1 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3 * final.c (final_1): Set insn_last_address as well as
6 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9 * explow.c (emit_stack_probe): Call validize_mem on memory location
10 before passing it to gen_probe_stack. Create address operand and
11 legitimize it for the probe_stack_address case.
13 2018-04-09 Jan Hubicka <jh@suse.cz>
16 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
17 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
18 * tree.c (free_lang_data_in_type): Fix handling of binfos;
20 (free_lang_data): Rebuild type inheritance graph.
22 2018-04-09 Martin Sebor <msebor@redhat.com>
24 * invoke.texi (-finline-small-functions): Mention other optimization
26 (-findirect-inlining, -fpartial-inlining): Same.
27 (-finline-functions-called-once): Same.
28 (-freorder-blocks-and-partition): Same.
30 2018-04-09 Jan Hubicka <jh@suse.cz>
33 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
34 jumps; choose last target that matches the criteria (i.e.
35 no partition changes for non-crossing jumps).
36 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
37 support for redirecting crossing jumps to non-crossing.
39 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
41 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
42 also for naked functions.
44 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
46 * config/arc/arc.md (add_shift): New pattern.
47 (add_shift2): Likewise.
48 (sub_shift): Likewise.
49 (sub_shift_cmp0_noout): Likewise.
50 (compare_si_ashiftsi): Likewise.
51 (xbfu_cmp0_noout): New combine pattern.
52 (xbfu_cmp0"): Likewise.
53 (movsi_set_cc_insn): Place the predicable variant first.
54 (commutative_binary_cmp0_noout): Remove clobber.
55 (commutative_binary_cmp0): New pattern.
56 (noncommutative_binary_cmp0): Likewise.
57 (noncommutative_binary_cmp0_noout): Likewise.
58 (noncommutative_binary_comparison_result_used): Removed.
59 (rsub_cmp0): New pattern.
60 (rsub_cmp0_noout): Likewise.
61 (extzvsi): Changed, keep only meaningful variants.
62 (SQH, SEZ): New iterators.
63 (SQH_postfix): New mode attribute.
64 (SEZ_prefix): New code attribute.
65 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
66 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
67 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
69 (noncommutative_operator): Check the availability of barrel
72 2018-04-09 Richard Biener <rguenther@suse.de>
74 PR tree-optimization/85284
75 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
76 Only use the niter constraining form of simple_iv when the exit
79 2018-04-09 Tom de Vries <tom@codesourcery.com>
82 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
83 (define_expand "*memory_barrier"): New define_expand.
84 (define_insn "memory_barrier"): New insn.
86 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
88 PR rtl-optimization/80463
89 PR rtl-optimization/83972
90 PR rtl-optimization/83480
92 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
93 correct producer for the insn.
94 (tidy_control_flow): Fixup seqnos in case of debug insns.
96 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
98 PR rtl-optimization/83913
100 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
101 different sched-times when merging exprs.
103 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
105 PR rtl-optimization/83962
107 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
108 tidy_fallthru_edge and tidy_control_flow.
110 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
112 PR rtl-optimization/83530
114 * sel-sched.c (force_next_insn): New global variable.
115 (remove_insn_for_debug): When force_next_insn is true, also leave only
116 next insn in the ready list.
117 (sel_sched_region): When the region wasn't scheduled, make another pass
118 over it with force_next_insn set to 1.
120 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
122 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
124 * config/nds32/constants.md (unspec_volatile_element): Add enum values
125 for interrupt control.
126 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
127 functions for interrupt control.
128 * config/nds32/nds32-intrinsic.md: Likewise.
129 * config/nds32/nds32_intrinsic.h: Likewise.
130 * config/nds32/nds32.h (nds32_builtins): Likewise.
132 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
134 * config/nds32/nds32.c (nds32_init_machine_status,
135 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
136 strict_aligned_p field.
137 (nds32_expand_to_rtl_hook): New function.
138 (TARGET_EXPAND_TO_RTL_HOOK): Define.
139 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
141 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
142 Chung-Ju Wu <jasonwucj@gmail.com>
144 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
145 * config/nds32/nds32-n7.md: New file.
146 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
147 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
149 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
150 * config/nds32/nds32.md (pipeline_model): Add n7.
151 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
152 * config/nds32/pipelines.md: Include n7 settings.
154 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
155 Chung-Ju Wu <jasonwucj@gmail.com>
157 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
158 * config/nds32/nds32-e8.md: New file.
159 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
160 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
162 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
163 * config/nds32/nds32.md (pipeline_model): Add e8.
164 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
165 * config/nds32/pipelines.md: Include e8 settings.
167 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
168 Chung-Ju Wu <jasonwucj@gmail.com>
170 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
171 * config/nds32/nds32-n8.md: New file.
172 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
173 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
175 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
176 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
177 * config/nds32/nds32.md (pipeline_model): Add n8.
178 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
179 * config/nds32/pipelines.md: Include n8 settings.
181 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
182 Chung-Ju Wu <jasonwucj@gmail.com>
184 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
185 * config/nds32/nds32-n9-2r1w.md: New file.
186 * config/nds32/nds32-n9-3r2w.md: New file.
187 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
188 nds32_register_ports): New or modify for cpu n9.
189 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
191 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
192 * config/nds32/nds32-utils.c: New file.
193 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
194 TARGET_MUL_SLOW): Define.
195 * config/nds32/nds32.md (pipeline_model): New attribute.
196 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
197 New options that support cpu n9.
198 * config/nds32/pipelines.md: Include n9 settings.
199 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
201 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
203 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
204 information if necessary.
205 (output_cond_branch_compare_zero): Likewise.
206 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
207 (nds32_target_alignment): Refine for alignment.
208 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
209 (FUNCTION_BOUNDARY): Modify.
210 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
212 * config/nds32/nds32.opt (malways-align, malign-functions): New.
214 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
216 * config/nds32/constants.md (unspec_volatile_element): Add values for
217 TLB operation and data prefetch.
218 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
219 functions for TLB operation and data prefetch.
220 * config/nds32/nds32-intrinsic.md: Likewise.
221 * config/nds32/nds32_intrinsic.h: Likewise.
222 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
223 (nds32_print_operand): Likewise.
224 * config/nds32/nds32.h (nds32_builtins): Likewise.
226 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
227 Andrew Pinski <pinsika@gcc.gnu.org>
230 * match.pd: Use constant_boolean_node of correct type instead of
231 boolean_true_node or boolean_false_node for simplifying
232 pointer comparisons to zero.
234 2018-04-07 Jakub Jelinek <jakub@redhat.com>
236 PR tree-optimization/80021
237 * tree.c (verify_type_variant): Make error call in verify_variant_match
238 translatable and remove final full stop.
240 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
242 * config/nds32/constants.md (unspec_volatile_element): Add
243 UNSPEC_VOLATILE_EH_RETURN.
244 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
245 nds32_output_stack_pop): Support dwarf exception handling process.
246 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
247 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
248 exception handling process.
249 (nds32_compute_stack_frame): Likewise.
250 (nds32_return_addr_rtx): Likewise.
251 (nds32_initial_elimination_offset): Likewise.
252 (nds32_expand_prologue): Likewise.
253 (nds32_expand_epilogue): Likewise.
254 (nds32_dynamic_chain_address): New function.
255 * config/nds32/nds32.h (machine_function): Add fields for dwarf
257 (DYNAMIC_CHAIN_ADDRESS): Define.
258 (EH_RETURN_DATA_REGNO): Define.
259 (EH_RETURN_STACKADJ_RTX): Define.
260 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
261 patterns for dwarf exception handling.
263 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
265 * config/nds32/nds32.h: Clean up obsolete macros.
267 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
269 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
270 Add enum values for particular instructions.
271 * config/nds32/nds32-intrinsic.c: Implementation of expanding
272 particular intrinsic functions.
273 * config/nds32/nds32-intrinsic.md: Likewise.
274 * config/nds32/nds32_intrinsic.h: Likewise.
275 * config/nds32/nds32.h (nds32_builtins): Likewise.
276 * config/nds32/nds32.md (type): Add pbsad and pbsada.
277 (btst, ave): New patterns for particular instructions.
279 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
281 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
282 Add enum values for atomic load/store and memory sync.
283 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
285 * config/nds32/nds32-intrinsic.md: Likewise.
286 * config/nds32/nds32_intrinsic.h: Likewise.
287 * config/nds32/nds32.h (nds32_builtins): Likewise.
289 2018-04-07 Jakub Jelinek <jakub@redhat.com>
291 PR tree-optimization/85257
292 * fold-const.c (native_encode_vector): If not all elts could fit
293 and off is -1, return 0 rather than offset.
294 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
295 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
296 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
297 adjust buffer in native_interpret_expr call.
299 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
301 * config/nds32/constants.md (unspec_volatile_element): Add cache
303 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
304 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
305 * config/nds32/nds32.c (nds32_cctl_names): New.
306 (nds32_print_operand): Handle cache control register names.
307 * config/nds32/nds32.h (nds32_builtins): New enum values.
308 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
310 * config/nds32/nds32.md (type): Add mmu.
311 * config/nds32/pipelines.md (simple_insn): Add mmu.
313 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
315 * config/nds32/nds32.md (type): Remove call.
316 * config/nds32/pipelines.md (simple_insn): Likewise.
318 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
320 * config/nds32/constants.md (unspec_volatile_element): Add
321 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
322 UNSPEC_VOLATILE_FMFCFG.
323 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
324 description for fmfcfg and fmfcsr.
325 (bdesc_1arg): Add fmtcsr.
326 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
327 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
328 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
329 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
330 unspec_fmfcfg): New patterns.
331 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
332 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
333 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
334 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
335 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
336 __nds32__fmfcfg): Define.
338 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
340 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
341 intrinsic register names.
342 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
343 intrinsic register enum values and macros.
345 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
347 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
348 for load/store addressing form.
349 (nds32_print_operand_address): Likewise.
351 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
354 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
355 based on LABEL_REF. Remove useless assertion.
356 (pic_address_needs_scratch): Fix formatting.
357 (sparc_legitimize_pic_address): Minor tweaks.
358 (sparc_delegitimize_address): Adjust assertion accordingly.
359 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
360 into symbolic_operand.
361 (movsi_high_pic_label_ref): Likewise.
362 (movsi_lo_sum_pic_label_ref): Likewise.
363 (movdi_pic_label_ref): Likewise.
364 (movdi_high_pic_label_ref): Likewise.
365 (movdi_lo_sum_pic_label_ref): Likewise.
367 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
369 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
370 custom LIB_SPEC setup.
372 2018-04-06 Ruslan Bukin <br@bsdpad.com>
373 Kito Cheng <kito.cheng@gmail.com>
375 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
376 * config/riscv/freebsd.h: New.
378 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
380 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
381 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
384 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
385 Kito Cheng <kito.cheng@gmail.com>
387 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
388 nds32_output_call, nds32_symbol_binds_local_p): New functions.
389 * config/nds32/nds32-protos.h (nds32_output_call,
390 nds32_output_return): Declare.
391 * config/nds32/nds32.md: Refine all the call and return patterns.
393 2018-04-06 Jakub Jelinek <jakub@redhat.com>
396 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
397 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
399 PR rtl-optimization/84872
400 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
401 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
404 2018-04-06 Tamar Christina <tamar.christina@arm.com>
406 * expr.c (copy_blkmode_to_reg): Revert 254862.
407 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
409 2018-04-06 Richard Biener <rguenther@suse.de>
412 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
413 after seeing a component reference with an adjacent field. Treat
414 refs to arrays at struct end of external decls similar to
415 refs to unconstrained commons.
417 2018-04-06 Jakub Jelinek <jakub@redhat.com>
420 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
421 look through SAVE_EXPRs with non-side-effects argument. Adjust
423 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
426 2018-04-06 Richard Biener <rguenther@suse.de>
429 * alias.c (find_base_term): New wrapper around find_base_term
430 unwinding CSELIB_VAL_PTR changes.
431 (find_base_term): Do not restore CSELIB_VAL_PTR during the
434 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
436 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
438 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
439 constant definitions.
440 ("nop"): lr 0,0 -> nopr r0
441 ("nop_lr0", "nop_lr1"): New insn definitions.
443 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
445 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
446 NDS32_V3PUSH_AVAILABLE_P macro.
448 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
449 Chung-Ju Wu <jasonwucj@gmail.com>
451 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
452 (nds32*-*-*): Add float and fpu_config into supported_defaults.
453 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
454 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
455 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
456 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
457 * config/nds32/constraints.md: New constraints and checking for hard
459 * config/nds32/iterators.md: New mode iterator and attribute for hard
461 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
463 * config/nds32/nds32-fpu.md: New file.
464 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
465 deal with hard float code generation.
466 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
468 (abi_type, float_reg_number): New enum type.
469 * config/nds32/nds32-predicates.c: New predicates for hard float.
470 * config/nds32/nds32-protos.h: Declare functions for hard float.
471 * config/nds32/nds32.c: Implementation for hard float configuration.
472 * config/nds32/nds32.h: Definitions for hard float configuration.
473 * config/nds32/nds32.md: Include hard float machine description and
474 modify patterns for hard float configuration.
475 * config/nds32/nds32.opt: New options for hard float configuration.
476 * config/nds32/predicates.md: New predicates for hard float
479 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
481 * common/config/nds32/nds32-common.c
482 (nds32_option_optimization_table): Enable -mreleax-hint by default.
484 2018-04-05 Jakub Jelinek <jakub@redhat.com>
487 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
488 CONSTRUCTOR_ELT (ctor, ...)->value.
490 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
493 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
495 2018-04-05 Tom de Vries <tom@codesourcery.com>
498 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
501 2018-04-05 Shiva Chen <shiva0217@gmail.com>
502 Kito Cheng <kito.cheng@gmail.com>
504 * config/nds32/constraints.md (U33): Fine-tune checking condition.
505 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
506 * config/nds32/nds32.h (nds32_16bit_address_type): Add
507 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
509 2018-04-05 Shiva Chen <shiva0217@gmail.com>
510 Kito Cheng <kito.cheng@gmail.com>
512 * config/nds32/constraints.md (Ufe): New memory constraint.
513 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
514 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
515 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
517 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
518 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
520 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
522 * config/nds32/nds32.md: Use optimize_size in the condition for
523 alu-shift instructions.
525 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
527 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
529 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
531 * config/nds32/nds32.md (negsi2): Refine pattern.
533 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
534 Chung-Ju Wu <jasonwucj@gmail.com>
536 * config/nds32/iterators.md (shift_rotate): New code iterator.
537 (shift): New code attribute.
538 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
539 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
540 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
541 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
543 (andsi3, *andsi3): Ditto.
544 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
545 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
546 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
547 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
548 nds32_ior_operand, nds32_xor_operand): New predicates.
550 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
552 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
553 (addsi3, subsi3): ... this.
555 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
557 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
559 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
561 * config/nds32/nds32.md: Adjust indention.
563 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
565 * config/nds32/nds32.md (feature): New attribute.
567 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
569 * config/nds32/nds32.md (subtype): New attribute.
571 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
574 * config/arm/arm-builtins.c (arm_expand_builtin): Change
575 expansion to perform a bitwise AND of the argument followed by a
576 boolean negation of the result.
578 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
580 PR rtl-optimization/84878
581 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
582 the basic block. Assert the use reference is not artificial and that
583 it has an associated insn.
585 2018-04-04 Michael Matz <matz@suse.de>
587 * builtins.c (compute_objsize): Pass correct operand
588 to array_at_struct_end_p.
590 2018-04-04 Richard Biener <rguenther@suse.de>
593 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
594 from contexts for DINFO_LEVEL_TERSE and below.
596 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
598 * config/nds32/nds32-doubleword.md (move_<mode>): Require
599 resiter_operand condition.
600 * config/nds32/nds32.md (*move<mode>): Ditto.
602 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
603 Monk Chiang <sh.chiang04@gmail.com>
605 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
607 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
609 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
611 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
612 Kito Cheng <kito.cheng@gmail.com>
614 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
615 nds32_cond_code_str, output_cond_branch,
616 output_cond_branch_compare_zero, nds32_expand_cbranch,
617 nds32_expand_cstore, nds32_expand_movcc,
618 nds32_output_cbranchsi4_equality_zero,
619 nds32_output_cbranchsi4_equality_reg,
620 nds32_output_cbranchsi4_equality_reg_or_const_int,
621 nds32_output_cbranchsi4_greater_less_zero: New functions.
622 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
623 nds32_expand_cstore, nds32_expand_movcc,
624 nds32_output_cbranchsi4_equality_zero,
625 nds32_output_cbranchsi4_equality_reg,
626 nds32_output_cbranchsi4_equality_reg_or_const_int,
627 nds32_output_cbranchsi4_greater_less_zero): Declare.
628 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
629 nds32_rimm11s_operand): New predicates.
630 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
631 * config/nds32/nds32.md: Rewrite all the branch and conditional move
634 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
636 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
637 * config/nds32/nds32.md: Ditto.
638 * config/nds32/pipelines.md: Ditto.
640 2018-04-04 Richard Biener <rguenther@suse.de>
642 PR tree-optimization/85168
643 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
644 propagating abnormals.
646 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
648 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
650 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
651 Kito Cheng <kito.cheng@gmail.com>
653 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
654 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
655 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
656 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
657 * config/nds32/nds32.md (sibcall_internal): New.
658 (sibcall_register): Remove.
659 (sibcall_immediate): Remove.
660 (sibcall_value_internal): New.
661 (sibcall_value_register): Remove.
662 (sibcall_value_immediate): Remove.
663 * config/nds32/predicates.md (nds32_general_register_operand): New.
664 (nds32_call_address_operand): New.
666 2018-04-03 Jakub Jelinek <jakub@redhat.com>
668 PR rtl-optimization/85167
669 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
670 bb_defs if *split_p, instead preinitialize it to NULL.
672 PR tree-optimization/85156
673 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
674 evaluating the argument multiple times.
676 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
678 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
680 (_mm_cvtpd_ps): Likewise.
681 (_mm_cvttpd_epi32): Likewise.
682 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
683 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
684 vector, pixel, and bool following altivec.h include.
686 2018-04-03 Martin Sebor <msebor@redhat.com>
688 * doc/extend.texi (Common Function Attributes): Clarify.
689 (const attribute): Likewise.
690 (pure attribute): Likewise.
692 2018-04-03 Jakub Jelinek <jakub@redhat.com>
695 * config/i386/i386.c (ix86_expand_vector_set): Use
696 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
698 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
700 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
701 instructions when changing rounding bits to preserve precision bits
702 in the x87 control word.
704 2018-04-03 Martin Liska <mliska@suse.cz>
706 PR tree-optimization/82491
707 * rtl.h (strip_offset_and_add): Replace += suboffset with
708 poly_uint64 () + suboffset.
710 2018-03-29 Martin Liska <mliska@suse.cz>
711 Martin Jambor <mjambor@suse.cz>
714 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
715 param_type is not an integral or pointer type.
717 2018-04-03 Richard Biener <rguenther@suse.de>
719 * sese.h (recompute_all_dominators): Remove.
721 2018-04-02 Martin Sebor <msebor@redhat.com>
723 * doc/invoke.texi (-Wrestrict): Fix typos.
725 2018-04-02 Jim Wilson <jimw@sifive.com>
727 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
728 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
729 (<optab>di3, <optab>si3_extend): Likewise.
730 (<optab>si3_mask, <optab>si3_mask_1): New.
731 (<optab>di3_mask, <optab>di3_mask_1): New.
732 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
733 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
734 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
736 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
738 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
741 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
743 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
744 (nds32_canonicalize_comparison): New function.
746 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
747 Kito Cheng <kito.cheng@gmail.com>
748 Kuan-Lin Chen <kuanlinchentw@gmail.com>
750 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
751 * config/nds32/constants.md (unspec_volatile_element): Add
752 UNSPEC_VOLATILE_RELAX_GROUP.
753 * config/nds32/nds32-relax-opt.c: New file.
754 * config/nds32/nds32-predicates.c
755 (nds32_symbol_load_store_p): New function.
756 * config/nds32/nds32-protos.h
757 (nds32_symbol_load_store_p): Declare function.
758 (make_pass_nds32_relax_opt): Declare new rtl pass function.
759 * config/nds32/nds32.c
760 (nds32_register_pass): New function to register pass.
761 (nds32_register_passes): New function to register passes.
762 * config/nds32/nds32.md (relax_group): New pattern.
763 * config/nds32/nds32.opt (mrelax-hint): New option.
764 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
766 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
768 * config/nds32/t-nds32: Modify files dependency.
770 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
772 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
773 (PROFILE_HOOK): Define its implementation.
775 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
777 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
778 type and 32-bit size.
780 2018-04-01 Jakub Jelinek <jakub@redhat.com>
783 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
784 (V_128_256): New mode iterator.
785 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
786 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
787 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
789 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
790 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
792 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
795 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
796 NaN inputs correctly.
798 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
801 * config/rs6000/vsx.md (??r): New mode attribute.
802 (*vsx_mov<mode>_64bit): Use it.
803 (*vsx_mov<mode>_32bit): Likewise.
805 2018-03-30 Martin Sebor <msebor@redhat.com>
807 PR tree-optimization/84818
808 * builtins.c (check_access): Use warning_n.
810 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
813 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
815 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
818 2018-03-30 Julia Koval <julia.koval@intel.com>
821 * x86-tune.def (movx, partial_reg_dependency): Enable for
824 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
827 * lra-constraints.c (process_alt_operands): Move setting
828 this_alternative_matches below.
830 2018-03-29 Martin Liska <mliska@suse.cz>
833 * doc/invoke.texi: Document how LTO works with debug info.
834 Describe auto-load support of binutils. Mention 'x86-64'
835 as valid option value of -march option.
837 2018-03-29 Jakub Jelinek <jakub@redhat.com>
839 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
842 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
843 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
844 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
847 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
850 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
851 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
852 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
853 for RS6000_BTM_POWERPC64.
854 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
855 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
856 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
861 2018-03-28 Carl Love <cel@us.ibm.com>
864 2017-09-27 Carl Love <cel@us.ibm.com>
866 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
867 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
868 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
871 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
873 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
874 instead of __vector bool.
875 (_mm_max_pu8): Likewise.
876 (_mm_min_pi16): Likewise.
878 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
881 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
885 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
887 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
888 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
889 (div_extend): Likewise.
890 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
892 (__builtin_divweuo): Likewise.
893 (__builtin_divdeo): Likewise.
894 (__builtin_divdeuo): Likewise.
896 2018-03-28 Jakub Jelinek <jakub@redhat.com>
899 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
900 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
902 PR tree-optimization/82004
903 * gimple-match-head.c (optimize_pow_to_exp): New function.
904 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
905 Don't fold to exp if optimize_pow_to_exp is false.
907 2018-03-28 Martin Liska <mliska@suse.cz>
910 * calls.c (initialize_argument_information): Fix trailing space.
911 * common.opt: Fix typo and provide better explanation for
912 -fsanitize-coverage option.
913 * config/i386/i386.opt: Fix typo.
915 2018-03-28 Jakub Jelinek <jakub@redhat.com>
916 Martin Liska <mliska@suse.cz>
919 * gimplify.c (asan_poison_variable): Don't do the check for
920 gimplify_omp_ctxp here.
921 (gimplify_decl_expr): Do it here.
922 (gimplify_target_expr): Likewise.
924 2018-03-28 Martin Liska <mliska@suse.cz>
927 * config/i386/i386.c (ix86_function_arg_advance): Do not call
928 chkp_type_bounds_count if MPX is not enabled.
930 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
932 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
934 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
937 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
938 function to create the function decl for complex long double
939 multiply and divide for -mabi=ieeelongdouble.
940 (init_float128_ieee): Call it.
942 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
945 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
946 -fcf-protection=branch -mibt.
947 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
949 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
952 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
954 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
957 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
958 extern array declarations.
960 2018-03-27 Richard Biener <rguenther@suse.de>
963 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
964 explicit single_use checks.
966 2018-03-27 Richard Biener <rguenther@suse.de>
968 PR tree-optimization/85082
969 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
972 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
974 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
975 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
976 Turn on fasynchronous-unwind-tables and funwind-tables.
978 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
981 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
982 (*bmi_blsr_<mode>_ccz): Ditto.
984 2018-03-26 Tom de Vries <tom@codesourcery.com>
986 PR tree-optimization/85063
987 * omp-general.c (offloading_function_p): New function. Factor out
989 * omp-offload.c (pass_omp_target_link::gate): ... here.
990 * omp-general.h (offloading_function_p): Declare.
991 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
992 with attribute omp declare target for offloading functions.
994 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
996 PR tree-optimization/84005
997 * tree-data-ref.h (get_base_for_alignment): Declare.
998 * tree-data-ref.c (get_base_for_alignment_1): New function.
999 (get_base_for_alignment): Likewise.
1000 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
1001 get_base_for_alignment to find a suitable base object, instead
1002 of always using drb->base_address.
1004 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1007 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
1008 known size by default.
1010 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
1013 * lra-constraints.c (process_alt_operands): Don't match BLKmode
1014 and non BLKmode operands.
1016 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1019 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
1020 Clean up attributes.
1022 2018-03-23 Richard Biener <rguenther@suse.de>
1025 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
1026 we are going to emit early debug for LTO.
1028 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1031 * function.c (match_asm_constraints_1): Don't optimize if input
1032 doesn't satisfy general_operand predicate for output's mode.
1035 * alias.c (write_dependence_p): Don't require for x_canonicalized
1036 non-VOIDmode if x has VOIDmode.
1039 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
1040 just don't try to optimize it rather than assert it never happens.
1042 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1044 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
1045 macro expansions for definition of ST_INTERNAL_<mode> and
1046 LD_INTERNAL_<mode> builtins.
1047 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
1049 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
1051 (altivec_expand_st_builtin): Likewise.
1052 (altivec_expand_builtin): Remove calls to deleted functions.
1053 (rs6000_address_for_altivec): Delete this function.
1054 * config/rs6000/vector.md: Remove expands for
1055 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
1057 2018-03-22 Sudakshina Das <sudi.das@arm.com>
1060 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
1061 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
1062 re-computing once computed.
1063 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
1064 (arm_init_machine_status): Initialize
1065 machine->static_chain_stack_bytes.
1067 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1070 * doc/extend.texi: Add four new prototypes for vec_ld.
1071 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
1072 definitions for more logical presentation.
1073 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
1074 entries for V1TI variants of __builtin_altivec_ld builtin.
1075 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
1076 handling of V1TI variant of LVX icode pattern.
1077 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
1078 (rs6000_gimple_fold_builtin): Likewise.
1079 (altivec_init_builtins): Add code to define
1080 __builtin_altivec_lvx_v1ti function.
1082 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1085 * function.c (match_asm_constraints_1): Don't do the optimization
1086 if input isn't a REG, SUBREG, MEM or constant.
1088 2018-03-22 Tom de Vries <tom@codesourcery.com>
1090 PR tree-optimization/84956
1091 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
1092 bb_has_abnormal_pred.
1094 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1097 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
1098 DECL_INITIAL (decl) to decl at the end.
1099 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
1102 2018-03-21 Joseph Myers <joseph@codesourcery.com>
1104 * doc/extend.texi (__builtin_tgmath): Document when complex
1105 integer types are treated as _Complex _Float64.
1107 2018-03-21 Tom de Vries <tom@codesourcery.com>
1109 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
1111 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1113 PR tree-optimization/84960
1114 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
1115 if it is ENTRY block, move them into single succ of ENTRY in that case.
1117 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
1119 PR tree-optimization/84811
1120 * poly-int.h (poly_span_traits): Remove the T3 parameter and
1121 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
1122 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
1123 (known_subrange_p): Update accordingly. Cast each value involved
1124 in the size comparison, rather than casting the result of the
1127 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1129 PR tree-optimization/84982
1130 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
1131 by flipping the least significant bit rather than all bits from
1132 bitpos to bitpos + bitsize - 1.
1134 2018-03-21 Nathan Sidwell <nathan@acm.org>
1136 * doc/extend.texi (Deprecated Features): Remove mention of
1137 long-deleted deprecations.
1139 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1142 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
1143 * configure: Regenerate.
1145 2018-03-21 Tom de Vries <tom@codesourcery.com>
1147 PR tree-optimization/83126
1148 * tree-parloops.c (num_phis): New function.
1149 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
1151 2018-03-21 Nathan Sidwell <nathan@acm.org>
1153 * doc/extend.texi (Deprecated Features): Update deprecated flags,
1154 mention anon-struct/union members and trailing attributes.
1156 2018-03-21 Bin Cheng <bin.cheng@arm.com>
1158 PR tree-optimization/84969
1159 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
1160 builtin memset partitions if they set different rhs values.
1162 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1164 PR rtl-optimization/84989
1165 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
1166 VEC_DUPLICATE with scalar result mode.
1168 2018-03-21 Martin Liska <mliska@suse.cz>
1171 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
1172 not intended return statement.
1174 2018-03-21 Martin Liska <mliska@suse.cz>
1177 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
1178 (chkp_find_bound_slots_1): Limit number of iterations.
1180 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
1183 * Minor grammar fixes for x86 options.
1185 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1188 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
1189 holding REG_CFA_RESTORE notes, instead turn them into a USE.
1191 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
1194 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
1195 (altivec_lvx_<mode>_1op): Likewise.
1196 (altivec_stvx_<mode>_2op): Likewise.
1197 (altivec_stvx_<mode>_1op): Likewise.
1198 (altivec_lvx_<VM2:mode>): New define_expand.
1199 (altivec_stvx_<VM2:mode>): Likewise.
1200 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
1201 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1202 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
1203 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1204 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
1205 (rs6000_gen_lvx): Likewise.
1206 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
1207 (altivec_expand_stv_builtin): Likewise.
1208 (altivec_expand_builtin): Likewise.
1209 * config/rs6000/vector.md: Likewise.
1211 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1214 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
1217 2018-03-20 Richard Biener <rguenther@suse.de>
1220 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
1221 sign-conversions as zero, fall back to standard scalar_stmt
1224 2018-03-20 Martin Liska <mliska@suse.cz>
1227 * predict.c (rebuild_frequencies): Handle case when we have
1228 PROFILE_ABSENT, but flag_guess_branch_prob is false.
1230 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1233 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
1234 flag_section_anchors.
1235 * varasm.c (use_blocks_for_decl_p): Remove hack for
1236 dw2_force_const_mem.
1239 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
1241 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
1242 be created, use lowpart_subreg of operands[0] rather than operands[0]
1244 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
1245 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
1246 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
1247 and n constraint instead of aarch64_shift_imm_di and Usd.
1248 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
1249 (*aarch64_<optab>_reg_minus<mode>3): ... this.
1251 2018-03-20 Sudakshina Das <sudi.das@arm.com>
1254 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
1255 to favor GPR over NEON registers.
1256 (<shift>di3_neon): Likewise.
1258 2018-03-20 Tom de Vries <tom@codesourcery.com>
1261 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
1262 (nvptx_process_pars): Emit bar.sync asap and alap.
1264 2018-03-20 Tom de Vries <tom@codesourcery.com>
1267 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
1268 seen_label if seen_label is already set.
1270 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1273 * config/i386/i386.c (fold_builtin_cpu): For features above 31
1274 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
1275 Use 1U instead of 1. Formatting fixes.
1278 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
1279 instead of TREE_TYPE (s1) for the return value.
1281 2018-03-19 Jakub Jelinek <jakub@redhat.com>
1283 PR tree-optimization/84946
1284 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
1285 bitsize + bitsize in poly_uint64 rather than poly_int64.
1288 * dwarf2asm.c: Include fold-const.c.
1289 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
1290 of decl rather than decl itself.
1292 PR rtl-optimization/84643
1293 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
1295 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
1298 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
1299 calling assemble_variable.
1301 2018-03-19 Sudakshina Das <sudi.das@arm.com>
1304 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
1305 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
1307 2018-03-19 Jim Wilson <jimw@sifive.com>
1310 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
1311 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
1312 (riscv_first_stack_step): Likewise.
1313 (riscv_option_override): Use STACK_BOUNDARY instead of
1315 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
1317 (BIGGEST_ALIGNMENT): Set to 128.
1318 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
1319 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
1322 2018-03-19 Richard Biener <rguenther@suse.de>
1324 PR tree-optimization/84933
1325 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
1326 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
1328 2018-03-19 Richard Biener <rguenther@suse.de>
1330 PR tree-optimization/84859
1331 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
1332 (cond_if_else_store_replacement): Perform sinking operation on
1333 single-store BBs regardless of MAX_STORES_TO_SINK setting.
1334 Generalize what a BB with a single eligible store is.
1336 2018-03-19 Richard Biener <rguenther@suse.de>
1338 PR tree-optimization/84929
1339 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
1340 chrec_is_positive against non-chrec arg.
1342 2018-03-19 Tamar Christina <tamar.christina@arm.com>
1345 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
1347 2018-03-18 Martin Liska <mliska@suse.cz>
1349 PR rtl-optimization/84635
1350 * regrename.c (build_def_use): Use matches_mode only when
1353 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
1355 PR tree-optimization/84913
1356 * tree-vect-loop.c (vectorizable_reduction): Don't try to
1357 vectorize chains of COND_EXPRs.
1359 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1361 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
1363 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1365 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
1367 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1369 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
1371 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1372 Kito Cheng <kito.cheng@gmail.com>
1374 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
1375 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
1376 (nds32_adjust_reg_alloc_order): New function.
1377 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
1379 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
1381 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
1382 nds32_print_operand, nds32_print_operand_address): Use
1383 HOST_WIDE_INT_PRINT_DEC instead.
1385 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1387 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
1389 2018-03-17 Jakub Jelinek <jakub@redhat.com>
1392 * config/i386/i386.c (initial_ix86_tune_features,
1393 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
1395 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
1396 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
1397 rather than 1u << ix86_tune. Formatting fix.
1398 (ix86_option_override_internal): Change ix86_arch_mask from
1399 unsigned int to unsigned HOST_WIDE_INT, initialize to
1400 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
1401 (ix86_function_specific_restore): Likewise.
1403 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1406 * postreload.c (reload_combine_recognize_pattern): Perform
1407 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
1408 truncate_int_for_mode the result for the destination's mode.
1411 * hsa-gen.c (mem_type_for_type): Fix comment typo.
1412 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
1414 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1417 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
1420 * lra-assigns.c (lra_split_hard_reg_for): Don't use
1421 regno_allocno_class_array and sorted_pseudos.
1422 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
1423 insns where regno is used.
1425 2018-03-16 Martin Liska <mliska@suse.cz>
1428 * multiple_target.c (create_dispatcher_calls): Redirect
1429 reference in the symbol table.
1431 2018-03-16 Martin Liska <mliska@suse.cz>
1434 * multiple_target.c (create_dispatcher_calls): Redirect also
1437 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1441 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
1442 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
1443 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
1445 2018-03-16 Julia Koval <julia.koval@intel.com>
1447 * doc/invoke.texi (Skylake Server): Add CLWB.
1448 Cannonlake): Remove CLWB.
1450 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1452 PR tree-optimization/84841
1453 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
1455 (FLOAT_ONE_CONST_TYPE): Define.
1456 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
1457 (sort_by_operand_rank): Put entries with higher constant_type last
1458 rather than first to match comments.
1460 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
1462 * config/nios2/nios2.md (movsi_internal): Fix thinko in
1465 2018-03-15 Jakub Jelinek <jakub@redhat.com>
1468 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
1469 check and use address of target always.
1471 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
1474 * config/i386/i386.c (indirect_thunk_needed): Update comments.
1475 (indirect_thunk_bnd_needed): Likewise.
1476 (indirect_thunks_used): Likewise.
1477 (indirect_thunks_bnd_used): Likewise.
1478 (indirect_return_needed): New.
1479 (indirect_return_bnd_needed): Likewise.
1480 (output_indirect_thunk_function): Add a bool argument for
1482 (output_indirect_thunk_function): Don't generate alias for
1483 function return thunk.
1484 (ix86_code_end): Call output_indirect_thunk_function to generate
1485 function return thunks.
1486 (ix86_output_function_return): Set indirect_return_bnd_needed
1487 and indirect_return_needed instead of indirect_thunk_bnd_needed
1488 and indirect_thunk_needed.
1490 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
1492 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
1493 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
1494 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
1496 2018-03-15 David Malcolm <dmalcolm@redhat.com>
1497 Paul Hua <paul.hua.gm@gmail.com>
1500 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
1502 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
1504 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
1505 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
1508 2018-03-15 Tamar Christina <tamar.christina@arm.com>
1511 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
1512 instead of GET_MODE_SIZE when comparing Units.
1514 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1517 * varasm.c (hash_section): Return an unchangeble hash value
1518 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
1519 Return !aarch64_can_use_per_function_literal_pools_p ().
1521 2018-03-15 Jakub Jelinek <jakub@redhat.com>
1524 * optabs.c (emit_conditional_move): Pass address of cmode's copy
1525 rather than address of cmode as last argument to prepare_cmp_insn.
1527 2018-03-15 Julia Koval <julia.koval@intel.com>
1529 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
1530 F_AVX512VNNI, F_AVX512BITALG): New.
1532 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
1535 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
1536 insn for floating-point loads and stores.
1538 2018-03-14 Carl Love <cel@us.ibm.com>
1540 * config/rs6000/rs6000-c.c: Add macro definitions for
1541 ALTIVEC_BUILTIN_VEC_PERMXOR.
1542 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
1543 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
1544 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
1545 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
1547 * config/doc/extend.texi: Add prototypes for vec_permxor.
1549 2018-03-14 David Malcolm <dmalcolm@redhat.com>
1552 * diagnostic-show-locus.c (class layout_point): Convert m_line
1553 from int to linenum_type.
1554 (line_span::comparator): Use linenum "compare" function when
1555 comparing line numbers.
1556 (test_line_span): New function.
1557 (layout_range::contains_point): Convert param "row" from int to
1559 (layout_range::intersects_line_p): Likewise.
1560 (layout::will_show_line_p): Likewise.
1561 (layout::print_source_line): Likewise.
1562 (layout::should_print_annotation_line_p): Likewise.
1563 (layout::print_annotation_line): Likewise.
1564 (layout::print_leading_fixits): Likewise.
1565 (layout::annotation_line_showed_range_p): Likewise.
1566 (struct line_corrections): Likewise for field m_row.
1567 (line_corrections::line_corrections): Likewise for param "row".
1568 (layout::print_trailing_fixits): Likewise.
1569 (layout::get_state_at_point): Likewise.
1570 (layout::get_x_bound_for_row): Likewise.
1571 (layout::print_line): Likewise.
1572 (diagnostic_show_locus): Likewise for locals "last_line" and
1574 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
1575 * input.c (selftest::test_linenum_comparisons): New function.
1576 (selftest::input_c_tests): Call it.
1577 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
1578 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
1579 * selftest.h (ASSERT_GT): New macro.
1580 (ASSERT_GT_AT): New macro.
1581 (ASSERT_LT): New macro.
1582 (ASSERT_LT_AT): New macro.
1584 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
1586 PR rtl-optimization/84780
1587 * combine.c (distribute_links): Don't make a link based on pc_rtx.
1589 2018-03-14 Martin Liska <mliska@suse.cz>
1591 * tree.c (record_node_allocation_statistics): Use
1592 get_stats_node_kind.
1593 (get_stats_node_kind): New function extracted from
1594 record_node_allocation_statistics.
1595 (free_node): Use get_stats_node_kind.
1597 2018-03-14 Richard Biener <rguenther@suse.de>
1599 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
1600 that the value-set of ANTIC_IN doesn't grow.
1603 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1605 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1606 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1608 2018-03-14 Julia Koval <julia.koval@intel.com>
1610 * config.gcc (icelake-client, icelake-server): New.
1612 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
1613 (initial_ix86_arch_features): Ditto.
1614 (PTA_SKYLAKE): Add SGX.
1615 (PTA_ICELAKE): Remove.
1616 (PTA_ICELAKE_CLIENT): New.
1617 (PTA_ICELAKE_SERVER): New.
1618 (ix86_option_override_internal): Split up icelake on icelake client and
1620 (get_builtin_code_for_version): Ditto.
1621 (fold_builtin_cpu): Ditto.
1622 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
1623 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
1624 * config/i386/i386.h (processor_type): Ditto.
1625 * doc/invoke.texi: Ditto.
1627 2018-03-14 Jakub Jelinek <jakub@redhat.com>
1630 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
1631 INTEGER_CST offset, add it together with bitpos / 8 and
1632 sign extend based on POINTER_SIZE.
1636 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
1639 * config/i386/constraints.md (Yc): New register constraint.
1640 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
1641 Use Yc constraint for alternative 2 of operand 0. Remove
1642 preferred_for_speed attribute.
1644 2018-03-14 Richard Biener <rguenther@suse.de>
1646 PR tree-optimization/84830
1647 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
1648 with the old one to avoid oscillations.
1650 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
1653 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
1655 (assign_by_spills): Return a flag of reload assignment failure.
1656 Do not process the reload assignment failures. Do not spill other
1657 reload pseudos if they has the same reg class. Update n if
1659 (lra_assign): Add a return arg. Set up from the result of
1660 assign_by_spills call.
1661 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1662 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1663 usage_insns if it is not NULL.
1664 (spill_hard_reg_in_range): New function.
1665 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1666 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1667 function prototypes.
1668 (lra_assign): Change prototype.
1669 * lra.c (lra): Add code to deal with fails by splitting hard reg
1672 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
1674 * config/riscv/riscv.opt (mrelax): New option.
1675 * config/riscv/riscv.c (riscv_file_start): Emit ".option
1676 "norelax" when riscv_mrelax is disabled.
1677 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
1679 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1682 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
1683 reassociation for int modes.
1685 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1687 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
1688 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
1690 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
1691 * config/aarch64/aarch64-sve.md
1692 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
1693 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
1694 (*extend<mode><Vwide>2): Rename to...
1695 (aarch64_sve_extend<mode><Vwide>2): ...this.
1696 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
1697 renaming the old pattern to...
1698 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
1700 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
1701 define_expand, renaming the old pattern to...
1702 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
1703 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
1704 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
1705 account when deciding which SVE instruction the optab should use.
1706 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
1708 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1710 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
1711 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
1712 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
1713 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
1714 (tlsdesc_small_<mode>): Turn a define_expand and use
1715 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
1716 (tlsdesc_small_advsimd_<mode>): ...this.
1717 (tlsdesc_small_sve_<mode>): New pattern.
1719 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1721 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
1722 (UNSPEC_UMUL_HIGHPART): New constants.
1723 (MUL_HIGHPART): New int iteraor.
1724 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
1725 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
1727 (*<su>mul<mode>3_highpart): New define_insn.
1729 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1732 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1735 2018-03-13 Martin Liska <mliska@suse.cz>
1738 * (sem_item_optimizer::sem_item_optimizer): Initialize new
1740 (sem_item_optimizer::~sem_item_optimizer): Release it.
1741 (sem_item_optimizer::merge_classes): Register variable aliases.
1742 (sem_item_optimizer::fixup_pt_set): New function.
1743 (sem_item_optimizer::fixup_points_to_sets): Likewise.
1744 * ipa-icf.h: Declare new variables and functions.
1746 2018-03-13 Jakub Jelinek <jakub@redhat.com>
1749 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
1750 integer_pow2p@2 and test integer_pow2p in condition.
1751 (A < 0 ? C : 0): Similarly for @1.
1754 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
1755 characters starting at p contain '\0' character, don't look beyond
1759 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
1760 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
1763 * reg-stack.c (change_stack): Change update_end var from int to
1764 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
1765 also call set_block_for_insn on the newly added insns and rescan.
1768 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
1769 on the last operand.
1772 * tree.c (stabilize_reference_1): Return save_expr (e) for
1773 STATEMENT_LIST even if it doesn't have side-effects.
1775 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
1777 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
1779 2018-03-12 Renlin Li <renlin.li@arm.com>
1781 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
1782 aarch64_output_scalar_simd_mov_immediate.
1784 2018-03-12 Martin Sebor <msebor@redhat.com>
1786 PR tree-optimization/83456
1787 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
1788 for perfectly overlapping calls to memcpy.
1789 (gimple_fold_builtin_memory_chk): Same.
1790 (gimple_fold_builtin_strcpy): Handle no-warning.
1791 (gimple_fold_builtin_stxcpy_chk): Same.
1792 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
1794 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
1796 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
1797 parameter. Use it for SFmode.
1798 (rs6000_function_arg_advance_1): Adjust.
1799 (rs6000_function_arg): Adjust.
1800 (rs6000_gimplify_va_arg): Pass false for that new parameter.
1802 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
1804 PR rtl-optimization/84169
1805 PR rtl-optimization/84780
1806 * combine.c (can_combine_p): Check for a 2-insn combination whether
1807 the destination register is used between the two insns, too.
1809 2018-03-12 Richard Biener <rguenther@suse.de>
1811 PR tree-optimization/84803
1812 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
1813 for refs DR analysis didn't process.
1815 2018-03-12 Richard Biener <rguenther@suse.de>
1817 PR tree-optimization/84777
1818 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
1819 force-vectorize loops ignore whether we are optimizing for size.
1821 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
1823 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
1824 (TARGET_MD_ASM_ADJUST): Define.
1826 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
1827 Kito Cheng <kito.cheng@gmail.com>
1828 Chung-Ju Wu <jasonwucj@gmail.com>
1830 * config/nds32/nds32.c (nds32_compute_stack_frame,
1831 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
1832 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
1833 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
1834 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
1835 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
1836 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
1837 * config/nds32/nds32.md (prologue, epilogue): Use macro
1838 NDS32_V3PUSH_AVAILABLE_P to do checking.
1840 2018-03-11 Jakub Jelinek <jakub@redhat.com>
1843 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
1844 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
1845 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
1846 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
1847 addition of most attributes on !orig_type_die or the attribute not
1848 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
1850 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1851 Chung-Ju Wu <jasonwucj@gmail.com>
1853 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
1855 * config/nds32/nds32.opt (mvh): New option.
1857 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1858 Chung-Ju Wu <jasonwucj@gmail.com>
1860 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
1862 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
1863 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
1866 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1867 Chung-Ju Wu <jasonwucj@gmail.com>
1869 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
1871 * config/nds32/nds32-multiple.md (strlensi): New pattern.
1872 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
1874 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
1875 Kito Cheng <kito.cheng@gmail.com>
1876 Chung-Ju Wu <jasonwucj@gmail.com>
1878 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
1879 UNSPEC_FFMISM and UNSPEC_FLMISM.
1880 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
1881 for ffb, ffmism and flmism.
1882 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
1883 (unspec_ffmism): Ditto.
1884 (unspec_flmism): Ditto.
1885 (nds32_expand_builtin_impl): Check if string extension is available.
1886 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
1887 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
1889 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
1892 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1895 * lra-assigns.c (assign_by_spills): Return a flag of reload
1896 assignment failure. Do not process the reload assignment
1897 failures. Do not spill other reload pseudos if they has the same
1899 (lra_assign): Add a return arg. Set up from the result of
1900 assign_by_spills call.
1901 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1902 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1903 usage_insns if it is not NULL.
1904 (spill_hard_reg_in_range): New function.
1905 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1906 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1907 function prototypes.
1908 (lra_assign): Change prototype.
1909 * lra.c (lra): Add code to deal with fails by splitting hard reg
1912 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
1915 * config/i386/i386.opt: Replace Enforcment with Enforcement.
1917 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
1920 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
1921 (dw_val_node): Add val_symbolic_view.
1922 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
1923 (symview_upper_bound): New.
1924 (new_line_info_table): Initialize symviews_since_reset.
1925 (dwarf2out_source_line): Count symviews_since_reset and set
1926 symview_upper_bound.
1927 (dw_val_equal_p): Handle symview.
1928 (add_AT_symview): New.
1929 (print_dw_val): Handle symview.
1930 (attr_checksum, attr_checksum_ordered): Likewise.
1931 (same_dw_val_p, size_of_die): Likewise.
1932 (value_format, output_die): Likewise.
1933 (add_high_low_attributes): Use add_AT_symview for entry_view.
1934 (dwarf2out_finish): Reset symview_upper_bound, clear
1937 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
1940 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
1941 Add strict argument and use it.
1942 (rs6000_split_multireg_move): Update for new strict argument.
1943 (mem_operand_gpr): Disallow all non-offsettable addresses.
1944 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
1946 2018-03-09 Jakub Jelinek <jakub@redhat.com>
1949 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
1950 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
1951 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
1954 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
1955 decl, use remap_type if we want to use the type.
1957 2018-03-09 Martin Sebor <msebor@redhat.com>
1959 PR tree-optimization/84526
1960 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1962 (builtin_access::generic_overlap): Be prepared to handle non-array
1965 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
1967 PR rtl-optimization/84682
1968 * lra-constraints.c (process_address_1): Check is_address flag
1969 for address constraints.
1970 (process_alt_operands): Likewise.
1971 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
1972 preprocess_constraints.
1973 * recog.h (preprocess_constraints): Add oploc parameter.
1975 * recog.c (preprocess_constraints): Test address_operand for
1976 CT_ADDRESS constraints.
1978 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1981 * lra-assigns.c (assign_by_spills): Return a flag of reload
1982 assignment failure. Do not process the reload assignment
1983 failures. Do not spill other reload pseudos if they has the same
1985 (lra_assign): Add a return arg. Set up from the result of
1986 assign_by_spills call.
1987 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1988 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1989 usage_insns if it is not NULL.
1990 (spill_hard_reg_in_range): New function.
1991 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1992 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1993 function prototypes.
1994 (lra_assign): Change prototype.
1995 * lra.c (lra): Add code to deal with fails by splitting hard reg
1998 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2001 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
2002 Accept complain bool parameter. Only emit errors if it is true.
2003 (arm_parse_cpu_option_name): Likewise.
2004 (arm_target_thumb_only): Adjust callers of the above.
2005 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
2006 prototype to take a default true bool parameter.
2007 (arm_parse_arch_option_name): Likewise.
2009 2018-03-09 David Malcolm <dmalcolm@redhat.com>
2010 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
2014 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
2015 * configure: Regenerate.
2016 * configure.ac ("linker --version-script option"): New.
2017 ("linker soname option"): New.
2019 2018-03-09 Richard Biener <rguenther@suse.de>
2021 PR tree-optimization/84775
2022 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
2023 immediate uses of predicate stmts and mark them modified.
2026 PR tree-optimization/84178
2027 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2029 (version_loop_for_if_conversion): Delay update_ssa call.
2030 (tree_if_conversion): Delay update_ssa until after predicate
2033 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
2036 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
2037 when the function accesses prior frames.
2039 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2042 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
2043 gen_llsym, otherwise call maybe_gen_llsym.
2046 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
2047 has ',' character inside of it.
2049 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2052 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
2053 as clobbering CC_REGNUM.
2055 2018-03-08 Richard Biener <rguenther@suse.de>
2058 * tree-scalar-evolution.c: Include tree-into-ssa.h.
2059 (follow_copies_to_constant): Do not follow SSA names registered
2062 2018-03-08 Richard Biener <rguenther@suse.de>
2064 PR tree-optimization/84178
2065 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2067 (version_loop_for_if_conversion): Delay update_ssa call.
2068 (tree_if_conversion): Delay update_ssa until after predicate
2071 2018-03-08 David Malcolm <dmalcolm@redhat.com>
2073 PR tree-optimization/84178
2074 * tree-if-conv.c (release_bb_predicate): Remove the
2075 the assertion that the stmts have NULL use_ops.
2076 Discard the statements, asserting that they haven't
2077 yet been added to a BB.
2079 2018-03-08 Richard Biener <rguenther@suse.de>
2081 PR tree-optimization/84746
2082 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
2083 (phi_translate): Pass in destination ANTIC_OUT set.
2084 (phi_translate_1): Likewise. For a simplified result lookup
2085 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
2086 (phi_translate_set): Adjust.
2087 (do_pre_regular_insertion): Likewise.
2088 (do_pre_partial_partial_insertion): Likewise.
2090 2018-03-08 Martin Liska <mliska@suse.cz>
2092 PR gcov-profile/84735
2093 * doc/gcov.texi: Document usage of profile files.
2094 * gcov-io.h: Document changes in the format.
2096 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
2100 * dwarf2out.c (struct dw_line_info_table): Update comments for
2102 (FORCE_RESET_NEXT_VIEW): New.
2103 (FORCE_RESETTING_VIEW_P): New.
2104 (RESETTING_VIEW_P): Check for -1 too.
2105 (ZERO_VIEW_P): Likewise.
2106 (new_line_info_table): Force-reset next view.
2107 (dwarf2out_begin_function): Likewise.
2108 (dwarf2out_source_line): Simplify zero_view_p initialization.
2109 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
2110 view directly. Omit view when omitting .loc at line 0.
2112 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2114 PR tree-optimization/84740
2115 * tree-switch-conversion.c (process_switch): Call build_constructors
2116 only if info.phi_count is non-zero.
2118 PR tree-optimization/84739
2119 * tree-tailcall.c (find_tail_calls): Check call arguments against
2120 DECL_ARGUMENTS (current_function_decl) rather than
2121 DECL_ARGUMENTS (func) when checking for tail recursion.
2123 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2125 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
2126 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
2127 Volker Reichelt's entry and add entries for people that perform
2128 GCC fuzzy testing and report numerous bugs.
2130 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
2133 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
2134 readonly data in sdata, if that is disabled.
2135 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
2136 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
2137 -mreadonly-in-sdata option.
2139 2018-03-07 Martin Sebor <msebor@redhat.com>
2141 PR tree-optimization/84468
2142 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
2143 basic block when looking for nul assignment.
2145 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
2148 * except.h (output_function_exception_table): Adjust prototype.
2149 * except.c (output_function_exception_table): Remove FNNAME parameter
2150 and add SECTION parameter. Ouput one part of the table at a time.
2151 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
2152 the first part of the exception table and emit unwind directives.
2153 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
2154 (i386_pe_seh_cold_init): Likewise.
2155 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
2156 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
2157 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
2158 (ix86_output_call_insn): Emit a nop in one more case for SEH.
2159 * config/i386/winnt.c: Include except.h.
2160 (struct seh_frame_state): Add reg_offset, after_prologue and
2161 in_cold_section fields.
2162 (i386_pe_seh_end_prologue): Set seh->after_prologue.
2163 (i386_pe_seh_cold_init): New function.
2164 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
2165 to seh->in_cold_section.
2166 (seh_emit_push): Record the offset of the push.
2167 (seh_emit_save): Record the offet of the save.
2168 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
2169 Test seh->after_prologue to disregard the epilogue.
2170 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
2171 (i386_pe_end_cold_function): New function.
2173 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2176 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
2177 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
2180 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
2182 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
2183 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
2186 * multiple_target.c: Include tree-inline.h and intl.h.
2187 (expand_target_clones): Diagnose and fail if node->definition and
2188 !tree_versionable_function_p (node->decl).
2190 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
2192 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
2194 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
2195 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
2196 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
2198 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2201 * combine.c (try_combine): Use reg_or_subregno instead of handling
2202 just paradoxical SUBREGs and REGs.
2204 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
2206 * config/arc/arc.c (arc_finalize_pic): Remove function.
2207 (arc_must_save_register): We use single base PIC register, remove
2208 checks to save/restore the PIC register.
2209 (arc_expand_prologue): Likewise.
2210 * config/arc/arc-protos.h (arc_set_default_type_attributes):
2212 (arc_verify_short): Likewise.
2213 (arc_attr_type): Likewise.
2214 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
2215 (walk_stores): Likewise.
2216 (arc_address_cost): Make it static.
2217 (arc_verify_short): Likewise.
2218 (branch_dest): Likewise.
2219 (arc_attr_type): Likewise.
2220 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
2221 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
2222 (arc_final_prescan_insn): Remove inserting the nops due to
2223 hardware hazards. It is done in reorg step.
2224 (insn_length_variant_t): Remove.
2225 (insn_length_parameters_t): Likewise.
2226 (arc_insn_length_parameters): Likewise.
2227 (arc_get_insn_variants): Likewise.
2228 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
2230 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2233 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
2236 PR tree-optimization/84687
2237 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
2239 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
2241 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2243 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
2244 Rename to ppc_speculation_barrier.
2245 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
2246 __builtin_ppc_speculation_barrier.
2248 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2251 * combine.c (combine_simplify_rtx): Don't try to simplify if
2252 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
2255 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
2257 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
2258 to 32 bytes when compiling for POWER9.
2260 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2263 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
2264 regparm >= 3 with no arg reg available also for calls with
2265 flag_force_indirect_call. Pass decl to ix86_function_regparm.
2268 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
2270 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
2272 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
2275 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
2277 2018-03-05 Richard Biener <rguenther@suse.de>
2279 PR tree-optimization/84486
2280 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
2281 When inserting a __builtin_assume_aligned call set the LHS
2282 SSA name alignment info accordingly.
2284 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
2286 PR tree-optimization/84114
2287 * config/aarch64/aarch64.c (aarch64_reassociation_width)
2288 Avoid reassociation of FLOAT_MODE addition.
2290 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
2292 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
2293 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
2294 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
2295 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
2296 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
2297 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
2298 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
2300 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
2301 __builtin_ia32_wbinvd): New builtins.
2302 (SPECIAL_ARGS2): New.
2303 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
2304 (SPECIAL_ARGS2): New.
2305 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
2306 (ix86_valid_target_attribute_inner_p): Ditto.
2307 (ix86_init_mmx_sse_builtins): Add special_args2.
2308 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
2309 TARGET_WBNOINVD_P): New.
2310 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
2311 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
2312 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
2313 * config/i386/immintrin.h (_wbinvd): New intrinsic.
2314 * config/i386/pconfigintrin.h: New file.
2315 * config/i386/wbnoinvdintrin.h: Ditto.
2316 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
2317 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
2319 2018-03-05 Richard Biener <rguenther@suse.de>
2321 PR tree-optimization/84670
2322 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2324 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2325 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2326 (compute_antic_aux): Only assert the number of values in ANTIC_IN
2327 doesn't grow if all successors (recursively) were visited at least
2330 2018-03-05 Richard Biener <rguenther@suse.de>
2332 PR tree-optimization/84650
2333 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
2334 if executed in the loop pipeline.
2336 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
2338 * doc/configfiles.texi (Configuration Files): Move info about
2339 conditionalizing $target-protos.h to...
2340 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
2341 differs from $target-protos.h.
2343 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
2344 Chung-Ju Wu <jasonwucj@gmail.com>
2346 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
2347 * config/nds32/nds32-multiple.md (setmemsi): Define.
2348 * config/nds32/nds32-memory-manipulation.c
2349 (nds32_gen_dup_4_byte_to_word_value): New.
2350 (emit_setmem_word_loop): New.
2351 (emit_setmem_byte_loop): New.
2352 (nds32_expand_setmem_loop): New.
2353 (nds32_expand_setmem_loop_v3m): New.
2354 (nds32_expand_setmem_unroll): New.
2355 (nds32_expand_setmem): New.
2357 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2358 Chung-Ju Wu <jasonwucj@gmail.com>
2360 * config/nds32/nds32-memory-manipulation.c
2361 (nds32_emit_load_store): New.
2362 (nds32_emit_post_inc_load_store): New.
2363 (nds32_emit_mem_move): New.
2364 (nds32_emit_mem_move_block): New.
2365 (nds32_expand_movmemsi_loop_unknown_size): New.
2366 (nds32_expand_movmemsi_loop_known_size): New.
2367 (nds32_expand_movmemsi_loop): New.
2368 (nds32_expand_movmemsi_unroll): New.
2369 (nds32_expand_movmemqi): Rename ...
2370 (nds32_expand_movmemsi): ... to this.
2371 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
2372 (movmemsi): ... to this.
2373 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
2374 (nds32_expand_movmemsi): ... to this.
2376 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2377 Monk Chiang <sh.chiang04@gmail.com>
2378 Chung-Ju Wu <jasonwucj@gmail.com>
2380 * config/nds32/nds32-protos.h
2381 (nds32_expand_load_multiple): New arguments.
2382 (nds32_expand_store_multiple): Ditto.
2383 (nds32_valid_multiple_load_store): Rename ...
2384 (nds32_valid_multiple_load_store_p): ... to this.
2385 * config/nds32/nds32-memory-manipulation.c
2386 (nds32_expand_load_multiple): Refine implementation.
2387 (nds32_expand_store_multiple): Ditto.
2388 * config/nds32/nds32-multiple.md
2389 (load_multiple): Update nds32_expand_load_multiple interface.
2390 (store_multiple): Update nds32_expand_store_multiple interface.
2391 * config/nds32/nds32-predicates.c
2392 (nds32_valid_multiple_load_store): Rename ...
2393 (nds32_valid_multiple_load_store_p): ... to this and refine
2395 * config/nds32/predicates.md
2396 (nds32_load_multiple_and_update_address_operation): New predicate.
2397 (nds32_store_multiple_and_update_address_operation): New predicate.
2399 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2400 Chung-Ju Wu <jasonwucj@gmail.com>
2402 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
2403 (combo): New attribute.
2404 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
2406 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
2408 * config/nds32/nds32.opt: Change -mcmodel= default value.
2410 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
2411 Monk Chiang <sh.chiang04@gmail.com>
2412 Chung-Ju Wu <jasonwucj@gmail.com>
2414 * config/nds32/constants.md (unspec_element): New enum.
2415 * config/nds32/constraints.md (Umw): New constraint.
2416 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
2417 * config/nds32/nds32-intrinsic.md: Likewise.
2418 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
2419 (nds32_valid_smw_lwm_base_p): New.
2420 (nds32_output_smw_single_word): New.
2421 (nds32_output_lmw_single_word): New.
2422 (nds32_expand_unaligned_load): New.
2423 (nds32_expand_unaligned_store): New.
2424 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
2425 (nds32_output_smw_single_word): Declare.
2426 (nds32_output_lmw_single_word): Declare.
2427 (nds32_expand_unaligned_load): Declare.
2428 (nds32_expand_unaligned_store): Declare.
2429 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
2430 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
2431 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
2432 NDS32_BUILTIN_UASTORE_DW.
2433 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
2436 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
2437 Kito Cheng <kito.cheng@gmail.com>
2438 Chung-Ju Wu <jasonwucj@gmail.com>
2440 * config/nds32/nds32-intrinsic.c
2441 (nds32_expand_builtin_null_ftype_reg): Delete.
2442 (nds32_expand_builtin_reg_ftype_imm): Ditto.
2443 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
2444 (nds32_read_argument): New.
2445 (nds32_legitimize_target): Ditto.
2446 (nds32_legitimize_argument): Ditto.
2447 (nds32_check_constant_argument): Ditto.
2448 (nds32_expand_unop_builtin): Ditto.
2449 (nds32_expand_unopimm_builtin): Ditto.
2450 (nds32_expand_binop_builtin): Ditto.
2451 (nds32_builtin_decl_impl): Ditto.
2452 (builtin_description): Ditto.
2453 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
2454 (nds32_init_builtins_impl): Ditto.
2455 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
2456 (nds32_builtin_decl): New.
2457 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
2458 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
2460 2018-03-02 Jeff Law <law@redhat.com>
2462 * reorg.c (stop_search_p): Handle DEBUG_INSN.
2463 (redundant_insn, fill_simple_delay_slots): Likewise.
2464 (fill_slots_from_thread): Likewise.
2465 * resource.c (mark_referenced_resources): Likewise.
2466 (mark_set_resources, find_dead_or_set_registers): Likewise.
2468 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2470 * substring-locations.h (format_warning_va): Formatting fix for
2472 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
2474 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
2475 * substring-locations.c: Include intl.h.
2476 (format_warning_va): Turned into small wrapper around
2477 format_warning_n_va, renamed to ...
2478 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
2479 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
2481 (format_warning_at_substring_n): New function.
2482 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
2483 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
2484 format_warning_at_substring with just a shorter name instead of
2485 const function pointer.
2486 (fmtwarn_n): New function.
2487 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
2488 appropriate, get rid of all the fmtstr temporaries, move conditionals
2489 with G_() wrapped string literals directly into fmtwarn arguments,
2490 cast dir.len to (int), formatting fixes.
2492 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
2494 * doc/invoke.texi: Remove "Cilk Plus" references.
2496 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2497 Richard Biener <rguenther@suse.de>
2500 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
2501 for error or warning attributes if CALL_FROM_THUNK_P is set.
2504 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2507 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
2508 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
2511 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
2512 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
2514 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
2515 instead of -1U in last predictors element's probability member.
2517 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2520 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2521 arguments if they are comparable.
2523 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
2525 PR tree-optimization/84634
2526 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
2527 masks and masked_loop_p with a single loop_masks, making sure it's
2528 null for bb vectorization.
2530 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
2532 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
2533 (vect_analyze_data_ref_access): Use loop->safe_len rather than
2534 loop->force_vectorize to check whether there is no alias.
2536 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2539 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
2541 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
2543 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
2544 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
2545 instead of a loop around prev_real_insn.
2546 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
2550 * config/i386/i386.c (ix86_print_operand): Use conditional
2551 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
2554 2018-03-02 Richard Biener <rguenther@suse.de>
2556 PR tree-optimization/84427
2557 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
2558 (bitmap_set_subtract_values): Rewrite to handle multiple
2561 (prune_clobbered_mems): Likewise.
2562 (phi_translate): Take edge instead of pred/phiblock.
2563 (phi_translate_1): Likewise.
2564 (phi_translate_set): Likewise. Insert all translated
2565 exprs for a value into the set, keeping possibly multiple
2566 expressions per value.
2567 (compute_antic_aux): Adjust for phi_translate changes.
2568 When intersecting union the expressions and prune those
2569 not in the final value set, keeping possibly multiple
2570 expressions per value. Do not use value-insertion
2571 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
2572 all expressions. Add verification that the value-sets
2573 only shrink during iteration.
2574 (compute_partial_antic_aux): Adjust for the phi_translate changes.
2575 (do_pre_regular_insertion): Likewise.
2576 (do_pre_partial_partial_insertion): Likewise.
2578 2018-03-02 Richard Biener <rguenther@suse.de>
2581 * config/darwin.c (saved_debug_info_level): New static global.
2582 (darwin_asm_lto_start): Disable debug info generation for LTO out.
2583 (darwin_asm_lto_end): Restore debug info generation settings.
2585 2018-03-01 Martin Liska <mliska@suse.cz>
2588 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
2591 2018-03-01 Richard Biener <rguenther@suse.de>
2594 * dwarf2out.c (gen_variable_die): Properly handle late VLA
2595 type annotation with LTO when debug was disabled at compile-time.
2597 2018-03-01 Matthew Fortune <mfortune@gmail.com>
2599 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
2601 (mips_final_postscan_insn): Likewise.
2603 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
2605 PR rtl-optimization/84528
2606 * alias.c (init_alias_target): Add commentary.
2607 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
2608 a unique base value if the frame pointer is not eliminated
2609 to the stack pointer.
2611 2018-03-01 Tom de Vries <tom@codesourcery.com>
2613 PR rtl-optimization/83327
2614 * lra-int.h (hard_regs_spilled_into): Declare.
2615 * lra.c (hard_regs_spilled_into): Define.
2616 (init_reg_info): Init hard_regs_spilled_into.
2617 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
2618 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
2619 (process_bb_lives): Handle hard_regs_spilled_into.
2620 (lra_create_live_ranges_1): Before doing liveness propagation, clear
2621 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
2623 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
2625 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
2626 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
2627 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
2628 * config/rs6000/aix72.h: New file.
2630 2018-02-28 Jakub Jelinek <jakub@redhat.com>
2632 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
2633 instead of warning_at with conditional singular and plural messages
2637 * stor-layout.c (update_alignment_for_field): For
2638 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
2639 && !DECL_PACKED (field), do the alignment update, just use
2640 only desired_align instead of MAX (type_align, desired_align)
2642 (place_field): Don't do known_align < desired_align handling
2643 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
2644 is non-NULL, instead do it after rli->prev_field handling and
2645 only if not within a bitfield word. For DECL_PACKED (field)
2646 use type_align of BITS_PER_UNIT.
2648 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
2650 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
2651 superfluous parentheses and trailing spaces.
2653 2018-02-28 Richard Biener <rguenther@suse.de>
2655 PR tree-optimization/84584
2656 * graphite-scop-detection.c (scop_detection::add_scop): Discard
2657 SCoPs with fake exit edge.
2659 2018-02-28 Martin Liska <mliska@suse.cz>
2662 * timevar.c (timer::print): Fix format to properly print 100%
2665 2018-02-28 Richard Biener <rguenther@suse.de>
2668 * genmatch.c (capture_info::walk_match): Do not mark
2669 captured expressions without operands as expr_p given
2670 they act more like predicates and should be subject to
2671 "lost tail" side-effect preserving.
2673 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
2675 PR rtl-optimization/81611
2676 * auto-inc-dec.c (attempt_change): Move dead note from
2677 mem_insn if it's the next use of regno
2678 (find_address): Take address use of reg holding
2679 non-incremented value. Add parm to limit search to the named
2681 (merge_in_block): Attempt to use a mem insn that is the next
2682 use of the original regno.
2684 2018-02-27 Martin Sebor <msebor@redhat.com>
2687 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
2688 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
2690 2018-02-27 Martin Sebor <msebor@redhat.com>
2692 PR translation/84207
2693 * diagnostic-core.h (warning_n, error_n, inform_n): Change
2694 n argument to unsigned HOST_WIDE_INT.
2695 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
2696 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
2697 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
2698 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
2700 2018-02-27 Richard Biener <rguenther@suse.de>
2702 PR tree-optimization/84512
2703 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
2704 Do not use the estimate returned from record_stmt_cost for
2705 the scalar iteration cost but sum properly using add_stmt_cost.
2707 2018-02-27 Richard Biener <rguenther@suse.de>
2709 PR tree-optimization/84466
2710 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2711 Adjust last change to less strictly validate use operands.
2713 2018-02-27 Martin Liska <mliska@suse.cz>
2715 PR gcov-profile/84548
2716 * gcov.c (process_file): Allow partial overlap and consider it
2717 also as group functions.
2718 (output_lines): Properly calculate range of lines for a group.
2720 2018-02-27 Martin Liska <mliska@suse.cz>
2722 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
2723 'ggc' suffixes. Change first column width.
2724 (timer::print): Fix formatting of the column.
2726 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
2728 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
2729 preserve inline entry blocks for the sake of debug inline
2730 entry point markers alone.
2731 (remove_unused_locals): Suggest in comments a better place to
2732 force the preservation of inline entry blocks that are
2733 otherwise unused, but do not preserve them.
2735 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2737 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
2739 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2742 * config/i386/constraints.md (Bs): Replace
2743 ix86_indirect_branch_register with
2744 TARGET_INDIRECT_BRANCH_REGISTER.
2746 * config/i386/i386.md (indirect_jump): Likewise.
2747 (tablejump): Likewise.
2748 (*sibcall_memory): Likewise.
2749 (*sibcall_value_memory): Likewise.
2750 Peepholes of indirect call and jump via memory: Likewise.
2751 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
2752 (*sibcall_value_GOT_32): Likewise.
2753 * config/i386/predicates.md (indirect_branch_operand): Likewise.
2754 (GOT_memory_operand): Likewise.
2755 (call_insn_operand): Likewise.
2756 (sibcall_insn_operand): Likewise.
2757 (GOT32_symbol_operand): Likewise.
2758 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
2760 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2762 PR rtl-optimization/83496
2763 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
2764 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
2765 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
2766 redundant insn, if any.
2767 (relax_delay_slots): Likewise.
2768 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
2770 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
2772 PR tree-optimization/83965
2773 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
2774 that grouped statements are part of a reduction chain. Return
2775 true if the statement is not marked as a reduction itself but
2777 (vect_recog_dot_prod_pattern): Don't check whether the statement
2778 is part of a group here.
2779 (vect_recog_sad_pattern): Likewise.
2780 (vect_recog_widen_sum_pattern): Likewise.
2782 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2785 * final.c (rest_of_clean_state): Also look for calls inside sequences.
2787 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2790 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
2792 (ix86_output_indirect_function_return): New prototype.
2793 (ix86_split_simple_return_pop_internal): Likewise.
2794 * config/i386/i386.c (indirect_return_via_cx): New.
2795 (indirect_return_via_cx_bnd): Likewise.
2796 (indirect_thunk_name): Handle return va CX_REG.
2797 (output_indirect_thunk_function): Create alias for
2798 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
2799 (ix86_output_indirect_jmp): Remove the bool argument.
2800 (ix86_output_indirect_function_return): New function.
2801 (ix86_split_simple_return_pop_internal): Likewise.
2802 * config/i386/i386.md (*indirect_jump): Don't pass false
2803 to ix86_output_indirect_jmp.
2804 (*tablejump_1): Likewise.
2805 (simple_return_pop_internal): Change it to define_insn_and_split.
2806 Call ix86_split_simple_return_pop_internal to split it for
2808 (simple_return_indirect_internal): Call
2809 ix86_output_indirect_function_return instead of
2810 ix86_output_indirect_jmp.
2812 2018-02-26 Jakub Jelinek <jakub@redhat.com>
2815 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
2816 memset and value initialization afterwards.
2818 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
2820 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
2822 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2825 * common/config/aarch64/aarch64-common.c
2826 (aarch_option_optimization_table[]): Switch
2827 off fomit-frame-pointer
2829 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
2830 Chung-Ju Wu <jasonwucj@gmail.com>
2832 * config/nds32/nds32-multiple.md (load_multiple): Disallow
2834 (store_multiple): Ditto.
2836 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
2838 * config.gcc: Add --with-cpu support for nds32 target.
2839 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
2840 * config/nds32/nds32.opt: Add -mcpu= option.
2842 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
2844 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
2845 isel=yes): Warn for these deprecated options.
2847 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
2849 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
2850 ISA_2_5_MASKS_EMBEDDED.
2852 2018-02-23 Jakub Jelinek <jakub@redhat.com>
2854 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
2855 p->max as pointers rather than using iterative_hash_expr.
2857 2018-02-23 Carl Love <cel@us.ibm.com>
2859 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
2860 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
2862 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
2863 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
2864 P8V_BUILTIN_VEC_VUNSIGNED2.
2866 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
2869 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
2870 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
2872 * lra-constraints.c (curr_insn_transform): Set up
2873 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
2875 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
2876 * lra-eliminations.c (spill_pseudos): Ditto.
2877 (process_insn_for_elimination): Ditto.
2878 * lra-lives.c (reg_early_clobber_p): Use the new macros.
2879 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
2880 LRA_NON_CLOBBERED_ALT.
2882 2018-02-22 Martin Sebor <msebor@redhat.com>
2884 PR tree-optimization/84480
2885 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
2886 to maybe_diag_stxncpy_trunc. Call it.
2887 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
2888 from gimple_fold_builtin_strcpy. Print inlining stack.
2889 (handle_builtin_stxncpy): Print inlining stack.
2890 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
2892 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
2895 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
2896 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
2897 and -fcheck-pointer-bounds are used together.
2898 (indirect_thunk_prefix): New enum.
2899 (indirect_thunk_need_prefix): New function.
2900 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
2901 "_nt" instead of "_bnd" for NOTRACK prefix.
2902 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
2903 (output_indirect_thunk_function): Likewise.
2905 (ix86_code_end): Update output_indirect_thunk_function calls.
2906 (ix86_output_indirect_branch_via_reg): Replace
2907 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
2908 (ix86_output_indirect_branch_via_push): Likewise.
2909 (ix86_output_function_return): Likewise.
2910 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
2911 incompatible with -fcf-protection=branch and
2912 -fcheck-pointer-bounds.
2914 2018-02-22 Steve Ellcey <sellcey@cavium.com>
2917 * config/aarch64/aarch64.c (aarch64_print_address_internal):
2918 Change gcc_assert call to output_operand_lossage.
2920 2018-02-22 Steve Ellcey <sellcey@cavium.com>
2922 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
2924 2018-02-22 DJ Delorie <dj@redhat.com>
2925 Sebastian Perta <sebastian.perta@renesas.com>
2926 Oleg Endo <olegendo@gcc.gnu.org>
2928 * config/rx/rx.c (rx_rtx_costs): New function.
2929 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
2931 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
2933 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
2935 2018-02-22 Martin Liska <mliska@suse.cz>
2938 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
2939 Add "native" as a possible value.
2941 2018-02-22 Martin Liska <mliska@suse.cz>
2944 * config/i386/i386.c (ix86_option_override_internal):
2945 Add "native" as a possible value for -march and -mtune.
2947 2018-02-22 Jakub Jelinek <jakub@redhat.com>
2950 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
2951 to all type variants.
2953 PR tree-optimization/84503
2954 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
2955 width as info->bitpos + info->bitsize - start.
2956 (merged_store_group::merge_overlapping): Simplify width computation.
2957 (check_no_overlap): New function.
2958 (imm_store_chain_info::try_coalesce_bswap): Compute expected
2959 start + width and last_order of the group, fail if check_no_overlap
2961 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
2962 to group if check_no_overlap fails.
2964 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2966 * config/rs6000/altivec.md: Delete contraint arguments to
2967 define_expand, define_split, and define_peephole2, and in
2968 define_insn_and_split if always unused.
2969 * config/rs6000/darwin.md: Ditto.
2970 * config/rs6000/dfp.md: Ditto.
2971 * config/rs6000/rs6000.md: Ditto.
2972 * config/rs6000/sync.md: Ditto.
2973 * config/rs6000/vector.md: Ditto.
2974 * config/rs6000/vsx.md: Ditto.
2976 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2978 * config/rs6000/altivec.md: Write output control strings as braced
2979 blocks instead of double-quoted strings.
2980 * config/rs6000/darwin.md: Ditto.
2981 * config/rs6000/rs6000.md: Ditto.
2982 * config/rs6000/vector.md: Ditto.
2983 * config/rs6000/vsx.md: Ditto.
2985 2018-02-21 Jason Merrill <jason@redhat.com>
2987 PR c++/84314 - ICE with templates and fastcall attribute.
2988 * attribs.c (build_type_attribute_qual_variant): Remove assert.
2990 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2992 * ipa-cp.c (determine_versionability): Fix comment typos.
2994 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2997 * ipa-cp.c (determine_versionability): Do not version functions caling
3000 2018-02-21 Martin Liska <mliska@suse.cz>
3003 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
3004 Add "native" as a possible value.
3005 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
3006 the macro when native cpu detection is available.
3008 2018-02-21 Martin Liska <mliska@suse.cz>
3011 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
3012 Add "native" as a possible value.
3013 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
3014 when native cpu detection is available.
3016 2018-02-21 Jakub Jelinek <jakub@redhat.com>
3017 Martin Sebor <msebor@redhat.com>
3019 PR tree-optimization/84478
3020 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
3022 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
3023 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
3024 support which is conservatively correct, for 2 only stay conservative
3025 for maxlen. Formatting and comment capitalization fixes. Add STRICT
3026 argument to the 2 argument get_range_strlen, adjust 6 arg
3027 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
3029 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
3030 (gimple_fold_builtin_strlen): Pass true as last argument to
3033 2018-02-20 Martin Sebor <msebor@redhat.com>
3036 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
3037 (builtin_memref::set_base_and_offset): Same. Handle inner references.
3038 (builtin_memref::builtin_memref): Factor out parts into
3039 set_base_and_offset and call it.
3041 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3044 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
3045 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
3046 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
3047 search at the associated MODE_INT.
3049 2018-02-20 Jeff Law <law@redhat.com>
3052 PR tree-optimization/81592
3054 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
3055 for range data rather than using global data.
3056 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
3057 range data rather than using global data.
3058 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
3059 pass it to children as needed.
3060 (struct directive::fmtresult): Similarly.
3061 (struct directive::set_width): Similarly.
3062 (struct directive::set_precision): Similarly.
3063 (format_integer, format_directive, parse_directive): Similarly.
3064 (format_none): Accept unnamed vr_values parameter.
3065 (format_percent, format_floating, format_character): Similarly.
3066 (format_string, format_plain): Similarly.
3067 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
3068 the EVRP range analyzer for range data rather than using global data.
3069 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
3070 gimple-ssa-evrp-analyze.h
3071 (class sprintf_dom_walker): Add after_dom_children member function.
3072 Add evrp_range_analyzer member.
3073 (sprintf_dom_walker::before_dom_children): Call into the EVRP
3074 range analyzer as needed.
3075 (sprintf_dom_walker::after_dom_children): New member function.
3076 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
3078 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
3079 (evrp_range_analyzer::pop_to_marker): Likewise.
3081 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3083 PR tree-optimization/84419
3084 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
3085 with the required type if its current type is compatible but
3088 2018-02-20 Jakub Jelinek <jakub@redhat.com>
3091 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
3092 after vectorization.
3094 2018-02-20 Martin Liska <mliska@suse.cz>
3097 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
3098 possible values if we don't have a hint.
3100 2018-02-20 Martin Liska <mliska@suse.cz>
3104 * final.c (shorten_branches): Build align_tab array with one
3106 * opts.c (finish_options): Add alignment option limit check.
3107 (MAX_CODE_ALIGN): Likewise.
3108 (MAX_CODE_ALIGN_VALUE): Likewise.
3109 * doc/invoke.texi: Document maximum allowed option value for
3110 all -falign-* options.
3112 2018-02-19 Jakub Jelinek <jakub@redhat.com>
3115 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
3116 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
3117 * var-tracking.c (emit_note_insn_var_location): Remove all references
3118 to NOTE_INSN_CALL_ARG_LOCATION.
3119 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
3120 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
3121 Use copy_rtx_if_shared.
3122 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
3123 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
3124 (dwarf2out_var_location): Remove handling of
3125 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
3127 * final.c (final_scan_insn): Remove all references to
3128 NOTE_INSN_CALL_ARG_LOCATION.
3129 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
3130 before dumping final insns.
3131 * except.c (emit_note_eh_region_end): Remove all references to
3132 NOTE_INSN_CALL_ARG_LOCATION.
3133 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
3134 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
3135 * config/arc/arc.c (hwloop_optimize): Likewise.
3136 * config/arm/arm.c (create_fix_barrier): Likewise.
3137 * config/s390/s390.c (s390_chunkify_start): Likewise.
3138 * config/sh/sh.c (find_barrier): Likewise.
3139 * config/i386/i386.c (rest_of_insert_endbranch,
3140 ix86_seh_fixup_eh_fallthru): Likewise.
3141 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
3142 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
3143 * config/frv/frv.c (frv_function_prologue): Likewise.
3144 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
3146 (note_outside_basic_block_p): Remove all references to
3147 NOTE_INSN_CALL_ARG_LOCATION.
3148 * gengtype.c (adjust_field_rtx_def): Likewise.
3149 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
3151 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
3152 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
3155 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
3158 PR tree-optimization/84452
3159 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
3160 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
3163 2018-02-19 Martin Liska <mliska@suse.cz>
3166 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
3168 2018-02-19 Martin Liska <mliska@suse.cz>
3169 Richard Sandiford <richard.sandiford@linaro.org>
3171 PR tree-optimization/82491
3172 * gimple-fold.c (get_base_constructor): Make earlier bail out
3175 2018-02-19 Carl Love <cel@us.ibm.com>
3177 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
3178 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
3180 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
3181 P8V_BUILTIN_VEC_NEG.
3183 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
3185 * config/rl78/rl78.md (movdf): New define expand.
3187 2018-02-19 Martin Liska <mliska@suse.cz>
3190 * doc/invoke.texi: Fix typo.
3191 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
3193 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
3195 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
3196 handle rs6000_single_float and rs6000_double_float specially for
3199 2018-02-16 Jeff Law <law@redhat.com>
3201 * config/rx/rx.c (add_pop_cfi_notes): New function.;
3204 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3207 * ipa-inline.c (inline_small_functions): Fix a typo.
3209 2018-02-16 Nathan Sidwell <nathan@acm.org>
3211 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
3213 2018-02-16 Carl Love <cel@us.ibm.com>
3215 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
3216 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
3217 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
3218 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
3219 expansion to P8V_BUILTIN_VEC_FLOAT2.
3221 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
3223 PR rtl-optimization/70023
3224 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
3225 src_regno into account.
3227 2018-02-16 Carl Love <cel@us.ibm.com>
3229 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
3230 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
3231 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
3232 * config/rs6000/rs6000.c: Remove case statements for
3233 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
3234 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
3235 and P9V_BUILTIN_VEC_VINSERT4B.
3236 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
3237 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
3238 * config/rs6000/vsx.md:
3239 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
3242 2018-02-16 Carl Love <cel@us.ibm.com>
3244 * config/rs6000/altivec.h: Add builtin names vec_extract4b
3246 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
3248 * config/rs6000/rs6000-c.c: Add the definitions for
3249 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
3250 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
3251 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
3252 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
3253 definition for insert4b and define insn *insert3b_internal.
3254 * doc/extend.texi: Add documentation for vec_extract4b.
3256 2018-02-16 Nathan Sidwell <nathan@acm.org>
3258 * doc/extend.texi (Backwards Compatibility): Mention friend
3259 injection. Note for-scope is deprecated.
3260 * doc/invoke.texi (-ffriend-injection): Deprecate.
3262 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
3264 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
3265 that moved to I2, also allow destinations that are a paradoxical
3266 subreg (instead of a normal reg).
3268 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
3271 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
3274 2018-02-16 Richard Biener <rguenther@suse.de>
3276 PR tree-optimization/84037
3277 PR tree-optimization/84016
3279 * config/i386/i386.c (ix86_builtin_vectorization_cost):
3280 Adjust vec_construct for the fact we need additional higher latency
3281 128bit inserts for AVX256 and AVX512 vector builds.
3282 (ix86_add_stmt_cost): Scale vector construction cost for
3285 2018-02-16 Richard Biener <rguenther@suse.de>
3287 PR tree-optimization/84417
3288 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
3289 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
3290 (non_rewritable_lvalue_p): Likewise, use poly-ints.
3292 2018-02-16 Martin Liska <mliska@suse.cz>
3295 * internal-fn.def (ASAN_CHECK): Set proper flags.
3296 (ASAN_MARK): Likewise.
3298 2018-02-16 Julia Koval <julia.koval@intel.com>
3300 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
3301 from PTA_CANNONLAKE.
3303 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3306 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
3307 Use ++iter rather than iter++ for std::list iterators.
3308 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
3309 defer deleting them until all nodes in the forest are processed. Do
3310 free even leaf nodes. Change to_process into auto_vec.
3313 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
3314 * vec.h (vec_default_construct): Use memset instead of placement new
3315 if BROKEN_VALUE_INITIALIZATION is defined.
3316 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
3317 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
3320 PR rtl-optimization/83723
3321 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
3322 * lra.c (lra_substitute_pseudo): Likewise. If true, use
3323 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
3325 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
3327 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
3329 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
3331 PR rtl-optimization/81443
3332 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
3333 from inner REGs to paradoxical SUBREGs.
3335 2018-02-16 Richard Biener <rguenther@suse.de>
3337 PR tree-optimization/84399
3338 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3339 For operands we can analyze at their definition make sure we can
3340 analyze them at each use as well.
3342 2018-02-16 Richard Biener <rguenther@suse.de>
3344 PR tree-optimization/84190
3345 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
3346 volatile accesses if the decl isn't volatile.
3348 2018-02-15 Jason Merrill <jason@redhat.com>
3350 PR c++/84314 - ICE with templates and fastcall attribute.
3351 * attribs.c (build_type_attribute_qual_variant): Don't clobber
3352 TYPE_CANONICAL on an existing type.
3354 2018-02-15 Jakub Jelinek <jakub@redhat.com>
3356 PR tree-optimization/84383
3357 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
3358 dstoff nor call operand_equal_p if dstbase is NULL.
3360 PR tree-optimization/84334
3361 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
3362 also a CONSTANT_CLASS_P, punt.
3364 2018-02-14 Jim Wilson <jimw@sifive.com>
3366 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
3367 first SMALL_OPERAND check. New local min_second_step. Move assert
3368 to where locals are set. Add TARGET_RVC support.
3369 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
3371 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
3373 * doc/invoke.texi: Correct -Wformat-overflow code sample.
3375 2018-02-14 Martin Sebor <msebor@redhat.com>
3377 PR tree-optimization/83698
3378 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
3379 arrays constrain the offset range to their bounds.
3380 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
3381 (builtin_access::overlap): Avoid setting the size of overlap if it's
3383 (maybe_diag_overlap): Also consider arrays when deciding what values
3384 of offsets to include in diagnostics.
3386 2018-02-14 Martin Sebor <msebor@redhat.com>
3389 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
3390 that correspond to the kind of a declaration.
3392 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
3395 * config/pa/pa.md: Load address of PIC label using the linkage table
3396 if the label is nonlocal.
3398 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
3400 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
3401 warning message if user requests -maltivec=be.
3402 * doc/invoke.texi: Document deprecation of -maltivec=be.
3404 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
3407 * config/rs6000/rs6000-c.c: Update definitions for
3408 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
3409 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
3411 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3414 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
3415 add _get_ssp intrinsics. Remove argument from
3416 __builtin_ia32_rdssp[d|q].
3417 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
3418 * config/i386/i386-builtin.def: Remove argument from
3419 __builtin_ia32_rdssp[d|q].
3420 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
3421 ix86_expand_special_args_builtin for _rdssp[d|q].
3422 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
3423 Clear register before usage.
3424 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
3425 Add documentation for new _get_ssp and _inc_ssp intrinsics.
3427 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
3429 PR tree-optimization/84357
3430 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
3431 operand 1 of an ARRAY_REF too.
3433 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
3436 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
3437 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3439 (set_of_reg): New struct.
3440 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
3441 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
3442 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3444 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
3445 Split into bitclr, bitset, bitinvert patterns if appropriate.
3446 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
3447 use rx_fuse_in_memory_bitop.
3448 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
3449 to named insn, correct maximum insn length.
3451 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
3454 * machmode.def: Define a complex mode for PARTIAL_INT.
3455 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
3457 * doc/rtl.texi: Document CSPImode.
3458 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
3460 (msp430_hard_regno_nregs_with_padding): Likewise.
3462 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
3465 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
3467 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
3469 PR rtl-optimization/84169
3470 * combine.c (try_combine): New variable split_i2i3. Set it to true if
3471 we generated a parallel as new i3 and we split that to new i2 and i3
3472 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
3473 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
3474 those to i2, not i1. Partially rewrite this scan code.
3476 2018-02-13 Jakub Jelinek <jakub@redhat.com>
3479 * stor-layout.c (place_field): For variable length fields, adjust
3480 offset_align afterwards not just based on the field's alignment,
3481 but also on the size.
3484 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
3485 of exps and logs in the use_exp2 case.
3487 2018-02-13 Jeff Law <law@redhat.com>
3489 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
3492 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
3495 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
3499 * common.opt (gas-loc-support, gas-locview-support): New.
3500 (ginline-points, ginternal-reset-location-views): New.
3501 * doc/invoke.texi: Document them. Use @itemx where intended.
3502 (gvariable-location-views): Adjust.
3503 * target.def (reset_location_view): New.
3504 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
3505 (TARGET_RESET_LOCATION_VIEW): New.
3506 * doc/tm.texi: Rebuilt.
3507 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
3508 (dwarf2out_default_as_locview_support): New.
3509 (output_asm_line_debug_info): Use option variables.
3510 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
3511 (output_loc_list): Likewise.
3512 (add_high_low_attributes): Check option variables.
3513 Don't output entry view attribute in strict mode.
3514 (gen_inlined_subroutine_die): Check option variables.
3515 (dwarf2out_inline_entry): Likewise.
3516 (init_sections_and_labels): Likewise.
3517 (dwarf2out_early_finish): Likewise.
3518 (maybe_reset_location_view): New, from...
3519 (dwarf2out_var_location): ... here. Call it.
3520 * debug.h (dwarf2out_default_as_loc_support): Declare.
3521 (dwarf2out_default_as_locview_support): Declare.
3522 * hooks.c (hook_int_rtx_insn_0): New.
3523 * hooks.h (hook_int_rtx_insn_0): Declare.
3524 * toplev.c (process_options): Take -gas-loc-support and
3525 -gas-locview-support from dwarf2out. Enable
3526 -gvariable-location-views by default only with locview
3527 assembler support. Enable -ginternal-reset-location-views by
3528 default only if the target defines the corresponding hook.
3529 Enable -ginline-points by default if location views are
3530 enabled; force it disabled if statement frontiers are
3532 * tree-inline.c (expand_call_inline): Check option variables.
3533 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
3535 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
3537 PR tree-optimization/84321
3538 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
3539 handling. Also check whether the anti-range contains any values
3540 that satisfy the mask; switch to a VR_RANGE if not.
3542 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
3545 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
3547 2018-02-13 Martin Jambor <mjambor@suse.cz>
3550 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
3551 of call statements, also set location of a load to a temporary.
3553 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
3555 * config/rl78/rl78.c (add_vector_labels): New function.
3556 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
3557 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
3558 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
3559 which checks that no arguments are passed.
3560 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
3561 * doc/extend.texi: Documentation for the new attribute.
3563 2018-02-13 Andreas Schwab <schwab@suse.de>
3565 * config/riscv/linux.h (CPP_SPEC): Define.
3567 2018-02-13 Jakub Jelinek <jakub@redhat.com>
3570 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
3571 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
3572 OPTION_MASK_ISA_AES as first argument to def_builtin_const
3573 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
3574 instead of OPTION_MASK_ISA_PCLMUL as first argument to
3575 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
3576 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
3577 temporarily for AES and PCLMUL builtins.
3579 PR tree-optimization/84339
3580 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
3581 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
3585 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
3586 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
3587 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
3589 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
3591 * omp-simd-clone.h: New file.
3592 * omp-simd-clone.c: Include omp-simd-clone.h.
3593 (expand_simd_clones): No longer static.
3594 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
3595 cgraph.h and omp-simd-clone.h.
3596 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
3597 (vect_recog_widen_shift_pattern): Formatting fix.
3598 (vect_pattern_recog_1): Don't check optab for calls.
3601 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
3602 operands[2] into a REG before using gen_lowpart on it.
3604 2018-02-12 Jeff Law <law@redhat.com>
3607 * config/sh/sh.c (find_barrier): Consider a sibling call
3610 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
3611 successfully back substituting a reg.
3613 2018-02-12 Richard Biener <rguenther@suse.de>
3615 PR tree-optimization/84037
3616 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
3617 parameter, move visited init to caller.
3618 (vect_slp_analyze_operations): Separate cost from validity
3619 check, initialize visited once for all instances.
3620 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
3622 * tree-vect-stmts.c (vect_model_simple_cost): Make early
3624 (vect_model_promotion_demotion_cost): Likewise.
3625 (vectorizable_bswap): Guard cost modeling with !slp_node
3626 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
3628 (vectorizable_call): Likewise.
3629 (vectorizable_conversion): Likewise.
3630 (vectorizable_assignment): Likewise.
3631 (vectorizable_shift): Likewise.
3632 (vectorizable_operation): Likewise.
3633 (vectorizable_store): Likewise.
3634 (vectorizable_load): Likewise.
3635 (vectorizable_condition): Likewise.
3636 (vectorizable_comparison): Likewise.
3638 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
3641 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
3642 (ASAN_MARK): Fix fnspec to account for return value, change pointer
3643 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
3645 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3648 * params.def (inline-min-speedup): Increase from 8 to 15.
3649 (max-inline-insns-auto): Decrease from 40 to 30.
3650 * ipa-split.c (consider_split): Add some buffer for function to
3651 be considered inlining candidate.
3652 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
3655 2018-02-12 Richard Biener <rguenther@suse.de>
3657 PR tree-optimization/84037
3658 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
3659 matched stmts if we cannot swap the non-matched ones.
3661 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3663 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
3664 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
3665 _mm_maskz_scalef_round_ss): New intrinsics.
3666 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
3667 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
3668 __builtin_ia32_scalefss_round): Remove.
3669 (__builtin_ia32_scalefsd_mask_round,
3670 __builtin_ia32_scalefss_mask_round): New intrinsics.
3671 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
3672 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
3673 ((match_operand:VF_128 2 "<round_nimm_predicate>"
3674 "<round_constraint>")): Changed to ...
3675 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
3676 "<round_scalar_constraint>")): ... this.
3677 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
3678 %0, %1, %2<round_op3>}"): Changed to ...
3679 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
3680 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
3681 %2<round_scalar_mask_op3>}"): ... this.
3682 * config/i386/subst.md (round_scalar_nimm_predicate): New.
3684 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3686 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
3687 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
3688 (_mm_maskz_sqrt_round_ss): New intrinsics.
3689 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
3690 (__builtin_ia32_sqrtsd_mask_round)
3691 (__builtin_ia32_sqrtss_mask_round): New builtins.
3692 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
3693 (__builtin_ia32_sqrtss_round): Remove.
3694 (__builtin_ia32_sqrtsd_mask_round)
3695 (__builtin_ia32_sqrtss_mask_round): New builtins.
3696 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
3697 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
3698 ((match_operand:VF_128 1 "vector_operand"
3699 "xBm,<round_constraint>")): Changed to ...
3700 ((match_operand:VF_128 1 "vector_operand"
3701 "xBm,<round_scalar_constraint>")): ... this.
3702 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
3703 %0, %2, %<iptr>1<round_op3>}): Changed to ...
3704 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
3705 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
3706 %<iptr>1<round_scalar_mask_op3>}): ... this.
3707 ((set_attr "prefix" "<round_prefix>")): Changed to ...
3708 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
3710 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
3713 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
3714 Cast vec_cmpeq result to correct type.
3715 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
3716 Cast vec_cmpgt result to correct type.
3718 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
3720 * final.c (final_scan_insn_1): Renamed from...
3721 (final_scan_insn): ... this. New wrapper, to recover
3722 seen from the outermost call in recursive ones.
3723 * config/sparc/sparc.c (output_return): Drop seen from call.
3724 (output_sibcall): Likewise.
3725 * config/visium/visium.c (output_branch): Likewise.
3727 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
3729 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
3732 2018-02-10 Alan Modra <amodra@gmail.com>
3735 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
3736 Specify LR as an input.
3738 2018-02-10 Jakub Jelinek <jakub@redhat.com>
3741 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
3742 remove_member_access_dummy_vars): New functions.
3743 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
3744 lower_omp_1, execute_lower_omp): Use them.
3746 PR rtl-optimization/84308
3747 * shrink-wrap.c (spread_components): Release todo vector.
3749 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
3751 PR rtl-optimization/57193
3752 * ira-color.c (struct allocno_color_data): Add member
3753 conflict_allocno_hard_prefs.
3754 (update_conflict_allocno_hard_prefs): New.
3755 (bucket_allocno_compare_func): Add a preference based on
3756 conflict_allocno_hard_prefs.
3757 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
3758 (color_allocnos): Remove a dead code. Initiate
3759 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
3761 2018-02-09 Jakub Jelinek <jakub@redhat.com>
3764 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
3765 constraint from =wa to wa. Avoid a subreg on the output operand,
3766 instead use a pseudo and subreg it in a move.
3767 (p9_xxbrd_<mode>): Changed to ...
3768 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
3769 (p9_xxbrd_v2df): New expander.
3770 (p9_xxbrw_<mode>): Changed to ...
3771 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
3772 (p9_xxbrw_v4sf): New expander.
3774 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
3776 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
3778 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
3781 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
3782 multiply in 32-bit mode.
3783 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
3784 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
3787 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
3789 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
3790 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
3791 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
3792 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
3794 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
3797 * dwarf2out.c (is_trivial_indirect_ref): New function.
3798 (dwarf2out_late_global_decl): Do not generate a location
3799 attribute for variables that have a non-trivial DECL_VALUE_EXPR
3800 and that are not defined in the current unit.
3802 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
3804 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
3805 instead of a libcall for UNORDERED.
3807 2018-02-09 Tamar Christina <tamar.christina@arm.com>
3810 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
3811 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
3813 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3816 * config/s390/s390.c (s390_set_current_function): Invoke
3817 s390_indirect_branch_settings also if fndecl didn't change.
3819 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
3821 * config/rs6000/rs6000.md (blockage): Set length to zero.
3823 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
3825 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
3827 2018-02-09 Jakub Jelinek <jakub@redhat.com>
3830 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
3831 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
3835 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
3836 PARALLEL incoming that failed vt_get_decl_and_offset check.
3839 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
3840 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
3842 (get_variable_section): For decls in named .bss* sections pass true as
3843 second argument to bss_initializer_p.
3845 2018-02-09 Marek Polacek <polacek@redhat.com>
3846 Jakub Jelinek <jakub@redhat.com>
3849 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
3850 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
3851 Sync some changes from cxx_fold_indirect_ref.
3853 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
3855 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
3857 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
3858 (BLOCK_INLINE_ENTRY_LABEL): New.
3859 (dwarf2out_var_location): Disregard inline entry markers.
3860 (inline_entry_data): New struct.
3861 (inline_entry_data_hasher): New hashtable type.
3862 (inline_entry_data_hasher::hash): New.
3863 (inline_entry_data_hasher::equal): New.
3864 (inline_entry_data_table): New variable.
3865 (add_high_low_attributes): Add DW_AT_entry_pc and
3866 DW_AT_GNU_entry_view attributes if a pending entry is found
3867 in inline_entry_data_table. Add old entry_pc attribute only
3868 if debug nonbinding markers are disabled.
3869 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
3870 markers are enabled.
3871 (block_within_block_p, dwarf2out_inline_entry): New.
3872 (dwarf2out_finish): Check that no entries remained in
3873 inline_entry_data_table.
3874 * final.c (reemit_insn_block_notes): Handle inline entry notes.
3875 (final_scan_insn, notice_source_line): Likewise.
3876 (rest_of_clean_state): Skip inline entry markers.
3877 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
3879 * gimple.c (gimple_build_debug_inline_entry): New.
3880 * gimple.h (enum gimple_debug_subcode): Add
3881 GIMPLE_DEBUG_INLINE_ENTRY.
3882 (gimple_build_debug_inline_entry): Declare.
3883 (gimple_debug_inline_entry_p): New.
3884 (gimple_debug_nonbind_marker_p): Adjust.
3885 * insn-notes.def (INLINE_ENTRY): New.
3886 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
3887 inline entry marker notes.
3888 (print_insn): Likewise.
3889 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
3890 (INSN_DEBUG_MARKER_KIND): Likewise.
3891 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
3892 * tree-inline.c (expand_call_inline): Build and insert
3893 debug_inline_entry stmt.
3894 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
3895 inline entry blocks early, if nonbind markers are enabled.
3896 (dump_scope_block): Dump fragment info.
3897 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
3898 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
3899 (gimple_build_debug_inline_entry): New.
3900 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
3901 Enable/disable inline entry points too.
3902 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
3903 (DEBUG_INSN): Describe inline entry markers.
3905 * common.opt (gvariable-location-views): New.
3906 (gvariable-location-views=incompat5): New.
3907 * config.in: Rebuilt.
3908 * configure: Rebuilt.
3909 * configure.ac: Test assembler for view support.
3910 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
3911 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
3912 * dwarf2out.c (var_loc_view): New typedef.
3913 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
3914 (dwarf2out_locviews_in_attribute): New.
3915 (dwarf2out_locviews_in_loclist): New.
3916 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
3917 (enum dw_line_info_opcode): Add LI_adv_address.
3918 (struct dw_line_info_table): Add view.
3919 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
3920 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
3921 (zero_view_p): New variable.
3922 (ZERO_VIEW_P): New macro.
3923 (output_asm_line_debug_info): New.
3924 (struct var_loc_node): Add view.
3925 (add_AT_view_list, AT_loc_list): New.
3926 (add_var_loc_to_decl): Add view param. Test it against last.
3927 (new_loc_list): Add view params. Record them.
3928 (AT_loc_list_ptr): Handle loc and view lists.
3929 (view_list_to_loc_list_val_node): New.
3930 (print_dw_val): Handle dw_val_class_view_list.
3931 (size_of_die): Likewise.
3932 (value_format): Likewise.
3933 (loc_list_has_views): New.
3934 (gen_llsym): Set vl_symbol too.
3935 (maybe_gen_llsym, skip_loc_list_entry): New.
3936 (dwarf2out_maybe_output_loclist_view_pair): New.
3937 (output_loc_list): Output view list or entries too.
3938 (output_view_list_offset): New.
3939 (output_die): Handle dw_val_class_view_list.
3940 (output_dwarf_version): New.
3941 (output_compilation_unit_header): Use it.
3942 (output_skeleton_debug_sections): Likewise.
3943 (output_rnglists, output_line_info): Likewise.
3944 (output_pubnames, output_aranges): Update version comments.
3945 (output_one_line_info_table): Output view numbers in asm comments.
3946 (dw_loc_list): Determine current endview, pass it to new_loc_list.
3947 Call maybe_gen_llsym.
3948 (loc_list_from_tree_1): Adjust.
3949 (add_AT_location_description): Create view list attribute if
3950 needed, check it's absent otherwise.
3951 (convert_cfa_to_fb_loc_list): Adjust.
3952 (maybe_emit_file): Call output_asm_line_debug_info for test.
3953 (dwarf2out_var_location): Reset views as needed. Precompute
3954 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
3955 attribute. Set view.
3956 (new_line_info_table): Reset next view.
3957 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
3958 (dwarf2out_source_line): Likewise. Output view resets and labels to
3959 the assembler, or select appropriate line info opcodes.
3960 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
3961 (optimize_string_length): Catch it. Adjust.
3962 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
3963 dw_val_class_view_list, and remove it if no longer needed.
3964 (hash_loc_list): Hash view numbers.
3965 (loc_list_hasher::equal): Compare them.
3966 (optimize_location_lists): Check whether a view list symbol is
3967 needed, and whether the locview attribute is present, and
3968 whether they match. Remove the locview attribute if no longer
3970 (index_location_lists): Call skip_loc_list_entry for test.
3971 (dwarf2out_finish): Call output_asm_line_debug_info for test.
3972 Use output_dwarf_version.
3973 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
3974 (struct dw_val_node): Add val_view_list.
3975 * final.c (SEEN_NEXT_VIEW): New.
3976 (set_next_view_needed): New.
3977 (clear_next_view_needed): New.
3978 (maybe_output_next_view): New.
3979 (final_start_function): Rename to...
3980 (final_start_function_1): ... this. Take pointer to FIRST,
3981 add SEEN parameter. Emit param bindings in the initial view.
3982 (final_start_function): Reintroduce SEEN-less interface.
3983 (final): Rename to...
3984 (final_1): ... this. Take SEEN parameter. Output final pending
3985 next view at the end.
3986 (final): Reintroduce seen-less interface.
3987 (final_scan_insn): Output pending next view before switching
3988 sections or ending a block. Mark the next view as needed when
3989 outputting variable locations. Notify debug backend of section
3990 changes, and of location view changes.
3991 (rest_of_handle_final): Adjust.
3992 * toplev.c (process_options): Autodetect value for debug variable
3993 location views option. Warn on incompat5 without -gdwarf-5.
3994 * doc/invoke.texi (gvariable-location-views): New.
3995 (gvariable-location-views=incompat5): New.
3996 (gno-variable-location-views): New.
3998 2018-02-08 David Malcolm <dmalcolm@redhat.com>
4000 PR tree-optimization/84136
4001 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
4002 that the result of find_edge is non-NULL.
4004 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4007 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
4008 storing integer register in SImode. Fix cost of 256 and 512
4009 byte aligned SSE register store.
4011 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4013 * config/i386/i386.c (ix86_multiplication_cost): Fix
4014 multiplication cost for TARGET_AVX512DQ.
4016 2018-02-08 Marek Polacek <polacek@redhat.com>
4018 PR tree-optimization/84238
4019 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
4022 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4024 PR tree-optimization/84265
4025 * tree-vect-stmts.c (vectorizable_store): Don't treat
4026 VMAT_CONTIGUOUS accesses as grouped.
4027 (vectorizable_load): Likewise.
4029 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4031 PR tree-optimization/81635
4032 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
4033 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
4034 (test_round_for_mask): New functions.
4035 (wide_int_cc_tests): Call test_round_for_mask.
4036 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
4037 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
4038 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
4039 range returned by get_range_info.
4041 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4044 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
4045 * symtab.c: Include builtins.h
4046 (symtab_node::output_to_lto_symbol_table_p): Move here
4047 from lto-streamer-out.c:output_symbol_p.
4048 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
4049 (output_symbol_p): Move all logic to symtab.c
4050 (produce_symtab): Update.
4052 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4054 * config/s390/s390-opts.h (enum indirect_branch): Define.
4055 * config/s390/s390-protos.h (s390_return_addr_from_memory)
4056 (s390_indirect_branch_via_thunk)
4057 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
4058 (enum s390_indirect_branch_type): Define.
4059 * config/s390/s390.c (struct s390_frame_layout, struct
4060 machine_function): Remove.
4061 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
4062 (indirect_branch_table_label_no, indirect_branch_table_name):
4064 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
4065 (enum s390_indirect_branch_option): Define.
4066 (s390_return_addr_from_memory): New function.
4067 (s390_handle_string_attribute): New function.
4068 (s390_attribute_table): Add new attribute handler.
4069 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
4070 (s390_indirect_branch_via_thunk): New function.
4071 (s390_indirect_branch_via_inline_thunk): New function.
4072 (s390_function_ok_for_sibcall): When jumping via thunk disallow
4073 sibling call optimization for non z10 compiles.
4074 (s390_emit_call): Force indirect branch target to be a single
4075 register. Add r1 clobber for non-z10 compiles.
4076 (s390_emit_epilogue): Emit return jump via return_use expander.
4077 (s390_reorg): Handle JUMP_INSNs as execute targets.
4078 (s390_option_override_internal): Perform validity checks for the
4079 new command line options.
4080 (s390_indirect_branch_attrvalue): New function.
4081 (s390_indirect_branch_settings): New function.
4082 (s390_set_current_function): Invoke s390_indirect_branch_settings.
4083 (s390_output_indirect_thunk_function): New function.
4084 (s390_code_end): Implement target hook.
4085 (s390_case_values_threshold): Implement target hook.
4086 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
4088 * config/s390/s390.h (struct s390_frame_layout)
4089 (struct machine_function): Move here from s390.c.
4090 (TARGET_INDIRECT_BRANCH_NOBP_RET)
4091 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
4092 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
4093 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
4094 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
4095 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
4096 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
4097 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
4098 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
4099 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
4100 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
4101 (mnemonic attribute): Add values which aren't recognized
4103 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
4104 pattern for branch conversion. Fix mnemonic attribute.
4105 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
4106 indirect branch via thunk if requested.
4107 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
4108 ("*indirect_jump"): Disable for branch conversion using out of
4110 ("indirect_jump_via_thunk<mode>_z10")
4111 ("indirect_jump_via_thunk<mode>")
4112 ("indirect_jump_via_inlinethunk<mode>_z10")
4113 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
4114 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
4115 ("casesi_jump_via_inlinethunk<mode>_z10")
4116 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
4117 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
4118 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
4119 ("*indirect2_jump"): Disable for branch conversion.
4120 ("casesi_jump"): Turn into expander and expand patterns for branch
4122 ("return_use"): New expander.
4123 ("*return"): Emit return via thunk and rename it to ...
4124 ("*return<mode>"): ... this one.
4125 * config/s390/s390.opt: Add new options and and enum for the
4128 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4130 * lra-constraints.c (match_reload): Unconditionally use
4131 gen_lowpart_SUBREG, rather than selecting between that
4132 and equivalent gen_rtx_SUBREG code.
4134 2018-02-08 Richard Biener <rguenther@suse.de>
4136 PR tree-optimization/84233
4137 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
4138 changed flag instead of boguously re-using phi_inserted.
4140 2018-02-08 Martin Jambor <mjambor@suse.cz>
4142 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
4143 static local variables.
4145 2018-02-08 Richard Biener <rguenther@suse.de>
4147 PR tree-optimization/84278
4148 * tree-vect-stmts.c (vectorizable_store): When looking for
4149 smaller vector types to perform grouped strided loads/stores
4150 make sure the mode is supported by the target.
4151 (vectorizable_load): Likewise.
4153 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4155 * config/aarch64/aarch64.c (aarch64_components_for_bb):
4156 Increase LDP/STP opportunities by adding adjacent callee-saves.
4158 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4160 PR rtl-optimization/84068
4161 PR rtl-optimization/83459
4162 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
4164 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
4166 PR tree-optimization/84224
4167 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
4168 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
4171 2018-02-07 Iain Sandoe <iain@codesourcery.com>
4174 * config/rs6000/altivec.md (*restore_world): Remove LR use.
4175 * config/rs6000/predicates.md (restore_world_operation): Adjust op
4176 count, remove one USE.
4178 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
4180 * doc/install.texi (Configuration): Document the
4181 --with-long-double-format={ibm,ieee} PowerPC configuration
4185 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
4186 Convert from define_expand to be define_insn_and_split. Rework
4187 float/double/_Float128 conversions to QI/HI/SImode to work with
4188 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
4189 conversions to QI/HImode types did a store and then a load to
4190 truncate the value. For conversions to VSX registers, don't split
4191 the insn, instead emit the code directly. Use the code iterator
4192 any_fix to combine signed and unsigned conversions.
4193 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
4194 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
4195 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4196 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
4197 (fix_<mode>di2_hw): Likewise.
4198 (fixuns_<mode>di2_hw): Likewise.
4199 (fix_<mode>si2_hw): Likewise.
4200 (fixuns_<mode>si2_hw): Likewise.
4201 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
4202 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4203 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
4204 fix<uns>_trunc<SFDF:mode>si2_p8.
4205 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
4207 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
4208 (fix<uns>_<mode>_mem): Likewise.
4209 (fctiw<u>z_<mode>_mem): Likewise.
4210 (fix<uns>_<mode>_mem): Likewise.
4211 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
4212 the register allocator from doing a direct move to the GPRs to do
4213 a store, and instead use the ISA 3.0 store byte/half-word from
4214 vector register instruction. For IEEE 128-bit floating point,
4215 also optimize stores of 32-bit ints.
4216 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
4218 2018-02-07 Alan Hayward <alan.hayward@arm.com>
4220 * genextract.c (push_pathstr_operand): New function to support
4222 (walk_rtx): Call push_pathstr_operand.
4223 (print_path): Support [a-zA-Z].
4225 2018-02-07 Richard Biener <rguenther@suse.de>
4227 PR tree-optimization/84037
4228 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
4229 (cse_and_gimplify_to_preheader): Declare.
4230 (vect_get_place_in_interleaving_chain): Likewise.
4231 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4233 (_loop_vec_info::~_loop_vec_info): Delete it.
4234 (cse_and_gimplify_to_preheader): New function.
4235 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
4236 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
4237 (vectorizable_load): Likewise. For grouped stores always base
4238 the IV on the first element.
4239 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
4240 condition before gimplifying.
4242 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4244 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
4245 *DIV_EXPR and *MOD_EXPR.
4247 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
4250 * config/i386/i386.c (ix86_option_override_internal): Mask out
4251 the CF_SET bit when checking -fcf-protection.
4253 2018-02-07 Tom de Vries <tom@codesourcery.com>
4256 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
4259 2018-02-07 Richard Biener <rguenther@suse.de>
4261 PR tree-optimization/84204
4262 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
4265 PR tree-optimization/84205
4266 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
4267 special-case isl_ast_op_zdiv_r.
4269 PR tree-optimization/84223
4270 * graphite-scop-detection.c (gather_bbs::before_dom_children):
4271 Only add conditions from within the region.
4272 (gather_bbs::after_dom_children): Adjust.
4274 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
4277 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
4278 * config/avr/avr.md: Only post-reload split REG-REG moves if
4279 either register is GENERAL_REG_P.
4281 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4283 PR tree-optimization/84235
4284 * tree-ssa-scopedtables.c
4285 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
4286 if the subtraction is performed in floating point type where NaNs are
4287 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
4288 build 1. Formatting fix.
4290 2018-02-06 Jakub Jelinek <jakub@redhat.com>
4293 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
4294 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
4295 and skip it regardless of bb boundaries. Use CALL_P macro,
4296 don't test INSN_P (insn) together with CALL_P or JUMP_P check
4297 unnecessarily, formatting fix.
4299 2018-02-06 Michael Collison <michael.collison@arm.com>
4301 * config/arm/thumb2.md:
4302 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
4303 (*thumb_mov_notscc): Ditto.
4305 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
4308 * config/rs6000/rs6000.md (su code attribute): Use "u" for
4309 unsigned_fix, not "s".
4311 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4313 * configure.ac (gcc_fn_eh_frame_ro): New function.
4314 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
4315 correct .eh_frame permissions.
4316 * configure: Regenerate.
4318 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
4320 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
4323 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4325 * config/rs6000/rs6000.c (rs6000_option_override_internal):
4326 Display warning message for -mno-speculate-indirect-jumps.
4328 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
4330 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
4332 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
4334 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
4336 PR tree-optimization/84225
4337 * tree-eh.c (find_trapping_overflow): Only call
4338 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
4340 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4343 * config/i386/i386.c: Reimplement the check of possible options
4344 -mibt/-mshstk conbination. Change error messages.
4345 * doc/invoke.texi: Fix a typo: remove extra '='.
4347 2018-02-06 Marek Polacek <polacek@redhat.com>
4349 PR tree-optimization/84228
4350 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
4352 2018-02-06 Tamar Christina <tamar.christina@arm.com>
4355 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
4356 emitted arch directives.
4357 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
4358 __ARM_FEATURE_COPROC before changing architectures.
4360 2018-02-06 Richard Biener <rguenther@suse.de>
4362 * config/i386/i386.c (print_reg): Fix typo.
4363 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
4365 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
4367 * configure: Regenerate.
4369 2018-02-05 Martin Sebor <msebor@redhat.com>
4371 PR tree-optimization/83369
4372 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
4375 2018-02-05 Martin Liska <mliska@suse.cz>
4377 * doc/invoke.texi: Cherry-pick upstream r323995.
4379 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
4381 * ira.c (ira_init_register_move_cost): Adjust comment.
4383 2018-02-05 Martin Liska <mliska@suse.cz>
4385 PR gcov-profile/84137
4386 * doc/gcov.texi: Fix typo in documentation.
4388 2018-02-05 Martin Liska <mliska@suse.cz>
4390 PR gcov-profile/83879
4391 * doc/gcov.texi: Document necessity of --dynamic-list-data when
4392 using dlopen functionality.
4394 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
4396 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
4397 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
4398 _mm_maskz_range_ss, _mm_mask_range_round_ss,
4399 _mm_maskz_range_round_ss): New intrinsics.
4400 (__builtin_ia32_rangesd128_round)
4401 (__builtin_ia32_rangess128_round): Remove.
4402 (__builtin_ia32_rangesd128_mask_round,
4403 __builtin_ia32_rangess128_mask_round): New builtins.
4404 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
4405 __builtin_ia32_rangess128_round): Remove.
4406 (__builtin_ia32_rangesd128_mask_round,
4407 __builtin_ia32_rangess128_mask_round): New builtins.
4408 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
4409 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
4410 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
4411 "<round_saeonly_constraint>")): Changed to ...
4412 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
4413 "<round_saeonly_scalar_constraint>")): ... this.
4414 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
4415 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
4416 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
4417 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
4418 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
4420 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
4422 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
4424 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
4425 Remove all values except native, 8540 and 8548.
4427 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
4429 * config/i386/i386.c (ix86_output_function_return): Pass
4430 INVALID_REGNUM, instead of -1, as invalid register number to
4431 indirect_thunk_name and output_indirect_thunk.
4433 2018-02-02 Julia Koval <julia.koval@intel.com>
4435 * config.gcc: Add -march=icelake.
4436 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
4437 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
4438 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
4439 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
4440 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
4441 (processor_target_table): Add icelake.
4442 (ix86_option_override_internal): Handle new PTAs.
4443 (get_builtin_code_for_version): Handle icelake.
4444 (M_INTEL_COREI7_ICELAKE): New.
4445 (fold_builtin_cpu): Handle icelake.
4446 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
4447 * doc/invoke.texi: Add -march=icelake.
4449 2018-02-02 Julia Koval <julia.koval@intel.com>
4451 * config/i386/i386.c (ix86_option_override_internal): Change flags type
4452 to wide_int_bitmask.
4453 * wide-int-bitmask.h: New.
4455 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4458 * config/i386/i386.md: Replace Pmode with word_mode in
4459 builtin_setjmp_setup and builtin_longjmp to support x32.
4461 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
4465 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
4467 (rs6000_supported_cpu_names): New static variable.
4468 (linux_cpu_translation_table): Likewise.
4469 (elf_platform) <cpu>: Define new static variable and use it.
4470 Translate kernel AT_PLATFORM name to canonical name if needed.
4471 Error if platform name is unknown.
4473 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
4476 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
4478 2018-02-01 Jeff Law <law@redhat.com>
4481 * config/i386/i386.c (release_scratch_register_on_entry): Add new
4482 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
4483 the scratch if RELEASE_VIA_POP is false.
4484 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
4485 If we have to save a temporary register, decrement SIZE appropriately.
4486 Pass new arguments to release_scratch_register_on_entry.
4487 (ix86_adjust_stack_and_probe): Likewise.
4488 (ix86_emit_probe_stack_range): Pass new arguments to
4489 release_scratch_register_on_entry.
4491 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
4493 PR rtl-optimization/84157
4494 * combine.c (change_zero_ext): Use REG_P predicate in
4495 front of HARD_REGISTER_P predicate.
4497 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
4499 * config/avr/avr.c (avr_option_override): Move disabling of
4500 -fdelete-null-pointer-checks to...
4501 * common/config/avr/avr-common.c (avr_option_optimization_table):
4504 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4506 PR tree-optimization/81635
4507 * tree-data-ref.c (split_constant_offset_1): For types that
4508 wrap on overflow, try to use range info to prove that wrapping
4511 2018-02-01 Renlin Li <renlin.li@arm.com>
4514 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
4516 (aarch64_register_move_cost): Likewise.
4517 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
4519 (REG_CLASS_NAMES): Likewise.
4520 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
4521 TAILCALL_ADDR_REGS. Remove IP registers.
4522 * config/aarch64/aarch64.md (Ucs): Update register constraint.
4524 2018-02-01 Richard Biener <rguenther@suse.de>
4526 * domwalk.h (dom_walker::dom_walker): Add additional constructor
4527 for specifying RPO order and allow NULL for that.
4528 * domwalk.c (dom_walker::dom_walker): Likewise.
4529 (dom_walker::walk): Handle NULL RPO order.
4530 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
4532 (rewrite_update_dom_walker): Likewise.
4533 (mark_def_dom_walker): Likewise.
4535 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4537 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
4538 (aarch64_maybe_expand_sve_subreg_move): Declare.
4539 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
4540 * config/aarch64/predicates.md (aarch64_any_register_operand): New
4542 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
4543 that are semantically a reverse operation.
4544 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
4545 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
4546 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
4548 (aarch64_can_change_mode_class): For big-endian, forbid changes
4549 between two SVE modes if they have different element sizes.
4551 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4553 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
4554 the TImode handling for big-endian targets.
4556 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4558 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
4559 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
4561 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
4562 Remove BSWAP handing for big-endian targets and use the form of
4563 LD1RQ appropariate for the mode.
4565 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4567 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
4568 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
4571 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4574 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
4575 check for operands that need to go through aarch64_sve_reload_be.
4577 2018-02-01 Jakub Jelinek <jakub@redhat.com>
4579 PR tree-optimization/81661
4580 PR tree-optimization/84117
4581 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
4582 * tree-eh.c: Include gimplify.h.
4583 (find_trapping_overflow, replace_trapping_overflow,
4584 rewrite_to_non_trapping_overflow): New functions.
4585 * tree-vect-loop.c: Include tree-eh.h.
4586 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
4587 * tree-data-ref.c: Include tree-eh.h.
4588 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
4590 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
4592 PR rtl-optimization/84123
4593 * combine.c (change_zero_ext): Check if hard register satisfies
4594 can_change_dest_mode before calling gen_lowpart_SUBREG.
4596 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
4599 * ira.c (ira_init_register_move_cost): Remove assert.
4601 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4603 PR rtl-optimization/84071
4604 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
4605 * doc/tm.texi: Regenerate.
4607 2018-01-31 Richard Biener <rguenther@suse.de>
4609 PR tree-optimization/84132
4610 * tree-data-ref.c (analyze_miv_subscript): Properly
4611 check whether evolution_function_is_affine_multivariate_p
4612 before calling gcd_of_steps_may_divide_p.
4614 2018-01-31 Julia Koval <julia.koval@intel.com>
4617 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
4618 * config/i386/i386.md (rdpid_rex64) New.
4619 (rdpid): Make 32bit only.
4621 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
4624 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
4625 an IDENTIFIER_NODE for FUNCTION_TYPE's.
4627 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4630 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4632 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
4634 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4636 PR rtl-optimization/84071
4637 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
4638 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
4640 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4642 * config/arc/arc.c (arc_handle_aux_attribute): New function.
4643 (arc_attribute_table): Add 'aux' attribute.
4644 (arc_in_small_data_p): Consider aux like variables.
4645 (arc_is_aux_reg_p): New function.
4646 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
4647 (arc_get_aux_arg): New function.
4648 (prepare_move_operands): Handle aux-register access.
4649 (arc_handle_aux_attribute): New function.
4650 * doc/extend.texi (ARC Variable attributes): Add subsection.
4652 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4654 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
4655 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
4656 (arc_attribute_table): Add 'uncached' attribute.
4657 (arc_print_operand): Print '.di' flag for uncached memory
4659 (arc_in_small_data_p): Do not consider for small data the uncached
4661 (arc_is_uncached_mem_p): New function.
4662 * config/arc/predicates.md (compact_store_memory_operand): Check
4663 for uncached memory accesses.
4664 (nonvol_nonimm_operand): Likewise.
4665 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
4667 2018-01-31 Jakub Jelinek <jakub@redhat.com>
4670 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
4671 falign-loops=): Add Optimization flag.
4673 2018-01-30 Jeff Law <law@redhat.com>
4676 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
4677 INT_REGISTERS_SAVED. Check it prior to calling
4678 get_scratch_register_on_entry.
4679 (ix86_adjust_stack_and_probe): Similarly.
4680 (ix86_emit_probe_stack_range): Similarly.
4681 (ix86_expand_prologue): Corresponding changes.
4683 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4686 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
4687 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
4689 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
4692 * lra-constraints.c (curr_insn_transform): Process AND in the
4695 2018-01-30 Jakub Jelinek <jakub@redhat.com>
4697 PR rtl-optimization/83986
4698 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
4699 dependence against last_pending_memory_flush in addition to
4702 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
4704 PR tree-optimization/81611
4705 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
4708 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4711 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
4714 2018-01-30 Richard Biener <rguenther@suse.de>
4715 Jakub Jelinek <jakub@redhat.com>
4717 PR tree-optimization/84111
4718 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
4719 inner loops added during recursion, as they don't have up-to-date
4722 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4725 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
4726 (can_inline_edge_by_limits_p): ... here.
4727 (can_early_inline_edge_p, check_callers,
4728 update_caller_keys, update_callee_keys, recursive_inlining,
4729 add_new_edges_to_heap, speculation_useful_p,
4730 inline_small_functions,
4731 inline_small_functions, flatten_function,
4732 inline_to_all_callers_1): Update.
4734 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4736 * profile-count.c (profile_count::combine_with_ipa_count): Handle
4739 2018-01-30 Richard Biener <rguenther@suse.de>
4741 PR tree-optimization/83008
4742 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
4743 invariant and constant vector uses in stmts when they need
4746 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4749 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
4750 * configure: Regenerate.
4752 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4754 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
4756 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
4757 Use gen_rtx_REG rather than gen_lowpart.
4759 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4761 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
4762 rather than 0 when creating partial subregs.
4764 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4766 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
4769 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
4772 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
4773 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
4774 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
4775 flags. This restores the settings used before the 2017-07-24.
4776 Turning off pre increment/decrement/modify allows IVOPTS to
4777 optimize DF/SF loops where the index is an int.
4779 2018-01-29 Richard Biener <rguenther@suse.de>
4780 Kelvin Nilsen <kelvin@gcc.gnu.org>
4783 * tree-vect-stmts.c (vectorizable_call): Don't call
4784 targetm.vectorize_builtin_md_vectorized_function if callee is
4787 2018-01-22 Carl Love <cel@us.ibm.com>
4789 * doc/extend.tex: Fix typo in second arg in
4790 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
4792 2018-01-29 Richard Biener <rguenther@suse.de>
4794 PR tree-optimization/84086
4795 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
4796 (flush_ssaname_freelist): When SSA names were released reset
4797 the SCEV hash table.
4799 2018-01-29 Richard Biener <rguenther@suse.de>
4801 PR tree-optimization/84057
4802 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
4803 removed paths when removing edges.
4805 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
4807 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
4808 -mfunction-return=@var{choice}.
4810 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
4813 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
4815 (layout::print_source_line): Likewise.
4816 (test_get_line_width_without_trailing_whitespace): Add test cases.
4818 2018-01-27 Jakub Jelinek <jakub@redhat.com>
4821 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
4824 2018-01-26 Jim Wilson <jimw@sifive.com>
4826 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
4828 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
4831 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4833 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
4834 and CMP + SUB-immediate -> SUBS.
4836 2018-01-26 Martin Sebor <msebor@redhat.com>
4838 PR tree-optimization/83896
4839 * tree-ssa-strlen.c (get_string_len): Rename...
4840 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
4841 Avoid assuming length is constant.
4842 (handle_char_store): Use HOST_WIDE_INT for string length.
4844 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
4847 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
4848 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
4850 2018-01-26 Richard Biener <rguenther@suse.de>
4852 PR rtl-optimization/84003
4853 * dse.c (record_store): Only record redundant stores when
4854 the earlier store aliases at least all accesses the later one does.
4856 2018-01-26 Jakub Jelinek <jakub@redhat.com>
4858 PR rtl-optimization/83985
4859 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
4860 REG_CFA_RESTORE insns.
4861 (delete_unmarked_insns): Don't ignore separate shrink wrapping
4862 REG_CFA_RESTORE insns here.
4865 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
4866 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
4868 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4870 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
4871 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
4872 (arc_init): Likewise.
4873 (arc_override_options): Likewise.
4874 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
4876 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
4878 * config/arc/arc.h (TARGET_DBNZ): Define.
4879 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
4880 properly set the tune attribute.
4881 (dbnz): Use TARGET_DBNZ guard.
4882 * config/arc/arc.opt (mtune): Add core3 option.
4884 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4886 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
4887 recognize new pic like addresses.
4888 (arc_delegitimize_address): Clean up.
4890 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4892 * config/arc/arc-arches.def: Option mrf16 valid for all
4894 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
4895 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
4896 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
4897 * config/arc/arc-tables.opt: Regenerate.
4898 * config/arc/arc.c (arc_conditional_register_usage): Handle
4899 reduced register file case.
4900 (arc_file_start): Set must have build attributes.
4901 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
4903 * config/arc/arc.opt (mrf16): Add new option.
4904 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
4905 * config/arc/genmultilib.awk: Handle new mrf16 option.
4906 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
4907 * config/arc/t-multilib: Regenerate.
4908 * doc/invoke.texi (ARC Options): Document mrf16 option.
4910 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4912 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
4913 * config/arc/arc.c (arc_handle_secure_attribute): New function.
4914 (arc_attribute_table): Add 'secure_call' attribute.
4915 (arc_print_operand): Print secure call operand.
4916 (arc_function_ok_for_sibcall): Don't optimize tail calls when
4918 (arc_is_secure_call_p): New function. * config/arc/arc.md
4919 (call_i): Add support for sjli instruction.
4920 (call_value_i): Likewise.
4921 * config/arc/constraints.md (Csc): New constraint.
4923 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4924 John Eric Martin <John.Martin@emmicro-us.com>
4926 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
4927 * config/arc/arc.c (_arc_jli_section): New struct.
4928 (arc_jli_section): New type.
4929 (rc_jli_sections): New static variable.
4930 (arc_handle_jli_attribute): New function.
4931 (arc_attribute_table): Add jli_always and jli_fixed attribute.
4932 (arc_file_end): New function.
4933 (TARGET_ASM_FILE_END): Define.
4934 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
4935 (arc_add_jli_section): New function.
4936 (jli_call_scan): Likewise.
4937 (arc_reorg): Call jli_call_scan.
4938 (arc_output_addsi): Remove 'S' from printing asm operand.
4939 (arc_is_jli_call_p): New function.
4940 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
4942 (movhi_insn): Likewise.
4943 (movsi_insn): Likewise.
4944 (movsi_set_cc_insn): Likewise.
4945 (loadqi_update): Likewise.
4946 (load_zeroextendqisi_update): Likewise.
4947 (load_signextendqisi_update): Likewise.
4948 (loadhi_update): Likewise.
4949 (load_zeroextendhisi_update): Likewise.
4950 (load_signextendhisi_update): Likewise.
4951 (loadsi_update): Likewise.
4952 (loadsf_update): Likewise.
4953 (movsicc_insn): Likewise.
4954 (bset_insn): Likewise.
4955 (bxor_insn): Likewise.
4956 (bclr_insn): Likewise.
4957 (bmsk_insn): Likewise.
4958 (bicsi3_insn): Likewise.
4959 (cmpsi_cc_c_insn): Likewise.
4960 (movsi_ne): Likewise.
4961 (movsi_cond_exec): Likewise.
4962 (clrsbsi2): Likewise.
4973 (call_i): Remove 'S' asm letter, add jli instruction.
4974 (call_value_i): Likewise.
4975 * config/arc/arc.op (mjli-always): New option.
4976 * config/arc/constraints.md (Cji): New constraint.
4977 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
4979 (subsf3_fpx): Likewise.
4980 (mulsf3_fpx): Likewise.
4981 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
4983 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
4985 * doc/invoke.texi (ARC): Document mjli-always option.
4987 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
4989 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
4990 avoid addition with 0 and use incw and decw where possible.
4992 2018-01-26 Richard Biener <rguenther@suse.de>
4994 PR tree-optimization/81082
4995 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
4996 association if it requires casting to unsigned.
4997 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
4998 from fold_plusminus_mult_expr to catch important cases late when
4999 range info is available.
5001 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5003 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
5004 * configure.ac (hidden_linkonce): New test.
5005 * configure: Regenerate.
5006 * config.in: Regenerate.
5008 2018-01-26 Julia Koval <julia.koval@intel.com>
5010 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
5011 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
5012 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
5013 _mm_mask_bitshuffle_epi64_mask): Fix type.
5014 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
5015 USI_FTYPE_V4DI_V4DI_USI): Remove.
5016 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
5017 __builtin_ia32_vpshufbitqmb256_mask,
5018 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
5019 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
5020 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
5022 2018-01-26 Alan Modra <amodra@gmail.com>
5025 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
5026 UNSPEC_VBPERMQ. Sort other unspecs.
5028 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
5030 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
5032 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
5035 * predict.c (drop_profile): Do not push/pop cfun; update also
5037 (handle_missing_profiles): Fix logic looking for zero profiles.
5039 2018-01-25 Jakub Jelinek <jakub@redhat.com>
5042 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
5043 on functions with #pragma omp declare simd or functions with simd
5045 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
5046 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
5047 Remove trailing \n from warning_at calls.
5049 2018-01-25 Tom de Vries <tom@codesourcery.com>
5052 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5053 for neutered workers.
5055 2018-01-24 Joseph Myers <joseph@codesourcery.com>
5058 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
5059 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
5061 2018-01-24 Jeff Law <law@redhat.com>
5064 * i386.c (get_probe_interval): Move to earlier point.
5065 (ix86_compute_frame_layout): If -fstack-clash-protection and
5066 the frame is larger than the probe interval, then use pushes
5067 to save registers rather than reg->mem moves.
5068 (ix86_expand_prologue): Remove conditional for int_registers_saved
5071 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
5074 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
5075 min/max for never referenced object.
5077 2018-01-24 Jakub Jelinek <jakub@redhat.com>
5080 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
5082 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
5083 attributes from DECL_ATTRIBUTES (decl) without affecting
5084 DECL_ATTRIBUTES (current_function_decl).
5085 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
5086 functions with non-NULL DECL_ABSTRACT_ORIGIN.
5088 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
5090 PR tree-optimization/83979
5091 * fold-const.c (fold_comparison): Use constant_boolean_node
5092 instead of boolean_{true,false}_node.
5094 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
5096 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
5099 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5101 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5102 Simplify the clause that sets the length attribute.
5103 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5104 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
5105 clause that sets the length attribute.
5106 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5108 2018-01-24 Tom de Vries <tom@codesourcery.com>
5111 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
5112 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
5113 Add strict parameter.
5114 (prevent_branch_around_nothing): Insert dummy insn between branch to
5115 label and label with no ptx insn inbetween.
5116 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
5118 2018-01-24 Tom de Vries <tom@codesourcery.com>
5121 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5122 for neutered threads in warp.
5123 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
5125 2018-01-24 Richard Biener <rguenther@suse.de>
5127 PR tree-optimization/83176
5128 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
5131 2018-01-24 Richard Biener <rguenther@suse.de>
5133 PR tree-optimization/82819
5134 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
5135 code generating pluses that are no-ops in the target precision.
5137 2018-01-24 Richard Biener <rguenther@suse.de>
5140 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
5142 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5144 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
5145 to merge probabilities.
5146 * predict.c (probably_never_executed): Also mark as cold functions
5147 with global 0 profile and guessed local profile.
5148 * profile-count.c (profile_probability::combine_with_count): New
5150 * profile-count.h (profile_probability::operator*,
5151 profile_probability::operator*=, profile_probability::operator/,
5152 profile_probability::operator/=): Reduce precision to adjusted
5153 and set value to guessed on contradictory divisions.
5154 (profile_probability::combine_with_freq): Remove.
5155 (profile_probability::combine_wiht_count): Declare.
5156 (profile_count::force_nonzero):: Set to adjusted.
5157 (profile_count::probability_in):: Set quality to adjusted.
5158 * tree-ssa-tail-merge.c (replace_block_by): Use
5161 2018-01-23 Andrew Waterman <andrew@sifive.com>
5162 Jim Wilson <jimw@sifive.com>
5164 * config/riscv/riscv.c (riscv_stack_boundary): New.
5165 (riscv_option_override): Set riscv_stack_boundary. Handle
5166 riscv_preferred_stack_boundary_arg.
5167 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
5168 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
5169 (STACK_BOUNDARY): Set to riscv_stack_boundary.
5170 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
5171 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
5172 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
5174 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
5177 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
5178 of struct ix86_frame.
5179 (ix86_expand_epilogue): Likewise. Add a local variable for
5180 the reg_save_offset field in struct ix86_frame.
5182 2018-01-23 Bin Cheng <bin.cheng@arm.com>
5184 PR tree-optimization/82604
5185 * tree-loop-distribution.c (enum partition_kind): New enum item
5186 PKIND_PARTIAL_MEMSET.
5187 (partition_builtin_p): Support above new enum item.
5188 (generate_code_for_partition): Ditto.
5189 (compute_access_range): Differentiate cases that equality can be
5190 proven at all loops, the innermost loops or no loops.
5191 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
5192 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
5193 (finalize_partitions, distribute_loop): Don't fuse partition of
5194 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
5195 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
5198 2018-01-23 Martin Liska <mliska@suse.cz>
5200 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
5201 order to ignore the predictor.
5202 (PRED_POLYMORPHIC_CALL): Likewise.
5203 (PRED_RECURSIVE_CALL): Likewise.
5205 2018-01-23 Martin Liska <mliska@suse.cz>
5207 * tree-profile.c (tree_profiling): Print function header to
5208 aware reader which function we are working on.
5209 * value-prof.c (gimple_find_values_to_profile): Do not print
5210 not interesting value histograms.
5212 2018-01-23 Martin Liska <mliska@suse.cz>
5214 * profile-count.h (enum profile_quality): Add
5215 profile_uninitialized as the first value. Do not number values
5216 as they are zero based.
5217 (profile_count::verify): Update sanity check.
5218 (profile_probability::verify): Likewise.
5220 2018-01-23 Nathan Sidwell <nathan@acm.org>
5222 * doc/invoke.texi (ffor-scope): Deprecate.
5224 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5226 PR tree-optimization/83510
5227 * domwalk.c (set_all_edges_as_executable): New function.
5228 (dom_walker::dom_walker): Convert bool param
5229 "skip_unreachable_blocks" to enum reachability. Move setup of
5230 edge flags to set_all_edges_as_executable and only do it when
5231 reachability is REACHABLE_BLOCKS.
5232 * domwalk.h (enum dom_walker::reachability): New enum.
5233 (dom_walker::dom_walker): Convert bool param
5234 "skip_unreachable_blocks" to enum reachability.
5235 (set_all_edges_as_executable): New decl.
5236 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
5237 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
5239 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
5240 but converting true to REACHABLE_BLOCKS.
5241 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
5243 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
5244 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
5245 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
5247 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
5248 if check_all_array_refs will be called.
5250 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5252 * tree.c (selftest::test_location_wrappers): Add more test
5255 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5257 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
5258 (selftest::test_bit_in_range): Likewise.
5260 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5263 * doc/sourcebuild.texi (vect_float): Say that the selector
5264 only describes the situation when -funsafe-math-optimizations is on.
5265 (vect_float_strict): Document.
5267 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5269 PR tree-optimization/83965
5270 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
5271 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
5272 instead of checking only for a reduction.
5273 (vect_recog_widen_sum_pattern): Likewise.
5275 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5277 * predict.c (probably_never_executed): Only use precise profile info.
5278 (compute_function_frequency): Skip after inlining hack since we now
5279 have quality checking.
5281 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5283 * profile-count.h (profile_probability::very_unlikely,
5284 profile_probability::unlikely, profile_probability::even): Set
5285 precision to guessed.
5287 2018-01-23 Richard Biener <rguenther@suse.de>
5289 PR tree-optimization/83963
5290 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
5291 Properly terminate dominator walk when crossing the exit edge not
5292 when visiting its source block.
5294 2018-01-23 Jakub Jelinek <jakub@redhat.com>
5297 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
5298 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
5300 2018-01-22 Jakub Jelinek <jakub@redhat.com>
5302 PR tree-optimization/83957
5303 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
5304 semicolon after for body surrounded by braces.
5306 PR tree-optimization/83081
5307 * profile-count.h (profile_probability::split): New method.
5308 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
5309 Use profile_probability::split.
5310 (do_compare_rtx_and_jump): Fix adjustment of probabilities
5311 when splitting a single conditional jump into 2.
5313 2018-01-22 David Malcolm <dmalcolm@redhat.com>
5315 PR tree-optimization/69452
5316 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
5319 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5321 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
5322 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
5323 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
5325 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5327 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
5328 * config/rl78/rl78.md (movdi): New define_expand.
5329 * config/rl78/rl78.c (rl78_split_movdi): New function.
5331 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
5334 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
5336 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
5337 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
5338 128-bit to produce an UNSPEC move to get the double word with the
5339 signbit and then a shift directly to do signbit.
5340 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
5341 implementation with a new version that just does either a direct
5342 move or a regular move. Move memory interface to separate insns.
5343 Move insns so they are next to the expander.
5344 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
5345 with signbit move. Split big and little endian case.
5346 (signbit<mode>2_dm_mem_le): Likewise.
5347 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
5348 (signbit<mode>2_dm2): Likewise.
5350 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5352 * config/rl78/rl78.md (anddi3): New define_expand.
5354 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5356 * config/rl78/rl78.md (umindi3): New define_expand.
5358 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5360 * config/rl78/rl78.md (smindi3): New define_expand.
5362 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5364 * config/rl78/rl78.md (smaxdi3): New define_expand.
5366 2018-01-22 Carl Love <cel@us.ibm.com>
5368 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
5369 LVX_V1TI): Add macro expansion.
5370 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
5371 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
5372 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
5373 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
5374 Change check to determine if the instruction is a byte reversing
5375 entry. Fix typo in comment.
5376 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
5377 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
5378 Add def_builtin calls for new builtins.
5379 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
5380 Add define_insn expansion.
5382 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5384 * config/rl78/rl78.md (umaxdi3): New define_expand.
5386 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5388 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
5389 for non-QImode registers.
5391 2018-01-22 Richard Biener <rguenther@suse.de>
5393 PR tree-optimization/83963
5394 * graphite-scop-detection.c (scop_detection::get_sese): Delay
5395 including the loop exit block.
5396 (scop_detection::merge_sese): Likewise.
5397 (scop_detection::add_scop): Do it here instead.
5399 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5401 * doc/sourcebuild.texi (arm_softfloat): Document.
5403 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
5406 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
5407 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
5408 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
5410 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5411 David Edelsohn <dje.gcc@gmail.com>
5414 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5415 Change "crset eq" to "crset 2".
5416 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5417 (*call_indirect_aix<mode>_nospec): Likewise.
5418 (*call_value_indirect_aix<mode>_nospec): Likewise.
5419 (*call_indirect_elfv2<mode>_nospec): Likewise.
5420 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
5421 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
5422 change assembly output from . to $.
5423 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5424 (indirect_jump<mode>_nospec): Change assembly output from . to $.
5425 (*tablejump<mode>_internal1_nospec): Likewise.
5427 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
5430 * config/sh/sh_optimize_sett_clrt.cc:
5431 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
5433 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
5435 PR tree-optimization/83940
5436 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
5437 offset_dt to vect_constant_def rather than vect_unknown_def_type.
5438 (vect_check_load_store_mask): Add a mask_dt_out parameter and
5439 use it to pass back the definition type.
5440 (vect_check_store_rhs): Likewise rhs_dt_out.
5441 (vect_build_gather_load_calls): Add a mask_dt argument and use
5442 it instead of a call to vect_is_simple_use.
5443 (vectorizable_store): Update calls to vect_check_load_store_mask
5444 and vect_check_store_rhs. Use the dt returned by the latter instead
5445 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
5446 instead of calls to vect_is_simple_use. Pass the scalar rather
5447 than the vector operand to vect_is_simple_use when handling
5448 second and subsequent copies of an rhs value.
5449 (vectorizable_load): Update calls to vect_check_load_store_mask
5450 and vect_build_gather_load_calls. Use the cached mask_dt and
5451 gs_info.offset_dt instead of calls to vect_is_simple_use.
5453 2018-01-20 Jakub Jelinek <jakub@redhat.com>
5456 * tree-emutls.c: Include gimplify.h.
5457 (lower_emutls_2): New function.
5458 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
5459 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
5460 it before further processing.
5463 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
5464 UINTVAL (trueop1) instead of INTVAL (op1).
5466 2018-01-19 Jakub Jelinek <jakub@redhat.com>
5470 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
5471 INCOMING_FRAME_SP_OFFSET if not defined.
5472 (scan_trace): Add ENTRY argument. If true and
5473 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
5474 emit a note to adjust the CFA offset.
5475 (create_cfi_notes): Adjust scan_trace callers.
5476 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
5477 INCOMING_FRAME_SP_OFFSET in the CIE.
5478 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
5479 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
5481 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
5482 * doc/tm.texi: Regenerated.
5484 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5486 PR rtl-optimization/83147
5487 * lra-constraints.c (remove_inheritance_pseudos): Use
5488 lra_substitute_pseudo_within_insn.
5490 2018-01-19 Tom de Vries <tom@codesourcery.com>
5491 Cesar Philippidis <cesar@codesourcery.com>
5494 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
5496 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
5499 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
5500 spaces for function labels.
5502 2018-01-19 Martin Liska <mliska@suse.cz>
5504 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
5505 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
5506 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
5507 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
5508 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
5509 (PRED_CONST_RETURN): Change from 69 to 65.
5510 (PRED_NULL_RETURN): Change from 91 to 71.
5511 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
5512 (PRED_LOOP_GUARD): Change from 66 to 73.
5514 2018-01-19 Martin Liska <mliska@suse.cz>
5516 * predict.c (predict_insn_def): Add new assert.
5517 (struct branch_predictor): Change type to signed integer.
5518 (test_prediction_value_range): Amend test to cover
5520 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
5521 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
5522 (PRED_LOOP_ITERATIONS_MAX): Likewise.
5523 (PRED_LOOP_IV_COMPARE): Likewise.
5524 * predict.h (PROB_UNINITIALIZED): Define new constant.
5526 2018-01-19 Martin Liska <mliska@suse.cz>
5528 * predict.c (dump_prediction): Add new format for
5529 analyze_brprob.py script which is enabled with -details
5531 * profile-count.h (precise_p): New function.
5533 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
5535 PR tree-optimization/83922
5536 * tree-vect-loop.c (vect_verify_full_masking): Return false if
5537 there are no statements that need masking.
5538 (vect_active_double_reduction_p): New function.
5539 (vect_analyze_loop_operations): Use it when handling phis that
5540 are not in the loop header.
5542 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
5544 PR tree-optimization/83914
5545 * tree-vect-loop.c (vectorizable_induction): Don't convert
5546 init_expr or apply the peeling adjustment for inductions
5547 that are nested within the vectorized loop.
5549 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5551 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
5554 2018-01-18 Jakub Jelinek <jakub@redhat.com>
5558 * function.h (gimplify_parameters): Add gimple_seq * argument.
5559 * function.c: Include gimple.h and options.h.
5560 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
5561 for the added local temporaries if needed.
5562 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
5563 if there are any parameter cleanups, wrap whole body into a
5564 try/finally with the cleanups.
5566 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
5569 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
5570 Use GET_MODE_CLASS for scalar floating point.
5572 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
5576 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
5577 Fix call of call_cgraph_insertion_hooks.
5579 2018-01-18 Martin Sebor <msebor@redhat.com>
5581 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
5583 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
5586 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
5589 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
5592 * common.opt: (-ffile-prefix-map): New option.
5593 * opts.c (common_handle_option): Defer it.
5594 * opts-global.c (handle_common_deferred_options): Handle it.
5595 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
5596 * file-prefix-map.h: New file.
5597 (remap_debug_filename, add_debug_prefix_map): ...here.
5598 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
5599 * final.c (debug_prefix_map, add_debug_prefix_map
5600 remap_debug_filename): Move to...
5601 * file-prefix-map.c: New file.
5602 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
5603 generalize, get rid of alloca(), use strrchr() instead of strchr().
5604 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
5605 Implement in terms of add_prefix_map().
5606 (remap_macro_filename, remap_debug_filename): Implement in term of
5608 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
5609 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
5610 * dbxout.c: Include file-prefix-map.h.
5611 * varasm.c: Likewise.
5612 * vmsdbgout.c: Likewise.
5613 * xcoffout.c: Likewise.
5614 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
5615 * doc/cppopts.texi (-fmacro-prefix-map): Document.
5616 * doc/invoke.texi (-ffile-prefix-map): Document.
5617 (-fdebug-prefix-map): Update description.
5619 2018-01-18 Martin Liska <mliska@suse.cz>
5621 * config/i386/i386.c (indirect_thunk_name): Document that also
5623 (output_indirect_thunk): Document why both instructions
5624 (pause and lfence) are generated.
5626 2018-01-18 Richard Biener <rguenther@suse.de>
5628 PR tree-optimization/83887
5629 * graphite-scop-detection.c
5630 (scop_detection::get_nearest_dom_with_single_entry): Remove.
5631 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
5632 (scop_detection::merge_sese): Re-implement with a flood-fill
5633 algorithm that properly finds a SESE region if it exists.
5635 2018-01-18 Jakub Jelinek <jakub@redhat.com>
5638 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
5639 pointer_diff optimizations use view_convert instead of convert.
5641 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5643 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5644 Generate different code for -mno-speculate-indirect-jumps.
5645 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5646 (*call_indirect_aix<mode>): Disable for
5647 -mno-speculate-indirect-jumps.
5648 (*call_indirect_aix<mode>_nospec): New define_insn.
5649 (*call_value_indirect_aix<mode>): Disable for
5650 -mno-speculate-indirect-jumps.
5651 (*call_value_indirect_aix<mode>_nospec): New define_insn.
5652 (*sibcall_nonlocal_sysv<mode>): Generate different code for
5653 -mno-speculate-indirect-jumps.
5654 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5656 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
5658 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
5659 long double type, set the flags for noting the default long double
5660 type, even if we don't pass or return a long double type.
5662 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
5665 * ipa-inline.c (flatten_function): Do not overwrite final inlining
5668 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
5670 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
5671 support for merge[hl].
5672 (fold_mergehl_helper): New helper function.
5673 (tree-vector-builder.h): New #include for tree_vector_builder usage.
5674 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
5675 (altivec_vmrglw_direct): Add xxmrglw insn.
5677 2018-01-17 Andrew Waterman <andrew@sifive.com>
5679 * config/riscv/riscv.c (riscv_conditional_register_usage): If
5680 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
5682 2018-01-17 David Malcolm <dmalcolm@redhat.com>
5685 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
5686 call the lto_location_cache before reading the
5687 DECL_SOURCE_LOCATION of the types.
5689 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
5690 Richard Sandiford <richard.sandiford@linaro.org>
5692 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
5693 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
5694 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
5695 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
5696 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
5698 * config/aarch64/constraints.md (aarch64_movti_operand):
5700 * config/aarch64/predicates.md (Uti): Add new constraint.
5702 2018-01-17 Carl Love <cel@us.ibm.com>
5704 * config/rs6000/vsx.md (define_expand xl_len_r,
5705 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
5706 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
5708 (define_expand, define_insn): Move the shift left from the
5709 define_insn to the define_expand for lxvl and stxvl instructions.
5710 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
5711 and XL_LEN_R definitions to PURE.
5713 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
5715 * config/i386/i386.c (indirect_thunk_name): Declare regno
5716 as unsigned int. Compare regno with INVALID_REGNUM.
5717 (output_indirect_thunk): Ditto.
5718 (output_indirect_thunk_function): Ditto.
5719 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
5720 in the call to output_indirect_thunk_function.
5722 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
5725 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
5726 rather than the size of inner_type to determine the stack slot size
5727 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
5729 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
5732 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
5735 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
5737 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
5738 endian Linux systems to optionally enable multilibs for selecting
5739 the long double type if the user configured an explicit type.
5740 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
5741 have no long double multilibs if not defined.
5742 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
5743 warn if the user used -mabi={ieee,ibm}longdouble and we built
5744 multilibs for long double.
5745 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
5746 appropriate multilib option.
5747 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
5749 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
5750 for building long double multilibs.
5751 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
5753 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
5755 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
5758 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
5760 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
5763 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
5766 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
5769 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5771 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
5772 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
5774 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
5776 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
5777 different rtl trees depending on TARGET_64BIT.
5778 (rs6000_gen_lvx): Likewise.
5780 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5782 * config/visium/visium.md (nop): Tweak comment.
5783 (hazard_nop): Likewise.
5785 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5787 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
5788 -mspeculate-indirect-jumps.
5789 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
5790 for -mno-speculate-indirect-jumps.
5791 (*call_indirect_elfv2<mode>_nospec): New define_insn.
5792 (*call_value_indirect_elfv2<mode>): Disable for
5793 -mno-speculate-indirect-jumps.
5794 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
5795 (indirect_jump): Emit different RTL for
5796 -mno-speculate-indirect-jumps.
5797 (*indirect_jump<mode>): Disable for
5798 -mno-speculate-indirect-jumps.
5799 (*indirect_jump<mode>_nospec): New define_insn.
5800 (tablejump): Emit different RTL for
5801 -mno-speculate-indirect-jumps.
5802 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
5803 (tablejumpsi_nospec): New define_expand.
5804 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
5805 (tablejumpdi_nospec): New define_expand.
5806 (*tablejump<mode>_internal1): Disable for
5807 -mno-speculate-indirect-jumps.
5808 (*tablejump<mode>_internal1_nospec): New define_insn.
5809 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
5812 2018-01-16 Artyom Skrobov tyomitch@gmail.com
5814 * caller-save.c (insert_save): Drop unnecessary parameter. All
5817 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5818 Richard Biener <rguenth@suse.de>
5821 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
5822 return early, inline manually is_gimple_sizepos. Make sure if we
5823 call gimplify_expr we don't end up with a gimple constant.
5824 * tree.c (variably_modified_type_p): Don't return true for
5825 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
5826 * gimplify.h (is_gimple_sizepos): Remove.
5828 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5830 PR tree-optimization/83857
5831 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
5832 vectorizable_live_operation for pure SLP statements.
5833 (vectorizable_live_operation): Handle PHIs.
5835 2018-01-16 Richard Biener <rguenther@suse.de>
5837 PR tree-optimization/83867
5838 * tree-vect-stmts.c (vect_transform_stmt): Precompute
5839 nested_in_vect_loop_p since the scalar stmt may get invalidated.
5841 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5844 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
5845 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
5846 If off is not INTEGER_CST, issue a may not be aligned warning
5847 rather than isn't aligned. Use isn%'t rather than isn't.
5848 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
5850 <case MULT_EXPR>: Improve the case when bottom and one of the
5851 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
5852 operand, in that case check if the other operand is multiple of
5853 bottom divided by the INTEGER_CST operand.
5855 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5858 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
5859 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
5860 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
5861 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
5862 * config/pa/pa.c (pa_function_arg_advance): Likewise.
5863 (pa_function_arg, pa_arg_partial_bytes): Likewise.
5864 (pa_function_arg_size): New function.
5866 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5868 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
5869 in a separate statement.
5871 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5873 PR tree-optimization/83847
5874 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
5875 group gathers and scatters.
5877 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5879 PR rtl-optimization/86620
5880 * params.def (max-sched-ready-insns): Bump minimum value to 1.
5882 PR rtl-optimization/83213
5883 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
5884 to last if both are JUMP_INSNs.
5886 PR tree-optimization/83843
5887 * gimple-ssa-store-merging.c
5888 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
5889 store_immediate_info for bswap/nop orig_stores.
5891 2018-01-15 Andrew Waterman <andrew@sifive.com>
5893 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
5895 <UDIV>: Increase cost if !TARGET_DIV.
5897 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
5899 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
5900 (define_attr "cr_logical_3op"): New.
5901 (cceq_ior_compare): Adjust.
5902 (cceq_ior_compare_complement): Adjust.
5903 (*cceq_rev_compare): Adjust.
5904 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
5905 (is_cracked_insn): Adjust.
5906 (insn_must_be_first_in_group): Adjust.
5907 * config/rs6000/40x.md: Adjust.
5908 * config/rs6000/440.md: Adjust.
5909 * config/rs6000/476.md: Adjust.
5910 * config/rs6000/601.md: Adjust.
5911 * config/rs6000/603.md: Adjust.
5912 * config/rs6000/6xx.md: Adjust.
5913 * config/rs6000/7450.md: Adjust.
5914 * config/rs6000/7xx.md: Adjust.
5915 * config/rs6000/8540.md: Adjust.
5916 * config/rs6000/cell.md: Adjust.
5917 * config/rs6000/e300c2c3.md: Adjust.
5918 * config/rs6000/e500mc.md: Adjust.
5919 * config/rs6000/e500mc64.md: Adjust.
5920 * config/rs6000/e5500.md: Adjust.
5921 * config/rs6000/e6500.md: Adjust.
5922 * config/rs6000/mpc.md: Adjust.
5923 * config/rs6000/power4.md: Adjust.
5924 * config/rs6000/power5.md: Adjust.
5925 * config/rs6000/power6.md: Adjust.
5926 * config/rs6000/power7.md: Adjust.
5927 * config/rs6000/power8.md: Adjust.
5928 * config/rs6000/power9.md: Adjust.
5929 * config/rs6000/rs64.md: Adjust.
5930 * config/rs6000/titan.md: Adjust.
5932 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5934 * config/i386/predicates.md (indirect_branch_operand): Rewrite
5935 ix86_indirect_branch_register logic.
5937 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5939 * config/i386/constraints.md (Bs): Update
5940 ix86_indirect_branch_register check. Don't check
5941 ix86_indirect_branch_register with GOT_memory_operand.
5943 * config/i386/predicates.md (GOT_memory_operand): Don't check
5944 ix86_indirect_branch_register here.
5945 (GOT32_symbol_operand): Likewise.
5947 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5949 * config/i386/predicates.md (constant_call_address_operand):
5950 Rewrite ix86_indirect_branch_register logic.
5951 (sibcall_insn_operand): Likewise.
5953 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5955 * config/i386/constraints.md (Bs): Replace
5956 ix86_indirect_branch_thunk_register with
5957 ix86_indirect_branch_register.
5959 * config/i386/i386.md (indirect_jump): Likewise.
5960 (tablejump): Likewise.
5961 (*sibcall_memory): Likewise.
5962 (*sibcall_value_memory): Likewise.
5963 Peepholes of indirect call and jump via memory: Likewise.
5964 * config/i386/i386.opt: Likewise.
5965 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5966 (GOT_memory_operand): Likewise.
5967 (call_insn_operand): Likewise.
5968 (sibcall_insn_operand): Likewise.
5969 (GOT32_symbol_operand): Likewise.
5971 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5974 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
5975 type rather than type addr's type points to.
5976 (expand_omp_atomic_mutex): Likewise.
5977 (expand_omp_atomic): Likewise.
5979 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5982 * config/i386/i386.c (output_indirect_thunk_function): Use
5983 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
5984 for __x86_return_thunk.
5986 2018-01-15 Richard Biener <rguenther@suse.de>
5989 * expmed.c (extract_bit_field_1): Fix typo.
5991 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5994 * config/arm/iterators.md (VF): New mode iterator.
5995 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
5996 Remove integer-related logic from pattern.
5997 (neon_vabd<mode>_3): Likewise.
5999 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6002 * common.opt (fstrict-overflow): No longer an alias.
6003 (fwrapv-pointer): New option.
6004 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
6005 also for pointer types based on flag_wrapv_pointer.
6006 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
6007 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
6008 opts->x_flag_wrapv got set.
6009 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
6010 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
6011 POINTER_TYPE_OVERFLOW_UNDEFINED.
6012 * match.pd: Likewise in address comparison pattern.
6013 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
6015 2018-01-15 Richard Biener <rguenther@suse.de>
6018 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
6019 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
6020 Reset type names to their identifier if their TYPE_DECL doesn't
6021 have linkage (and thus is used for ODR and devirt).
6022 (save_debug_info_for_decl): Remove.
6023 (save_debug_info_for_type): Likewise.
6024 (add_tree_to_fld_list): Adjust.
6025 * tree-pretty-print.c (dump_generic_node): Make dumping of
6026 type names more robust.
6028 2018-01-15 Richard Biener <rguenther@suse.de>
6030 * BASE-VER: Bump to 8.0.1.
6032 2018-01-14 Martin Sebor <msebor@redhat.com>
6035 * builtins.c (check_access): Avoid warning when the no-warning bit
6038 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
6040 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
6041 * ira-color (allocno_hard_regs_compare): Likewise.
6043 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
6046 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
6047 Use .pushsection/.popsection.
6049 2018-01-14 Martin Sebor <msebor@redhat.com>
6052 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
6054 2018-01-14 Jakub Jelinek <jakub@redhat.com>
6056 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
6057 entry from extra_headers.
6058 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
6059 extra_headers, make the list bitwise identical to the i?86-*-* one.
6061 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6063 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
6064 -mcmodel=large with -mindirect-branch=thunk,
6065 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
6066 -mfunction-return=thunk-extern.
6067 * doc/invoke.texi: Document -mcmodel=large is incompatible with
6068 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
6069 -mfunction-return=thunk and -mfunction-return=thunk-extern.
6071 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6073 * config/i386/i386.c (print_reg): Print the name of the full
6074 integer register without '%'.
6075 (ix86_print_operand): Handle 'V'.
6076 * doc/extend.texi: Document 'V' modifier.
6078 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6080 * config/i386/constraints.md (Bs): Disallow memory operand for
6081 -mindirect-branch-register.
6083 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6084 (GOT_memory_operand): Likewise.
6085 (call_insn_operand): Likewise.
6086 (sibcall_insn_operand): Likewise.
6087 (GOT32_symbol_operand): Likewise.
6088 * config/i386/i386.md (indirect_jump): Call convert_memory_address
6089 for -mindirect-branch-register.
6090 (tablejump): Likewise.
6091 (*sibcall_memory): Likewise.
6092 (*sibcall_value_memory): Likewise.
6093 Disallow peepholes of indirect call and jump via memory for
6094 -mindirect-branch-register.
6095 (*call_pop): Replace m with Bw.
6096 (*call_value_pop): Likewise.
6097 (*sibcall_pop_memory): Replace m with Bs.
6098 * config/i386/i386.opt (mindirect-branch-register): New option.
6099 * doc/invoke.texi: Document -mindirect-branch-register option.
6101 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6103 * config/i386/i386-protos.h (ix86_output_function_return): New.
6104 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
6105 set function_return_type.
6106 (indirect_thunk_name): Add ret_p to indicate thunk for function
6108 (output_indirect_thunk_function): Pass false to
6109 indirect_thunk_name.
6110 (ix86_output_indirect_branch_via_reg): Likewise.
6111 (ix86_output_indirect_branch_via_push): Likewise.
6112 (output_indirect_thunk_function): Create alias for function
6113 return thunk if regno < 0.
6114 (ix86_output_function_return): New function.
6115 (ix86_handle_fndecl_attribute): Handle function_return.
6116 (ix86_attribute_table): Add function_return.
6117 * config/i386/i386.h (machine_function): Add
6118 function_return_type.
6119 * config/i386/i386.md (simple_return_internal): Use
6120 ix86_output_function_return.
6121 (simple_return_internal_long): Likewise.
6122 * config/i386/i386.opt (mfunction-return=): New option.
6123 (indirect_branch): Mention -mfunction-return=.
6124 * doc/extend.texi: Document function_return function attribute.
6125 * doc/invoke.texi: Document -mfunction-return= option.
6127 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6129 * config/i386/i386-opts.h (indirect_branch): New.
6130 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
6131 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
6132 with local indirect jump when converting indirect call and jump.
6133 (ix86_set_indirect_branch_type): New.
6134 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
6135 (indirectlabelno): New.
6136 (indirect_thunk_needed): Likewise.
6137 (indirect_thunk_bnd_needed): Likewise.
6138 (indirect_thunks_used): Likewise.
6139 (indirect_thunks_bnd_used): Likewise.
6140 (INDIRECT_LABEL): Likewise.
6141 (indirect_thunk_name): Likewise.
6142 (output_indirect_thunk): Likewise.
6143 (output_indirect_thunk_function): Likewise.
6144 (ix86_output_indirect_branch_via_reg): Likewise.
6145 (ix86_output_indirect_branch_via_push): Likewise.
6146 (ix86_output_indirect_branch): Likewise.
6147 (ix86_output_indirect_jmp): Likewise.
6148 (ix86_code_end): Call output_indirect_thunk_function if needed.
6149 (ix86_output_call_insn): Call ix86_output_indirect_branch if
6151 (ix86_handle_fndecl_attribute): Handle indirect_branch.
6152 (ix86_attribute_table): Add indirect_branch.
6153 * config/i386/i386.h (machine_function): Add indirect_branch_type
6154 and has_local_indirect_jump.
6155 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
6157 (tablejump): Likewise.
6158 (*indirect_jump): Use ix86_output_indirect_jmp.
6159 (*tablejump_1): Likewise.
6160 (simple_return_indirect_internal): Likewise.
6161 * config/i386/i386.opt (mindirect-branch=): New option.
6162 (indirect_branch): New.
6165 (thunk-inline): Likewise.
6166 (thunk-extern): Likewise.
6167 * doc/extend.texi: Document indirect_branch function attribute.
6168 * doc/invoke.texi: Document -mindirect-branch= option.
6170 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
6173 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
6175 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
6177 * ipa-inline.c (want_inline_small_function_p): Return false if
6178 inlining has already failed with CIF_FINAL_ERROR.
6179 (update_caller_keys): Call want_inline_small_function_p before
6181 (update_callee_keys): Likewise.
6183 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6185 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
6187 (rs6000_quadword_masked_address_p): Likewise.
6188 (quad_aligned_load_p): Likewise.
6189 (quad_aligned_store_p): Likewise.
6190 (const_load_sequence_p): Add comment to describe the outer-most loop.
6191 (mimic_memory_attributes_and_flags): New function.
6192 (rs6000_gen_stvx): Likewise.
6193 (replace_swapped_aligned_store): Likewise.
6194 (rs6000_gen_lvx): Likewise.
6195 (replace_swapped_aligned_load): Likewise.
6196 (replace_swapped_load_constant): Capitalize argument name in
6197 comment describing this function.
6198 (rs6000_analyze_swaps): Add a third pass to search for vector loads
6199 and stores that access quad-word aligned addresses and replace
6200 with stvx or lvx instructions when appropriate.
6201 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
6202 New function prototype.
6203 (rs6000_quadword_masked_address_p): Likewise.
6204 (rs6000_gen_lvx): Likewise.
6205 (rs6000_gen_stvx): Likewise.
6206 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
6207 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
6208 when memory address is aligned.
6209 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
6210 this split to select lvx instruction when memory address is aligned.
6211 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
6212 instruction when memory address is aligned.
6213 (*vsx_le_perm_load_v16qi): Likewise.
6214 (four unnamed splitters): Modify to select the stvx instruction
6215 when memory is aligned.
6217 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
6219 * predict.c (determine_unlikely_bbs): Handle correctly BBs
6220 which appears in the queue multiple times.
6222 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6223 Alan Hayward <alan.hayward@arm.com>
6224 David Sherwood <david.sherwood@arm.com>
6226 * tree-vectorizer.h (vec_lower_bound): New structure.
6227 (_loop_vec_info): Add check_nonzero and lower_bounds.
6228 (LOOP_VINFO_CHECK_NONZERO): New macro.
6229 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
6230 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
6231 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
6232 fields. Make seg_len the distance travelled, not including the
6234 (dr_direction_indicator): Declare.
6235 (dr_zero_step_indicator): Likewise.
6236 (dr_known_forward_stride_p): Likewise.
6237 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
6239 (runtime_alias_check_p): Allow runtime alias checks with
6241 (operator ==): Compare access_size and align.
6242 (prune_runtime_alias_test_list): Rework for new distinction between
6243 the access_size and seg_len.
6244 (create_intersect_range_checks_index): Likewise. Cope with polynomial
6246 (get_segment_min_max): New function.
6247 (create_intersect_range_checks): Use it.
6248 (dr_step_indicator): New function.
6249 (dr_direction_indicator): Likewise.
6250 (dr_zero_step_indicator): Likewise.
6251 (dr_known_forward_stride_p): Likewise.
6252 * tree-loop-distribution.c (data_ref_segment_size): Return
6253 DR_STEP * (niters - 1).
6254 (compute_alias_check_pairs): Update call to the dr_with_seg_len
6256 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
6257 (vect_preserves_scalar_order_p): New function, split out from...
6258 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
6259 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
6260 (vect_vfa_access_size): New function.
6261 (vect_vfa_align): Likewise.
6262 (vect_compile_time_alias): Take access_size_a and access_b arguments.
6263 (dump_lower_bound): New function.
6264 (vect_check_lower_bound): Likewise.
6265 (vect_small_gap_p): Likewise.
6266 (vectorizable_with_step_bound_p): Likewise.
6267 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
6268 depencies if the vectorization factor is 1. Convert the checks
6269 for nonzero steps into checks on the bounds of DR_STEP. Try using
6270 a bunds check for variable steps if the minimum required step is
6271 relatively small. Update calls to the dr_with_seg_len
6272 constructor and to vect_compile_time_alias.
6273 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
6275 (vect_loop_versioning): Call it.
6276 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
6278 (vect_estimate_min_profitable_iters): Account for any bounds checks.
6280 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6281 Alan Hayward <alan.hayward@arm.com>
6282 David Sherwood <david.sherwood@arm.com>
6284 * doc/sourcebuild.texi (vect_scatter_store): Document.
6285 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
6287 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
6289 * genopinit.c (main): Add supports_vec_scatter_store and
6290 supports_vec_scatter_store_cached to target_optabs.
6291 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
6292 IFN_MASK_SCATTER_STORE.
6293 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
6295 * internal-fn.h (internal_store_fn_p): Declare.
6296 (internal_fn_stored_value_index): Likewise.
6297 * internal-fn.c (scatter_store_direct): New macro.
6298 (expand_scatter_store_optab_fn): New function.
6299 (direct_scatter_store_optab_supported_p): New macro.
6300 (internal_store_fn_p): New function.
6301 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
6302 IFN_MASK_SCATTER_STORE.
6303 (internal_fn_mask_index): Likewise.
6304 (internal_fn_stored_value_index): New function.
6305 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
6307 * optabs-query.h (supports_vec_scatter_store_p): Declare.
6308 * optabs-query.c (supports_vec_scatter_store_p): New function.
6309 * tree-vectorizer.h (vect_get_store_rhs): Declare.
6310 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
6311 true for scatter stores.
6312 (vect_gather_scatter_fn_p): Handle scatter stores too.
6313 (vect_check_gather_scatter): Consider using scatter stores if
6314 supports_vec_scatter_store_p.
6315 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
6317 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6318 internal_fn_stored_value_index.
6319 (check_load_store_masking): Handle scatter stores too.
6320 (vect_get_store_rhs): Make public.
6321 (vectorizable_call): Use internal_store_fn_p.
6322 (vectorizable_store): Handle scatter store internal functions.
6323 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
6324 when deciding whether the end of the group has been reached.
6325 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
6326 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
6327 (mask_scatter_store<mode>): New insns.
6329 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6330 Alan Hayward <alan.hayward@arm.com>
6331 David Sherwood <david.sherwood@arm.com>
6333 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
6334 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
6335 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
6337 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
6338 Use vect_truncate_gather_scatter_offset if we can't treat the
6339 operation as a normal gather load or scatter store.
6340 (get_group_load_store_type): Take the gather_scatter_info
6341 as argument. Try using a gather load or scatter store for
6342 single-element groups.
6343 (get_load_store_type): Update calls to get_group_load_store_type
6344 and vect_use_strided_gather_scatters_p.
6346 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6347 Alan Hayward <alan.hayward@arm.com>
6348 David Sherwood <david.sherwood@arm.com>
6350 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
6351 optional tree argument.
6352 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
6354 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
6355 but continue to use the current value as a fallback.
6356 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
6357 to compare the updates.
6358 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
6359 (get_load_store_type): Use it when handling a strided access.
6360 (vect_get_strided_load_store_ops): New function.
6361 (vect_get_data_ptr_increment): Likewise.
6362 (vectorizable_load): Handle strided gather loads. Always pass
6363 a step to vect_create_data_ref_ptr and bump_vector_ptr.
6365 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6366 Alan Hayward <alan.hayward@arm.com>
6367 David Sherwood <david.sherwood@arm.com>
6369 * doc/md.texi (gather_load@var{m}): Document.
6370 (mask_gather_load@var{m}): Likewise.
6371 * genopinit.c (main): Add supports_vec_gather_load and
6372 supports_vec_gather_load_cached to target_optabs.
6373 * optabs-tree.c (init_tree_optimization_optabs): Use
6374 ggc_cleared_alloc to allocate target_optabs.
6375 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
6376 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
6378 * internal-fn.h (internal_load_fn_p): Declare.
6379 (internal_gather_scatter_fn_p): Likewise.
6380 (internal_fn_mask_index): Likewise.
6381 (internal_gather_scatter_fn_supported_p): Likewise.
6382 * internal-fn.c (gather_load_direct): New macro.
6383 (expand_gather_load_optab_fn): New function.
6384 (direct_gather_load_optab_supported_p): New macro.
6385 (direct_internal_fn_optab): New function.
6386 (internal_load_fn_p): Likewise.
6387 (internal_gather_scatter_fn_p): Likewise.
6388 (internal_fn_mask_index): Likewise.
6389 (internal_gather_scatter_fn_supported_p): Likewise.
6390 * optabs-query.c (supports_at_least_one_mode_p): New function.
6391 (supports_vec_gather_load_p): Likewise.
6392 * optabs-query.h (supports_vec_gather_load_p): Declare.
6393 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
6394 and memory_type field.
6395 (NUM_PATTERNS): Bump to 15.
6396 * tree-vect-data-refs.c: Include internal-fn.h.
6397 (vect_gather_scatter_fn_p): New function.
6398 (vect_describe_gather_scatter_call): Likewise.
6399 (vect_check_gather_scatter): Try using internal functions for
6400 gather loads. Recognize existing calls to a gather load function.
6401 (vect_analyze_data_refs): Consider using gather loads if
6402 supports_vec_gather_load_p.
6403 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
6404 (vect_get_gather_scatter_offset_type): Likewise.
6405 (vect_convert_mask_for_vectype): Likewise.
6406 (vect_add_conversion_to_patterm): Likewise.
6407 (vect_try_gather_scatter_pattern): Likewise.
6408 (vect_recog_gather_scatter_pattern): New pattern recognizer.
6409 (vect_vect_recog_func_ptrs): Add it.
6410 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6411 internal_fn_mask_index and internal_gather_scatter_fn_p.
6412 (check_load_store_masking): Take the gather_scatter_info as an
6413 argument and handle gather loads.
6414 (vect_get_gather_scatter_ops): New function.
6415 (vectorizable_call): Check internal_load_fn_p.
6416 (vectorizable_load): Likewise. Handle gather load internal
6418 (vectorizable_store): Update call to check_load_store_masking.
6419 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
6420 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
6421 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
6422 (aarch64_gather_scale_operand_d): New predicates.
6423 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
6424 (mask_gather_load<mode>): New insns.
6426 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6427 Alan Hayward <alan.hayward@arm.com>
6428 David Sherwood <david.sherwood@arm.com>
6430 * optabs.def (fold_left_plus_optab): New optab.
6431 * doc/md.texi (fold_left_plus_@var{m}): Document.
6432 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
6433 * internal-fn.c (fold_left_direct): Define.
6434 (expand_fold_left_optab_fn): Likewise.
6435 (direct_fold_left_optab_supported_p): Likewise.
6436 * fold-const-call.c (fold_const_fold_left): New function.
6437 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
6438 * tree-parloops.c (valid_reduction_p): New function.
6439 (gather_scalar_reductions): Use it.
6440 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
6441 (vect_finish_replace_stmt): Declare.
6442 * tree-vect-loop.c (fold_left_reduction_fn): New function.
6443 (needs_fold_left_reduction_p): New function, split out from...
6444 (vect_is_simple_reduction): ...here. Accept reductions that
6445 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
6446 (vect_force_simple_reduction): Also store the reduction type in
6447 the assignment's STMT_VINFO_REDUC_TYPE.
6448 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
6449 (merge_with_identity): New function.
6450 (vect_expand_fold_left): Likewise.
6451 (vectorize_fold_left_reduction): Likewise.
6452 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
6453 scalar phi in place for it. Check for target support and reject
6454 cases that would reassociate the operation. Defer the transform
6455 phase to vectorize_fold_left_reduction.
6456 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
6457 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
6458 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
6460 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6462 * tree-if-conv.c (predicate_mem_writes): Remove redundant
6463 call to ifc_temp_var.
6465 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6466 Alan Hayward <alan.hayward@arm.com>
6467 David Sherwood <david.sherwood@arm.com>
6469 * target.def (legitimize_address_displacement): Take the original
6470 offset as a poly_int.
6471 * targhooks.h (default_legitimize_address_displacement): Update
6473 * targhooks.c (default_legitimize_address_displacement): Likewise.
6474 * doc/tm.texi: Regenerate.
6475 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
6476 as an argument, moving assert of ad->disp == ad->disp_term to...
6477 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
6478 Try calling targetm.legitimize_address_displacement before expanding
6479 the address rather than afterwards, and adjust for the new interface.
6480 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
6481 Match the new hook interface. Handle SVE addresses.
6482 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
6485 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6487 * Makefile.in (OBJS): Add early-remat.o.
6488 * target.def (select_early_remat_modes): New hook.
6489 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
6490 * doc/tm.texi: Regenerate.
6491 * targhooks.h (default_select_early_remat_modes): Declare.
6492 * targhooks.c (default_select_early_remat_modes): New function.
6493 * timevar.def (TV_EARLY_REMAT): New timevar.
6494 * passes.def (pass_early_remat): New pass.
6495 * tree-pass.h (make_pass_early_remat): Declare.
6496 * early-remat.c: New file.
6497 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
6499 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
6501 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6502 Alan Hayward <alan.hayward@arm.com>
6503 David Sherwood <david.sherwood@arm.com>
6505 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
6506 vfm1 with a bound_epilog parameter.
6507 (vect_do_peeling): Update calls accordingly, and move the prologue
6508 call earlier in the function. Treat the base bound_epilog as 0 for
6509 fully-masked loops and retain vf - 1 for other loops. Add 1 to
6510 this base when peeling for gaps.
6511 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
6512 with fully-masked loops.
6513 (vect_estimate_min_profitable_iters): Handle the single peeled
6514 iteration in that case.
6516 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6517 Alan Hayward <alan.hayward@arm.com>
6518 David Sherwood <david.sherwood@arm.com>
6520 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
6521 single-element interleaving even if the size is not a power of 2.
6522 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
6523 accesses for single-element interleaving if the group size is
6526 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6527 Alan Hayward <alan.hayward@arm.com>
6528 David Sherwood <david.sherwood@arm.com>
6530 * doc/md.texi (fold_extract_last_@var{m}): Document.
6531 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
6532 * optabs.def (fold_extract_last_optab): New optab.
6533 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
6534 * internal-fn.c (fold_extract_direct): New macro.
6535 (expand_fold_extract_optab_fn): Likewise.
6536 (direct_fold_extract_optab_supported_p): Likewise.
6537 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
6538 * tree-vect-loop.c (vect_model_reduction_cost): Handle
6539 EXTRACT_LAST_REDUCTION.
6540 (get_initial_def_for_reduction): Do not create an initial vector
6541 for EXTRACT_LAST_REDUCTION reductions.
6542 (vectorizable_reduction): Leave the scalar phi in place for
6543 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
6544 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
6545 epilogue code for EXTRACT_LAST_REDUCTION and defer the
6546 transform phase to vectorizable_condition.
6547 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
6549 (vect_finish_stmt_generation): ...here.
6550 (vect_finish_replace_stmt): New function.
6551 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
6552 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
6554 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
6556 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6557 Alan Hayward <alan.hayward@arm.com>
6558 David Sherwood <david.sherwood@arm.com>
6560 * doc/md.texi (extract_last_@var{m}): Document.
6561 * optabs.def (extract_last_optab): New optab.
6562 * internal-fn.def (EXTRACT_LAST): New internal function.
6563 * internal-fn.c (cond_unary_direct): New macro.
6564 (expand_cond_unary_optab_fn): Likewise.
6565 (direct_cond_unary_optab_supported_p): Likewise.
6566 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
6567 loops using EXTRACT_LAST.
6568 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
6569 (extract_last_<mode>): ...this optab.
6570 (vec_extract<mode><Vel>): Update accordingly.
6572 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6573 Alan Hayward <alan.hayward@arm.com>
6574 David Sherwood <david.sherwood@arm.com>
6576 * target.def (empty_mask_is_expensive): New hook.
6577 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
6578 * doc/tm.texi: Regenerate.
6579 * targhooks.h (default_empty_mask_is_expensive): Declare.
6580 * targhooks.c (default_empty_mask_is_expensive): New function.
6581 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
6582 if the target says that empty masks are expensive.
6583 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
6585 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
6587 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6588 Alan Hayward <alan.hayward@arm.com>
6589 David Sherwood <david.sherwood@arm.com>
6591 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
6592 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
6593 (vect_use_loop_mask_for_alignment_p): New function.
6594 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
6595 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
6596 niters_skip argument. Make sure that the first niters_skip elements
6597 of the first iteration are inactive.
6598 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
6599 Update call to vect_set_loop_masks_directly.
6600 (get_misalign_in_elems): New function, split out from...
6601 (vect_gen_prolog_loop_niters): ...here.
6602 (vect_update_init_of_dr): Take a code argument that specifies whether
6603 the adjustment should be added or subtracted.
6604 (vect_update_init_of_drs): Likewise.
6605 (vect_prepare_for_masked_peels): New function.
6606 (vect_do_peeling): Skip prologue peeling if we're using a mask
6607 instead. Update call to vect_update_inits_of_drs.
6608 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6610 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
6611 alignment. Do not include the number of peeled iterations in
6612 the minimum threshold in that case.
6613 (vectorizable_induction): Adjust the start value down by
6614 LOOP_VINFO_MASK_SKIP_NITERS iterations.
6615 (vect_transform_loop): Call vect_prepare_for_masked_peels.
6616 Take the number of skipped iterations into account when calculating
6618 * tree-vect-stmts.c (vect_gen_while_not): New function.
6620 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6621 Alan Hayward <alan.hayward@arm.com>
6622 David Sherwood <david.sherwood@arm.com>
6624 * doc/sourcebuild.texi (vect_fully_masked): Document.
6625 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
6627 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
6629 (vect_analyze_loop_2): ...here. Don't check the vectorization
6630 factor against the number of loop iterations if the loop is
6633 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6634 Alan Hayward <alan.hayward@arm.com>
6635 David Sherwood <david.sherwood@arm.com>
6637 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
6638 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
6639 (dump_groups): Update accordingly.
6640 (iv_use::mem_type): New member variable.
6641 (address_p): New function.
6642 (record_use): Add a mem_type argument and initialize the new
6644 (record_group_use): Add a mem_type argument. Use address_p.
6645 Remove obsolete null checks of base_object. Update call to record_use.
6646 (find_interesting_uses_op): Update call to record_group_use.
6647 (find_interesting_uses_cond): Likewise.
6648 (find_interesting_uses_address): Likewise.
6649 (get_mem_type_for_internal_fn): New function.
6650 (find_address_like_use): Likewise.
6651 (find_interesting_uses_stmt): Try find_address_like_use before
6652 calling find_interesting_uses_op.
6653 (addr_offset_valid_p): Use the iv mem_type field as the type
6654 of the addressed memory.
6655 (add_autoinc_candidates): Likewise.
6656 (get_address_cost): Likewise.
6657 (split_small_address_groups_p): Use address_p.
6658 (split_address_groups): Likewise.
6659 (add_iv_candidate_for_use): Likewise.
6660 (autoinc_possible_for_pair): Likewise.
6661 (rewrite_groups): Likewise.
6662 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
6663 (determine_group_iv_cost): Update after split of USE_ADDRESS.
6664 (get_alias_ptr_type_for_ptr_address): New function.
6665 (rewrite_use_address): Rewrite address uses in calls that were
6666 identified by find_address_like_use.
6668 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6669 Alan Hayward <alan.hayward@arm.com>
6670 David Sherwood <david.sherwood@arm.com>
6672 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
6674 * gimple-expr.h (is_gimple_addressable: Likewise.
6675 * gimple-expr.c (is_gimple_address): Likewise.
6676 * internal-fn.c (expand_call_mem_ref): New function.
6677 (expand_mask_load_optab_fn): Use it.
6678 (expand_mask_store_optab_fn): Likewise.
6680 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6681 Alan Hayward <alan.hayward@arm.com>
6682 David Sherwood <david.sherwood@arm.com>
6684 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
6685 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
6686 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
6687 (cond_umax@var{mode}): Document.
6688 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
6689 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
6690 (cond_umin_optab, cond_umax_optab): New optabs.
6691 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
6692 (COND_IOR, COND_XOR): New internal functions.
6693 * internal-fn.h (get_conditional_internal_fn): Declare.
6694 * internal-fn.c (cond_binary_direct): New macro.
6695 (expand_cond_binary_optab_fn): Likewise.
6696 (direct_cond_binary_optab_supported_p): Likewise.
6697 (get_conditional_internal_fn): New function.
6698 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
6699 Cope with reduction statements that are vectorized as calls rather
6701 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
6702 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
6703 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
6704 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
6705 (UNSPEC_COND_EOR): New unspecs.
6706 (optab): Add mappings for them.
6707 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
6708 (sve_int_op, sve_fp_op): New int attributes.
6710 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6711 Alan Hayward <alan.hayward@arm.com>
6712 David Sherwood <david.sherwood@arm.com>
6714 * optabs.def (while_ult_optab): New optab.
6715 * doc/md.texi (while_ult@var{m}@var{n}): Document.
6716 * internal-fn.def (WHILE_ULT): New internal function.
6717 * internal-fn.h (direct_internal_fn_supported_p): New override
6718 that takes two types as argument.
6719 * internal-fn.c (while_direct): New macro.
6720 (expand_while_optab_fn): New function.
6721 (convert_optab_supported_p): Likewise.
6722 (direct_while_optab_supported_p): New macro.
6723 * wide-int.h (wi::udiv_ceil): New function.
6724 * tree-vectorizer.h (rgroup_masks): New structure.
6725 (vec_loop_masks): New typedef.
6726 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
6728 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
6729 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
6730 (vect_max_vf): New function.
6731 (slpeel_make_loop_iterate_ntimes): Delete.
6732 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
6733 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
6734 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
6735 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
6736 internal-fn.h, stor-layout.h and optabs-query.h.
6737 (vect_set_loop_mask): New function.
6738 (add_preheader_seq): Likewise.
6739 (add_header_seq): Likewise.
6740 (interleave_supported_p): Likewise.
6741 (vect_maybe_permute_loop_masks): Likewise.
6742 (vect_set_loop_masks_directly): Likewise.
6743 (vect_set_loop_condition_masked): Likewise.
6744 (vect_set_loop_condition_unmasked): New function, split out from
6745 slpeel_make_loop_iterate_ntimes.
6746 (slpeel_make_loop_iterate_ntimes): Rename to..
6747 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
6748 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
6749 (vect_do_peeling): Update call accordingly.
6750 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
6752 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6753 mask_compare_type, can_fully_mask_p and fully_masked_p.
6754 (release_vec_loop_masks): New function.
6755 (_loop_vec_info): Use it to free the loop masks.
6756 (can_produce_all_loop_masks_p): New function.
6757 (vect_get_max_nscalars_per_iter): Likewise.
6758 (vect_verify_full_masking): Likewise.
6759 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
6760 retries, and free the mask rgroups before retrying. Check loop-wide
6761 reasons for disallowing fully-masked loops. Make the final decision
6762 about whether use a fully-masked loop or not.
6763 (vect_estimate_min_profitable_iters): Do not assume that peeling
6764 for the number of iterations will be needed for fully-masked loops.
6765 (vectorizable_reduction): Disable fully-masked loops.
6766 (vectorizable_live_operation): Likewise.
6767 (vect_halve_mask_nunits): New function.
6768 (vect_double_mask_nunits): Likewise.
6769 (vect_record_loop_mask): Likewise.
6770 (vect_get_loop_mask): Likewise.
6771 (vect_transform_loop): Handle the case in which the final loop
6772 iteration might handle a partial vector. Call vect_set_loop_condition
6773 instead of slpeel_make_loop_iterate_ntimes.
6774 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
6775 (check_load_store_masking): New function.
6776 (prepare_load_store_mask): Likewise.
6777 (vectorizable_store): Handle fully-masked loops.
6778 (vectorizable_load): Likewise.
6779 (supportable_widening_operation): Use vect_halve_mask_nunits for
6781 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
6782 (vect_gen_while): New function.
6783 * config/aarch64/aarch64.md (umax<mode>3): New expander.
6784 (aarch64_uqdec<mode>): New insn.
6786 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6787 Alan Hayward <alan.hayward@arm.com>
6788 David Sherwood <david.sherwood@arm.com>
6790 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
6791 (reduc_xor_scal_optab): New optabs.
6792 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
6793 (reduc_xor_scal_@var{m}): Document.
6794 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
6795 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
6797 * fold-const-call.c (fold_const_call): Handle them.
6798 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
6799 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
6800 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
6801 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
6802 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
6803 (UNSPEC_XORV): New unspecs.
6804 (optab): Add entries for them.
6805 (BITWISEV): New int iterator.
6806 (bit_reduc_op): New int attributes.
6808 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6809 Alan Hayward <alan.hayward@arm.com>
6810 David Sherwood <david.sherwood@arm.com>
6812 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
6813 * internal-fn.def (VEC_SHL_INSERT): New internal function.
6814 * optabs.def (vec_shl_insert_optab): New optab.
6815 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
6816 (duplicate_and_interleave): Likewise.
6817 * tree-vect-loop.c: Include internal-fn.h.
6818 (neutral_op_for_slp_reduction): New function, split out from
6819 get_initial_defs_for_reduction.
6820 (get_initial_def_for_reduction): Handle option 2 for variable-length
6821 vectors by loading the neutral value into a vector and then shifting
6822 the initial value into element 0.
6823 (get_initial_defs_for_reduction): Replace the code argument with
6824 the neutral value calculated by neutral_op_for_slp_reduction.
6825 Use gimple_build_vector for constant-length vectors.
6826 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
6827 but the first group_size elements have a neutral value.
6828 Use duplicate_and_interleave otherwise.
6829 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
6830 Update call to get_initial_defs_for_reduction. Handle SLP
6831 reductions for variable-length vectors by creating one vector
6832 result for each scalar result, with the elements associated
6833 with other scalar results stubbed out with the neutral value.
6834 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
6835 Require IFN_VEC_SHL_INSERT for double reductions on
6836 variable-length vectors, or SLP reductions that have
6837 a neutral value. Require can_duplicate_and_interleave_p
6838 support for variable-length unchained SLP reductions if there
6839 is no neutral value, such as for MIN/MAX reductions. Also require
6840 the number of vector elements to be a multiple of the number of
6841 SLP statements when doing variable-length unchained SLP reductions.
6842 Update call to vect_create_epilog_for_reduction.
6843 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
6844 and remove initial values.
6845 (duplicate_and_interleave): Make public.
6846 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
6847 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
6849 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6850 Alan Hayward <alan.hayward@arm.com>
6851 David Sherwood <david.sherwood@arm.com>
6853 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
6854 (can_duplicate_and_interleave_p): New function.
6855 (vect_get_and_check_slp_defs): Take the vector of statements
6856 rather than just the current one. Remove excess parentheses.
6857 Restriction rejectinon of vect_constant_def and vect_external_def
6858 for variable-length vectors to boolean types, or types for which
6859 can_duplicate_and_interleave_p is false.
6860 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
6861 (duplicate_and_interleave): New function.
6862 (vect_get_constant_vectors): Use gimple_build_vector for
6863 constant-length vectors and suitable variable-length constant
6864 vectors. Use duplicate_and_interleave for other variable-length
6865 vectors. Don't defer the update when inserting new statements.
6867 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6868 Alan Hayward <alan.hayward@arm.com>
6869 David Sherwood <david.sherwood@arm.com>
6871 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
6872 min_profitable_iters doesn't go negative.
6874 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6875 Alan Hayward <alan.hayward@arm.com>
6876 David Sherwood <david.sherwood@arm.com>
6878 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
6879 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
6880 * optabs.def (vec_mask_load_lanes_optab): New optab.
6881 (vec_mask_store_lanes_optab): Likewise.
6882 * internal-fn.def (MASK_LOAD_LANES): New internal function.
6883 (MASK_STORE_LANES): Likewise.
6884 * internal-fn.c (mask_load_lanes_direct): New macro.
6885 (mask_store_lanes_direct): Likewise.
6886 (expand_mask_load_optab_fn): Handle masked operations.
6887 (expand_mask_load_lanes_optab_fn): New macro.
6888 (expand_mask_store_optab_fn): Handle masked operations.
6889 (expand_mask_store_lanes_optab_fn): New macro.
6890 (direct_mask_load_lanes_optab_supported_p): Likewise.
6891 (direct_mask_store_lanes_optab_supported_p): Likewise.
6892 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
6894 (vect_load_lanes_supported): Likewise.
6895 * tree-vect-data-refs.c (strip_conversion): New function.
6896 (can_group_stmts_p): Likewise.
6897 (vect_analyze_data_ref_accesses): Use it instead of checking
6898 for a pair of assignments.
6899 (vect_store_lanes_supported): Take a masked_p parameter.
6900 (vect_load_lanes_supported): Likewise.
6901 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
6902 vect_store_lanes_supported and vect_load_lanes_supported.
6903 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6904 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
6905 parameter. Don't allow gaps for masked accesses.
6906 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
6907 and vect_load_lanes_supported.
6908 (get_load_store_type): Take a masked_p parameter and update
6909 call to get_group_load_store_type.
6910 (vectorizable_store): Update call to get_load_store_type.
6911 Handle IFN_MASK_STORE_LANES.
6912 (vectorizable_load): Update call to get_load_store_type.
6913 Handle IFN_MASK_LOAD_LANES.
6915 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6916 Alan Hayward <alan.hayward@arm.com>
6917 David Sherwood <david.sherwood@arm.com>
6919 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
6921 * config/aarch64/aarch64-protos.h
6922 (aarch64_sve_struct_memory_operand_p): Declare.
6923 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
6924 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
6925 (VPRED, vpred): Handle SVE structure modes.
6926 * config/aarch64/constraints.md (Utx): New constraint.
6927 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
6928 (aarch64_sve_struct_nonimmediate_operand): New predicates.
6929 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
6930 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
6931 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
6932 structure modes. Split into pieces after RA.
6933 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
6934 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
6936 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
6937 SVE structure modes.
6938 (aarch64_classify_address): Likewise.
6939 (sizetochar): Move earlier in file.
6940 (aarch64_print_operand): Handle SVE register lists.
6941 (aarch64_array_mode): New function.
6942 (aarch64_sve_struct_memory_operand_p): Likewise.
6943 (TARGET_ARRAY_MODE): Redefine.
6945 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6946 Alan Hayward <alan.hayward@arm.com>
6947 David Sherwood <david.sherwood@arm.com>
6949 * target.def (array_mode): New target hook.
6950 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
6951 * doc/tm.texi: Regenerate.
6952 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
6953 * hooks.c (hook_optmode_mode_uhwi_none): New function.
6954 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
6956 * stor-layout.c (mode_for_array): Likewise. Support polynomial
6959 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6960 Alan Hayward <alan.hayward@arm.com>
6961 David Sherwood <david.sherwood@arm.com>
6963 * fold-const.c (fold_binary_loc): Check the argument types
6964 rather than the result type when testing for a vector operation.
6966 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6968 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
6969 * doc/tm.texi: Regenerate.
6971 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6972 Alan Hayward <alan.hayward@arm.com>
6973 David Sherwood <david.sherwood@arm.com>
6975 * doc/invoke.texi (-msve-vector-bits=): Document new option.
6976 (sve): Document new AArch64 extension.
6977 * doc/md.texi (w): Extend the description of the AArch64
6978 constraint to include SVE vectors.
6979 (Upl, Upa): Document new AArch64 predicate constraints.
6980 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
6982 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
6983 (msve-vector-bits=): New option.
6984 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
6985 SVE when these are disabled.
6986 (sve): New extension.
6987 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
6988 modes. Adjust their number of units based on aarch64_sve_vg.
6989 (MAX_BITSIZE_MODE_ANY_MODE): Define.
6990 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
6991 aarch64_addr_query_type.
6992 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
6993 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
6994 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
6995 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
6996 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
6997 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
6998 (aarch64_simd_imm_zero_p): Delete.
6999 (aarch64_check_zero_based_sve_index_immediate): Declare.
7000 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7001 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7002 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7003 (aarch64_sve_float_mul_immediate_p): Likewise.
7004 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7006 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
7007 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
7008 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
7009 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
7010 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
7011 (aarch64_regmode_natural_size): Likewise.
7012 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
7013 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
7015 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
7016 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
7017 for VG and the SVE predicate registers.
7018 (V_ALIASES): Add a "z"-prefixed alias.
7019 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
7020 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
7021 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
7022 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
7023 (REG_CLASS_NAMES): Add entries for them.
7024 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
7025 and the predicate registers.
7026 (aarch64_sve_vg): Declare.
7027 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
7028 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
7029 (REGMODE_NATURAL_SIZE): Define.
7030 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
7032 * config/aarch64/aarch64.c: Include cfgrtl.h.
7033 (simd_immediate_info): Add a constructor for series vectors,
7034 and an associated step field.
7035 (aarch64_sve_vg): New variable.
7036 (aarch64_dbx_register_number): Handle VG and the predicate registers.
7037 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
7038 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
7039 (VEC_ANY_DATA, VEC_STRUCT): New constants.
7040 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
7041 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
7042 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
7043 (aarch64_get_mask_mode): New functions.
7044 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
7045 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
7046 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
7047 predicate modes and predicate registers. Explicitly restrict
7048 GPRs to modes of 16 bytes or smaller. Only allow FP registers
7049 to store a vector mode if it is recognized by
7050 aarch64_classify_vector_mode.
7051 (aarch64_regmode_natural_size): New function.
7052 (aarch64_hard_regno_caller_save_mode): Return the original mode
7054 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
7055 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
7056 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
7057 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
7059 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
7060 does not overlap dest if the function is frame-related. Handle
7062 (aarch64_split_add_offset): New function.
7063 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
7064 them aarch64_add_offset.
7065 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
7066 and update call to aarch64_sub_sp.
7067 (aarch64_add_cfa_expression): New function.
7068 (aarch64_expand_prologue): Pass extra temporary registers to the
7069 functions above. Handle the case in which we need to emit new
7070 DW_CFA_expressions for registers that were originally saved
7071 relative to the stack pointer, but now have to be expressed
7072 relative to the frame pointer.
7073 (aarch64_output_mi_thunk): Pass extra temporary registers to the
7075 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
7076 IP0 and IP1 values for SVE frames.
7077 (aarch64_expand_vec_series): New function.
7078 (aarch64_expand_sve_widened_duplicate): Likewise.
7079 (aarch64_expand_sve_const_vector): Likewise.
7080 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
7081 Handle SVE constants. Use emit_move_insn to move a force_const_mem
7082 into the register, rather than emitting a SET directly.
7083 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
7084 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
7085 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
7086 (offset_9bit_signed_scaled_p): New functions.
7087 (aarch64_replicate_bitmask_imm): New function.
7088 (aarch64_bitmask_imm): Use it.
7089 (aarch64_cannot_force_const_mem): Reject expressions involving
7090 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
7091 (aarch64_classify_index): Handle SVE indices, by requiring
7092 a plain register index with a scale that matches the element size.
7093 (aarch64_classify_address): Handle SVE addresses. Assert that
7094 the mode of the address is VOIDmode or an integer mode.
7095 Update call to aarch64_classify_symbol.
7096 (aarch64_classify_symbolic_expression): Update call to
7097 aarch64_classify_symbol.
7098 (aarch64_const_vec_all_in_range_p): New function.
7099 (aarch64_print_vector_float_operand): Likewise.
7100 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
7101 "vN" for FP registers with SVE modes. Handle (const ...) vectors
7102 and the FP immediates 1.0 and 0.5.
7103 (aarch64_print_address_internal): Handle SVE addresses.
7104 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
7105 (aarch64_regno_regclass): Handle predicate registers.
7106 (aarch64_secondary_reload): Handle big-endian reloads of SVE
7108 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
7109 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
7110 (aarch64_convert_sve_vector_bits): New function.
7111 (aarch64_override_options): Use it to handle -msve-vector-bits=.
7112 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7114 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
7115 Handle SVE vector and predicate modes. Accept VL-based constants
7116 that need only one temporary register, and VL offsets that require
7117 no temporary registers.
7118 (aarch64_conditional_register_usage): Mark the predicate registers
7119 as fixed if SVE isn't available.
7120 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
7121 Return true for SVE vector and predicate modes.
7122 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
7123 rather than an unsigned int. Handle SVE modes.
7124 (aarch64_preferred_simd_mode): Update call accordingly. Handle
7126 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
7128 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7129 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7130 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7131 (aarch64_sve_float_mul_immediate_p): New functions.
7132 (aarch64_sve_valid_immediate): New function.
7133 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
7134 Explicitly reject structure modes. Check for INDEX constants.
7135 Handle PTRUE and PFALSE constants.
7136 (aarch64_check_zero_based_sve_index_immediate): New function.
7137 (aarch64_simd_imm_zero_p): Delete.
7138 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
7139 vector modes. Accept constants in the range of CNT[BHWD].
7140 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
7141 ask for an Advanced SIMD mode.
7142 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
7143 (aarch64_simd_vector_alignment): Handle SVE predicates.
7144 (aarch64_vectorize_preferred_vector_alignment): New function.
7145 (aarch64_simd_vector_alignment_reachable): Use it instead of
7147 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
7148 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
7150 (MAX_VECT_LEN): Delete.
7151 (expand_vec_perm_d): Add a vec_flags field.
7152 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
7153 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
7154 (aarch64_evpc_ext): Don't apply a big-endian lane correction
7156 (aarch64_evpc_rev): Rename to...
7157 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
7158 (aarch64_evpc_rev_global): New function.
7159 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
7160 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
7162 (aarch64_evpc_sve_tbl): New function.
7163 (aarch64_expand_vec_perm_const_1): Update after rename of
7164 aarch64_evpc_rev. Handle SVE permutes too, trying
7165 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
7166 than aarch64_evpc_tbl.
7167 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
7168 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
7169 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
7170 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
7171 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
7172 (aarch64_expand_sve_vcond): New functions.
7173 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
7174 of aarch64_vector_mode_p.
7175 (aarch64_dwarf_poly_indeterminate_value): New function.
7176 (aarch64_compute_pressure_classes): Likewise.
7177 (aarch64_can_change_mode_class): Likewise.
7178 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
7179 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
7180 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
7181 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
7182 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
7183 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
7184 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
7185 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
7187 (Dn, Dl, Dr): Accept const as well as const_vector.
7188 (Dz): Likewise. Compare against CONST0_RTX.
7189 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
7190 of "vector" where appropriate.
7191 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
7192 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
7193 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
7194 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
7195 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
7196 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
7197 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
7198 (v_int_equiv): Extend to SVE modes.
7199 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
7201 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
7202 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
7203 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
7204 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
7205 (SVE_COND_FP_CMP): New int iterators.
7206 (perm_hilo): Handle the new unpack unspecs.
7207 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
7209 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
7210 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
7211 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
7212 (aarch64_equality_operator, aarch64_constant_vector_operand)
7213 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
7214 (aarch64_sve_nonimmediate_operand): Likewise.
7215 (aarch64_sve_general_operand): Likewise.
7216 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
7217 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
7218 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
7219 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
7220 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
7221 (aarch64_sve_float_arith_immediate): Likewise.
7222 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
7223 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
7224 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
7225 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
7226 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
7227 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
7228 (aarch64_sve_float_arith_operand): Likewise.
7229 (aarch64_sve_float_arith_with_sub_operand): Likewise.
7230 (aarch64_sve_float_mul_operand): Likewise.
7231 (aarch64_sve_vec_perm_operand): Likewise.
7232 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
7233 (aarch64_mov_operand): Accept const_poly_int and const_vector.
7234 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
7235 as well as const_vector.
7236 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
7237 in file. Use CONST0_RTX and CONSTM1_RTX.
7238 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
7239 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
7240 Use aarch64_simd_imm_zero.
7241 * config/aarch64/aarch64-sve.md: New file.
7242 * config/aarch64/aarch64.md: Include it.
7243 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
7244 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
7245 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
7246 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
7247 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
7248 (sve): New attribute.
7249 (enabled): Disable instructions with the sve attribute unless
7251 (movqi, movhi): Pass CONST_POLY_INT operaneds through
7252 aarch64_expand_mov_immediate.
7253 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
7254 CNT[BHSD] immediates.
7255 (movti): Split CONST_POLY_INT moves into two halves.
7256 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
7257 Split additions that need a temporary here if the destination
7258 is the stack pointer.
7259 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
7260 (*add<mode>3_poly_1): New instruction.
7261 (set_clobber_cc): New expander.
7263 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7265 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
7266 parameter and use it instead of GET_MODE_SIZE (innermode). Use
7267 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
7268 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
7269 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
7270 Change innermode from fixed_mode_size to machine_mode.
7271 (simplify_subreg): Update call accordingly. Handle a constant-sized
7272 subreg of a variable-length CONST_VECTOR.
7274 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7275 Alan Hayward <alan.hayward@arm.com>
7276 David Sherwood <david.sherwood@arm.com>
7278 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
7279 (add_offset_to_base): New function, split out from...
7280 (create_mem_ref): ...here. When handling a scale other than 1,
7281 check first whether the address is valid without the offset.
7282 Add it into the base if so, leaving the index and scale as-is.
7284 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7287 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
7288 fold_for_warn before checking if arg2 is INTEGER_CST.
7290 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
7292 * config/rs6000/predicates.md (load_multiple_operation): Delete.
7293 (store_multiple_operation): Delete.
7294 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
7295 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
7296 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
7297 guarded by TARGET_STRING.
7298 (rs6000_output_load_multiple): Delete.
7299 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
7300 OPTION_MASK_STRING / TARGET_STRING handling.
7301 (print_operand) <'N', 'O'>: Add comment that these are unused now.
7302 (const rs6000_opt_masks) <"string">: Change mask to 0.
7303 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
7304 (MASK_STRING): Delete.
7305 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
7307 (load_multiple): Delete.
7314 (store_multiple): Delete.
7321 (movmemsi_8reg): Delete.
7322 (corresponding unnamed define_insn): Delete.
7323 (movmemsi_6reg): Delete.
7324 (corresponding unnamed define_insn): Delete.
7325 (movmemsi_4reg): Delete.
7326 (corresponding unnamed define_insn): Delete.
7327 (movmemsi_2reg): Delete.
7328 (corresponding unnamed define_insn): Delete.
7329 (movmemsi_1reg): Delete.
7330 (corresponding unnamed define_insn): Delete.
7331 * config/rs6000/rs6000.opt (mno-string): New.
7332 (mstring): Replace by deprecation warning stub.
7333 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
7335 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7337 * regrename.c (regrename_do_replace): If replacing the same
7338 reg multiple times, try to reuse last created gen_raw_REG.
7341 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
7342 main to workaround a bug in GDB.
7344 2018-01-12 Tom de Vries <tom@codesourcery.com>
7347 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
7349 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
7351 PR rtl-optimization/80481
7352 * ira-color.c (get_cap_member): New function.
7353 (allocnos_conflict_by_live_ranges_p): Use it.
7354 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
7355 (setup_slot_coalesced_allocno_live_ranges): Ditto.
7357 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
7360 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
7361 (*saddl_se_1): Ditto.
7363 (*ssubl_se_1): Ditto.
7365 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7367 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
7368 rather than wi::to_widest for DR_INITs.
7369 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
7370 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
7371 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
7373 (vect_analyze_group_access_1): Note that here.
7375 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7377 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
7378 polynomial type sizes.
7380 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7382 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
7383 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
7384 (gimple_add_tmp_var): Likewise.
7386 2018-01-12 Martin Liska <mliska@suse.cz>
7388 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
7389 (gimple_alloc_sizes): Likewise.
7390 (dump_gimple_statistics): Use PRIu64 in printf format.
7391 * gimple.h: Change uint64_t to int.
7393 2018-01-12 Martin Liska <mliska@suse.cz>
7395 * tree-core.h: Use uint64_t instead of int.
7396 * tree.c (tree_node_counts): Likewise.
7397 (tree_node_sizes): Likewise.
7398 (dump_tree_statistics): Use PRIu64 in printf format.
7400 2018-01-12 Martin Liska <mliska@suse.cz>
7402 * Makefile.in: As qsort_chk is implemented in vec.c, add
7403 vec.o to linkage of gencfn-macros.
7404 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
7405 passing the info to record_node_allocation_statistics.
7406 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
7408 * ggc-common.c (struct ggc_usage): Add operator== and use
7409 it in operator< and compare function.
7410 * mem-stats.h (struct mem_usage): Likewise.
7411 * vec.c (struct vec_usage): Remove operator< and compare
7412 function. Can be simply inherited.
7414 2018-01-12 Martin Jambor <mjambor@suse.cz>
7417 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
7418 * tree-ssa-math-opts.c: Include domwalk.h.
7419 (convert_mult_to_fma_1): New function.
7420 (fma_transformation_info): New type.
7421 (fma_deferring_state): Likewise.
7422 (cancel_fma_deferring): New function.
7423 (result_of_phi): Likewise.
7424 (last_fma_candidate_feeds_initial_phi): Likewise.
7425 (convert_mult_to_fma): Added deferring logic, split actual
7426 transformation to convert_mult_to_fma_1.
7427 (math_opts_dom_walker): New type.
7428 (math_opts_dom_walker::after_dom_children): New method, body moved
7429 here from pass_optimize_widening_mul::execute, added deferring logic
7431 (pass_optimize_widening_mul::execute): Moved most of code to
7432 math_opts_dom_walker::after_dom_children.
7433 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
7434 * config/i386/i386.c (ix86_option_override_internal): Added
7435 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
7437 2018-01-12 Richard Biener <rguenther@suse.de>
7440 * dwarf2out.c (gen_variable_die): Do not reset old_die for
7441 inline instance vars.
7443 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
7446 * config/rx/rx.c (rx_is_restricted_memory_address):
7449 2018-01-12 Richard Biener <rguenther@suse.de>
7451 PR tree-optimization/80846
7452 * target.def (split_reduction): New target hook.
7453 * targhooks.c (default_split_reduction): New function.
7454 * targhooks.h (default_split_reduction): Declare.
7455 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
7456 target requests first reduce vectors by combining low and high
7458 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
7459 (get_vectype_for_scalar_type_and_size): Export.
7460 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
7461 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
7462 * doc/tm.texi: Regenerate.
7463 * config/i386/i386.c (ix86_split_reduction): Implement
7464 TARGET_VECTORIZE_SPLIT_REDUCTION.
7466 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7469 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
7470 in PIC mode except for TARGET_VXWORKS_RTP.
7471 * config/sparc/sparc.c: Include cfgrtl.h.
7472 (TARGET_INIT_PIC_REG): Define.
7473 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
7474 (sparc_pic_register_p): New predicate.
7475 (sparc_legitimate_address_p): Use it.
7476 (sparc_legitimize_pic_address): Likewise.
7477 (sparc_delegitimize_address): Likewise.
7478 (sparc_mode_dependent_address_p): Likewise.
7479 (gen_load_pcrel_sym): Remove 4th parameter.
7480 (load_got_register): Adjust call to above. Remove obsolete stuff.
7481 (sparc_expand_prologue): Do not call load_got_register here.
7482 (sparc_flat_expand_prologue): Likewise.
7483 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
7484 (sparc_use_pseudo_pic_reg): New function.
7485 (sparc_init_pic_reg): Likewise.
7486 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7487 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
7489 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
7491 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
7492 Add item for branch_cost.
7494 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7496 PR rtl-optimization/83565
7497 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
7498 not extend the result to a larger mode for rotate operations.
7499 (num_sign_bit_copies1): Likewise.
7501 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7504 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
7506 Use values-Xc.o for -pedantic.
7507 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
7509 2018-01-12 Martin Liska <mliska@suse.cz>
7512 * ipa-devirt.c (final_warning_record::grow_type_warnings):
7514 (possible_polymorphic_call_targets): Use it.
7515 (ipa_devirt): Likewise.
7517 2018-01-12 Martin Liska <mliska@suse.cz>
7519 * profile-count.h (enum profile_quality): Use 0 as invalid
7520 enum value of profile_quality.
7522 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
7524 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
7525 -mext-string options.
7527 2018-01-12 Richard Biener <rguenther@suse.de>
7529 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
7530 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
7531 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
7533 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
7535 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
7537 * configure.ac (--with-long-double-format): Add support for the
7538 configuration option to change the default long double format on
7540 * config.gcc (powerpc*-linux*-*): Likewise.
7541 * configure: Regenerate.
7542 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
7543 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
7544 used without modification.
7546 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7548 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
7549 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
7550 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
7551 MISC_BUILTIN_SPEC_BARRIER.
7552 (rs6000_init_builtins): Likewise.
7553 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
7555 (speculation_barrier): New define_insn.
7556 * doc/extend.texi: Document __builtin_speculation_barrier.
7558 2018-01-11 Jakub Jelinek <jakub@redhat.com>
7561 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
7562 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
7563 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
7565 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
7566 integral modes instead of "ss" and "sd".
7567 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
7568 vectors with 32-bit and 64-bit elements.
7569 (vecdupssescalarmodesuffix): New mode attribute.
7570 (vec_dup<mode>): Use it.
7572 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
7575 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
7576 frame if argument is passed on stack.
7578 2018-01-11 Jakub Jelinek <jakub@redhat.com>
7581 * ree.c (combine_reaching_defs): Optimize also
7582 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
7583 reg2=any_extend(exp); reg1=reg2;, formatting fix.
7585 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
7588 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
7590 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
7593 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
7594 after they are computed.
7596 2018-01-11 Bin Cheng <bin.cheng@arm.com>
7598 PR tree-optimization/83695
7599 * gimple-loop-linterchange.cc
7600 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
7601 reset cached scev information after interchange.
7602 (pass_linterchange::execute): Remove call to scev_reset_htab.
7604 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7606 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
7607 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
7608 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
7609 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
7610 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
7611 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
7612 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
7613 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
7614 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
7615 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
7616 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
7617 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
7618 (V_lane_reg): Likewise.
7619 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
7621 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
7622 (vfmal_lane_low<mode>_intrinsic,
7623 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
7624 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
7625 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
7626 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
7627 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
7628 vfmsl_lane_high<mode>_intrinsic): New define_insns.
7630 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7632 * config/arm/arm-cpus.in (fp16fml): New feature.
7633 (ALL_SIMD): Add fp16fml.
7634 (armv8.2-a): Add fp16fml as an option.
7635 (armv8.3-a): Likewise.
7636 (armv8.4-a): Add fp16fml as part of fp16.
7637 * config/arm/arm.h (TARGET_FP16FML): Define.
7638 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
7640 * config/arm/arm-modes.def (V2HF): Define.
7641 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
7642 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
7643 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
7644 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
7645 vfmsl_low, vfmsl_high): New set of builtins.
7646 * config/arm/iterators.md (PLUSMINUS): New code iterator.
7647 (vfml_op): New code attribute.
7648 (VFMLHALVES): New int iterator.
7649 (VFML, VFMLSEL): New mode attributes.
7650 (V_reg): Define mapping for V2HF.
7651 (V_hi, V_lo): New mode attributes.
7652 (VF_constraint): Likewise.
7653 (vfml_half, vfml_half_selector): New int attributes.
7654 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
7656 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
7657 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
7659 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
7660 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
7661 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
7662 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
7664 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
7665 Document new effective target and option set.
7667 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7669 * config/arm/arm-cpus.in (armv8_4): New feature.
7670 (ARMv8_4a): New fgroup.
7671 (armv8.4-a): New arch.
7672 * config/arm/arm-tables.opt: Regenerate.
7673 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
7674 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
7675 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
7676 Add matching rules for -march=armv8.4-a and extensions.
7677 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
7679 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
7682 * config/rx/rx.md (BW): New mode attribute.
7683 (sync_lock_test_and_setsi): Add mode suffix to insn output.
7685 2018-01-11 Richard Biener <rguenther@suse.de>
7687 PR tree-optimization/83435
7688 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
7689 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
7690 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
7692 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7693 Alan Hayward <alan.hayward@arm.com>
7694 David Sherwood <david.sherwood@arm.com>
7696 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
7698 (aarch64_classify_address): Initialize it. Track polynomial offsets.
7699 (aarch64_print_address_internal): Use it to check for a zero offset.
7701 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7702 Alan Hayward <alan.hayward@arm.com>
7703 David Sherwood <david.sherwood@arm.com>
7705 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
7706 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
7707 Return a poly_int64 rather than a HOST_WIDE_INT.
7708 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
7709 rather than a HOST_WIDE_INT.
7710 * config/aarch64/aarch64.h (aarch64_frame): Protect with
7711 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
7712 hard_fp_offset, frame_size, initial_adjust, callee_offset and
7713 final_offset from HOST_WIDE_INT to poly_int64.
7714 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
7715 to_constant when getting the number of units in an Advanced SIMD
7717 (aarch64_builtin_vectorized_function): Check for a constant number
7719 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
7721 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
7722 attribute instead of GET_MODE_NUNITS.
7723 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
7724 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
7725 GET_MODE_SIZE for fixed-size registers.
7726 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
7727 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
7728 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
7729 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
7730 (aarch64_print_operand, aarch64_print_address_internal)
7731 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
7732 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
7733 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
7734 Handle polynomial GET_MODE_SIZE.
7735 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
7736 wider than SImode without modification.
7737 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
7738 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
7739 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
7740 passing and returning SVE modes.
7741 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
7742 rather than GEN_INT.
7743 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
7744 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
7745 (aarch64_allocate_and_probe_stack_space): Likewise.
7746 (aarch64_layout_frame): Cope with polynomial offsets.
7747 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
7748 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
7750 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
7751 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
7752 poly_int64 rather than a HOST_WIDE_INT.
7753 (aarch64_get_separate_components, aarch64_process_components)
7754 (aarch64_expand_prologue, aarch64_expand_epilogue)
7755 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
7756 (aarch64_anchor_offset): New function, split out from...
7757 (aarch64_legitimize_address): ...here.
7758 (aarch64_builtin_vectorization_cost): Handle polynomial
7759 TYPE_VECTOR_SUBPARTS.
7760 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
7762 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
7763 number of elements from the PARALLEL rather than the mode.
7764 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
7765 rather than GET_MODE_BITSIZE.
7766 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
7767 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
7768 (aarch64_expand_vec_perm_const_1): Handle polynomial
7769 d->perm.length () and d->perm elements.
7770 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
7771 Apply to_constant to d->perm elements.
7772 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
7773 polynomial CONST_VECTOR_NUNITS.
7774 (aarch64_move_pointer): Take amount as a poly_int64 rather
7776 (aarch64_progress_pointer): Avoid temporary variable.
7777 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
7778 the mode attribute instead of GET_MODE.
7780 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7781 Alan Hayward <alan.hayward@arm.com>
7782 David Sherwood <david.sherwood@arm.com>
7784 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
7785 x exists before using it.
7786 (aarch64_add_constant_internal): Rename to...
7787 (aarch64_add_offset_1): ...this. Replace regnum with separate
7788 src and dest rtxes. Handle the case in which they're different,
7789 including when the offset is zero. Replace scratchreg with an rtx.
7790 Use 2 additions if there is no spare register into which we can
7791 move a 16-bit constant.
7792 (aarch64_add_constant): Delete.
7793 (aarch64_add_offset): Replace reg with separate src and dest
7794 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
7795 Use aarch64_add_offset_1.
7796 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
7797 an rtx rather than an int. Take the delta as a poly_int64
7798 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
7799 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
7800 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
7801 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
7802 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
7804 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
7805 aarch64_add_constant.
7807 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7809 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
7810 Use scalar_float_mode.
7812 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7814 * config/aarch64/aarch64-simd.md
7815 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
7816 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
7817 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
7818 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
7819 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
7820 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
7821 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
7822 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
7823 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
7824 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
7826 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7829 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
7830 targ_options->x_arm_arch_string is non NULL.
7832 2018-01-11 Tamar Christina <tamar.christina@arm.com>
7834 * config/aarch64/aarch64.h
7835 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
7837 2018-01-11 Sudakshina Das <sudi.das@arm.com>
7840 * expmed.c (emit_store_flag_force): Swap if const op0
7841 and change VOIDmode to mode of op0.
7843 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7845 PR rtl-optimization/83761
7846 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
7847 than bytes to mode_for_size.
7849 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7852 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
7853 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
7856 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7859 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
7860 when in layout mode.
7861 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
7862 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
7865 2018-01-10 Michael Collison <michael.collison@arm.com>
7867 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
7868 * config/aarch64/aarch64-option-extension.def: Add
7869 AARCH64_OPT_EXTENSION of 'fp16fml'.
7870 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7871 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
7872 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
7873 * config/aarch64/constraints.md (Ui7): New constraint.
7874 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
7875 (VFMLA_SEL_W): Ditto.
7878 (VFMLA16_LOW): New int iterator.
7879 (VFMLA16_HIGH): Ditto.
7880 (UNSPEC_FMLAL): New unspec.
7881 (UNSPEC_FMLSL): Ditto.
7882 (UNSPEC_FMLAL2): Ditto.
7883 (UNSPEC_FMLSL2): Ditto.
7884 (f16mac): New code attribute.
7885 * config/aarch64/aarch64-simd-builtins.def
7886 (aarch64_fmlal_lowv2sf): Ditto.
7887 (aarch64_fmlsl_lowv2sf): Ditto.
7888 (aarch64_fmlalq_lowv4sf): Ditto.
7889 (aarch64_fmlslq_lowv4sf): Ditto.
7890 (aarch64_fmlal_highv2sf): Ditto.
7891 (aarch64_fmlsl_highv2sf): Ditto.
7892 (aarch64_fmlalq_highv4sf): Ditto.
7893 (aarch64_fmlslq_highv4sf): Ditto.
7894 (aarch64_fmlal_lane_lowv2sf): Ditto.
7895 (aarch64_fmlsl_lane_lowv2sf): Ditto.
7896 (aarch64_fmlal_laneq_lowv2sf): Ditto.
7897 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
7898 (aarch64_fmlalq_lane_lowv4sf): Ditto.
7899 (aarch64_fmlsl_lane_lowv4sf): Ditto.
7900 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
7901 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
7902 (aarch64_fmlal_lane_highv2sf): Ditto.
7903 (aarch64_fmlsl_lane_highv2sf): Ditto.
7904 (aarch64_fmlal_laneq_highv2sf): Ditto.
7905 (aarch64_fmlsl_laneq_highv2sf): Ditto.
7906 (aarch64_fmlalq_lane_highv4sf): Ditto.
7907 (aarch64_fmlsl_lane_highv4sf): Ditto.
7908 (aarch64_fmlalq_laneq_highv4sf): Ditto.
7909 (aarch64_fmlsl_laneq_highv4sf): Ditto.
7910 * config/aarch64/aarch64-simd.md:
7911 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
7912 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7913 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
7914 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7915 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
7916 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
7917 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
7918 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
7919 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
7920 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
7921 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
7922 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
7923 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
7924 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
7925 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
7926 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
7927 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
7928 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
7929 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
7930 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
7931 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
7932 (vfmlsl_low_u32): Ditto.
7933 (vfmlalq_low_u32): Ditto.
7934 (vfmlslq_low_u32): Ditto.
7935 (vfmlal_high_u32): Ditto.
7936 (vfmlsl_high_u32): Ditto.
7937 (vfmlalq_high_u32): Ditto.
7938 (vfmlslq_high_u32): Ditto.
7939 (vfmlal_lane_low_u32): Ditto.
7940 (vfmlsl_lane_low_u32): Ditto.
7941 (vfmlal_laneq_low_u32): Ditto.
7942 (vfmlsl_laneq_low_u32): Ditto.
7943 (vfmlalq_lane_low_u32): Ditto.
7944 (vfmlslq_lane_low_u32): Ditto.
7945 (vfmlalq_laneq_low_u32): Ditto.
7946 (vfmlslq_laneq_low_u32): Ditto.
7947 (vfmlal_lane_high_u32): Ditto.
7948 (vfmlsl_lane_high_u32): Ditto.
7949 (vfmlal_laneq_high_u32): Ditto.
7950 (vfmlsl_laneq_high_u32): Ditto.
7951 (vfmlalq_lane_high_u32): Ditto.
7952 (vfmlslq_lane_high_u32): Ditto.
7953 (vfmlalq_laneq_high_u32): Ditto.
7954 (vfmlslq_laneq_high_u32): Ditto.
7955 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
7956 (AARCH64_FL_FOR_ARCH8_4): New.
7957 (AARCH64_ISA_F16FML): New ISA flag.
7958 (TARGET_F16FML): New feature flag for fp16fml.
7959 (doc/invoke.texi): Document new fp16fml option.
7961 2018-01-10 Michael Collison <michael.collison@arm.com>
7963 * config/aarch64/aarch64-builtins.c:
7964 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
7965 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7966 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
7967 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
7968 (AARCH64_ISA_SHA3): New ISA flag.
7969 (TARGET_SHA3): New feature flag for sha3.
7970 * config/aarch64/iterators.md (sha512_op): New int attribute.
7971 (CRYPTO_SHA512): New int iterator.
7972 (UNSPEC_SHA512H): New unspec.
7973 (UNSPEC_SHA512H2): Ditto.
7974 (UNSPEC_SHA512SU0): Ditto.
7975 (UNSPEC_SHA512SU1): Ditto.
7976 * config/aarch64/aarch64-simd-builtins.def
7977 (aarch64_crypto_sha512hqv2di): New builtin.
7978 (aarch64_crypto_sha512h2qv2di): Ditto.
7979 (aarch64_crypto_sha512su0qv2di): Ditto.
7980 (aarch64_crypto_sha512su1qv2di): Ditto.
7981 (aarch64_eor3qv8hi): Ditto.
7982 (aarch64_rax1qv2di): Ditto.
7983 (aarch64_xarqv2di): Ditto.
7984 (aarch64_bcaxqv8hi): Ditto.
7985 * config/aarch64/aarch64-simd.md:
7986 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
7987 (aarch64_crypto_sha512su0qv2di): Ditto.
7988 (aarch64_crypto_sha512su1qv2di): Ditto.
7989 (aarch64_eor3qv8hi): Ditto.
7990 (aarch64_rax1qv2di): Ditto.
7991 (aarch64_xarqv2di): Ditto.
7992 (aarch64_bcaxqv8hi): Ditto.
7993 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
7994 (vsha512h2q_u64): Ditto.
7995 (vsha512su0q_u64): Ditto.
7996 (vsha512su1q_u64): Ditto.
7997 (veor3q_u16): Ditto.
7998 (vrax1q_u64): Ditto.
8000 (vbcaxq_u16): Ditto.
8001 * config/arm/types.md (crypto_sha512): New type attribute.
8002 (crypto_sha3): Ditto.
8003 (doc/invoke.texi): Document new sha3 option.
8005 2018-01-10 Michael Collison <michael.collison@arm.com>
8007 * config/aarch64/aarch64-builtins.c:
8008 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
8009 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8010 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
8011 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
8012 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
8013 (AARCH64_ISA_SM4): New ISA flag.
8014 (TARGET_SM4): New feature flag for sm4.
8015 * config/aarch64/aarch64-simd-builtins.def
8016 (aarch64_sm3ss1qv4si): Ditto.
8017 (aarch64_sm3tt1aq4si): Ditto.
8018 (aarch64_sm3tt1bq4si): Ditto.
8019 (aarch64_sm3tt2aq4si): Ditto.
8020 (aarch64_sm3tt2bq4si): Ditto.
8021 (aarch64_sm3partw1qv4si): Ditto.
8022 (aarch64_sm3partw2qv4si): Ditto.
8023 (aarch64_sm4eqv4si): Ditto.
8024 (aarch64_sm4ekeyqv4si): Ditto.
8025 * config/aarch64/aarch64-simd.md:
8026 (aarch64_sm3ss1qv4si): Ditto.
8027 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
8028 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
8029 (aarch64_sm4eqv4si): Ditto.
8030 (aarch64_sm4ekeyqv4si): Ditto.
8031 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
8032 (sm3part_op): Ditto.
8033 (CRYPTO_SM3TT): Ditto.
8034 (CRYPTO_SM3PART): Ditto.
8035 (UNSPEC_SM3SS1): New unspec.
8036 (UNSPEC_SM3TT1A): Ditto.
8037 (UNSPEC_SM3TT1B): Ditto.
8038 (UNSPEC_SM3TT2A): Ditto.
8039 (UNSPEC_SM3TT2B): Ditto.
8040 (UNSPEC_SM3PARTW1): Ditto.
8041 (UNSPEC_SM3PARTW2): Ditto.
8042 (UNSPEC_SM4E): Ditto.
8043 (UNSPEC_SM4EKEY): Ditto.
8044 * config/aarch64/constraints.md (Ui2): New constraint.
8045 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
8046 * config/arm/types.md (crypto_sm3): New type attribute.
8047 (crypto_sm4): Ditto.
8048 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
8049 (vsm3tt1aq_u32): Ditto.
8050 (vsm3tt1bq_u32): Ditto.
8051 (vsm3tt2aq_u32): Ditto.
8052 (vsm3tt2bq_u32): Ditto.
8053 (vsm3partw1q_u32): Ditto.
8054 (vsm3partw2q_u32): Ditto.
8055 (vsm4eq_u32): Ditto.
8056 (vsm4ekeyq_u32): Ditto.
8057 (doc/invoke.texi): Document new sm4 option.
8059 2018-01-10 Michael Collison <michael.collison@arm.com>
8061 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
8062 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
8063 (AARCH64_FL_FOR_ARCH8_4): New.
8064 (AARCH64_FL_V8_4): New flag.
8065 (doc/invoke.texi): Document new armv8.4-a option.
8067 2018-01-10 Michael Collison <michael.collison@arm.com>
8069 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8070 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
8071 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
8072 * config/aarch64/aarch64-option-extension.def: Add
8073 AARCH64_OPT_EXTENSION of 'sha2'.
8074 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
8075 (crypto): Disable sha2 and aes if crypto disabled.
8076 (crypto): Enable aes and sha2 if enabled.
8077 (simd): Disable sha2 and aes if simd disabled.
8078 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
8080 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
8081 (TARGET_SHA2): New feature flag for sha2.
8082 (TARGET_AES): New feature flag for aes.
8083 * config/aarch64/aarch64-simd.md:
8084 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
8085 conditional on TARGET_AES.
8086 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
8087 (aarch64_crypto_sha1hsi): Make pattern conditional
8089 (aarch64_crypto_sha1hv4si): Ditto.
8090 (aarch64_be_crypto_sha1hv4si): Ditto.
8091 (aarch64_crypto_sha1su1v4si): Ditto.
8092 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
8093 (aarch64_crypto_sha1su0v4si): Ditto.
8094 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
8095 (aarch64_crypto_sha256su0v4si): Ditto.
8096 (aarch64_crypto_sha256su1v4si): Ditto.
8097 (doc/invoke.texi): Document new aes and sha2 options.
8099 2018-01-10 Martin Sebor <msebor@redhat.com>
8101 PR tree-optimization/83781
8102 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
8105 2018-01-11 Martin Sebor <msebor@gmail.com>
8106 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8108 PR tree-optimization/83501
8109 PR tree-optimization/81703
8111 * tree-ssa-strlen.c (get_string_cst): Rename...
8112 (get_string_len): ...to this. Handle global constants.
8113 (handle_char_store): Adjust.
8115 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
8116 Jim Wilson <jimw@sifive.com>
8118 * config/riscv/riscv-protos.h (riscv_output_return): New.
8119 * config/riscv/riscv.c (struct machine_function): New naked_p field.
8120 (riscv_attribute_table, riscv_output_return),
8121 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
8122 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
8123 (riscv_compute_frame_info): Only compute frame->mask if not a naked
8125 (riscv_expand_prologue): Add early return for naked function.
8126 (riscv_expand_epilogue): Likewise.
8127 (riscv_function_ok_for_sibcall): Return false for naked function.
8128 (riscv_set_current_function): New.
8129 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
8130 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
8131 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
8132 * doc/extend.texi (RISC-V Function Attributes): New.
8134 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
8136 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
8137 check for 128-bit long double before checking TCmode.
8138 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
8139 128-bit long doubles before checking TFmode or TCmode.
8140 (FLOAT128_IBM_P): Likewise.
8142 2018-01-10 Martin Sebor <msebor@redhat.com>
8144 PR tree-optimization/83671
8145 * builtins.c (c_strlen): Unconditionally return zero for the empty
8147 Use -Warray-bounds for warnings.
8148 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
8149 for non-constant array indices with COMPONENT_REF, arrays of
8150 arrays, and pointers to arrays.
8151 (gimple_fold_builtin_strlen): Determine and set length range for
8152 non-constant character arrays.
8154 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
8157 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
8160 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
8162 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
8164 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8167 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
8168 VECTOR_MEM_ALTIVEC_OR_VSX_P.
8169 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
8170 indexed_or_indirect_operand predicate.
8171 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
8172 (*vsx_le_perm_load_v8hi): Likewise.
8173 (*vsx_le_perm_load_v16qi): Likewise.
8174 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
8175 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
8176 (*vsx_le_perm_store_v8hi): Likewise.
8177 (*vsx_le_perm_store_v16qi): Likewise.
8178 (eight unnamed splitters): Likewise.
8180 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8182 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
8183 * config/rs6000/emmintrin.h: Likewise.
8184 * config/rs6000/mmintrin.h: Likewise.
8185 * config/rs6000/xmmintrin.h: Likewise.
8187 2018-01-10 David Malcolm <dmalcolm@redhat.com>
8190 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
8192 * tree.c (tree_nop_conversion): Return true for location wrapper
8194 (maybe_wrap_with_location): New function.
8195 (selftest::check_strip_nops): New function.
8196 (selftest::test_location_wrappers): New function.
8197 (selftest::tree_c_tests): Call it.
8198 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
8199 (maybe_wrap_with_location): New decl.
8200 (EXPR_LOCATION_WRAPPER_P): New macro.
8201 (location_wrapper_p): New inline function.
8202 (tree_strip_any_location_wrapper): New inline function.
8204 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
8207 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
8208 stack_realign_offset for the largest alignment of stack slot
8210 (ix86_find_max_used_stack_alignment): New function.
8211 (ix86_finalize_stack_frame_flags): Use it. Set
8212 max_used_stack_alignment if we don't realign stack.
8213 * config/i386/i386.h (machine_function): Add
8214 max_used_stack_alignment.
8216 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
8218 * config/arm/arm.opt (-mbranch-cost): New option.
8219 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
8222 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
8225 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
8226 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
8228 2018-01-10 Richard Biener <rguenther@suse.de>
8231 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
8232 early out so it also covers the case where we have a non-NULL
8235 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8237 PR tree-optimization/83753
8238 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
8239 for non-strided grouped accesses if the number of elements is 1.
8241 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8244 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
8245 * i386.h (TARGET_USE_GATHER): Define.
8246 * x86-tune.def (X86_TUNE_USE_GATHER): New.
8248 2018-01-10 Martin Liska <mliska@suse.cz>
8251 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
8252 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
8254 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
8255 CLEANUP_NO_PARTITIONING is not set.
8257 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8259 * doc/rtl.texi: Remove documentation of (const ...) wrappers
8260 for vectors, as a partial revert of r254296.
8261 * rtl.h (const_vec_p): Delete.
8262 (const_vec_duplicate_p): Don't test for vector CONSTs.
8263 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
8264 * expmed.c (make_tree): Likewise.
8267 * common.md (E, F): Use CONSTANT_P instead of checking for
8269 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
8270 checking for CONST_VECTOR.
8272 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8275 * predict.c (force_edge_cold): Handle in more sane way edges
8278 2018-01-09 Carl Love <cel@us.ibm.com>
8280 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
8282 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
8283 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
8284 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
8285 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
8286 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
8287 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
8288 * config/rs6000/rs6000-protos.h: Add extern defition for
8289 rs6000_generate_float2_double_code.
8290 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
8292 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
8293 (float2_v2df): Add define_expand.
8295 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
8298 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
8299 op_mode in the force_to_mode call.
8301 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8303 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
8304 instead of checking each element individually.
8305 (aarch64_evpc_uzp): Likewise.
8306 (aarch64_evpc_zip): Likewise.
8307 (aarch64_evpc_ext): Likewise.
8308 (aarch64_evpc_rev): Likewise.
8309 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
8310 instead of checking each element individually. Return true without
8312 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
8313 whether all selected elements come from the same input, instead of
8314 checking each element individually. Remove calls to gen_rtx_REG,
8315 start_sequence and end_sequence and instead assert that no rtl is
8318 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8320 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
8321 order of HIGH and CONST checks.
8323 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8325 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
8326 if the destination isn't an SSA_NAME.
8328 2018-01-09 Richard Biener <rguenther@suse.de>
8330 PR tree-optimization/83668
8331 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
8333 (canonicalize_loop_form): ... here, renamed from ...
8334 (canonicalize_loop_closed_ssa_form): ... this and amended to
8335 swap successor edges for loop exit blocks to make us use
8336 the RPO order we need for initial schedule generation.
8338 2018-01-09 Joseph Myers <joseph@codesourcery.com>
8340 PR tree-optimization/64811
8341 * match.pd: When optimizing comparisons with Inf, avoid
8342 introducing or losing exceptions from comparisons with NaN.
8344 2018-01-09 Martin Liska <mliska@suse.cz>
8347 * asan.c (shadow_mem_size): Add gcc_assert.
8349 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
8351 Don't save registers in main().
8354 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
8355 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
8356 * config/avr/avr.c (avr_set_current_function): Don't error if
8357 naked, OS_task or OS_main are specified at the same time.
8358 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
8360 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
8362 * common/config/avr/avr-common.c (avr_option_optimization_table):
8363 Switch on -mmain-is-OS_task for optimizing compilations.
8365 2018-01-09 Richard Biener <rguenther@suse.de>
8367 PR tree-optimization/83572
8368 * graphite.c: Include cfganal.h.
8369 (graphite_transform_loops): Connect infinite loops to exit
8370 and remove fake edges at the end.
8372 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8374 * ipa-inline.c (edge_badness): Revert accidental checkin.
8376 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8379 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
8380 symbols; not inline clones.
8382 2018-01-09 Jakub Jelinek <jakub@redhat.com>
8385 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
8386 hard registers. Formatting fixes.
8388 PR preprocessor/83722
8389 * gcc.c (try_generate_repro): Pass
8390 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
8391 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
8394 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
8395 Kito Cheng <kito.cheng@gmail.com>
8397 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
8398 (riscv_leaf_function_p): Delete.
8399 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
8401 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8403 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
8405 (do_ifelse): New function.
8406 (do_isel): New function.
8407 (do_sub3): New function.
8408 (do_add3): New function.
8409 (do_load_mask_compare): New function.
8410 (do_overlap_load_compare): New function.
8411 (expand_compare_loop): New function.
8412 (expand_block_compare): Call expand_compare_loop() when appropriate.
8413 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
8415 (-mblock-compare-inline-loop-limit): New option.
8417 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8420 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
8421 Reverse order of second and third operands in first alternative.
8422 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
8423 of first and second elements in UNSPEC_VPERMR vector.
8424 (altivec_expand_vec_perm_le): Likewise.
8426 2018-01-08 Jeff Law <law@redhat.com>
8428 PR rtl-optimizatin/81308
8429 * tree-switch-conversion.c (cfg_altered): New file scoped static.
8430 (process_switch): If group_case_labels makes a change, then set
8432 (pass_convert_switch::execute): If a switch is converted, then
8433 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
8435 PR rtl-optimization/81308
8436 * recog.c (split_all_insns): Conditionally cleanup the CFG after
8439 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
8441 PR target/83663 - Revert r255946
8442 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
8443 generation for cases where splatting a value is not useful.
8444 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
8445 across a vec_duplicate and a paradoxical subreg forming a vector
8446 mode to a vec_concat.
8448 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8450 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
8451 -march=armv8.3-a variants.
8452 * config/arm/t-multilib: Likewise.
8453 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
8455 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8457 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
8459 (cceq_ior_compare_complement): Give it a name so I can use it, and
8460 change boolean_or_operator predicate to boolean_operator so it can
8461 be used to generate a crand.
8462 (eqne): New code iterator.
8463 (bd/bd_neg): New code_attrs.
8464 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
8465 a single define_insn.
8466 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
8467 decrement (bdnzt/bdnzf/bdzt/bdzf).
8468 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
8469 with the new names of the branch decrement patterns, and added the
8470 names of the branch decrement conditional patterns.
8472 2018-01-08 Richard Biener <rguenther@suse.de>
8474 PR tree-optimization/83563
8475 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
8478 2018-01-08 Richard Biener <rguenther@suse.de>
8481 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
8483 2018-01-08 Richard Biener <rguenther@suse.de>
8485 PR tree-optimization/83685
8486 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
8487 references to abnormals.
8489 2018-01-08 Richard Biener <rguenther@suse.de>
8492 * dwarf2out.c (output_indirect_strings): Handle empty
8493 skeleton_debug_str_hash.
8494 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
8496 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
8498 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
8499 (emit_store_direct): Likewise.
8500 (arc_trampoline_adjust_address): Likewise.
8501 (arc_asm_trampoline_template): New function.
8502 (arc_initialize_trampoline): Use asm_trampoline_template.
8503 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
8504 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
8505 * config/arc/arc.md (flush_icache): Delete pattern.
8507 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
8509 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
8510 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
8513 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
8516 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
8517 by not USED_FOR_TARGET.
8518 (make_pass_resolve_sw_modes): Likewise.
8520 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
8522 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
8525 2018-01-08 Richard Biener <rguenther@suse.de>
8528 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
8530 2018-01-08 Richard Biener <rguenther@suse.de>
8533 * match.pd ((t * 2) / 2) -> t): Add missing :c.
8535 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
8538 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
8539 basic blocks with a small number of successors.
8540 (convert_control_dep_chain_into_preds): Improve handling of
8542 (dump_predicates): Split apart into...
8543 (dump_pred_chain): ...here...
8544 (dump_pred_info): ...and here.
8545 (can_one_predicate_be_invalidated_p): Add debugging printfs.
8546 (can_chain_union_be_invalidated_p): Improve check for invalidation
8548 (uninit_uses_cannot_happen): Avoid unnecessary if
8549 convert_control_dep_chain_into_preds yielded nothing.
8551 2018-01-06 Martin Sebor <msebor@redhat.com>
8553 PR tree-optimization/83640
8554 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
8555 subtracting negative offset from size.
8556 (builtin_access::overlap): Adjust offset bounds of the access to fall
8557 within the size of the object if possible.
8559 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
8561 PR rtl-optimization/83699
8562 * expmed.c (extract_bit_field_1): Restrict the vector usage of
8563 extract_bit_field_as_subreg to cases in which the extracted
8564 value is also a vector.
8566 * lra-constraints.c (process_alt_operands): Test for the equivalence
8567 substitutions when detecting a possible reload cycle.
8569 2018-01-06 Jakub Jelinek <jakub@redhat.com>
8572 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
8573 by default if flag_selective_schedling{,2}. Formatting fixes.
8575 PR rtl-optimization/83682
8576 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
8577 if it has non-VECTOR_MODE element mode.
8578 (vec_duplicate_p): Likewise.
8581 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
8582 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
8584 2018-01-05 Jakub Jelinek <jakub@redhat.com>
8587 * config/i386/i386-builtin.def
8588 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
8589 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
8590 Require also OPTION_MASK_ISA_AVX512F in addition to
8591 OPTION_MASK_ISA_GFNI.
8592 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
8593 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
8594 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
8595 to OPTION_MASK_ISA_GFNI.
8596 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
8597 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
8598 OPTION_MASK_ISA_AVX512BW.
8599 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
8600 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
8601 addition to OPTION_MASK_ISA_GFNI.
8602 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
8603 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
8604 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
8605 to OPTION_MASK_ISA_GFNI.
8606 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
8607 a requirement for all ISAs rather than any of them with a few
8609 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
8611 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
8612 bitmasks to be enabled with 3 exceptions, instead of requiring any
8613 enabled ISA with lots of exceptions.
8614 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
8615 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
8616 Change avx512bw in isa attribute to avx512f.
8617 * config/i386/sgxintrin.h: Add license boilerplate.
8618 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
8619 to __AVX512F__ and __AVX512VL to __AVX512VL__.
8620 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
8621 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
8623 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
8624 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
8625 temporarily sse2 rather than sse if not enabled already.
8628 * config/i386/sse.md (VI248_VLBW): Rename to ...
8629 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
8630 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
8631 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
8632 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
8633 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
8634 mode iterator instead of VI248_VLBW.
8636 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
8638 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
8639 (record_modified): Skip clobbers; add debug output.
8640 (param_change_prob): Use sreal frequencies.
8642 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8644 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
8645 punt for user-aligned variables.
8647 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8649 * tree-chrec.c (chrec_contains_symbols): Return true for
8652 2018-01-05 Sudakshina Das <sudi.das@arm.com>
8655 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
8656 of (x|y) == x for BICS pattern.
8658 2018-01-05 Jakub Jelinek <jakub@redhat.com>
8660 PR tree-optimization/83605
8661 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
8662 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
8665 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
8667 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
8668 * config/epiphany/rtems.h: New file.
8670 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8671 Uros Bizjak <ubizjak@gmail.com>
8674 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
8675 QIreg_operand instead of register_operand predicate.
8676 * config/i386/i386.c (ix86_rop_should_change_byte_p,
8677 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
8678 comments instead of -fmitigate[-_]rop.
8680 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8683 * cgraphunit.c (symbol_table::compile): Switch to text_section
8684 before calling assembly_start debug hook.
8685 * run-rtl-passes.c (run_rtl_passes): Likewise.
8688 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8690 * tree-vrp.c (extract_range_from_binary_expr_1): Check
8691 range_int_cst_p rather than !symbolic_range_p before calling
8692 extract_range_from_multiplicative_op_1.
8694 2018-01-04 Jeff Law <law@redhat.com>
8696 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
8697 redundant test in assertion.
8699 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8701 * doc/rtl.texi: Document machine_mode wrapper classes.
8703 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8705 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
8708 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8710 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
8711 the VEC_PERM_EXPR fold to fail.
8713 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8716 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
8717 to switched_sections.
8719 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8722 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
8725 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
8728 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
8729 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
8731 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8734 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
8735 is BLKmode and bitpos not zero or mode change is needed.
8737 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8740 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
8743 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
8746 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
8747 instead of MULT rtx. Update all corresponding splitters.
8749 (*ssub<modesuffix>): Ditto.
8751 (*cmp_sadd_di): Update split patterns.
8752 (*cmp_sadd_si): Ditto.
8753 (*cmp_sadd_sidi): Ditto.
8754 (*cmp_ssub_di): Ditto.
8755 (*cmp_ssub_si): Ditto.
8756 (*cmp_ssub_sidi): Ditto.
8757 * config/alpha/predicates.md (const23_operand): New predicate.
8758 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
8759 Look for ASHIFT, not MULT inner operand.
8760 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
8762 2018-01-04 Martin Liska <mliska@suse.cz>
8764 PR gcov-profile/83669
8765 * gcov.c (output_intermediate_file): Add version to intermediate
8767 * doc/gcov.texi: Document new field 'version' in intermediate
8768 file format. Fix location of '-k' option of gcov command.
8770 2018-01-04 Martin Liska <mliska@suse.cz>
8773 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
8775 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8777 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
8779 2018-01-03 Martin Sebor <msebor@redhat.com>
8781 PR tree-optimization/83655
8782 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
8783 checking calls with invalid arguments.
8785 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8787 * tree-vect-stmts.c (vect_get_store_rhs): New function.
8788 (vectorizable_mask_load_store): Delete.
8789 (vectorizable_call): Return false for masked loads and stores.
8790 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
8791 instead of gimple_assign_rhs1.
8792 (vectorizable_load): Handle IFN_MASK_LOAD.
8793 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
8795 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8797 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
8799 (vectorizable_mask_load_store): ...here.
8800 (vectorizable_load): ...and here.
8802 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8804 * tree-vect-stmts.c (vect_build_all_ones_mask)
8805 (vect_build_zero_merge_argument): New functions, split out from...
8806 (vectorizable_load): ...here.
8808 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8810 * tree-vect-stmts.c (vect_check_store_rhs): New function,
8812 (vectorizable_mask_load_store): ...here.
8813 (vectorizable_store): ...and here.
8815 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8817 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
8819 (vectorizable_mask_load_store): ...here.
8821 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8823 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
8824 (vect_model_store_cost): Take a vec_load_store_type instead of a
8826 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
8827 (vect_model_store_cost): Take a vec_load_store_type instead of a
8829 (vectorizable_mask_load_store): Update accordingly.
8830 (vectorizable_store): Likewise.
8831 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
8833 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8835 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
8836 IFN_MASK_LOAD calls here rather than...
8837 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
8839 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8840 Alan Hayward <alan.hayward@arm.com>
8841 David Sherwood <david.sherwood@arm.com>
8843 * expmed.c (extract_bit_field_1): For vector extracts,
8844 fall back to extract_bit_field_as_subreg if vec_extract
8847 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8848 Alan Hayward <alan.hayward@arm.com>
8849 David Sherwood <david.sherwood@arm.com>
8851 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
8852 they are variable or constant sized.
8853 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
8854 slots for constant-sized data.
8856 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8857 Alan Hayward <alan.hayward@arm.com>
8858 David Sherwood <david.sherwood@arm.com>
8860 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
8861 handling COND_EXPRs with boolean comparisons, try to find a better
8862 basis for the mask type than the boolean itself.
8864 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8866 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
8867 is calculated and how it can be overridden.
8868 * genmodes.c (max_bitsize_mode_any_mode): New variable.
8869 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
8871 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
8874 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8875 Alan Hayward <alan.hayward@arm.com>
8876 David Sherwood <david.sherwood@arm.com>
8878 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
8879 Remove the mode argument.
8880 (aarch64_simd_valid_immediate): Remove the mode and inverse
8882 * config/aarch64/iterators.md (bitsize): New iterator.
8883 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
8884 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
8885 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
8886 aarch64_simd_valid_immediate.
8887 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
8888 (aarch64_reg_or_bic_imm): Likewise.
8889 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
8890 with an insn_type enum and msl with a modifier_type enum.
8891 Replace element_width with a scalar_mode. Change the shift
8892 to unsigned int. Add constructors for scalar_float_mode and
8893 scalar_int_mode elements.
8894 (aarch64_vect_float_const_representable_p): Delete.
8895 (aarch64_can_const_movi_rtx_p)
8896 (aarch64_simd_scalar_immediate_valid_for_move)
8897 (aarch64_simd_make_constant): Update call to
8898 aarch64_simd_valid_immediate.
8899 (aarch64_advsimd_valid_immediate_hs): New function.
8900 (aarch64_advsimd_valid_immediate): Likewise.
8901 (aarch64_simd_valid_immediate): Remove mode and inverse
8902 arguments. Rewrite to use the above. Use const_vec_duplicate_p
8903 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
8904 and aarch64_float_const_representable_p on the result.
8905 (aarch64_output_simd_mov_immediate): Remove mode argument.
8906 Update call to aarch64_simd_valid_immediate and use of
8907 simd_immediate_info.
8908 (aarch64_output_scalar_simd_mov_immediate): Update call
8911 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8912 Alan Hayward <alan.hayward@arm.com>
8913 David Sherwood <david.sherwood@arm.com>
8915 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
8916 (mode_nunits): Likewise CONST_MODE_NUNITS.
8917 * machmode.def (ADJUST_NUNITS): Document.
8918 * genmodes.c (mode_data::need_nunits_adj): New field.
8919 (blank_mode): Update accordingly.
8920 (adj_nunits): New variable.
8921 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
8923 (emit_mode_size_inline): Set need_bytesize_adj for all modes
8924 listed in adj_nunits.
8925 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
8926 listed in adj_nunits. Don't emit case statements for such modes.
8927 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
8928 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
8929 nothing if adj_nunits is nonnull.
8930 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
8931 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
8932 (emit_mode_fbit): Update use of print_maybe_const_decl.
8933 (emit_move_size): Likewise. Treat the array as non-const
8935 (emit_mode_adjustments): Handle adj_nunits.
8937 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8939 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
8940 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
8941 (VECTOR_MODES): Use it.
8942 (make_vector_modes): Take the prefix as an argument.
8944 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8945 Alan Hayward <alan.hayward@arm.com>
8946 David Sherwood <david.sherwood@arm.com>
8948 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
8949 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
8950 for MODE_VECTOR_BOOL.
8951 * machmode.def (VECTOR_BOOL_MODE): Document.
8952 * genmodes.c (VECTOR_BOOL_MODE): New macro.
8953 (make_vector_bool_mode): New function.
8954 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
8956 * lto-streamer-in.c (lto_input_mode_table): Likewise.
8957 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
8959 * stor-layout.c (int_mode_for_mode): Likewise.
8960 * tree.c (build_vector_type_for_mode): Likewise.
8961 * varasm.c (output_constant_pool_2): Likewise.
8962 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
8963 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
8964 for MODE_VECTOR_BOOL.
8965 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
8966 of mode class checks.
8967 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
8968 instead of a list of mode class checks.
8969 (expand_vector_scalar_condition): Likewise.
8970 (type_for_widest_vector_mode): Handle BImode as an inner mode.
8972 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8973 Alan Hayward <alan.hayward@arm.com>
8974 David Sherwood <david.sherwood@arm.com>
8976 * machmode.h (mode_size): Change from unsigned short to
8978 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
8979 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8980 or if measurement_type is not polynomial.
8981 (fixed_size_mode::includes_p): Check for constant-sized modes.
8982 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
8983 return a poly_uint16 rather than an unsigned short.
8984 (emit_mode_size): Change the type of mode_size from unsigned short
8985 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
8986 (emit_mode_adjustments): Cope with polynomial vector sizes.
8987 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8989 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8991 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
8992 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
8993 * caller-save.c (setup_save_areas): Likewise.
8994 (replace_reg_with_saved_mem): Likewise.
8995 * calls.c (emit_library_call_value_1): Likewise.
8996 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
8997 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
8998 (gen_lowpart_for_combine): Likewise.
8999 * convert.c (convert_to_integer_1): Likewise.
9000 * cse.c (equiv_constant, cse_insn): Likewise.
9001 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
9002 (cselib_subst_to_values): Likewise.
9003 * dce.c (word_dce_process_block): Likewise.
9004 * df-problems.c (df_word_lr_mark_ref): Likewise.
9005 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
9006 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
9007 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
9008 (rtl_for_decl_location): Likewise.
9009 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
9010 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
9011 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
9012 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
9013 (expand_expr_real_1): Likewise.
9014 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
9015 (pad_below): Likewise.
9016 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9017 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
9018 * ira.c (get_subreg_tracking_sizes): Likewise.
9019 * ira-build.c (ira_create_allocno_objects): Likewise.
9020 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
9021 (ira_sort_regnos_for_alter_reg): Likewise.
9022 * ira-costs.c (record_operand_costs): Likewise.
9023 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
9024 (resolve_simple_move): Likewise.
9025 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
9026 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
9027 (lra_constraints): Likewise.
9028 (CONST_POOL_OK_P): Reject variable-sized modes.
9029 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
9030 (add_pseudo_to_slot, lra_spill): Likewise.
9031 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9032 * optabs-query.c (get_best_extraction_insn): Likewise.
9033 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9034 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
9035 (expand_mult_highpart, valid_multiword_target_p): Likewise.
9036 * recog.c (offsettable_address_addr_space_p): Likewise.
9037 * regcprop.c (maybe_mode_change): Likewise.
9038 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
9039 * regrename.c (build_def_use): Likewise.
9040 * regstat.c (dump_reg_info): Likewise.
9041 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
9042 (find_reloads, find_reloads_subreg_address): Likewise.
9043 * reload1.c (eliminate_regs_1): Likewise.
9044 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
9045 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
9046 (simplify_binary_operation_1, simplify_subreg): Likewise.
9047 * targhooks.c (default_function_arg_padding): Likewise.
9048 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
9049 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
9050 (verify_gimple_assign_ternary): Likewise.
9051 * tree-inline.c (estimate_move_cost): Likewise.
9052 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9053 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
9054 (get_address_cost_ainc): Likewise.
9055 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
9056 (vect_supportable_dr_alignment): Likewise.
9057 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9058 (vectorizable_reduction): Likewise.
9059 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
9060 (vectorizable_operation, vectorizable_load): Likewise.
9061 * tree.c (build_same_sized_truth_vector_type): Likewise.
9062 * valtrack.c (cleanup_auto_inc_dec): Likewise.
9063 * var-tracking.c (emit_note_insn_var_location): Likewise.
9064 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
9065 (ADDR_VEC_ALIGN): Likewise.
9067 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9068 Alan Hayward <alan.hayward@arm.com>
9069 David Sherwood <david.sherwood@arm.com>
9071 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
9073 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9074 or if measurement_type is polynomial.
9075 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
9076 * combine.c (make_extraction): Likewise.
9077 * dse.c (find_shift_sequence): Likewise.
9078 * dwarf2out.c (mem_loc_descriptor): Likewise.
9079 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
9080 (extract_bit_field, extract_low_bits): Likewise.
9081 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
9082 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
9083 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
9084 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
9085 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9086 * reload.c (find_reloads): Likewise.
9087 * reload1.c (alter_reg): Likewise.
9088 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
9089 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
9090 * tree-if-conv.c (predicate_mem_writes): Likewise.
9091 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
9092 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
9093 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
9094 * valtrack.c (dead_debug_insert_temp): Likewise.
9095 * varasm.c (mergeable_constant_section): Likewise.
9096 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
9098 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9099 Alan Hayward <alan.hayward@arm.com>
9100 David Sherwood <david.sherwood@arm.com>
9102 * expr.c (expand_assignment): Cope with polynomial mode sizes
9103 when assigning to a CONCAT.
9105 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9106 Alan Hayward <alan.hayward@arm.com>
9107 David Sherwood <david.sherwood@arm.com>
9109 * machmode.h (mode_precision): Change from unsigned short to
9111 (mode_to_precision): Return a poly_uint16 rather than an unsigned
9113 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
9114 or if measurement_type is not polynomial.
9115 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
9116 in which the mode is already known to be a scalar_int_mode.
9117 * genmodes.c (emit_mode_precision): Change the type of mode_precision
9118 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
9120 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9121 for GET_MODE_PRECISION.
9122 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9123 for GET_MODE_PRECISION.
9124 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
9126 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
9127 (expand_field_assignment, make_extraction): Likewise.
9128 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
9129 (get_last_value): Likewise.
9130 * convert.c (convert_to_integer_1): Likewise.
9131 * cse.c (cse_insn): Likewise.
9132 * expr.c (expand_expr_real_1): Likewise.
9133 * lra-constraints.c (simplify_operand_subreg): Likewise.
9134 * optabs-query.c (can_atomic_load_p): Likewise.
9135 * optabs.c (expand_atomic_load): Likewise.
9136 (expand_atomic_store): Likewise.
9137 * ree.c (combine_reaching_defs): Likewise.
9138 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
9139 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
9140 * tree.h (type_has_mode_precision_p): Likewise.
9141 * ubsan.c (instrument_si_overflow): Likewise.
9143 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9144 Alan Hayward <alan.hayward@arm.com>
9145 David Sherwood <david.sherwood@arm.com>
9147 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
9148 polynomial numbers of units.
9149 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
9150 (valid_vector_subparts_p): New function.
9151 (build_vector_type): Remove temporary shim and take the number
9152 of units as a poly_uint64 rather than an int.
9153 (build_opaque_vector_type): Take the number of units as a
9154 poly_uint64 rather than an int.
9155 * tree.c (build_vector_from_ctor): Handle polynomial
9156 TYPE_VECTOR_SUBPARTS.
9157 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
9158 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
9159 (build_vector_from_val): If the number of units is variable,
9160 use build_vec_duplicate_cst for constant operands and
9161 VEC_DUPLICATE_EXPR otherwise.
9162 (make_vector_type): Remove temporary is_constant ().
9163 (build_vector_type, build_opaque_vector_type): Take the number of
9164 units as a poly_uint64 rather than an int.
9165 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
9167 * cfgexpand.c (expand_debug_expr): Likewise.
9168 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
9169 (store_constructor, expand_expr_real_1): Likewise.
9170 (const_scalar_mask_from_tree): Likewise.
9171 * fold-const-call.c (fold_const_reduction): Likewise.
9172 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
9173 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
9174 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
9175 (fold_relational_const): Likewise.
9176 (native_interpret_vector): Likewise. Change the size from an
9177 int to an unsigned int.
9178 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
9179 TYPE_VECTOR_SUBPARTS.
9180 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
9181 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
9182 duplicating a non-constant operand into a variable-length vector.
9183 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
9184 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
9185 * ipa-icf.c (sem_variable::equals): Likewise.
9186 * match.pd: Likewise.
9187 * omp-simd-clone.c (simd_clone_subparts): Likewise.
9188 * print-tree.c (print_node): Likewise.
9189 * stor-layout.c (layout_type): Likewise.
9190 * targhooks.c (default_builtin_vectorization_cost): Likewise.
9191 * tree-cfg.c (verify_gimple_comparison): Likewise.
9192 (verify_gimple_assign_binary): Likewise.
9193 (verify_gimple_assign_ternary): Likewise.
9194 (verify_gimple_assign_single): Likewise.
9195 * tree-pretty-print.c (dump_generic_node): Likewise.
9196 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9197 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
9198 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
9199 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
9200 (vect_shift_permute_load_chain): Likewise.
9201 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
9202 (expand_vector_condition, optimize_vector_constructor): Likewise.
9203 (lower_vec_perm, get_compute_type): Likewise.
9204 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9205 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
9206 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
9207 (vect_recog_mask_conversion_pattern): Likewise.
9208 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
9209 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
9210 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9211 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
9212 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
9213 (vectorizable_shift, vectorizable_operation, vectorizable_store)
9214 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
9215 (supportable_widening_operation): Likewise.
9216 (supportable_narrowing_operation): Likewise.
9217 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
9219 * varasm.c (output_constant): Likewise.
9221 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9222 Alan Hayward <alan.hayward@arm.com>
9223 David Sherwood <david.sherwood@arm.com>
9225 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
9226 so that both the length == 3 and length != 3 cases set up their
9227 own permute vectors. Add comments explaining why we know the
9228 number of elements is constant.
9229 (vect_permute_load_chain): Likewise.
9231 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9232 Alan Hayward <alan.hayward@arm.com>
9233 David Sherwood <david.sherwood@arm.com>
9235 * machmode.h (mode_nunits): Change from unsigned char to
9237 (ONLY_FIXED_SIZE_MODES): New macro.
9238 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
9239 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
9240 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
9242 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
9243 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
9244 or if measurement_type is not polynomial.
9245 * genmodes.c (ZERO_COEFFS): New macro.
9246 (emit_mode_nunits_inline): Make mode_nunits_inline return a
9248 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
9249 Use ZERO_COEFFS when emitting initializers.
9250 * data-streamer.h (bp_pack_poly_value): New function.
9251 (bp_unpack_poly_value): Likewise.
9252 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9253 for GET_MODE_NUNITS.
9254 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9255 for GET_MODE_NUNITS.
9256 * tree.c (make_vector_type): Remove temporary shim and make
9257 the real function take the number of units as a poly_uint64
9259 (build_vector_type_for_mode): Handle polynomial nunits.
9260 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
9261 * emit-rtl.c (const_vec_series_p_1): Likewise.
9262 (gen_rtx_CONST_VECTOR): Likewise.
9263 * fold-const.c (test_vec_duplicate_folding): Likewise.
9264 * genrecog.c (validate_pattern): Likewise.
9265 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
9266 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9267 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
9268 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
9269 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
9270 * rtlanal.c (subreg_get_info): Likewise.
9271 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9272 (vect_grouped_load_supported): Likewise.
9273 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
9274 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
9275 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
9276 (simplify_const_unary_operation, simplify_binary_operation_1)
9277 (simplify_const_binary_operation, simplify_ternary_operation)
9278 (test_vector_ops_duplicate, test_vector_ops): Likewise.
9279 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
9280 instead of CONST_VECTOR_NUNITS.
9281 * varasm.c (output_constant_pool_2): Likewise.
9282 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
9283 explicit-encoded elements in the XVEC for variable-length vectors.
9285 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9287 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
9289 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9290 Alan Hayward <alan.hayward@arm.com>
9291 David Sherwood <david.sherwood@arm.com>
9293 * coretypes.h (fixed_size_mode): Declare.
9294 (fixed_size_mode_pod): New typedef.
9295 * builtins.h (target_builtins::x_apply_args_mode)
9296 (target_builtins::x_apply_result_mode): Change type to
9297 fixed_size_mode_pod.
9298 * builtins.c (apply_args_size, apply_result_size, result_vector)
9299 (expand_builtin_apply_args_1, expand_builtin_apply)
9300 (expand_builtin_return): Update accordingly.
9302 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9304 * cse.c (hash_rtx_cb): Hash only the encoded elements.
9305 * cselib.c (cselib_hash_rtx): Likewise.
9306 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
9307 CONST_VECTOR encoding.
9309 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9310 Jeff Law <law@redhat.com>
9313 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
9314 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
9315 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
9316 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
9319 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
9320 explicitly probe *sp in a noreturn function if there were any callee
9321 register saves or frame pointer is needed.
9323 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9326 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
9327 BLKmode for ternary, binary or unary expressions.
9330 * var-tracking.c (delete_vta_debug_insn): New inline function.
9331 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
9332 insns from get_insns () to NULL instead of each bb separately.
9333 Use delete_vta_debug_insn. No longer static.
9334 (vt_debug_insns_local, variable_tracking_main_1): Adjust
9335 delete_vta_debug_insns callers.
9336 * rtl.h (delete_vta_debug_insns): Declare.
9337 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
9338 instead of variable_tracking_main.
9340 2018-01-03 Martin Sebor <msebor@redhat.com>
9342 PR tree-optimization/83603
9343 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
9344 arguments past the endof the argument list in functions declared
9345 without a prototype.
9346 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
9347 Avoid checking when arguments are null.
9349 2018-01-03 Martin Sebor <msebor@redhat.com>
9352 * doc/extend.texi (attribute const): Fix a typo.
9353 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
9354 issuing -Wsuggest-attribute for void functions.
9356 2018-01-03 Martin Sebor <msebor@redhat.com>
9358 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
9359 offset_int::from instead of wide_int::to_shwi.
9360 (maybe_diag_overlap): Remove assertion.
9361 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
9362 * gimple-ssa-sprintf.c (format_directive): Same.
9363 (parse_directive): Same.
9364 (sprintf_dom_walker::compute_format_length): Same.
9365 (try_substitute_return_value): Same.
9367 2018-01-03 Jeff Law <law@redhat.com>
9370 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
9371 non-constant residual for zero at runtime and avoid probing in
9372 that case. Reorganize code for trailing problem to mirror handling
9375 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9377 PR tree-optimization/83501
9378 * tree-ssa-strlen.c (get_string_cst): New.
9379 (handle_char_store): Call get_string_cst.
9381 2018-01-03 Martin Liska <mliska@suse.cz>
9383 PR tree-optimization/83593
9384 * tree-ssa-strlen.c: Include tree-cfg.h.
9385 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
9386 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
9387 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
9389 (strlen_dom_walker::before_dom_children): Call
9390 gimple_purge_dead_eh_edges. Dump tranformation with details
9392 (strlen_dom_walker::before_dom_children): Update call by adding
9393 new argument cleanup_eh.
9394 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
9396 2018-01-03 Martin Liska <mliska@suse.cz>
9399 * cif-code.def (VARIADIC_THUNK): New enum value.
9400 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
9403 2018-01-03 Jan Beulich <jbeulich@suse.com>
9405 * sse.md (mov<mode>_internal): Tighten condition for when to use
9406 vmovdqu<ssescalarsize> for TI and OI modes.
9408 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9410 Update copyright years.
9412 2018-01-03 Martin Liska <mliska@suse.cz>
9415 * ipa-visibility.c (function_and_variable_visibility): Skip
9416 functions with noipa attribure.
9418 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9420 * gcc.c (process_command): Update copyright notice dates.
9421 * gcov-dump.c (print_version): Ditto.
9422 * gcov.c (print_version): Ditto.
9423 * gcov-tool.c (print_version): Ditto.
9424 * gengtype.c (create_file): Ditto.
9425 * doc/cpp.texi: Bump @copying's copyright year.
9426 * doc/cppinternals.texi: Ditto.
9427 * doc/gcc.texi: Ditto.
9428 * doc/gccint.texi: Ditto.
9429 * doc/gcov.texi: Ditto.
9430 * doc/install.texi: Ditto.
9431 * doc/invoke.texi: Ditto.
9433 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9435 * vector-builder.h (vector_builder::m_full_nelts): Change from
9436 unsigned int to poly_uint64.
9437 (vector_builder::full_nelts): Update prototype accordingly.
9438 (vector_builder::new_vector): Likewise.
9439 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
9440 (vector_builder::operator ==): Likewise.
9441 (vector_builder::finalize): Likewise.
9442 * int-vector-builder.h (int_vector_builder::int_vector_builder):
9443 Take the number of elements as a poly_uint64 rather than an
9445 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
9446 from unsigned int to poly_uint64.
9447 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
9448 (vec_perm_indices::new_vector): Likewise.
9449 (vec_perm_indices::length): Likewise.
9450 (vec_perm_indices::nelts_per_input): Likewise.
9451 (vec_perm_indices::input_nelts): Likewise.
9452 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
9453 number of elements per input as a poly_uint64 rather than an
9454 unsigned int. Use the original encoding for variable-length
9455 vectors, rather than clamping each individual element.
9456 For the second and subsequent elements in each pattern,
9457 clamp the step and base before clamping their sum.
9458 (vec_perm_indices::series_p): Handle polynomial element counts.
9459 (vec_perm_indices::all_in_range_p): Likewise.
9460 (vec_perm_indices_to_tree): Likewise.
9461 (vec_perm_indices_to_rtx): Likewise.
9462 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
9463 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
9464 (tree_vector_builder::new_binary_operation): Handle polynomial
9465 element counts. Return false if we need to know the number
9466 of elements at compile time.
9467 * fold-const.c (fold_vec_perm): Punt if the number of elements
9468 isn't known at compile time.
9470 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9472 * vec-perm-indices.h (vec_perm_builder): Change element type
9473 from HOST_WIDE_INT to poly_int64.
9474 (vec_perm_indices::element_type): Update accordingly.
9475 (vec_perm_indices::clamp): Handle polynomial element_types.
9476 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9477 (vec_perm_indices::all_in_range_p): Likewise.
9478 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
9480 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
9481 polynomial vec_perm_indices element types.
9482 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
9483 * fold-const.c (fold_vec_perm): Likewise.
9484 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
9485 * tree-vect-generic.c (lower_vec_perm): Likewise.
9486 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9487 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
9488 element type to HOST_WIDE_INT.
9490 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9491 Alan Hayward <alan.hayward@arm.com>
9492 David Sherwood <david.sherwood@arm.com>
9494 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
9495 rather than an int. Use plus_constant.
9496 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
9497 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
9499 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9500 Alan Hayward <alan.hayward@arm.com>
9501 David Sherwood <david.sherwood@arm.com>
9503 * calls.c (emit_call_1, expand_call): Change struct_value_size from
9504 a HOST_WIDE_INT to a poly_int64.
9506 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9507 Alan Hayward <alan.hayward@arm.com>
9508 David Sherwood <david.sherwood@arm.com>
9510 * calls.c (load_register_parameters): Cope with polynomial
9511 mode sizes. Require a constant size for BLKmode parameters
9512 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
9513 forces a parameter to be padded at the lsb end in order to
9514 fill a complete number of words, require the parameter size
9515 to be ordered wrt UNITS_PER_WORD.
9517 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9518 Alan Hayward <alan.hayward@arm.com>
9519 David Sherwood <david.sherwood@arm.com>
9521 * reload1.c (spill_stack_slot_width): Change element type
9522 from unsigned int to poly_uint64_pod.
9523 (alter_reg): Treat mode sizes as polynomial.
9525 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9526 Alan Hayward <alan.hayward@arm.com>
9527 David Sherwood <david.sherwood@arm.com>
9529 * reload.c (complex_word_subreg_p): New function.
9530 (reload_inner_reg_of_subreg, push_reload): Use it.
9532 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9533 Alan Hayward <alan.hayward@arm.com>
9534 David Sherwood <david.sherwood@arm.com>
9536 * lra-constraints.c (process_alt_operands): Reject matched
9537 operands whose sizes aren't ordered.
9538 (match_reload): Refer to this check here.
9540 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9541 Alan Hayward <alan.hayward@arm.com>
9542 David Sherwood <david.sherwood@arm.com>
9544 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
9545 that the mode size is in the set {1, 2, 4, 8, 16}.
9547 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9548 Alan Hayward <alan.hayward@arm.com>
9549 David Sherwood <david.sherwood@arm.com>
9551 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
9552 Use plus_constant instead of gen_rtx_PLUS.
9554 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9555 Alan Hayward <alan.hayward@arm.com>
9556 David Sherwood <david.sherwood@arm.com>
9558 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
9559 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
9560 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
9561 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
9562 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
9563 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
9564 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
9565 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
9566 * config/i386/i386.c (ix86_push_rounding): ...this new function.
9567 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
9569 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
9570 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
9571 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
9572 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
9573 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
9574 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
9575 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
9576 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
9577 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
9578 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
9580 * expr.c (emit_move_resolve_push): Treat the input and result
9581 of PUSH_ROUNDING as a poly_int64.
9582 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
9583 (emit_push_insn): Likewise.
9584 * lra-eliminations.c (mark_not_eliminable): Likewise.
9585 * recog.c (push_operand): Likewise.
9586 * reload1.c (elimination_effects): Likewise.
9587 * rtlanal.c (nonzero_bits1): Likewise.
9588 * calls.c (store_one_arg): Likewise. Require the padding to be
9589 known at compile time.
9591 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9592 Alan Hayward <alan.hayward@arm.com>
9593 David Sherwood <david.sherwood@arm.com>
9595 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
9596 Use plus_constant instead of gen_rtx_PLUS.
9598 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9599 Alan Hayward <alan.hayward@arm.com>
9600 David Sherwood <david.sherwood@arm.com>
9602 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
9605 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9606 Alan Hayward <alan.hayward@arm.com>
9607 David Sherwood <david.sherwood@arm.com>
9609 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
9610 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
9611 via stack temporaries. Treat the mode size as polynomial too.
9613 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9614 Alan Hayward <alan.hayward@arm.com>
9615 David Sherwood <david.sherwood@arm.com>
9617 * expr.c (expand_expr_real_2): When handling conversions involving
9618 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
9619 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
9620 as a poly_uint64 too.
9622 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9623 Alan Hayward <alan.hayward@arm.com>
9624 David Sherwood <david.sherwood@arm.com>
9626 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
9628 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9629 Alan Hayward <alan.hayward@arm.com>
9630 David Sherwood <david.sherwood@arm.com>
9632 * combine.c (can_change_dest_mode): Handle polynomial
9633 REGMODE_NATURAL_SIZE.
9634 * expmed.c (store_bit_field_1): Likewise.
9635 * expr.c (store_constructor): Likewise.
9636 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
9637 and polynomial REGMODE_NATURAL_SIZE.
9638 (gen_lowpart_common): Likewise.
9639 * reginfo.c (record_subregs_of_mode): Likewise.
9640 * rtlanal.c (read_modify_subreg_p): Likewise.
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 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
9647 numbers of elements.
9649 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9650 Alan Hayward <alan.hayward@arm.com>
9651 David Sherwood <david.sherwood@arm.com>
9653 * match.pd: Cope with polynomial numbers of vector elements.
9655 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9656 Alan Hayward <alan.hayward@arm.com>
9657 David Sherwood <david.sherwood@arm.com>
9659 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
9660 in a POINTER_PLUS_EXPR.
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 * omp-simd-clone.c (simd_clone_subparts): New function.
9667 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
9668 (ipa_simd_modify_function_body): Likewise.
9670 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9671 Alan Hayward <alan.hayward@arm.com>
9672 David Sherwood <david.sherwood@arm.com>
9674 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
9675 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
9676 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
9677 (expand_vector_condition, vector_element): Likewise.
9678 (subparts_gt): New function.
9679 (get_compute_type): Use subparts_gt.
9680 (count_type_subparts): Delete.
9681 (expand_vector_operations_1): Use subparts_gt instead of
9682 count_type_subparts.
9684 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9685 Alan Hayward <alan.hayward@arm.com>
9686 David Sherwood <david.sherwood@arm.com>
9688 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
9689 (vect_compile_time_alias): ...this new function. Do the calculation
9690 on poly_ints rather than trees.
9691 (vect_prune_runtime_alias_test_list): Update call accordingly.
9693 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9694 Alan Hayward <alan.hayward@arm.com>
9695 David Sherwood <david.sherwood@arm.com>
9697 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
9699 (vect_schedule_slp_instance): Likewise.
9701 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9702 Alan Hayward <alan.hayward@arm.com>
9703 David Sherwood <david.sherwood@arm.com>
9705 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
9706 constant and extern definitions for variable-length vectors.
9707 (vect_get_constant_vectors): Note that the number of units
9708 is known to be constant.
9710 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9711 Alan Hayward <alan.hayward@arm.com>
9712 David Sherwood <david.sherwood@arm.com>
9714 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
9715 of units as polynomial. Choose between WIDE and NARROW based
9718 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9719 Alan Hayward <alan.hayward@arm.com>
9720 David Sherwood <david.sherwood@arm.com>
9722 * tree-vect-stmts.c (simd_clone_subparts): New function.
9723 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
9725 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9726 Alan Hayward <alan.hayward@arm.com>
9727 David Sherwood <david.sherwood@arm.com>
9729 * tree-vect-stmts.c (vectorizable_call): Treat the number of
9730 vectors as polynomial. Use build_index_vector for
9733 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9734 Alan Hayward <alan.hayward@arm.com>
9735 David Sherwood <david.sherwood@arm.com>
9737 * tree-vect-stmts.c (get_load_store_type): Treat the number of
9738 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
9739 for variable-length vectors.
9740 (vectorizable_mask_load_store): Treat the number of units as
9741 polynomial, asserting that it is constant if the condition has
9742 already been enforced.
9743 (vectorizable_store, vectorizable_load): Likewise.
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-loop.c (vectorizable_live_operation): Treat the number
9750 of units as polynomial. Punt if we can't tell at compile time
9751 which vector contains the final result.
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-loop.c (vectorizable_induction): Treat the number
9758 of units as polynomial. Punt on SLP inductions. Use an integer
9759 VEC_SERIES_EXPR for variable-length integer reductions. Use a
9760 cast of such a series for variable-length floating-point
9763 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9764 Alan Hayward <alan.hayward@arm.com>
9765 David Sherwood <david.sherwood@arm.com>
9767 * tree.h (build_index_vector): Declare.
9768 * tree.c (build_index_vector): New function.
9769 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
9770 of units as polynomial, forcibly converting it to a constant if
9771 vectorizable_reduction has already enforced the condition.
9772 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
9773 to create a {1,2,3,...} vector.
9774 (vectorizable_reduction): Treat the number of units as polynomial.
9775 Choose vectype_in based on the largest scalar element size rather
9776 than the smallest number of units. Enforce the restrictions
9779 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9780 Alan Hayward <alan.hayward@arm.com>
9781 David Sherwood <david.sherwood@arm.com>
9783 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
9784 number of units as polynomial.
9786 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9787 Alan Hayward <alan.hayward@arm.com>
9788 David Sherwood <david.sherwood@arm.com>
9790 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
9791 * target.def (autovectorize_vector_sizes): Return the vector sizes
9792 by pointer, using vector_sizes rather than a bitmask.
9793 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
9794 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
9795 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
9797 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
9798 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
9799 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
9800 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
9801 * omp-general.c (omp_max_vf): Likewise.
9802 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9803 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
9804 * tree-vect-loop.c (vect_analyze_loop): Likewise.
9805 * tree-vect-slp.c (vect_slp_bb): Likewise.
9806 * doc/tm.texi: Regenerate.
9807 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
9809 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
9810 the vector size as a poly_uint64 rather than an unsigned int.
9811 (current_vector_size): Change from an unsigned int to a poly_uint64.
9812 (get_vectype_for_scalar_type): Update accordingly.
9813 * tree.h (build_truth_vector_type): Take the size and number of
9814 units as a poly_uint64 rather than an unsigned int.
9815 (build_vector_type): Add a temporary overload that takes
9816 the number of units as a poly_uint64 rather than an unsigned int.
9817 * tree.c (make_vector_type): Likewise.
9818 (build_truth_vector_type): Take the number of units as a poly_uint64
9819 rather than an unsigned int.
9821 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9822 Alan Hayward <alan.hayward@arm.com>
9823 David Sherwood <david.sherwood@arm.com>
9825 * target.def (get_mask_mode): Take the number of units and length
9826 as poly_uint64s rather than unsigned ints.
9827 * targhooks.h (default_get_mask_mode): Update accordingly.
9828 * targhooks.c (default_get_mask_mode): Likewise.
9829 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
9830 * doc/tm.texi: Regenerate.
9832 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9833 Alan Hayward <alan.hayward@arm.com>
9834 David Sherwood <david.sherwood@arm.com>
9836 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
9837 * omp-general.c (omp_max_vf): Likewise.
9838 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
9839 (expand_omp_simd): Handle polynomial safelen.
9840 * omp-low.c (omplow_simd_context): Add a default constructor.
9841 (omplow_simd_context::max_vf): Change from int to poly_uint64.
9842 (lower_rec_simd_input_clauses): Update accordingly.
9843 (lower_rec_input_clauses): Likewise.
9845 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9846 Alan Hayward <alan.hayward@arm.com>
9847 David Sherwood <david.sherwood@arm.com>
9849 * tree-vectorizer.h (vect_nunits_for_cost): New function.
9850 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
9851 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
9852 (vect_analyze_slp_cost): Likewise.
9853 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
9854 (vect_model_load_cost): Likewise.
9856 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9857 Alan Hayward <alan.hayward@arm.com>
9858 David Sherwood <david.sherwood@arm.com>
9860 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
9861 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
9862 from an unsigned int * to a poly_uint64_pod *.
9863 (calculate_unrolling_factor): New function.
9864 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
9866 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9867 Alan Hayward <alan.hayward@arm.com>
9868 David Sherwood <david.sherwood@arm.com>
9870 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
9871 from an unsigned int to a poly_uint64.
9872 (_loop_vec_info::slp_unrolling_factor): Likewise.
9873 (_loop_vec_info::vectorization_factor): Change from an int
9875 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
9876 (vect_get_num_vectors): New function.
9877 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
9878 (vect_get_num_copies): Use vect_get_num_vectors.
9879 (vect_analyze_data_ref_dependences): Change max_vf from an int *
9880 to an unsigned int *.
9881 (vect_analyze_data_refs): Change min_vf from an int * to a
9883 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9884 than an unsigned HOST_WIDE_INT.
9885 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
9886 (vect_analyze_data_ref_dependence): Change max_vf from an int *
9887 to an unsigned int *.
9888 (vect_analyze_data_ref_dependences): Likewise.
9889 (vect_compute_data_ref_alignment): Handle polynomial vf.
9890 (vect_enhance_data_refs_alignment): Likewise.
9891 (vect_prune_runtime_alias_test_list): Likewise.
9892 (vect_shift_permute_load_chain): Likewise.
9893 (vect_supportable_dr_alignment): Likewise.
9894 (dependence_distance_ge_vf): Take the vectorization factor as a
9895 poly_uint64 rather than an unsigned HOST_WIDE_INT.
9896 (vect_analyze_data_refs): Change min_vf from an int * to a
9898 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
9899 vfm1 as a poly_uint64 rather than an int. Make the same change
9900 for the returned bound_scalar.
9901 (vect_gen_vector_loop_niters): Handle polynomial vf.
9902 (vect_do_peeling): Likewise. Update call to
9903 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
9904 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
9906 * tree-vect-loop.c (vect_determine_vectorization_factor)
9907 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
9908 (vect_get_known_peeling_cost): Likewise.
9909 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
9910 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
9911 (vect_transform_loop): Likewise. Use the lowest possible VF when
9912 updating the upper bounds of the loop.
9913 (vect_min_worthwhile_factor): Make static. Return an unsigned int
9915 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
9916 polynomial unroll factors.
9917 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
9918 (vect_make_slp_decision): Likewise.
9919 (vect_supported_load_permutation_p): Likewise, and polynomial
9921 (vect_analyze_slp_cost): Handle polynomial vf.
9922 (vect_slp_analyze_node_operations): Likewise.
9923 (vect_slp_analyze_bb_1): Likewise.
9924 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9925 than an unsigned HOST_WIDE_INT.
9926 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
9927 (vectorizable_load): Handle polynomial vf.
9928 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
9930 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
9932 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9933 Alan Hayward <alan.hayward@arm.com>
9934 David Sherwood <david.sherwood@arm.com>
9936 * match.pd: Handle bit operations involving three constants
9937 and try to fold one pair.
9939 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9941 * tree-vect-loop-manip.c: Include gimple-fold.h.
9942 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
9943 niters_maybe_zero parameters. Handle other cases besides a step of 1.
9944 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
9945 Add a path that uses a step of VF instead of 1, but disable it
9947 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
9948 and niters_no_overflow parameters. Update calls to
9949 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
9950 Create a new SSA name if the latter choses to use a ste other
9951 than zero, and return it via niters_vector_mult_vf_var.
9952 * tree-vect-loop.c (vect_transform_loop): Update calls to
9953 vect_do_peeling, vect_gen_vector_loop_niters and
9954 slpeel_make_loop_iterate_ntimes.
9955 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
9956 (vect_gen_vector_loop_niters): Update declarations after above changes.
9958 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
9960 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
9961 128-bit round to integer instructions.
9962 (ceil<mode>2): Likewise.
9963 (btrunc<mode>2): Likewise.
9964 (round<mode>2): Likewise.
9966 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9968 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
9969 unaligned VSX load/store on P8/P9.
9970 (expand_block_clear): Allow the use of unaligned VSX
9971 load/store on P8/P9.
9973 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9975 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
9977 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
9978 swap associated with both a load and a store.
9980 2018-01-02 Andrew Waterman <andrew@sifive.com>
9982 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
9983 * config/riscv/riscv.md (clear_cache): Use it.
9985 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
9987 * web.c: Remove out-of-date comment.
9989 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9991 * expr.c (fixup_args_size_notes): Check that any existing
9992 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
9993 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
9994 (emit_single_push_insn): ...here.
9996 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9998 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
9999 (const_vector_encoded_nelts): New function.
10000 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
10001 (const_vector_int_elt, const_vector_elt): Declare.
10002 * emit-rtl.c (const_vector_int_elt_1): New function.
10003 (const_vector_elt): Likewise.
10004 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
10005 of CONST_VECTOR_ELT.
10007 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10009 * expr.c: Include rtx-vector-builder.h.
10010 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
10011 directly on the tree encoding.
10012 (const_vector_from_tree): Likewise.
10013 * optabs.c: Include rtx-vector-builder.h.
10014 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
10015 sequence of "u" values.
10016 * vec-perm-indices.c: Include rtx-vector-builder.h.
10017 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
10018 directly on the vec_perm_indices encoding.
10020 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10022 * doc/rtl.texi (const_vector): Describe new encoding scheme.
10023 * Makefile.in (OBJS): Add rtx-vector-builder.o.
10024 * rtx-vector-builder.h: New file.
10025 * rtx-vector-builder.c: Likewise.
10026 * rtl.h (rtx_def::u2): Add a const_vector field.
10027 (CONST_VECTOR_NPATTERNS): New macro.
10028 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
10029 (CONST_VECTOR_DUPLICATE_P): Likewise.
10030 (CONST_VECTOR_STEPPED_P): Likewise.
10031 (CONST_VECTOR_ENCODED_ELT): Likewise.
10032 (const_vec_duplicate_p): Check for a duplicated vector encoding.
10033 (unwrap_const_vec_duplicate): Likewise.
10034 (const_vec_series_p): Check for a non-duplicated vector encoding.
10035 Say that the function only returns true for integer vectors.
10036 * emit-rtl.c: Include rtx-vector-builder.h.
10037 (gen_const_vec_duplicate_1): Delete.
10038 (gen_const_vector): Call gen_const_vec_duplicate instead of
10039 gen_const_vec_duplicate_1.
10040 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
10041 (gen_const_vec_duplicate): Use rtx_vector_builder.
10042 (gen_const_vec_series): Likewise.
10043 (gen_rtx_CONST_VECTOR): Likewise.
10044 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
10045 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10046 Build a new vector rather than modifying a CONST_VECTOR in-place.
10047 (handle_special_swappables): Update call accordingly.
10048 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
10049 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10050 Build a new vector rather than modifying a CONST_VECTOR in-place.
10051 (handle_special_swappables): Update call accordingly.
10053 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10055 * simplify-rtx.c (simplify_const_binary_operation): Use
10056 CONST_VECTOR_ELT instead of XVECEXP.
10058 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10060 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
10061 the selector elements to be different from the data elements
10062 if the selector is a VECTOR_CST.
10063 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
10064 ssizetype for the selector.
10066 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10068 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
10069 before testing each element individually.
10070 * tree-vect-generic.c (lower_vec_perm): Likewise.
10072 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10074 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
10075 * selftest-run-tests.c (selftest::run_tests): Call it.
10076 * vector-builder.h (vector_builder::operator ==): New function.
10077 (vector_builder::operator !=): Likewise.
10078 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
10079 (vec_perm_indices::all_from_input_p): New function.
10080 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10081 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
10082 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
10083 instead of reading the VECTOR_CST directly. Detect whether both
10084 vector inputs are the same before constructing the vec_perm_indices,
10085 and update the number of inputs argument accordingly. Use the
10086 utility functions added above. Only construct sel2 if we need to.
10088 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10090 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
10091 the broadcast of the low byte.
10092 (expand_mult_highpart): Use an explicit encoding for the permutes.
10093 * optabs-query.c (can_mult_highpart_p): Likewise.
10094 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
10095 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10096 (vectorizable_bswap): Likewise.
10097 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
10098 explicit encoding for the power-of-2 permutes.
10099 (vect_permute_store_chain): Likewise.
10100 (vect_grouped_load_supported): Likewise.
10101 (vect_permute_load_chain): Likewise.
10103 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10105 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
10106 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
10107 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
10108 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10109 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
10110 (vect_gen_perm_mask_any): Likewise.
10112 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10114 * int-vector-builder.h: New file.
10115 * vec-perm-indices.h: Include int-vector-builder.h.
10116 (vec_perm_indices): Redefine as an int_vector_builder.
10117 (auto_vec_perm_indices): Delete.
10118 (vec_perm_builder): Redefine as a stand-alone class.
10119 (vec_perm_indices::vec_perm_indices): New function.
10120 (vec_perm_indices::clamp): Likewise.
10121 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
10122 (vec_perm_indices::new_vector): New function.
10123 (vec_perm_indices::new_expanded_vector): Update for new
10124 vec_perm_indices class.
10125 (vec_perm_indices::rotate_inputs): New function.
10126 (vec_perm_indices::all_in_range_p): Operate directly on the
10127 encoded form, without computing elided elements.
10128 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
10129 encoding. Update for new vec_perm_indices class.
10130 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
10131 the given vec_perm_builder.
10132 (expand_vec_perm_var): Update vec_perm_builder constructor.
10133 (expand_mult_highpart): Use vec_perm_builder instead of
10134 auto_vec_perm_indices.
10135 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
10136 vec_perm_indices instead of auto_vec_perm_indices. Use a single
10137 or double series encoding as appropriate.
10138 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
10139 vec_perm_indices instead of auto_vec_perm_indices.
10140 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10141 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10142 (vect_permute_store_chain): Likewise.
10143 (vect_grouped_load_supported): Likewise.
10144 (vect_permute_load_chain): Likewise.
10145 (vect_shift_permute_load_chain): Likewise.
10146 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10147 (vect_transform_slp_perm_load): Likewise.
10148 (vect_schedule_slp_instance): Likewise.
10149 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10150 (vectorizable_mask_load_store): Likewise.
10151 (vectorizable_bswap): Likewise.
10152 (vectorizable_store): Likewise.
10153 (vectorizable_load): Likewise.
10154 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
10155 vec_perm_indices instead of auto_vec_perm_indices. Use
10156 tree_to_vec_perm_builder to read the vector from a tree.
10157 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
10158 vec_perm_builder instead of a vec_perm_indices.
10159 (have_whole_vector_shift): Use vec_perm_builder and
10160 vec_perm_indices instead of auto_vec_perm_indices. Leave the
10161 truncation to calc_vec_perm_mask_for_shift.
10162 (vect_create_epilog_for_reduction): Likewise.
10163 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
10164 from auto_vec_perm_indices to vec_perm_indices.
10165 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10166 instead of changing individual elements.
10167 (aarch64_vectorize_vec_perm_const): Use new_vector to install
10168 the vector in d.perm.
10169 * config/arm/arm.c (expand_vec_perm_d::perm): Change
10170 from auto_vec_perm_indices to vec_perm_indices.
10171 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10172 instead of changing individual elements.
10173 (arm_vectorize_vec_perm_const): Use new_vector to install
10174 the vector in d.perm.
10175 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
10176 Update vec_perm_builder constructor.
10177 (rs6000_expand_interleave): Likewise.
10178 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
10179 (rs6000_expand_interleave): Likewise.
10181 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10183 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
10184 to qimode could truncate the indices.
10185 * optabs.c (expand_vec_perm_var): Likewise.
10187 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10189 * Makefile.in (OBJS): Add vec-perm-indices.o.
10190 * vec-perm-indices.h: New file.
10191 * vec-perm-indices.c: Likewise.
10192 * target.h (vec_perm_indices): Replace with a forward class
10194 (auto_vec_perm_indices): Move to vec-perm-indices.h.
10195 * optabs.h: Include vec-perm-indices.h.
10196 (expand_vec_perm): Delete.
10197 (selector_fits_mode_p, expand_vec_perm_var): Declare.
10198 (expand_vec_perm_const): Declare.
10199 * target.def (vec_perm_const_ok): Replace with...
10200 (vec_perm_const): ...this new hook.
10201 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
10202 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
10203 * doc/tm.texi: Regenerate.
10204 * optabs.def (vec_perm_const): Delete.
10205 * doc/md.texi (vec_perm_const): Likewise.
10206 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
10207 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
10208 expand_vec_perm for constant permutation vectors. Assert that
10209 the mode of variable permutation vectors is the integer equivalent
10210 of the mode that is being permuted.
10211 * optabs-query.h (selector_fits_mode_p): Declare.
10212 * optabs-query.c: Include vec-perm-indices.h.
10213 (selector_fits_mode_p): New function.
10214 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
10215 is defined, instead of checking whether the vec_perm_const_optab
10216 exists. Use targetm.vectorize.vec_perm_const instead of
10217 targetm.vectorize.vec_perm_const_ok. Check whether the indices
10218 fit in the vector mode before using a variable permute.
10219 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
10220 vec_perm_indices instead of an rtx.
10221 (expand_vec_perm): Replace with...
10222 (expand_vec_perm_const): ...this new function. Take the selector
10223 as a vec_perm_indices rather than an rtx. Also take the mode of
10224 the selector. Update call to shift_amt_for_vec_perm_mask.
10225 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
10226 Use vec_perm_indices::new_expanded_vector to expand the original
10227 selector into bytes. Check whether the indices fit in the vector
10228 mode before using a variable permute.
10229 (expand_vec_perm_var): Make global.
10230 (expand_mult_highpart): Use expand_vec_perm_const.
10231 * fold-const.c: Includes vec-perm-indices.h.
10232 * tree-ssa-forwprop.c: Likewise.
10233 * tree-vect-data-refs.c: Likewise.
10234 * tree-vect-generic.c: Likewise.
10235 * tree-vect-loop.c: Likewise.
10236 * tree-vect-slp.c: Likewise.
10237 * tree-vect-stmts.c: Likewise.
10238 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
10240 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
10241 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
10242 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
10243 (aarch64_vectorize_vec_perm_const): ...this new function.
10244 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10245 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10246 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
10247 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
10248 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10249 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10250 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
10252 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
10253 check for NEON modes.
10254 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
10255 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
10256 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
10257 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
10259 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
10260 the old VEC_PERM_CONST conditions.
10261 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
10262 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
10263 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
10264 (ia64_vectorize_vec_perm_const_ok): Merge into...
10265 (ia64_vectorize_vec_perm_const): ...this new function.
10266 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
10267 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
10268 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
10269 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
10270 * config/mips/mips.c (mips_expand_vec_perm_const)
10271 (mips_vectorize_vec_perm_const_ok): Merge into...
10272 (mips_vectorize_vec_perm_const): ...this new function.
10273 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
10274 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
10275 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
10276 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
10277 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
10278 (rs6000_expand_vec_perm_const): Delete.
10279 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
10281 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10282 (altivec_expand_vec_perm_const_le): Take each operand individually.
10283 Operate on constant selectors rather than rtxes.
10284 (altivec_expand_vec_perm_const): Likewise. Update call to
10285 altivec_expand_vec_perm_const_le.
10286 (rs6000_expand_vec_perm_const): Delete.
10287 (rs6000_vectorize_vec_perm_const_ok): Delete.
10288 (rs6000_vectorize_vec_perm_const): New function.
10289 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10290 an element count and rtx array.
10291 (rs6000_expand_extract_even): Update call accordingly.
10292 (rs6000_expand_interleave): Likewise.
10293 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
10294 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
10295 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
10296 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
10297 (rs6000_expand_vec_perm_const): Delete.
10298 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10299 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10300 (altivec_expand_vec_perm_const_le): Take each operand individually.
10301 Operate on constant selectors rather than rtxes.
10302 (altivec_expand_vec_perm_const): Likewise. Update call to
10303 altivec_expand_vec_perm_const_le.
10304 (rs6000_expand_vec_perm_const): Delete.
10305 (rs6000_vectorize_vec_perm_const_ok): Delete.
10306 (rs6000_vectorize_vec_perm_const): New function. Remove stray
10307 reference to the SPE evmerge intructions.
10308 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10309 an element count and rtx array.
10310 (rs6000_expand_extract_even): Update call accordingly.
10311 (rs6000_expand_interleave): Likewise.
10312 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
10313 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
10315 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10317 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10319 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
10320 vector mode and that that mode matches the mode of the data
10322 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
10323 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
10324 directly using expand_vec_perm_1 when forcing selectors into
10326 (expand_vec_perm_var): New function, split out from expand_vec_perm.
10328 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10330 * optabs-query.h (can_vec_perm_p): Delete.
10331 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
10332 * optabs-query.c (can_vec_perm_p): Split into...
10333 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
10334 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
10335 particular selector is valid.
10336 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10337 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10338 (vect_grouped_load_supported): Likewise.
10339 (vect_shift_permute_load_chain): Likewise.
10340 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10341 (vect_transform_slp_perm_load): Likewise.
10342 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10343 (vectorizable_bswap): Likewise.
10344 (vect_gen_perm_mask_checked): Likewise.
10345 * fold-const.c (fold_ternary_loc): Likewise. Don't take
10346 implementations of variable permutation vectors into account
10347 when deciding which selector to use.
10348 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
10349 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
10350 with a false third argument.
10351 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
10352 to test whether the constant selector is valid and can_vec_perm_var_p
10353 to test whether a variable selector is valid.
10355 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10357 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
10358 * optabs-query.c (can_vec_perm_p): Likewise.
10359 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
10360 instead of vec_perm_indices.
10361 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
10362 (vect_gen_perm_mask_checked): Likewise,
10363 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
10364 (vect_gen_perm_mask_checked): Likewise,
10366 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10368 * optabs-query.h (qimode_for_vec_perm): Declare.
10369 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
10370 (qimode_for_vec_perm): ...this new function.
10371 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
10373 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10375 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
10376 does not have a conditional at the top.
10378 2018-01-02 Richard Biener <rguenther@suse.de>
10380 * ipa-inline.c (big_speedup_p): Fix expression.
10382 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10385 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
10388 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10392 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
10393 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
10394 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
10395 cond_taken_branch_cost 3->4.
10397 2018-01-01 Jakub Jelinek <jakub@redhat.com>
10399 PR tree-optimization/83581
10400 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
10401 TODO_cleanup_cfg if any changes have been made.
10403 PR middle-end/83608
10404 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
10405 convert_modes if target mode has the right side, but different mode
10408 PR middle-end/83609
10409 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
10410 last argument when extracting from CONCAT. If either from_real or
10411 from_imag is NULL, use expansion through memory. If result is not
10412 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
10413 the parts directly to inner mode, if even that fails, use expansion
10416 PR middle-end/83623
10417 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
10418 check for bswap in mode rather than HImode and use that in expand_unop
10421 Copyright (C) 2018 Free Software Foundation, Inc.
10423 Copying and distribution of this file, with or without modification,
10424 are permitted in any medium without royalty provided the copyright
10425 notice and this notice are preserved.