@suse.cz>
7415 * tree-core.h: Use uint64_t instead of int.
7416 * tree.c (tree_node_counts): Likewise.
7417 (tree_node_sizes): Likewise.
7418 (dump_tree_statistics): Use PRIu64 in printf format.
7420 2018-01-12 Martin Liska <mliska@suse.cz>
7422 * Makefile.in: As qsort_chk is implemented in vec.c, add
7423 vec.o to linkage of gencfn-macros.
7424 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
7425 passing the info to record_node_allocation_statistics.
7426 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
7428 * ggc-common.c (struct ggc_usage): Add operator== and use
7429 it in operator< and compare function.
7430 * mem-stats.h (struct mem_usage): Likewise.
7431 * vec.c (struct vec_usage): Remove operator< and compare
7432 function. Can be simply inherited.
7434 2018-01-12 Martin Jambor <mjambor@suse.cz>
7437 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
7438 * tree-ssa-math-opts.c: Include domwalk.h.
7439 (convert_mult_to_fma_1): New function.
7440 (fma_transformation_info): New type.
7441 (fma_deferring_state): Likewise.
7442 (cancel_fma_deferring): New function.
7443 (result_of_phi): Likewise.
7444 (last_fma_candidate_feeds_initial_phi): Likewise.
7445 (convert_mult_to_fma): Added deferring logic, split actual
7446 transformation to convert_mult_to_fma_1.
7447 (math_opts_dom_walker): New type.
7448 (math_opts_dom_walker::after_dom_children): New method, body moved
7449 here from pass_optimize_widening_mul::execute, added deferring logic
7451 (pass_optimize_widening_mul::execute): Moved most of code to
7452 math_opts_dom_walker::after_dom_children.
7453 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
7454 * config/i386/i386.c (ix86_option_override_internal): Added
7455 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
7457 2018-01-12 Richard Biener <rguenther@suse.de>
7460 * dwarf2out.c (gen_variable_die): Do not reset old_die for
7461 inline instance vars.
7463 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
7466 * config/rx/rx.c (rx_is_restricted_memory_address):
7469 2018-01-12 Richard Biener <rguenther@suse.de>
7471 PR tree-optimization/80846
7472 * target.def (split_reduction): New target hook.
7473 * targhooks.c (default_split_reduction): New function.
7474 * targhooks.h (default_split_reduction): Declare.
7475 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
7476 target requests first reduce vectors by combining low and high
7478 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
7479 (get_vectype_for_scalar_type_and_size): Export.
7480 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
7481 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
7482 * doc/tm.texi: Regenerate.
7483 * config/i386/i386.c (ix86_split_reduction): Implement
7484 TARGET_VECTORIZE_SPLIT_REDUCTION.
7486 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7489 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
7490 in PIC mode except for TARGET_VXWORKS_RTP.
7491 * config/sparc/sparc.c: Include cfgrtl.h.
7492 (TARGET_INIT_PIC_REG): Define.
7493 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
7494 (sparc_pic_register_p): New predicate.
7495 (sparc_legitimate_address_p): Use it.
7496 (sparc_legitimize_pic_address): Likewise.
7497 (sparc_delegitimize_address): Likewise.
7498 (sparc_mode_dependent_address_p): Likewise.
7499 (gen_load_pcrel_sym): Remove 4th parameter.
7500 (load_got_register): Adjust call to above. Remove obsolete stuff.
7501 (sparc_expand_prologue): Do not call load_got_register here.
7502 (sparc_flat_expand_prologue): Likewise.
7503 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
7504 (sparc_use_pseudo_pic_reg): New function.
7505 (sparc_init_pic_reg): Likewise.
7506 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7507 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
7509 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
7511 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
7512 Add item for branch_cost.
7514 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7516 PR rtl-optimization/83565
7517 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
7518 not extend the result to a larger mode for rotate operations.
7519 (num_sign_bit_copies1): Likewise.
7521 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7524 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
7526 Use values-Xc.o for -pedantic.
7527 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
7529 2018-01-12 Martin Liska <mliska@suse.cz>
7532 * ipa-devirt.c (final_warning_record::grow_type_warnings):
7534 (possible_polymorphic_call_targets): Use it.
7535 (ipa_devirt): Likewise.
7537 2018-01-12 Martin Liska <mliska@suse.cz>
7539 * profile-count.h (enum profile_quality): Use 0 as invalid
7540 enum value of profile_quality.
7542 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
7544 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
7545 -mext-string options.
7547 2018-01-12 Richard Biener <rguenther@suse.de>
7549 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
7550 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
7551 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
7553 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
7555 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
7557 * configure.ac (--with-long-double-format): Add support for the
7558 configuration option to change the default long double format on
7560 * config.gcc (powerpc*-linux*-*): Likewise.
7561 * configure: Regenerate.
7562 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
7563 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
7564 used without modification.
7566 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7568 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
7569 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
7570 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
7571 MISC_BUILTIN_SPEC_BARRIER.
7572 (rs6000_init_builtins): Likewise.
7573 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
7575 (speculation_barrier): New define_insn.
7576 * doc/extend.texi: Document __builtin_speculation_barrier.
7578 2018-01-11 Jakub Jelinek <jakub@redhat.com>
7581 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
7582 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
7583 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
7585 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
7586 integral modes instead of "ss" and "sd".
7587 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
7588 vectors with 32-bit and 64-bit elements.
7589 (vecdupssescalarmodesuffix): New mode attribute.
7590 (vec_dup<mode>): Use it.
7592 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
7595 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
7596 frame if argument is passed on stack.
7598 2018-01-11 Jakub Jelinek <jakub@redhat.com>
7601 * ree.c (combine_reaching_defs): Optimize also
7602 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
7603 reg2=any_extend(exp); reg1=reg2;, formatting fix.
7605 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
7608 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
7610 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
7613 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
7614 after they are computed.
7616 2018-01-11 Bin Cheng <bin.cheng@arm.com>
7618 PR tree-optimization/83695
7619 * gimple-loop-linterchange.cc
7620 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
7621 reset cached scev information after interchange.
7622 (pass_linterchange::execute): Remove call to scev_reset_htab.
7624 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7626 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
7627 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
7628 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
7629 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
7630 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
7631 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
7632 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
7633 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
7634 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
7635 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
7636 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
7637 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
7638 (V_lane_reg): Likewise.
7639 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
7641 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
7642 (vfmal_lane_low<mode>_intrinsic,
7643 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
7644 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
7645 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
7646 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
7647 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
7648 vfmsl_lane_high<mode>_intrinsic): New define_insns.
7650 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7652 * config/arm/arm-cpus.in (fp16fml): New feature.
7653 (ALL_SIMD): Add fp16fml.
7654 (armv8.2-a): Add fp16fml as an option.
7655 (armv8.3-a): Likewise.
7656 (armv8.4-a): Add fp16fml as part of fp16.
7657 * config/arm/arm.h (TARGET_FP16FML): Define.
7658 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
7660 * config/arm/arm-modes.def (V2HF): Define.
7661 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
7662 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
7663 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
7664 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
7665 vfmsl_low, vfmsl_high): New set of builtins.
7666 * config/arm/iterators.md (PLUSMINUS): New code iterator.
7667 (vfml_op): New code attribute.
7668 (VFMLHALVES): New int iterator.
7669 (VFML, VFMLSEL): New mode attributes.
7670 (V_reg): Define mapping for V2HF.
7671 (V_hi, V_lo): New mode attributes.
7672 (VF_constraint): Likewise.
7673 (vfml_half, vfml_half_selector): New int attributes.
7674 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
7676 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
7677 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
7679 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
7680 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
7681 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
7682 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
7684 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
7685 Document new effective target and option set.
7687 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7689 * config/arm/arm-cpus.in (armv8_4): New feature.
7690 (ARMv8_4a): New fgroup.
7691 (armv8.4-a): New arch.
7692 * config/arm/arm-tables.opt: Regenerate.
7693 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
7694 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
7695 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
7696 Add matching rules for -march=armv8.4-a and extensions.
7697 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
7699 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
7702 * config/rx/rx.md (BW): New mode attribute.
7703 (sync_lock_test_and_setsi): Add mode suffix to insn output.
7705 2018-01-11 Richard Biener <rguenther@suse.de>
7707 PR tree-optimization/83435
7708 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
7709 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
7710 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
7712 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7713 Alan Hayward <alan.hayward@arm.com>
7714 David Sherwood <david.sherwood@arm.com>
7716 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
7718 (aarch64_classify_address): Initialize it. Track polynomial offsets.
7719 (aarch64_print_address_internal): Use it to check for a zero offset.
7721 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7722 Alan Hayward <alan.hayward@arm.com>
7723 David Sherwood <david.sherwood@arm.com>
7725 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
7726 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
7727 Return a poly_int64 rather than a HOST_WIDE_INT.
7728 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
7729 rather than a HOST_WIDE_INT.
7730 * config/aarch64/aarch64.h (aarch64_frame): Protect with
7731 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
7732 hard_fp_offset, frame_size, initial_adjust, callee_offset and
7733 final_offset from HOST_WIDE_INT to poly_int64.
7734 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
7735 to_constant when getting the number of units in an Advanced SIMD
7737 (aarch64_builtin_vectorized_function): Check for a constant number
7739 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
7741 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
7742 attribute instead of GET_MODE_NUNITS.
7743 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
7744 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
7745 GET_MODE_SIZE for fixed-size registers.
7746 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
7747 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
7748 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
7749 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
7750 (aarch64_print_operand, aarch64_print_address_internal)
7751 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
7752 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
7753 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
7754 Handle polynomial GET_MODE_SIZE.
7755 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
7756 wider than SImode without modification.
7757 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
7758 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
7759 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
7760 passing and returning SVE modes.
7761 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
7762 rather than GEN_INT.
7763 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
7764 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
7765 (aarch64_allocate_and_probe_stack_space): Likewise.
7766 (aarch64_layout_frame): Cope with polynomial offsets.
7767 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
7768 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
7770 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
7771 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
7772 poly_int64 rather than a HOST_WIDE_INT.
7773 (aarch64_get_separate_components, aarch64_process_components)
7774 (aarch64_expand_prologue, aarch64_expand_epilogue)
7775 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
7776 (aarch64_anchor_offset): New function, split out from...
7777 (aarch64_legitimize_address): ...here.
7778 (aarch64_builtin_vectorization_cost): Handle polynomial
7779 TYPE_VECTOR_SUBPARTS.
7780 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
7782 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
7783 number of elements from the PARALLEL rather than the mode.
7784 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
7785 rather than GET_MODE_BITSIZE.
7786 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
7787 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
7788 (aarch64_expand_vec_perm_const_1): Handle polynomial
7789 d->perm.length () and d->perm elements.
7790 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
7791 Apply to_constant to d->perm elements.
7792 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
7793 polynomial CONST_VECTOR_NUNITS.
7794 (aarch64_move_pointer): Take amount as a poly_int64 rather
7796 (aarch64_progress_pointer): Avoid temporary variable.
7797 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
7798 the mode attribute instead of GET_MODE.
7800 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7801 Alan Hayward <alan.hayward@arm.com>
7802 David Sherwood <david.sherwood@arm.com>
7804 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
7805 x exists before using it.
7806 (aarch64_add_constant_internal): Rename to...
7807 (aarch64_add_offset_1): ...this. Replace regnum with separate
7808 src and dest rtxes. Handle the case in which they're different,
7809 including when the offset is zero. Replace scratchreg with an rtx.
7810 Use 2 additions if there is no spare register into which we can
7811 move a 16-bit constant.
7812 (aarch64_add_constant): Delete.
7813 (aarch64_add_offset): Replace reg with separate src and dest
7814 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
7815 Use aarch64_add_offset_1.
7816 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
7817 an rtx rather than an int. Take the delta as a poly_int64
7818 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
7819 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
7820 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
7821 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
7822 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
7824 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
7825 aarch64_add_constant.
7827 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7829 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
7830 Use scalar_float_mode.
7832 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7834 * config/aarch64/aarch64-simd.md
7835 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
7836 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
7837 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
7838 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
7839 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
7840 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
7841 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
7842 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
7843 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
7844 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
7846 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7849 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
7850 targ_options->x_arm_arch_string is non NULL.
7852 2018-01-11 Tamar Christina <tamar.christina@arm.com>
7854 * config/aarch64/aarch64.h
7855 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
7857 2018-01-11 Sudakshina Das <sudi.das@arm.com>
7860 * expmed.c (emit_store_flag_force): Swap if const op0
7861 and change VOIDmode to mode of op0.
7863 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7865 PR rtl-optimization/83761
7866 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
7867 than bytes to mode_for_size.
7869 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7872 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
7873 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
7876 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7879 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
7880 when in layout mode.
7881 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
7882 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
7885 2018-01-10 Michael Collison <michael.collison@arm.com>
7887 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
7888 * config/aarch64/aarch64-option-extension.def: Add
7889 AARCH64_OPT_EXTENSION of 'fp16fml'.
7890 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7891 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
7892 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
7893 * config/aarch64/constraints.md (Ui7): New constraint.
7894 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
7895 (VFMLA_SEL_W): Ditto.
7898 (VFMLA16_LOW): New int iterator.
7899 (VFMLA16_HIGH): Ditto.
7900 (UNSPEC_FMLAL): New unspec.
7901 (UNSPEC_FMLSL): Ditto.
7902 (UNSPEC_FMLAL2): Ditto.
7903 (UNSPEC_FMLSL2): Ditto.
7904 (f16mac): New code attribute.
7905 * config/aarch64/aarch64-simd-builtins.def
7906 (aarch64_fmlal_lowv2sf): Ditto.
7907 (aarch64_fmlsl_lowv2sf): Ditto.
7908 (aarch64_fmlalq_lowv4sf): Ditto.
7909 (aarch64_fmlslq_lowv4sf): Ditto.
7910 (aarch64_fmlal_highv2sf): Ditto.
7911 (aarch64_fmlsl_highv2sf): Ditto.
7912 (aarch64_fmlalq_highv4sf): Ditto.
7913 (aarch64_fmlslq_highv4sf): Ditto.
7914 (aarch64_fmlal_lane_lowv2sf): Ditto.
7915 (aarch64_fmlsl_lane_lowv2sf): Ditto.
7916 (aarch64_fmlal_laneq_lowv2sf): Ditto.
7917 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
7918 (aarch64_fmlalq_lane_lowv4sf): Ditto.
7919 (aarch64_fmlsl_lane_lowv4sf): Ditto.
7920 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
7921 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
7922 (aarch64_fmlal_lane_highv2sf): Ditto.
7923 (aarch64_fmlsl_lane_highv2sf): Ditto.
7924 (aarch64_fmlal_laneq_highv2sf): Ditto.
7925 (aarch64_fmlsl_laneq_highv2sf): Ditto.
7926 (aarch64_fmlalq_lane_highv4sf): Ditto.
7927 (aarch64_fmlsl_lane_highv4sf): Ditto.
7928 (aarch64_fmlalq_laneq_highv4sf): Ditto.
7929 (aarch64_fmlsl_laneq_highv4sf): Ditto.
7930 * config/aarch64/aarch64-simd.md:
7931 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
7932 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7933 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
7934 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7935 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
7936 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
7937 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
7938 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
7939 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
7940 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
7941 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
7942 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
7943 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
7944 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
7945 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
7946 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
7947 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
7948 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
7949 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
7950 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
7951 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
7952 (vfmlsl_low_u32): Ditto.
7953 (vfmlalq_low_u32): Ditto.
7954 (vfmlslq_low_u32): Ditto.
7955 (vfmlal_high_u32): Ditto.
7956 (vfmlsl_high_u32): Ditto.
7957 (vfmlalq_high_u32): Ditto.
7958 (vfmlslq_high_u32): Ditto.
7959 (vfmlal_lane_low_u32): Ditto.
7960 (vfmlsl_lane_low_u32): Ditto.
7961 (vfmlal_laneq_low_u32): Ditto.
7962 (vfmlsl_laneq_low_u32): Ditto.
7963 (vfmlalq_lane_low_u32): Ditto.
7964 (vfmlslq_lane_low_u32): Ditto.
7965 (vfmlalq_laneq_low_u32): Ditto.
7966 (vfmlslq_laneq_low_u32): Ditto.
7967 (vfmlal_lane_high_u32): Ditto.
7968 (vfmlsl_lane_high_u32): Ditto.
7969 (vfmlal_laneq_high_u32): Ditto.
7970 (vfmlsl_laneq_high_u32): Ditto.
7971 (vfmlalq_lane_high_u32): Ditto.
7972 (vfmlslq_lane_high_u32): Ditto.
7973 (vfmlalq_laneq_high_u32): Ditto.
7974 (vfmlslq_laneq_high_u32): Ditto.
7975 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
7976 (AARCH64_FL_FOR_ARCH8_4): New.
7977 (AARCH64_ISA_F16FML): New ISA flag.
7978 (TARGET_F16FML): New feature flag for fp16fml.
7979 (doc/invoke.texi): Document new fp16fml option.
7981 2018-01-10 Michael Collison <michael.collison@arm.com>
7983 * config/aarch64/aarch64-builtins.c:
7984 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
7985 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7986 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
7987 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
7988 (AARCH64_ISA_SHA3): New ISA flag.
7989 (TARGET_SHA3): New feature flag for sha3.
7990 * config/aarch64/iterators.md (sha512_op): New int attribute.
7991 (CRYPTO_SHA512): New int iterator.
7992 (UNSPEC_SHA512H): New unspec.
7993 (UNSPEC_SHA512H2): Ditto.
7994 (UNSPEC_SHA512SU0): Ditto.
7995 (UNSPEC_SHA512SU1): Ditto.
7996 * config/aarch64/aarch64-simd-builtins.def
7997 (aarch64_crypto_sha512hqv2di): New builtin.
7998 (aarch64_crypto_sha512h2qv2di): Ditto.
7999 (aarch64_crypto_sha512su0qv2di): Ditto.
8000 (aarch64_crypto_sha512su1qv2di): Ditto.
8001 (aarch64_eor3qv8hi): Ditto.
8002 (aarch64_rax1qv2di): Ditto.
8003 (aarch64_xarqv2di): Ditto.
8004 (aarch64_bcaxqv8hi): Ditto.
8005 * config/aarch64/aarch64-simd.md:
8006 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
8007 (aarch64_crypto_sha512su0qv2di): Ditto.
8008 (aarch64_crypto_sha512su1qv2di): Ditto.
8009 (aarch64_eor3qv8hi): Ditto.
8010 (aarch64_rax1qv2di): Ditto.
8011 (aarch64_xarqv2di): Ditto.
8012 (aarch64_bcaxqv8hi): Ditto.
8013 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
8014 (vsha512h2q_u64): Ditto.
8015 (vsha512su0q_u64): Ditto.
8016 (vsha512su1q_u64): Ditto.
8017 (veor3q_u16): Ditto.
8018 (vrax1q_u64): Ditto.
8020 (vbcaxq_u16): Ditto.
8021 * config/arm/types.md (crypto_sha512): New type attribute.
8022 (crypto_sha3): Ditto.
8023 (doc/invoke.texi): Document new sha3 option.
8025 2018-01-10 Michael Collison <michael.collison@arm.com>
8027 * config/aarch64/aarch64-builtins.c:
8028 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
8029 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8030 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
8031 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
8032 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
8033 (AARCH64_ISA_SM4): New ISA flag.
8034 (TARGET_SM4): New feature flag for sm4.
8035 * config/aarch64/aarch64-simd-builtins.def
8036 (aarch64_sm3ss1qv4si): Ditto.
8037 (aarch64_sm3tt1aq4si): Ditto.
8038 (aarch64_sm3tt1bq4si): Ditto.
8039 (aarch64_sm3tt2aq4si): Ditto.
8040 (aarch64_sm3tt2bq4si): Ditto.
8041 (aarch64_sm3partw1qv4si): Ditto.
8042 (aarch64_sm3partw2qv4si): Ditto.
8043 (aarch64_sm4eqv4si): Ditto.
8044 (aarch64_sm4ekeyqv4si): Ditto.
8045 * config/aarch64/aarch64-simd.md:
8046 (aarch64_sm3ss1qv4si): Ditto.
8047 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
8048 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
8049 (aarch64_sm4eqv4si): Ditto.
8050 (aarch64_sm4ekeyqv4si): Ditto.
8051 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
8052 (sm3part_op): Ditto.
8053 (CRYPTO_SM3TT): Ditto.
8054 (CRYPTO_SM3PART): Ditto.
8055 (UNSPEC_SM3SS1): New unspec.
8056 (UNSPEC_SM3TT1A): Ditto.
8057 (UNSPEC_SM3TT1B): Ditto.
8058 (UNSPEC_SM3TT2A): Ditto.
8059 (UNSPEC_SM3TT2B): Ditto.
8060 (UNSPEC_SM3PARTW1): Ditto.
8061 (UNSPEC_SM3PARTW2): Ditto.
8062 (UNSPEC_SM4E): Ditto.
8063 (UNSPEC_SM4EKEY): Ditto.
8064 * config/aarch64/constraints.md (Ui2): New constraint.
8065 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
8066 * config/arm/types.md (crypto_sm3): New type attribute.
8067 (crypto_sm4): Ditto.
8068 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
8069 (vsm3tt1aq_u32): Ditto.
8070 (vsm3tt1bq_u32): Ditto.
8071 (vsm3tt2aq_u32): Ditto.
8072 (vsm3tt2bq_u32): Ditto.
8073 (vsm3partw1q_u32): Ditto.
8074 (vsm3partw2q_u32): Ditto.
8075 (vsm4eq_u32): Ditto.
8076 (vsm4ekeyq_u32): Ditto.
8077 (doc/invoke.texi): Document new sm4 option.
8079 2018-01-10 Michael Collison <michael.collison@arm.com>
8081 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
8082 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
8083 (AARCH64_FL_FOR_ARCH8_4): New.
8084 (AARCH64_FL_V8_4): New flag.
8085 (doc/invoke.texi): Document new armv8.4-a option.
8087 2018-01-10 Michael Collison <michael.collison@arm.com>
8089 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8090 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
8091 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
8092 * config/aarch64/aarch64-option-extension.def: Add
8093 AARCH64_OPT_EXTENSION of 'sha2'.
8094 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
8095 (crypto): Disable sha2 and aes if crypto disabled.
8096 (crypto): Enable aes and sha2 if enabled.
8097 (simd): Disable sha2 and aes if simd disabled.
8098 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
8100 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
8101 (TARGET_SHA2): New feature flag for sha2.
8102 (TARGET_AES): New feature flag for aes.
8103 * config/aarch64/aarch64-simd.md:
8104 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
8105 conditional on TARGET_AES.
8106 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
8107 (aarch64_crypto_sha1hsi): Make pattern conditional
8109 (aarch64_crypto_sha1hv4si): Ditto.
8110 (aarch64_be_crypto_sha1hv4si): Ditto.
8111 (aarch64_crypto_sha1su1v4si): Ditto.
8112 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
8113 (aarch64_crypto_sha1su0v4si): Ditto.
8114 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
8115 (aarch64_crypto_sha256su0v4si): Ditto.
8116 (aarch64_crypto_sha256su1v4si): Ditto.
8117 (doc/invoke.texi): Document new aes and sha2 options.
8119 2018-01-10 Martin Sebor <msebor@redhat.com>
8121 PR tree-optimization/83781
8122 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
8125 2018-01-11 Martin Sebor <msebor@gmail.com>
8126 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8128 PR tree-optimization/83501
8129 PR tree-optimization/81703
8131 * tree-ssa-strlen.c (get_string_cst): Rename...
8132 (get_string_len): ...to this. Handle global constants.
8133 (handle_char_store): Adjust.
8135 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
8136 Jim Wilson <jimw@sifive.com>
8138 * config/riscv/riscv-protos.h (riscv_output_return): New.
8139 * config/riscv/riscv.c (struct machine_function): New naked_p field.
8140 (riscv_attribute_table, riscv_output_return),
8141 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
8142 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
8143 (riscv_compute_frame_info): Only compute frame->mask if not a naked
8145 (riscv_expand_prologue): Add early return for naked function.
8146 (riscv_expand_epilogue): Likewise.
8147 (riscv_function_ok_for_sibcall): Return false for naked function.
8148 (riscv_set_current_function): New.
8149 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
8150 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
8151 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
8152 * doc/extend.texi (RISC-V Function Attributes): New.
8154 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
8156 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
8157 check for 128-bit long double before checking TCmode.
8158 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
8159 128-bit long doubles before checking TFmode or TCmode.
8160 (FLOAT128_IBM_P): Likewise.
8162 2018-01-10 Martin Sebor <msebor@redhat.com>
8164 PR tree-optimization/83671
8165 * builtins.c (c_strlen): Unconditionally return zero for the empty
8167 Use -Warray-bounds for warnings.
8168 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
8169 for non-constant array indices with COMPONENT_REF, arrays of
8170 arrays, and pointers to arrays.
8171 (gimple_fold_builtin_strlen): Determine and set length range for
8172 non-constant character arrays.
8174 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
8177 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
8180 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
8182 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
8184 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8187 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
8188 VECTOR_MEM_ALTIVEC_OR_VSX_P.
8189 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
8190 indexed_or_indirect_operand predicate.
8191 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
8192 (*vsx_le_perm_load_v8hi): Likewise.
8193 (*vsx_le_perm_load_v16qi): Likewise.
8194 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
8195 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
8196 (*vsx_le_perm_store_v8hi): Likewise.
8197 (*vsx_le_perm_store_v16qi): Likewise.
8198 (eight unnamed splitters): Likewise.
8200 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8202 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
8203 * config/rs6000/emmintrin.h: Likewise.
8204 * config/rs6000/mmintrin.h: Likewise.
8205 * config/rs6000/xmmintrin.h: Likewise.
8207 2018-01-10 David Malcolm <dmalcolm@redhat.com>
8210 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
8212 * tree.c (tree_nop_conversion): Return true for location wrapper
8214 (maybe_wrap_with_location): New function.
8215 (selftest::check_strip_nops): New function.
8216 (selftest::test_location_wrappers): New function.
8217 (selftest::tree_c_tests): Call it.
8218 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
8219 (maybe_wrap_with_location): New decl.
8220 (EXPR_LOCATION_WRAPPER_P): New macro.
8221 (location_wrapper_p): New inline function.
8222 (tree_strip_any_location_wrapper): New inline function.
8224 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
8227 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
8228 stack_realign_offset for the largest alignment of stack slot
8230 (ix86_find_max_used_stack_alignment): New function.
8231 (ix86_finalize_stack_frame_flags): Use it. Set
8232 max_used_stack_alignment if we don't realign stack.
8233 * config/i386/i386.h (machine_function): Add
8234 max_used_stack_alignment.
8236 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
8238 * config/arm/arm.opt (-mbranch-cost): New option.
8239 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
8242 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
8245 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
8246 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
8248 2018-01-10 Richard Biener <rguenther@suse.de>
8251 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
8252 early out so it also covers the case where we have a non-NULL
8255 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8257 PR tree-optimization/83753
8258 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
8259 for non-strided grouped accesses if the number of elements is 1.
8261 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8264 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
8265 * i386.h (TARGET_USE_GATHER): Define.
8266 * x86-tune.def (X86_TUNE_USE_GATHER): New.
8268 2018-01-10 Martin Liska <mliska@suse.cz>
8271 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
8272 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
8274 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
8275 CLEANUP_NO_PARTITIONING is not set.
8277 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8279 * doc/rtl.texi: Remove documentation of (const ...) wrappers
8280 for vectors, as a partial revert of r254296.
8281 * rtl.h (const_vec_p): Delete.
8282 (const_vec_duplicate_p): Don't test for vector CONSTs.
8283 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
8284 * expmed.c (make_tree): Likewise.
8287 * common.md (E, F): Use CONSTANT_P instead of checking for
8289 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
8290 checking for CONST_VECTOR.
8292 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8295 * predict.c (force_edge_cold): Handle in more sane way edges
8298 2018-01-09 Carl Love <cel@us.ibm.com>
8300 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
8302 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
8303 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
8304 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
8305 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
8306 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
8307 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
8308 * config/rs6000/rs6000-protos.h: Add extern defition for
8309 rs6000_generate_float2_double_code.
8310 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
8312 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
8313 (float2_v2df): Add define_expand.
8315 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
8318 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
8319 op_mode in the force_to_mode call.
8321 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8323 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
8324 instead of checking each element individually.
8325 (aarch64_evpc_uzp): Likewise.
8326 (aarch64_evpc_zip): Likewise.
8327 (aarch64_evpc_ext): Likewise.
8328 (aarch64_evpc_rev): Likewise.
8329 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
8330 instead of checking each element individually. Return true without
8332 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
8333 whether all selected elements come from the same input, instead of
8334 checking each element individually. Remove calls to gen_rtx_REG,
8335 start_sequence and end_sequence and instead assert that no rtl is
8338 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8340 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
8341 order of HIGH and CONST checks.
8343 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8345 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
8346 if the destination isn't an SSA_NAME.
8348 2018-01-09 Richard Biener <rguenther@suse.de>
8350 PR tree-optimization/83668
8351 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
8353 (canonicalize_loop_form): ... here, renamed from ...
8354 (canonicalize_loop_closed_ssa_form): ... this and amended to
8355 swap successor edges for loop exit blocks to make us use
8356 the RPO order we need for initial schedule generation.
8358 2018-01-09 Joseph Myers <joseph@codesourcery.com>
8360 PR tree-optimization/64811
8361 * match.pd: When optimizing comparisons with Inf, avoid
8362 introducing or losing exceptions from comparisons with NaN.
8364 2018-01-09 Martin Liska <mliska@suse.cz>
8367 * asan.c (shadow_mem_size): Add gcc_assert.
8369 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
8371 Don't save registers in main().
8374 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
8375 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
8376 * config/avr/avr.c (avr_set_current_function): Don't error if
8377 naked, OS_task or OS_main are specified at the same time.
8378 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
8380 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
8382 * common/config/avr/avr-common.c (avr_option_optimization_table):
8383 Switch on -mmain-is-OS_task for optimizing compilations.
8385 2018-01-09 Richard Biener <rguenther@suse.de>
8387 PR tree-optimization/83572
8388 * graphite.c: Include cfganal.h.
8389 (graphite_transform_loops): Connect infinite loops to exit
8390 and remove fake edges at the end.
8392 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8394 * ipa-inline.c (edge_badness): Revert accidental checkin.
8396 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8399 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
8400 symbols; not inline clones.
8402 2018-01-09 Jakub Jelinek <jakub@redhat.com>
8405 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
8406 hard registers. Formatting fixes.
8408 PR preprocessor/83722
8409 * gcc.c (try_generate_repro): Pass
8410 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
8411 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
8414 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
8415 Kito Cheng <kito.cheng@gmail.com>
8417 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
8418 (riscv_leaf_function_p): Delete.
8419 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
8421 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8423 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
8425 (do_ifelse): New function.
8426 (do_isel): New function.
8427 (do_sub3): New function.
8428 (do_add3): New function.
8429 (do_load_mask_compare): New function.
8430 (do_overlap_load_compare): New function.
8431 (expand_compare_loop): New function.
8432 (expand_block_compare): Call expand_compare_loop() when appropriate.
8433 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
8435 (-mblock-compare-inline-loop-limit): New option.
8437 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8440 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
8441 Reverse order of second and third operands in first alternative.
8442 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
8443 of first and second elements in UNSPEC_VPERMR vector.
8444 (altivec_expand_vec_perm_le): Likewise.
8446 2018-01-08 Jeff Law <law@redhat.com>
8448 PR rtl-optimizatin/81308
8449 * tree-switch-conversion.c (cfg_altered): New file scoped static.
8450 (process_switch): If group_case_labels makes a change, then set
8452 (pass_convert_switch::execute): If a switch is converted, then
8453 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
8455 PR rtl-optimization/81308
8456 * recog.c (split_all_insns): Conditionally cleanup the CFG after
8459 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
8461 PR target/83663 - Revert r255946
8462 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
8463 generation for cases where splatting a value is not useful.
8464 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
8465 across a vec_duplicate and a paradoxical subreg forming a vector
8466 mode to a vec_concat.
8468 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8470 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
8471 -march=armv8.3-a variants.
8472 * config/arm/t-multilib: Likewise.
8473 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
8475 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8477 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
8479 (cceq_ior_compare_complement): Give it a name so I can use it, and
8480 change boolean_or_operator predicate to boolean_operator so it can
8481 be used to generate a crand.
8482 (eqne): New code iterator.
8483 (bd/bd_neg): New code_attrs.
8484 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
8485 a single define_insn.
8486 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
8487 decrement (bdnzt/bdnzf/bdzt/bdzf).
8488 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
8489 with the new names of the branch decrement patterns, and added the
8490 names of the branch decrement conditional patterns.
8492 2018-01-08 Richard Biener <rguenther@suse.de>
8494 PR tree-optimization/83563
8495 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
8498 2018-01-08 Richard Biener <rguenther@suse.de>
8501 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
8503 2018-01-08 Richard Biener <rguenther@suse.de>
8505 PR tree-optimization/83685
8506 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
8507 references to abnormals.
8509 2018-01-08 Richard Biener <rguenther@suse.de>
8512 * dwarf2out.c (output_indirect_strings): Handle empty
8513 skeleton_debug_str_hash.
8514 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
8516 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
8518 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
8519 (emit_store_direct): Likewise.
8520 (arc_trampoline_adjust_address): Likewise.
8521 (arc_asm_trampoline_template): New function.
8522 (arc_initialize_trampoline): Use asm_trampoline_template.
8523 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
8524 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
8525 * config/arc/arc.md (flush_icache): Delete pattern.
8527 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
8529 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
8530 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
8533 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
8536 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
8537 by not USED_FOR_TARGET.
8538 (make_pass_resolve_sw_modes): Likewise.
8540 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
8542 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
8545 2018-01-08 Richard Biener <rguenther@suse.de>
8548 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
8550 2018-01-08 Richard Biener <rguenther@suse.de>
8553 * match.pd ((t * 2) / 2) -> t): Add missing :c.
8555 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
8558 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
8559 basic blocks with a small number of successors.
8560 (convert_control_dep_chain_into_preds): Improve handling of
8562 (dump_predicates): Split apart into...
8563 (dump_pred_chain): ...here...
8564 (dump_pred_info): ...and here.
8565 (can_one_predicate_be_invalidated_p): Add debugging printfs.
8566 (can_chain_union_be_invalidated_p): Improve check for invalidation
8568 (uninit_uses_cannot_happen): Avoid unnecessary if
8569 convert_control_dep_chain_into_preds yielded nothing.
8571 2018-01-06 Martin Sebor <msebor@redhat.com>
8573 PR tree-optimization/83640
8574 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
8575 subtracting negative offset from size.
8576 (builtin_access::overlap): Adjust offset bounds of the access to fall
8577 within the size of the object if possible.
8579 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
8581 PR rtl-optimization/83699
8582 * expmed.c (extract_bit_field_1): Restrict the vector usage of
8583 extract_bit_field_as_subreg to cases in which the extracted
8584 value is also a vector.
8586 * lra-constraints.c (process_alt_operands): Test for the equivalence
8587 substitutions when detecting a possible reload cycle.
8589 2018-01-06 Jakub Jelinek <jakub@redhat.com>
8592 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
8593 by default if flag_selective_schedling{,2}. Formatting fixes.
8595 PR rtl-optimization/83682
8596 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
8597 if it has non-VECTOR_MODE element mode.
8598 (vec_duplicate_p): Likewise.
8601 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
8602 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
8604 2018-01-05 Jakub Jelinek <jakub@redhat.com>
8607 * config/i386/i386-builtin.def
8608 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
8609 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
8610 Require also OPTION_MASK_ISA_AVX512F in addition to
8611 OPTION_MASK_ISA_GFNI.
8612 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
8613 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
8614 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
8615 to OPTION_MASK_ISA_GFNI.
8616 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
8617 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
8618 OPTION_MASK_ISA_AVX512BW.
8619 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
8620 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
8621 addition to OPTION_MASK_ISA_GFNI.
8622 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
8623 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
8624 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
8625 to OPTION_MASK_ISA_GFNI.
8626 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
8627 a requirement for all ISAs rather than any of them with a few
8629 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
8631 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
8632 bitmasks to be enabled with 3 exceptions, instead of requiring any
8633 enabled ISA with lots of exceptions.
8634 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
8635 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
8636 Change avx512bw in isa attribute to avx512f.
8637 * config/i386/sgxintrin.h: Add license boilerplate.
8638 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
8639 to __AVX512F__ and __AVX512VL to __AVX512VL__.
8640 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
8641 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
8643 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
8644 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
8645 temporarily sse2 rather than sse if not enabled already.
8648 * config/i386/sse.md (VI248_VLBW): Rename to ...
8649 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
8650 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
8651 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
8652 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
8653 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
8654 mode iterator instead of VI248_VLBW.
8656 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
8658 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
8659 (record_modified): Skip clobbers; add debug output.
8660 (param_change_prob): Use sreal frequencies.
8662 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8664 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
8665 punt for user-aligned variables.
8667 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8669 * tree-chrec.c (chrec_contains_symbols): Return true for
8672 2018-01-05 Sudakshina Das <sudi.das@arm.com>
8675 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
8676 of (x|y) == x for BICS pattern.
8678 2018-01-05 Jakub Jelinek <jakub@redhat.com>
8680 PR tree-optimization/83605
8681 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
8682 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
8685 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
8687 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
8688 * config/epiphany/rtems.h: New file.
8690 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8691 Uros Bizjak <ubizjak@gmail.com>
8694 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
8695 QIreg_operand instead of register_operand predicate.
8696 * config/i386/i386.c (ix86_rop_should_change_byte_p,
8697 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
8698 comments instead of -fmitigate[-_]rop.
8700 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8703 * cgraphunit.c (symbol_table::compile): Switch to text_section
8704 before calling assembly_start debug hook.
8705 * run-rtl-passes.c (run_rtl_passes): Likewise.
8708 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8710 * tree-vrp.c (extract_range_from_binary_expr_1): Check
8711 range_int_cst_p rather than !symbolic_range_p before calling
8712 extract_range_from_multiplicative_op_1.
8714 2018-01-04 Jeff Law <law@redhat.com>
8716 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
8717 redundant test in assertion.
8719 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8721 * doc/rtl.texi: Document machine_mode wrapper classes.
8723 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8725 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
8728 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8730 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
8731 the VEC_PERM_EXPR fold to fail.
8733 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8736 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
8737 to switched_sections.
8739 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8742 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
8745 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
8748 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
8749 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
8751 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8754 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
8755 is BLKmode and bitpos not zero or mode change is needed.
8757 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8760 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
8763 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
8766 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
8767 instead of MULT rtx. Update all corresponding splitters.
8769 (*ssub<modesuffix>): Ditto.
8771 (*cmp_sadd_di): Update split patterns.
8772 (*cmp_sadd_si): Ditto.
8773 (*cmp_sadd_sidi): Ditto.
8774 (*cmp_ssub_di): Ditto.
8775 (*cmp_ssub_si): Ditto.
8776 (*cmp_ssub_sidi): Ditto.
8777 * config/alpha/predicates.md (const23_operand): New predicate.
8778 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
8779 Look for ASHIFT, not MULT inner operand.
8780 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
8782 2018-01-04 Martin Liska <mliska@suse.cz>
8784 PR gcov-profile/83669
8785 * gcov.c (output_intermediate_file): Add version to intermediate
8787 * doc/gcov.texi: Document new field 'version' in intermediate
8788 file format. Fix location of '-k' option of gcov command.
8790 2018-01-04 Martin Liska <mliska@suse.cz>
8793 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
8795 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8797 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
8799 2018-01-03 Martin Sebor <msebor@redhat.com>
8801 PR tree-optimization/83655
8802 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
8803 checking calls with invalid arguments.
8805 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8807 * tree-vect-stmts.c (vect_get_store_rhs): New function.
8808 (vectorizable_mask_load_store): Delete.
8809 (vectorizable_call): Return false for masked loads and stores.
8810 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
8811 instead of gimple_assign_rhs1.
8812 (vectorizable_load): Handle IFN_MASK_LOAD.
8813 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
8815 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8817 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
8819 (vectorizable_mask_load_store): ...here.
8820 (vectorizable_load): ...and here.
8822 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8824 * tree-vect-stmts.c (vect_build_all_ones_mask)
8825 (vect_build_zero_merge_argument): New functions, split out from...
8826 (vectorizable_load): ...here.
8828 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8830 * tree-vect-stmts.c (vect_check_store_rhs): New function,
8832 (vectorizable_mask_load_store): ...here.
8833 (vectorizable_store): ...and here.
8835 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8837 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
8839 (vectorizable_mask_load_store): ...here.
8841 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8843 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
8844 (vect_model_store_cost): Take a vec_load_store_type instead of a
8846 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
8847 (vect_model_store_cost): Take a vec_load_store_type instead of a
8849 (vectorizable_mask_load_store): Update accordingly.
8850 (vectorizable_store): Likewise.
8851 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
8853 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8855 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
8856 IFN_MASK_LOAD calls here rather than...
8857 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
8859 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8860 Alan Hayward <alan.hayward@arm.com>
8861 David Sherwood <david.sherwood@arm.com>
8863 * expmed.c (extract_bit_field_1): For vector extracts,
8864 fall back to extract_bit_field_as_subreg if vec_extract
8867 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8868 Alan Hayward <alan.hayward@arm.com>
8869 David Sherwood <david.sherwood@arm.com>
8871 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
8872 they are variable or constant sized.
8873 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
8874 slots for constant-sized data.
8876 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8877 Alan Hayward <alan.hayward@arm.com>
8878 David Sherwood <david.sherwood@arm.com>
8880 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
8881 handling COND_EXPRs with boolean comparisons, try to find a better
8882 basis for the mask type than the boolean itself.
8884 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8886 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
8887 is calculated and how it can be overridden.
8888 * genmodes.c (max_bitsize_mode_any_mode): New variable.
8889 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
8891 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
8894 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8895 Alan Hayward <alan.hayward@arm.com>
8896 David Sherwood <david.sherwood@arm.com>
8898 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
8899 Remove the mode argument.
8900 (aarch64_simd_valid_immediate): Remove the mode and inverse
8902 * config/aarch64/iterators.md (bitsize): New iterator.
8903 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
8904 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
8905 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
8906 aarch64_simd_valid_immediate.
8907 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
8908 (aarch64_reg_or_bic_imm): Likewise.
8909 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
8910 with an insn_type enum and msl with a modifier_type enum.
8911 Replace element_width with a scalar_mode. Change the shift
8912 to unsigned int. Add constructors for scalar_float_mode and
8913 scalar_int_mode elements.
8914 (aarch64_vect_float_const_representable_p): Delete.
8915 (aarch64_can_const_movi_rtx_p)
8916 (aarch64_simd_scalar_immediate_valid_for_move)
8917 (aarch64_simd_make_constant): Update call to
8918 aarch64_simd_valid_immediate.
8919 (aarch64_advsimd_valid_immediate_hs): New function.
8920 (aarch64_advsimd_valid_immediate): Likewise.
8921 (aarch64_simd_valid_immediate): Remove mode and inverse
8922 arguments. Rewrite to use the above. Use const_vec_duplicate_p
8923 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
8924 and aarch64_float_const_representable_p on the result.
8925 (aarch64_output_simd_mov_immediate): Remove mode argument.
8926 Update call to aarch64_simd_valid_immediate and use of
8927 simd_immediate_info.
8928 (aarch64_output_scalar_simd_mov_immediate): Update call
8931 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8932 Alan Hayward <alan.hayward@arm.com>
8933 David Sherwood <david.sherwood@arm.com>
8935 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
8936 (mode_nunits): Likewise CONST_MODE_NUNITS.
8937 * machmode.def (ADJUST_NUNITS): Document.
8938 * genmodes.c (mode_data::need_nunits_adj): New field.
8939 (blank_mode): Update accordingly.
8940 (adj_nunits): New variable.
8941 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
8943 (emit_mode_size_inline): Set need_bytesize_adj for all modes
8944 listed in adj_nunits.
8945 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
8946 listed in adj_nunits. Don't emit case statements for such modes.
8947 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
8948 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
8949 nothing if adj_nunits is nonnull.
8950 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
8951 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
8952 (emit_mode_fbit): Update use of print_maybe_const_decl.
8953 (emit_move_size): Likewise. Treat the array as non-const
8955 (emit_mode_adjustments): Handle adj_nunits.
8957 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8959 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
8960 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
8961 (VECTOR_MODES): Use it.
8962 (make_vector_modes): Take the prefix as an argument.
8964 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8965 Alan Hayward <alan.hayward@arm.com>
8966 David Sherwood <david.sherwood@arm.com>
8968 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
8969 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
8970 for MODE_VECTOR_BOOL.
8971 * machmode.def (VECTOR_BOOL_MODE): Document.
8972 * genmodes.c (VECTOR_BOOL_MODE): New macro.
8973 (make_vector_bool_mode): New function.
8974 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
8976 * lto-streamer-in.c (lto_input_mode_table): Likewise.
8977 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
8979 * stor-layout.c (int_mode_for_mode): Likewise.
8980 * tree.c (build_vector_type_for_mode): Likewise.
8981 * varasm.c (output_constant_pool_2): Likewise.
8982 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
8983 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
8984 for MODE_VECTOR_BOOL.
8985 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
8986 of mode class checks.
8987 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
8988 instead of a list of mode class checks.
8989 (expand_vector_scalar_condition): Likewise.
8990 (type_for_widest_vector_mode): Handle BImode as an inner mode.
8992 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8993 Alan Hayward <alan.hayward@arm.com>
8994 David Sherwood <david.sherwood@arm.com>
8996 * machmode.h (mode_size): Change from unsigned short to
8998 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
8999 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9000 or if measurement_type is not polynomial.
9001 (fixed_size_mode::includes_p): Check for constant-sized modes.
9002 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
9003 return a poly_uint16 rather than an unsigned short.
9004 (emit_mode_size): Change the type of mode_size from unsigned short
9005 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
9006 (emit_mode_adjustments): Cope with polynomial vector sizes.
9007 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9009 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9011 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
9012 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
9013 * caller-save.c (setup_save_areas): Likewise.
9014 (replace_reg_with_saved_mem): Likewise.
9015 * calls.c (emit_library_call_value_1): Likewise.
9016 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
9017 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
9018 (gen_lowpart_for_combine): Likewise.
9019 * convert.c (convert_to_integer_1): Likewise.
9020 * cse.c (equiv_constant, cse_insn): Likewise.
9021 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
9022 (cselib_subst_to_values): Likewise.
9023 * dce.c (word_dce_process_block): Likewise.
9024 * df-problems.c (df_word_lr_mark_ref): Likewise.
9025 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
9026 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
9027 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
9028 (rtl_for_decl_location): Likewise.
9029 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
9030 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
9031 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
9032 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
9033 (expand_expr_real_1): Likewise.
9034 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
9035 (pad_below): Likewise.
9036 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9037 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
9038 * ira.c (get_subreg_tracking_sizes): Likewise.
9039 * ira-build.c (ira_create_allocno_objects): Likewise.
9040 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
9041 (ira_sort_regnos_for_alter_reg): Likewise.
9042 * ira-costs.c (record_operand_costs): Likewise.
9043 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
9044 (resolve_simple_move): Likewise.
9045 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
9046 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
9047 (lra_constraints): Likewise.
9048 (CONST_POOL_OK_P): Reject variable-sized modes.
9049 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
9050 (add_pseudo_to_slot, lra_spill): Likewise.
9051 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9052 * optabs-query.c (get_best_extraction_insn): Likewise.
9053 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9054 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
9055 (expand_mult_highpart, valid_multiword_target_p): Likewise.
9056 * recog.c (offsettable_address_addr_space_p): Likewise.
9057 * regcprop.c (maybe_mode_change): Likewise.
9058 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
9059 * regrename.c (build_def_use): Likewise.
9060 * regstat.c (dump_reg_info): Likewise.
9061 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
9062 (find_reloads, find_reloads_subreg_address): Likewise.
9063 * reload1.c (eliminate_regs_1): Likewise.
9064 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
9065 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
9066 (simplify_binary_operation_1, simplify_subreg): Likewise.
9067 * targhooks.c (default_function_arg_padding): Likewise.
9068 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
9069 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
9070 (verify_gimple_assign_ternary): Likewise.
9071 * tree-inline.c (estimate_move_cost): Likewise.
9072 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9073 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
9074 (get_address_cost_ainc): Likewise.
9075 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
9076 (vect_supportable_dr_alignment): Likewise.
9077 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9078 (vectorizable_reduction): Likewise.
9079 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
9080 (vectorizable_operation, vectorizable_load): Likewise.
9081 * tree.c (build_same_sized_truth_vector_type): Likewise.
9082 * valtrack.c (cleanup_auto_inc_dec): Likewise.
9083 * var-tracking.c (emit_note_insn_var_location): Likewise.
9084 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
9085 (ADDR_VEC_ALIGN): Likewise.
9087 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9088 Alan Hayward <alan.hayward@arm.com>
9089 David Sherwood <david.sherwood@arm.com>
9091 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
9093 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9094 or if measurement_type is polynomial.
9095 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
9096 * combine.c (make_extraction): Likewise.
9097 * dse.c (find_shift_sequence): Likewise.
9098 * dwarf2out.c (mem_loc_descriptor): Likewise.
9099 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
9100 (extract_bit_field, extract_low_bits): Likewise.
9101 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
9102 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
9103 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
9104 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
9105 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9106 * reload.c (find_reloads): Likewise.
9107 * reload1.c (alter_reg): Likewise.
9108 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
9109 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
9110 * tree-if-conv.c (predicate_mem_writes): Likewise.
9111 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
9112 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
9113 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
9114 * valtrack.c (dead_debug_insert_temp): Likewise.
9115 * varasm.c (mergeable_constant_section): Likewise.
9116 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
9118 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9119 Alan Hayward <alan.hayward@arm.com>
9120 David Sherwood <david.sherwood@arm.com>
9122 * expr.c (expand_assignment): Cope with polynomial mode sizes
9123 when assigning to a CONCAT.
9125 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9126 Alan Hayward <alan.hayward@arm.com>
9127 David Sherwood <david.sherwood@arm.com>
9129 * machmode.h (mode_precision): Change from unsigned short to
9131 (mode_to_precision): Return a poly_uint16 rather than an unsigned
9133 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
9134 or if measurement_type is not polynomial.
9135 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
9136 in which the mode is already known to be a scalar_int_mode.
9137 * genmodes.c (emit_mode_precision): Change the type of mode_precision
9138 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
9140 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9141 for GET_MODE_PRECISION.
9142 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9143 for GET_MODE_PRECISION.
9144 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
9146 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
9147 (expand_field_assignment, make_extraction): Likewise.
9148 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
9149 (get_last_value): Likewise.
9150 * convert.c (convert_to_integer_1): Likewise.
9151 * cse.c (cse_insn): Likewise.
9152 * expr.c (expand_expr_real_1): Likewise.
9153 * lra-constraints.c (simplify_operand_subreg): Likewise.
9154 * optabs-query.c (can_atomic_load_p): Likewise.
9155 * optabs.c (expand_atomic_load): Likewise.
9156 (expand_atomic_store): Likewise.
9157 * ree.c (combine_reaching_defs): Likewise.
9158 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
9159 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
9160 * tree.h (type_has_mode_precision_p): Likewise.
9161 * ubsan.c (instrument_si_overflow): Likewise.
9163 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9164 Alan Hayward <alan.hayward@arm.com>
9165 David Sherwood <david.sherwood@arm.com>
9167 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
9168 polynomial numbers of units.
9169 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
9170 (valid_vector_subparts_p): New function.
9171 (build_vector_type): Remove temporary shim and take the number
9172 of units as a poly_uint64 rather than an int.
9173 (build_opaque_vector_type): Take the number of units as a
9174 poly_uint64 rather than an int.
9175 * tree.c (build_vector_from_ctor): Handle polynomial
9176 TYPE_VECTOR_SUBPARTS.
9177 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
9178 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
9179 (build_vector_from_val): If the number of units is variable,
9180 use build_vec_duplicate_cst for constant operands and
9181 VEC_DUPLICATE_EXPR otherwise.
9182 (make_vector_type): Remove temporary is_constant ().
9183 (build_vector_type, build_opaque_vector_type): Take the number of
9184 units as a poly_uint64 rather than an int.
9185 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
9187 * cfgexpand.c (expand_debug_expr): Likewise.
9188 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
9189 (store_constructor, expand_expr_real_1): Likewise.
9190 (const_scalar_mask_from_tree): Likewise.
9191 * fold-const-call.c (fold_const_reduction): Likewise.
9192 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
9193 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
9194 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
9195 (fold_relational_const): Likewise.
9196 (native_interpret_vector): Likewise. Change the size from an
9197 int to an unsigned int.
9198 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
9199 TYPE_VECTOR_SUBPARTS.
9200 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
9201 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
9202 duplicating a non-constant operand into a variable-length vector.
9203 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
9204 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
9205 * ipa-icf.c (sem_variable::equals): Likewise.
9206 * match.pd: Likewise.
9207 * omp-simd-clone.c (simd_clone_subparts): Likewise.
9208 * print-tree.c (print_node): Likewise.
9209 * stor-layout.c (layout_type): Likewise.
9210 * targhooks.c (default_builtin_vectorization_cost): Likewise.
9211 * tree-cfg.c (verify_gimple_comparison): Likewise.
9212 (verify_gimple_assign_binary): Likewise.
9213 (verify_gimple_assign_ternary): Likewise.
9214 (verify_gimple_assign_single): Likewise.
9215 * tree-pretty-print.c (dump_generic_node): Likewise.
9216 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9217 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
9218 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
9219 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
9220 (vect_shift_permute_load_chain): Likewise.
9221 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
9222 (expand_vector_condition, optimize_vector_constructor): Likewise.
9223 (lower_vec_perm, get_compute_type): Likewise.
9224 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9225 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
9226 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
9227 (vect_recog_mask_conversion_pattern): Likewise.
9228 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
9229 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
9230 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9231 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
9232 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
9233 (vectorizable_shift, vectorizable_operation, vectorizable_store)
9234 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
9235 (supportable_widening_operation): Likewise.
9236 (supportable_narrowing_operation): Likewise.
9237 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
9239 * varasm.c (output_constant): Likewise.
9241 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9242 Alan Hayward <alan.hayward@arm.com>
9243 David Sherwood <david.sherwood@arm.com>
9245 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
9246 so that both the length == 3 and length != 3 cases set up their
9247 own permute vectors. Add comments explaining why we know the
9248 number of elements is constant.
9249 (vect_permute_load_chain): Likewise.
9251 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9252 Alan Hayward <alan.hayward@arm.com>
9253 David Sherwood <david.sherwood@arm.com>
9255 * machmode.h (mode_nunits): Change from unsigned char to
9257 (ONLY_FIXED_SIZE_MODES): New macro.
9258 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
9259 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
9260 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
9262 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
9263 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
9264 or if measurement_type is not polynomial.
9265 * genmodes.c (ZERO_COEFFS): New macro.
9266 (emit_mode_nunits_inline): Make mode_nunits_inline return a
9268 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
9269 Use ZERO_COEFFS when emitting initializers.
9270 * data-streamer.h (bp_pack_poly_value): New function.
9271 (bp_unpack_poly_value): Likewise.
9272 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9273 for GET_MODE_NUNITS.
9274 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9275 for GET_MODE_NUNITS.
9276 * tree.c (make_vector_type): Remove temporary shim and make
9277 the real function take the number of units as a poly_uint64
9279 (build_vector_type_for_mode): Handle polynomial nunits.
9280 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
9281 * emit-rtl.c (const_vec_series_p_1): Likewise.
9282 (gen_rtx_CONST_VECTOR): Likewise.
9283 * fold-const.c (test_vec_duplicate_folding): Likewise.
9284 * genrecog.c (validate_pattern): Likewise.
9285 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
9286 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9287 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
9288 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
9289 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
9290 * rtlanal.c (subreg_get_info): Likewise.
9291 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9292 (vect_grouped_load_supported): Likewise.
9293 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
9294 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
9295 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
9296 (simplify_const_unary_operation, simplify_binary_operation_1)
9297 (simplify_const_binary_operation, simplify_ternary_operation)
9298 (test_vector_ops_duplicate, test_vector_ops): Likewise.
9299 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
9300 instead of CONST_VECTOR_NUNITS.
9301 * varasm.c (output_constant_pool_2): Likewise.
9302 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
9303 explicit-encoded elements in the XVEC for variable-length vectors.
9305 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9307 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
9309 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9310 Alan Hayward <alan.hayward@arm.com>
9311 David Sherwood <david.sherwood@arm.com>
9313 * coretypes.h (fixed_size_mode): Declare.
9314 (fixed_size_mode_pod): New typedef.
9315 * builtins.h (target_builtins::x_apply_args_mode)
9316 (target_builtins::x_apply_result_mode): Change type to
9317 fixed_size_mode_pod.
9318 * builtins.c (apply_args_size, apply_result_size, result_vector)
9319 (expand_builtin_apply_args_1, expand_builtin_apply)
9320 (expand_builtin_return): Update accordingly.
9322 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9324 * cse.c (hash_rtx_cb): Hash only the encoded elements.
9325 * cselib.c (cselib_hash_rtx): Likewise.
9326 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
9327 CONST_VECTOR encoding.
9329 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9330 Jeff Law <law@redhat.com>
9333 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
9334 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
9335 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
9336 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
9339 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
9340 explicitly probe *sp in a noreturn function if there were any callee
9341 register saves or frame pointer is needed.
9343 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9346 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
9347 BLKmode for ternary, binary or unary expressions.
9350 * var-tracking.c (delete_vta_debug_insn): New inline function.
9351 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
9352 insns from get_insns () to NULL instead of each bb separately.
9353 Use delete_vta_debug_insn. No longer static.
9354 (vt_debug_insns_local, variable_tracking_main_1): Adjust
9355 delete_vta_debug_insns callers.
9356 * rtl.h (delete_vta_debug_insns): Declare.
9357 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
9358 instead of variable_tracking_main.
9360 2018-01-03 Martin Sebor <msebor@redhat.com>
9362 PR tree-optimization/83603
9363 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
9364 arguments past the endof the argument list in functions declared
9365 without a prototype.
9366 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
9367 Avoid checking when arguments are null.
9369 2018-01-03 Martin Sebor <msebor@redhat.com>
9372 * doc/extend.texi (attribute const): Fix a typo.
9373 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
9374 issuing -Wsuggest-attribute for void functions.
9376 2018-01-03 Martin Sebor <msebor@redhat.com>
9378 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
9379 offset_int::from instead of wide_int::to_shwi.
9380 (maybe_diag_overlap): Remove assertion.
9381 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
9382 * gimple-ssa-sprintf.c (format_directive): Same.
9383 (parse_directive): Same.
9384 (sprintf_dom_walker::compute_format_length): Same.
9385 (try_substitute_return_value): Same.
9387 2018-01-03 Jeff Law <law@redhat.com>
9390 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
9391 non-constant residual for zero at runtime and avoid probing in
9392 that case. Reorganize code for trailing problem to mirror handling
9395 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9397 PR tree-optimization/83501
9398 * tree-ssa-strlen.c (get_string_cst): New.
9399 (handle_char_store): Call get_string_cst.
9401 2018-01-03 Martin Liska <mliska@suse.cz>
9403 PR tree-optimization/83593
9404 * tree-ssa-strlen.c: Include tree-cfg.h.
9405 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
9406 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
9407 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
9409 (strlen_dom_walker::before_dom_children): Call
9410 gimple_purge_dead_eh_edges. Dump tranformation with details
9412 (strlen_dom_walker::before_dom_children): Update call by adding
9413 new argument cleanup_eh.
9414 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
9416 2018-01-03 Martin Liska <mliska@suse.cz>
9419 * cif-code.def (VARIADIC_THUNK): New enum value.
9420 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
9423 2018-01-03 Jan Beulich <jbeulich@suse.com>
9425 * sse.md (mov<mode>_internal): Tighten condition for when to use
9426 vmovdqu<ssescalarsize> for TI and OI modes.
9428 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9430 Update copyright years.
9432 2018-01-03 Martin Liska <mliska@suse.cz>
9435 * ipa-visibility.c (function_and_variable_visibility): Skip
9436 functions with noipa attribure.
9438 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9440 * gcc.c (process_command): Update copyright notice dates.
9441 * gcov-dump.c (print_version): Ditto.
9442 * gcov.c (print_version): Ditto.
9443 * gcov-tool.c (print_version): Ditto.
9444 * gengtype.c (create_file): Ditto.
9445 * doc/cpp.texi: Bump @copying's copyright year.
9446 * doc/cppinternals.texi: Ditto.
9447 * doc/gcc.texi: Ditto.
9448 * doc/gccint.texi: Ditto.
9449 * doc/gcov.texi: Ditto.
9450 * doc/install.texi: Ditto.
9451 * doc/invoke.texi: Ditto.
9453 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9455 * vector-builder.h (vector_builder::m_full_nelts): Change from
9456 unsigned int to poly_uint64.
9457 (vector_builder::full_nelts): Update prototype accordingly.
9458 (vector_builder::new_vector): Likewise.
9459 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
9460 (vector_builder::operator ==): Likewise.
9461 (vector_builder::finalize): Likewise.
9462 * int-vector-builder.h (int_vector_builder::int_vector_builder):
9463 Take the number of elements as a poly_uint64 rather than an
9465 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
9466 from unsigned int to poly_uint64.
9467 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
9468 (vec_perm_indices::new_vector): Likewise.
9469 (vec_perm_indices::length): Likewise.
9470 (vec_perm_indices::nelts_per_input): Likewise.
9471 (vec_perm_indices::input_nelts): Likewise.
9472 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
9473 number of elements per input as a poly_uint64 rather than an
9474 unsigned int. Use the original encoding for variable-length
9475 vectors, rather than clamping each individual element.
9476 For the second and subsequent elements in each pattern,
9477 clamp the step and base before clamping their sum.
9478 (vec_perm_indices::series_p): Handle polynomial element counts.
9479 (vec_perm_indices::all_in_range_p): Likewise.
9480 (vec_perm_indices_to_tree): Likewise.
9481 (vec_perm_indices_to_rtx): Likewise.
9482 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
9483 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
9484 (tree_vector_builder::new_binary_operation): Handle polynomial
9485 element counts. Return false if we need to know the number
9486 of elements at compile time.
9487 * fold-const.c (fold_vec_perm): Punt if the number of elements
9488 isn't known at compile time.
9490 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9492 * vec-perm-indices.h (vec_perm_builder): Change element type
9493 from HOST_WIDE_INT to poly_int64.
9494 (vec_perm_indices::element_type): Update accordingly.
9495 (vec_perm_indices::clamp): Handle polynomial element_types.
9496 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9497 (vec_perm_indices::all_in_range_p): Likewise.
9498 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
9500 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
9501 polynomial vec_perm_indices element types.
9502 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
9503 * fold-const.c (fold_vec_perm): Likewise.
9504 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
9505 * tree-vect-generic.c (lower_vec_perm): Likewise.
9506 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9507 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
9508 element type to HOST_WIDE_INT.
9510 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9511 Alan Hayward <alan.hayward@arm.com>
9512 David Sherwood <david.sherwood@arm.com>
9514 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
9515 rather than an int. Use plus_constant.
9516 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
9517 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
9519 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9520 Alan Hayward <alan.hayward@arm.com>
9521 David Sherwood <david.sherwood@arm.com>
9523 * calls.c (emit_call_1, expand_call): Change struct_value_size from
9524 a HOST_WIDE_INT to a poly_int64.
9526 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9527 Alan Hayward <alan.hayward@arm.com>
9528 David Sherwood <david.sherwood@arm.com>
9530 * calls.c (load_register_parameters): Cope with polynomial
9531 mode sizes. Require a constant size for BLKmode parameters
9532 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
9533 forces a parameter to be padded at the lsb end in order to
9534 fill a complete number of words, require the parameter size
9535 to be ordered wrt UNITS_PER_WORD.
9537 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9538 Alan Hayward <alan.hayward@arm.com>
9539 David Sherwood <david.sherwood@arm.com>
9541 * reload1.c (spill_stack_slot_width): Change element type
9542 from unsigned int to poly_uint64_pod.
9543 (alter_reg): Treat mode sizes as polynomial.
9545 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9546 Alan Hayward <alan.hayward@arm.com>
9547 David Sherwood <david.sherwood@arm.com>
9549 * reload.c (complex_word_subreg_p): New function.
9550 (reload_inner_reg_of_subreg, push_reload): Use it.
9552 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9553 Alan Hayward <alan.hayward@arm.com>
9554 David Sherwood <david.sherwood@arm.com>
9556 * lra-constraints.c (process_alt_operands): Reject matched
9557 operands whose sizes aren't ordered.
9558 (match_reload): Refer to this check here.
9560 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9561 Alan Hayward <alan.hayward@arm.com>
9562 David Sherwood <david.sherwood@arm.com>
9564 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
9565 that the mode size is in the set {1, 2, 4, 8, 16}.
9567 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9568 Alan Hayward <alan.hayward@arm.com>
9569 David Sherwood <david.sherwood@arm.com>
9571 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
9572 Use plus_constant instead of gen_rtx_PLUS.
9574 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9575 Alan Hayward <alan.hayward@arm.com>
9576 David Sherwood <david.sherwood@arm.com>
9578 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
9579 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
9580 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
9581 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
9582 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
9583 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
9584 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
9585 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
9586 * config/i386/i386.c (ix86_push_rounding): ...this new function.
9587 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
9589 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
9590 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
9591 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
9592 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
9593 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
9594 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
9595 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
9596 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
9597 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
9598 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
9600 * expr.c (emit_move_resolve_push): Treat the input and result
9601 of PUSH_ROUNDING as a poly_int64.
9602 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
9603 (emit_push_insn): Likewise.
9604 * lra-eliminations.c (mark_not_eliminable): Likewise.
9605 * recog.c (push_operand): Likewise.
9606 * reload1.c (elimination_effects): Likewise.
9607 * rtlanal.c (nonzero_bits1): Likewise.
9608 * calls.c (store_one_arg): Likewise. Require the padding to be
9609 known at compile time.
9611 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9612 Alan Hayward <alan.hayward@arm.com>
9613 David Sherwood <david.sherwood@arm.com>
9615 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
9616 Use plus_constant instead of gen_rtx_PLUS.
9618 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9619 Alan Hayward <alan.hayward@arm.com>
9620 David Sherwood <david.sherwood@arm.com>
9622 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
9625 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9626 Alan Hayward <alan.hayward@arm.com>
9627 David Sherwood <david.sherwood@arm.com>
9629 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
9630 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
9631 via stack temporaries. Treat the mode size as polynomial too.
9633 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9634 Alan Hayward <alan.hayward@arm.com>
9635 David Sherwood <david.sherwood@arm.com>
9637 * expr.c (expand_expr_real_2): When handling conversions involving
9638 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
9639 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
9640 as a poly_uint64 too.
9642 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9643 Alan Hayward <alan.hayward@arm.com>
9644 David Sherwood <david.sherwood@arm.com>
9646 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
9648 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9649 Alan Hayward <alan.hayward@arm.com>
9650 David Sherwood <david.sherwood@arm.com>
9652 * combine.c (can_change_dest_mode): Handle polynomial
9653 REGMODE_NATURAL_SIZE.
9654 * expmed.c (store_bit_field_1): Likewise.
9655 * expr.c (store_constructor): Likewise.
9656 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
9657 and polynomial REGMODE_NATURAL_SIZE.
9658 (gen_lowpart_common): Likewise.
9659 * reginfo.c (record_subregs_of_mode): Likewise.
9660 * rtlanal.c (read_modify_subreg_p): Likewise.
9662 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9663 Alan Hayward <alan.hayward@arm.com>
9664 David Sherwood <david.sherwood@arm.com>
9666 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
9667 numbers of elements.
9669 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9670 Alan Hayward <alan.hayward@arm.com>
9671 David Sherwood <david.sherwood@arm.com>
9673 * match.pd: Cope with polynomial numbers of vector elements.
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 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
9680 in a POINTER_PLUS_EXPR.
9682 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9683 Alan Hayward <alan.hayward@arm.com>
9684 David Sherwood <david.sherwood@arm.com>
9686 * omp-simd-clone.c (simd_clone_subparts): New function.
9687 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
9688 (ipa_simd_modify_function_body): Likewise.
9690 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9691 Alan Hayward <alan.hayward@arm.com>
9692 David Sherwood <david.sherwood@arm.com>
9694 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
9695 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
9696 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
9697 (expand_vector_condition, vector_element): Likewise.
9698 (subparts_gt): New function.
9699 (get_compute_type): Use subparts_gt.
9700 (count_type_subparts): Delete.
9701 (expand_vector_operations_1): Use subparts_gt instead of
9702 count_type_subparts.
9704 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9705 Alan Hayward <alan.hayward@arm.com>
9706 David Sherwood <david.sherwood@arm.com>
9708 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
9709 (vect_compile_time_alias): ...this new function. Do the calculation
9710 on poly_ints rather than trees.
9711 (vect_prune_runtime_alias_test_list): Update call accordingly.
9713 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9714 Alan Hayward <alan.hayward@arm.com>
9715 David Sherwood <david.sherwood@arm.com>
9717 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
9719 (vect_schedule_slp_instance): Likewise.
9721 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9722 Alan Hayward <alan.hayward@arm.com>
9723 David Sherwood <david.sherwood@arm.com>
9725 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
9726 constant and extern definitions for variable-length vectors.
9727 (vect_get_constant_vectors): Note that the number of units
9728 is known to be constant.
9730 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9731 Alan Hayward <alan.hayward@arm.com>
9732 David Sherwood <david.sherwood@arm.com>
9734 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
9735 of units as polynomial. Choose between WIDE and NARROW based
9738 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9739 Alan Hayward <alan.hayward@arm.com>
9740 David Sherwood <david.sherwood@arm.com>
9742 * tree-vect-stmts.c (simd_clone_subparts): New function.
9743 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
9745 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9746 Alan Hayward <alan.hayward@arm.com>
9747 David Sherwood <david.sherwood@arm.com>
9749 * tree-vect-stmts.c (vectorizable_call): Treat the number of
9750 vectors as polynomial. Use build_index_vector for
9753 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9754 Alan Hayward <alan.hayward@arm.com>
9755 David Sherwood <david.sherwood@arm.com>
9757 * tree-vect-stmts.c (get_load_store_type): Treat the number of
9758 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
9759 for variable-length vectors.
9760 (vectorizable_mask_load_store): Treat the number of units as
9761 polynomial, asserting that it is constant if the condition has
9762 already been enforced.
9763 (vectorizable_store, vectorizable_load): Likewise.
9765 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9766 Alan Hayward <alan.hayward@arm.com>
9767 David Sherwood <david.sherwood@arm.com>
9769 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
9770 of units as polynomial. Punt if we can't tell at compile time
9771 which vector contains the final result.
9773 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9774 Alan Hayward <alan.hayward@arm.com>
9775 David Sherwood <david.sherwood@arm.com>
9777 * tree-vect-loop.c (vectorizable_induction): Treat the number
9778 of units as polynomial. Punt on SLP inductions. Use an integer
9779 VEC_SERIES_EXPR for variable-length integer reductions. Use a
9780 cast of such a series for variable-length floating-point
9783 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9784 Alan Hayward <alan.hayward@arm.com>
9785 David Sherwood <david.sherwood@arm.com>
9787 * tree.h (build_index_vector): Declare.
9788 * tree.c (build_index_vector): New function.
9789 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
9790 of units as polynomial, forcibly converting it to a constant if
9791 vectorizable_reduction has already enforced the condition.
9792 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
9793 to create a {1,2,3,...} vector.
9794 (vectorizable_reduction): Treat the number of units as polynomial.
9795 Choose vectype_in based on the largest scalar element size rather
9796 than the smallest number of units. Enforce the restrictions
9799 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9800 Alan Hayward <alan.hayward@arm.com>
9801 David Sherwood <david.sherwood@arm.com>
9803 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
9804 number of units as polynomial.
9806 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9807 Alan Hayward <alan.hayward@arm.com>
9808 David Sherwood <david.sherwood@arm.com>
9810 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
9811 * target.def (autovectorize_vector_sizes): Return the vector sizes
9812 by pointer, using vector_sizes rather than a bitmask.
9813 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
9814 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
9815 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
9817 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
9818 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
9819 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
9820 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
9821 * omp-general.c (omp_max_vf): Likewise.
9822 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9823 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
9824 * tree-vect-loop.c (vect_analyze_loop): Likewise.
9825 * tree-vect-slp.c (vect_slp_bb): Likewise.
9826 * doc/tm.texi: Regenerate.
9827 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
9829 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
9830 the vector size as a poly_uint64 rather than an unsigned int.
9831 (current_vector_size): Change from an unsigned int to a poly_uint64.
9832 (get_vectype_for_scalar_type): Update accordingly.
9833 * tree.h (build_truth_vector_type): Take the size and number of
9834 units as a poly_uint64 rather than an unsigned int.
9835 (build_vector_type): Add a temporary overload that takes
9836 the number of units as a poly_uint64 rather than an unsigned int.
9837 * tree.c (make_vector_type): Likewise.
9838 (build_truth_vector_type): Take the number of units as a poly_uint64
9839 rather than an unsigned int.
9841 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9842 Alan Hayward <alan.hayward@arm.com>
9843 David Sherwood <david.sherwood@arm.com>
9845 * target.def (get_mask_mode): Take the number of units and length
9846 as poly_uint64s rather than unsigned ints.
9847 * targhooks.h (default_get_mask_mode): Update accordingly.
9848 * targhooks.c (default_get_mask_mode): Likewise.
9849 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
9850 * doc/tm.texi: Regenerate.
9852 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9853 Alan Hayward <alan.hayward@arm.com>
9854 David Sherwood <david.sherwood@arm.com>
9856 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
9857 * omp-general.c (omp_max_vf): Likewise.
9858 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
9859 (expand_omp_simd): Handle polynomial safelen.
9860 * omp-low.c (omplow_simd_context): Add a default constructor.
9861 (omplow_simd_context::max_vf): Change from int to poly_uint64.
9862 (lower_rec_simd_input_clauses): Update accordingly.
9863 (lower_rec_input_clauses): Likewise.
9865 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9866 Alan Hayward <alan.hayward@arm.com>
9867 David Sherwood <david.sherwood@arm.com>
9869 * tree-vectorizer.h (vect_nunits_for_cost): New function.
9870 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
9871 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
9872 (vect_analyze_slp_cost): Likewise.
9873 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
9874 (vect_model_load_cost): Likewise.
9876 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9877 Alan Hayward <alan.hayward@arm.com>
9878 David Sherwood <david.sherwood@arm.com>
9880 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
9881 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
9882 from an unsigned int * to a poly_uint64_pod *.
9883 (calculate_unrolling_factor): New function.
9884 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
9886 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9887 Alan Hayward <alan.hayward@arm.com>
9888 David Sherwood <david.sherwood@arm.com>
9890 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
9891 from an unsigned int to a poly_uint64.
9892 (_loop_vec_info::slp_unrolling_factor): Likewise.
9893 (_loop_vec_info::vectorization_factor): Change from an int
9895 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
9896 (vect_get_num_vectors): New function.
9897 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
9898 (vect_get_num_copies): Use vect_get_num_vectors.
9899 (vect_analyze_data_ref_dependences): Change max_vf from an int *
9900 to an unsigned int *.
9901 (vect_analyze_data_refs): Change min_vf from an int * to a
9903 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9904 than an unsigned HOST_WIDE_INT.
9905 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
9906 (vect_analyze_data_ref_dependence): Change max_vf from an int *
9907 to an unsigned int *.
9908 (vect_analyze_data_ref_dependences): Likewise.
9909 (vect_compute_data_ref_alignment): Handle polynomial vf.
9910 (vect_enhance_data_refs_alignment): Likewise.
9911 (vect_prune_runtime_alias_test_list): Likewise.
9912 (vect_shift_permute_load_chain): Likewise.
9913 (vect_supportable_dr_alignment): Likewise.
9914 (dependence_distance_ge_vf): Take the vectorization factor as a
9915 poly_uint64 rather than an unsigned HOST_WIDE_INT.
9916 (vect_analyze_data_refs): Change min_vf from an int * to a
9918 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
9919 vfm1 as a poly_uint64 rather than an int. Make the same change
9920 for the returned bound_scalar.
9921 (vect_gen_vector_loop_niters): Handle polynomial vf.
9922 (vect_do_peeling): Likewise. Update call to
9923 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
9924 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
9926 * tree-vect-loop.c (vect_determine_vectorization_factor)
9927 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
9928 (vect_get_known_peeling_cost): Likewise.
9929 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
9930 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
9931 (vect_transform_loop): Likewise. Use the lowest possible VF when
9932 updating the upper bounds of the loop.
9933 (vect_min_worthwhile_factor): Make static. Return an unsigned int
9935 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
9936 polynomial unroll factors.
9937 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
9938 (vect_make_slp_decision): Likewise.
9939 (vect_supported_load_permutation_p): Likewise, and polynomial
9941 (vect_analyze_slp_cost): Handle polynomial vf.
9942 (vect_slp_analyze_node_operations): Likewise.
9943 (vect_slp_analyze_bb_1): Likewise.
9944 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9945 than an unsigned HOST_WIDE_INT.
9946 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
9947 (vectorizable_load): Handle polynomial vf.
9948 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
9950 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
9952 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9953 Alan Hayward <alan.hayward@arm.com>
9954 David Sherwood <david.sherwood@arm.com>
9956 * match.pd: Handle bit operations involving three constants
9957 and try to fold one pair.
9959 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9961 * tree-vect-loop-manip.c: Include gimple-fold.h.
9962 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
9963 niters_maybe_zero parameters. Handle other cases besides a step of 1.
9964 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
9965 Add a path that uses a step of VF instead of 1, but disable it
9967 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
9968 and niters_no_overflow parameters. Update calls to
9969 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
9970 Create a new SSA name if the latter choses to use a ste other
9971 than zero, and return it via niters_vector_mult_vf_var.
9972 * tree-vect-loop.c (vect_transform_loop): Update calls to
9973 vect_do_peeling, vect_gen_vector_loop_niters and
9974 slpeel_make_loop_iterate_ntimes.
9975 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
9976 (vect_gen_vector_loop_niters): Update declarations after above changes.
9978 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
9980 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
9981 128-bit round to integer instructions.
9982 (ceil<mode>2): Likewise.
9983 (btrunc<mode>2): Likewise.
9984 (round<mode>2): Likewise.
9986 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9988 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
9989 unaligned VSX load/store on P8/P9.
9990 (expand_block_clear): Allow the use of unaligned VSX
9991 load/store on P8/P9.
9993 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9995 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
9997 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
9998 swap associated with both a load and a store.
10000 2018-01-02 Andrew Waterman <andrew@sifive.com>
10002 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
10003 * config/riscv/riscv.md (clear_cache): Use it.
10005 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
10007 * web.c: Remove out-of-date comment.
10009 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10011 * expr.c (fixup_args_size_notes): Check that any existing
10012 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
10013 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
10014 (emit_single_push_insn): ...here.
10016 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10018 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
10019 (const_vector_encoded_nelts): New function.
10020 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
10021 (const_vector_int_elt, const_vector_elt): Declare.
10022 * emit-rtl.c (const_vector_int_elt_1): New function.
10023 (const_vector_elt): Likewise.
10024 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
10025 of CONST_VECTOR_ELT.
10027 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10029 * expr.c: Include rtx-vector-builder.h.
10030 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
10031 directly on the tree encoding.
10032 (const_vector_from_tree): Likewise.
10033 * optabs.c: Include rtx-vector-builder.h.
10034 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
10035 sequence of "u" values.
10036 * vec-perm-indices.c: Include rtx-vector-builder.h.
10037 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
10038 directly on the vec_perm_indices encoding.
10040 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10042 * doc/rtl.texi (const_vector): Describe new encoding scheme.
10043 * Makefile.in (OBJS): Add rtx-vector-builder.o.
10044 * rtx-vector-builder.h: New file.
10045 * rtx-vector-builder.c: Likewise.
10046 * rtl.h (rtx_def::u2): Add a const_vector field.
10047 (CONST_VECTOR_NPATTERNS): New macro.
10048 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
10049 (CONST_VECTOR_DUPLICATE_P): Likewise.
10050 (CONST_VECTOR_STEPPED_P): Likewise.
10051 (CONST_VECTOR_ENCODED_ELT): Likewise.
10052 (const_vec_duplicate_p): Check for a duplicated vector encoding.
10053 (unwrap_const_vec_duplicate): Likewise.
10054 (const_vec_series_p): Check for a non-duplicated vector encoding.
10055 Say that the function only returns true for integer vectors.
10056 * emit-rtl.c: Include rtx-vector-builder.h.
10057 (gen_const_vec_duplicate_1): Delete.
10058 (gen_const_vector): Call gen_const_vec_duplicate instead of
10059 gen_const_vec_duplicate_1.
10060 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
10061 (gen_const_vec_duplicate): Use rtx_vector_builder.
10062 (gen_const_vec_series): Likewise.
10063 (gen_rtx_CONST_VECTOR): Likewise.
10064 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
10065 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10066 Build a new vector rather than modifying a CONST_VECTOR in-place.
10067 (handle_special_swappables): Update call accordingly.
10068 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
10069 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10070 Build a new vector rather than modifying a CONST_VECTOR in-place.
10071 (handle_special_swappables): Update call accordingly.
10073 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10075 * simplify-rtx.c (simplify_const_binary_operation): Use
10076 CONST_VECTOR_ELT instead of XVECEXP.
10078 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10080 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
10081 the selector elements to be different from the data elements
10082 if the selector is a VECTOR_CST.
10083 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
10084 ssizetype for the selector.
10086 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10088 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
10089 before testing each element individually.
10090 * tree-vect-generic.c (lower_vec_perm): Likewise.
10092 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10094 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
10095 * selftest-run-tests.c (selftest::run_tests): Call it.
10096 * vector-builder.h (vector_builder::operator ==): New function.
10097 (vector_builder::operator !=): Likewise.
10098 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
10099 (vec_perm_indices::all_from_input_p): New function.
10100 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10101 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
10102 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
10103 instead of reading the VECTOR_CST directly. Detect whether both
10104 vector inputs are the same before constructing the vec_perm_indices,
10105 and update the number of inputs argument accordingly. Use the
10106 utility functions added above. Only construct sel2 if we need to.
10108 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10110 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
10111 the broadcast of the low byte.
10112 (expand_mult_highpart): Use an explicit encoding for the permutes.
10113 * optabs-query.c (can_mult_highpart_p): Likewise.
10114 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
10115 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10116 (vectorizable_bswap): Likewise.
10117 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
10118 explicit encoding for the power-of-2 permutes.
10119 (vect_permute_store_chain): Likewise.
10120 (vect_grouped_load_supported): Likewise.
10121 (vect_permute_load_chain): Likewise.
10123 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10125 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
10126 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
10127 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
10128 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10129 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
10130 (vect_gen_perm_mask_any): Likewise.
10132 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10134 * int-vector-builder.h: New file.
10135 * vec-perm-indices.h: Include int-vector-builder.h.
10136 (vec_perm_indices): Redefine as an int_vector_builder.
10137 (auto_vec_perm_indices): Delete.
10138 (vec_perm_builder): Redefine as a stand-alone class.
10139 (vec_perm_indices::vec_perm_indices): New function.
10140 (vec_perm_indices::clamp): Likewise.
10141 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
10142 (vec_perm_indices::new_vector): New function.
10143 (vec_perm_indices::new_expanded_vector): Update for new
10144 vec_perm_indices class.
10145 (vec_perm_indices::rotate_inputs): New function.
10146 (vec_perm_indices::all_in_range_p): Operate directly on the
10147 encoded form, without computing elided elements.
10148 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
10149 encoding. Update for new vec_perm_indices class.
10150 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
10151 the given vec_perm_builder.
10152 (expand_vec_perm_var): Update vec_perm_builder constructor.
10153 (expand_mult_highpart): Use vec_perm_builder instead of
10154 auto_vec_perm_indices.
10155 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
10156 vec_perm_indices instead of auto_vec_perm_indices. Use a single
10157 or double series encoding as appropriate.
10158 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
10159 vec_perm_indices instead of auto_vec_perm_indices.
10160 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10161 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10162 (vect_permute_store_chain): Likewise.
10163 (vect_grouped_load_supported): Likewise.
10164 (vect_permute_load_chain): Likewise.
10165 (vect_shift_permute_load_chain): Likewise.
10166 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10167 (vect_transform_slp_perm_load): Likewise.
10168 (vect_schedule_slp_instance): Likewise.
10169 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10170 (vectorizable_mask_load_store): Likewise.
10171 (vectorizable_bswap): Likewise.
10172 (vectorizable_store): Likewise.
10173 (vectorizable_load): Likewise.
10174 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
10175 vec_perm_indices instead of auto_vec_perm_indices. Use
10176 tree_to_vec_perm_builder to read the vector from a tree.
10177 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
10178 vec_perm_builder instead of a vec_perm_indices.
10179 (have_whole_vector_shift): Use vec_perm_builder and
10180 vec_perm_indices instead of auto_vec_perm_indices. Leave the
10181 truncation to calc_vec_perm_mask_for_shift.
10182 (vect_create_epilog_for_reduction): Likewise.
10183 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
10184 from auto_vec_perm_indices to vec_perm_indices.
10185 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10186 instead of changing individual elements.
10187 (aarch64_vectorize_vec_perm_const): Use new_vector to install
10188 the vector in d.perm.
10189 * config/arm/arm.c (expand_vec_perm_d::perm): Change
10190 from auto_vec_perm_indices to vec_perm_indices.
10191 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10192 instead of changing individual elements.
10193 (arm_vectorize_vec_perm_const): Use new_vector to install
10194 the vector in d.perm.
10195 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
10196 Update vec_perm_builder constructor.
10197 (rs6000_expand_interleave): Likewise.
10198 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
10199 (rs6000_expand_interleave): Likewise.
10201 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10203 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
10204 to qimode could truncate the indices.
10205 * optabs.c (expand_vec_perm_var): Likewise.
10207 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10209 * Makefile.in (OBJS): Add vec-perm-indices.o.
10210 * vec-perm-indices.h: New file.
10211 * vec-perm-indices.c: Likewise.
10212 * target.h (vec_perm_indices): Replace with a forward class
10214 (auto_vec_perm_indices): Move to vec-perm-indices.h.
10215 * optabs.h: Include vec-perm-indices.h.
10216 (expand_vec_perm): Delete.
10217 (selector_fits_mode_p, expand_vec_perm_var): Declare.
10218 (expand_vec_perm_const): Declare.
10219 * target.def (vec_perm_const_ok): Replace with...
10220 (vec_perm_const): ...this new hook.
10221 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
10222 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
10223 * doc/tm.texi: Regenerate.
10224 * optabs.def (vec_perm_const): Delete.
10225 * doc/md.texi (vec_perm_const): Likewise.
10226 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
10227 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
10228 expand_vec_perm for constant permutation vectors. Assert that
10229 the mode of variable permutation vectors is the integer equivalent
10230 of the mode that is being permuted.
10231 * optabs-query.h (selector_fits_mode_p): Declare.
10232 * optabs-query.c: Include vec-perm-indices.h.
10233 (selector_fits_mode_p): New function.
10234 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
10235 is defined, instead of checking whether the vec_perm_const_optab
10236 exists. Use targetm.vectorize.vec_perm_const instead of
10237 targetm.vectorize.vec_perm_const_ok. Check whether the indices
10238 fit in the vector mode before using a variable permute.
10239 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
10240 vec_perm_indices instead of an rtx.
10241 (expand_vec_perm): Replace with...
10242 (expand_vec_perm_const): ...this new function. Take the selector
10243 as a vec_perm_indices rather than an rtx. Also take the mode of
10244 the selector. Update call to shift_amt_for_vec_perm_mask.
10245 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
10246 Use vec_perm_indices::new_expanded_vector to expand the original
10247 selector into bytes. Check whether the indices fit in the vector
10248 mode before using a variable permute.
10249 (expand_vec_perm_var): Make global.
10250 (expand_mult_highpart): Use expand_vec_perm_const.
10251 * fold-const.c: Includes vec-perm-indices.h.
10252 * tree-ssa-forwprop.c: Likewise.
10253 * tree-vect-data-refs.c: Likewise.
10254 * tree-vect-generic.c: Likewise.
10255 * tree-vect-loop.c: Likewise.
10256 * tree-vect-slp.c: Likewise.
10257 * tree-vect-stmts.c: Likewise.
10258 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
10260 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
10261 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
10262 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
10263 (aarch64_vectorize_vec_perm_const): ...this new function.
10264 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10265 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10266 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
10267 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
10268 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10269 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10270 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
10272 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
10273 check for NEON modes.
10274 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
10275 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
10276 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
10277 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
10279 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
10280 the old VEC_PERM_CONST conditions.
10281 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
10282 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
10283 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
10284 (ia64_vectorize_vec_perm_const_ok): Merge into...
10285 (ia64_vectorize_vec_perm_const): ...this new function.
10286 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
10287 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
10288 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
10289 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
10290 * config/mips/mips.c (mips_expand_vec_perm_const)
10291 (mips_vectorize_vec_perm_const_ok): Merge into...
10292 (mips_vectorize_vec_perm_const): ...this new function.
10293 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
10294 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
10295 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
10296 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
10297 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
10298 (rs6000_expand_vec_perm_const): Delete.
10299 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
10301 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10302 (altivec_expand_vec_perm_const_le): Take each operand individually.
10303 Operate on constant selectors rather than rtxes.
10304 (altivec_expand_vec_perm_const): Likewise. Update call to
10305 altivec_expand_vec_perm_const_le.
10306 (rs6000_expand_vec_perm_const): Delete.
10307 (rs6000_vectorize_vec_perm_const_ok): Delete.
10308 (rs6000_vectorize_vec_perm_const): New function.
10309 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10310 an element count and rtx array.
10311 (rs6000_expand_extract_even): Update call accordingly.
10312 (rs6000_expand_interleave): Likewise.
10313 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
10314 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
10315 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
10316 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
10317 (rs6000_expand_vec_perm_const): Delete.
10318 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10319 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10320 (altivec_expand_vec_perm_const_le): Take each operand individually.
10321 Operate on constant selectors rather than rtxes.
10322 (altivec_expand_vec_perm_const): Likewise. Update call to
10323 altivec_expand_vec_perm_const_le.
10324 (rs6000_expand_vec_perm_const): Delete.
10325 (rs6000_vectorize_vec_perm_const_ok): Delete.
10326 (rs6000_vectorize_vec_perm_const): New function. Remove stray
10327 reference to the SPE evmerge intructions.
10328 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10329 an element count and rtx array.
10330 (rs6000_expand_extract_even): Update call accordingly.
10331 (rs6000_expand_interleave): Likewise.
10332 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
10333 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
10335 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10337 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10339 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
10340 vector mode and that that mode matches the mode of the data
10342 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
10343 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
10344 directly using expand_vec_perm_1 when forcing selectors into
10346 (expand_vec_perm_var): New function, split out from expand_vec_perm.
10348 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10350 * optabs-query.h (can_vec_perm_p): Delete.
10351 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
10352 * optabs-query.c (can_vec_perm_p): Split into...
10353 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
10354 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
10355 particular selector is valid.
10356 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10357 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10358 (vect_grouped_load_supported): Likewise.
10359 (vect_shift_permute_load_chain): Likewise.
10360 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10361 (vect_transform_slp_perm_load): Likewise.
10362 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10363 (vectorizable_bswap): Likewise.
10364 (vect_gen_perm_mask_checked): Likewise.
10365 * fold-const.c (fold_ternary_loc): Likewise. Don't take
10366 implementations of variable permutation vectors into account
10367 when deciding which selector to use.
10368 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
10369 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
10370 with a false third argument.
10371 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
10372 to test whether the constant selector is valid and can_vec_perm_var_p
10373 to test whether a variable selector is valid.
10375 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10377 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
10378 * optabs-query.c (can_vec_perm_p): Likewise.
10379 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
10380 instead of vec_perm_indices.
10381 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
10382 (vect_gen_perm_mask_checked): Likewise,
10383 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
10384 (vect_gen_perm_mask_checked): Likewise,
10386 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10388 * optabs-query.h (qimode_for_vec_perm): Declare.
10389 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
10390 (qimode_for_vec_perm): ...this new function.
10391 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
10393 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10395 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
10396 does not have a conditional at the top.
10398 2018-01-02 Richard Biener <rguenther@suse.de>
10400 * ipa-inline.c (big_speedup_p): Fix expression.
10402 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10405 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
10408 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10412 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
10413 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
10414 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
10415 cond_taken_branch_cost 3->4.
10417 2018-01-01 Jakub Jelinek <jakub@redhat.com>
10419 PR tree-optimization/83581
10420 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
10421 TODO_cleanup_cfg if any changes have been made.
10423 PR middle-end/83608
10424 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
10425 convert_modes if target mode has the right side, but different mode
10428 PR middle-end/83609
10429 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
10430 last argument when extracting from CONCAT. If either from_real or
10431 from_imag is NULL, use expansion through memory. If result is not
10432 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
10433 the parts directly to inner mode, if even that fails, use expansion
10436 PR middle-end/83623
10437 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
10438 check for bswap in mode rather than HImode and use that in expand_unop
10441 Copyright (C) 2018 Free Software Foundation, Inc.
10443 Copying and distribution of this file, with or without modification,
10444 are permitted in any medium without royalty provided the copyright
10445 notice and this notice are preserved.