1 2018-04-09 Jan Hubicka <jh@suse.cz>
4 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
5 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
6 * tree.c (free_lang_data_in_type): Fix handling of binfos;
8 (free_lang_data): Rebuild type inheritance graph.
10 2018-04-09 Martin Sebor <msebor@redhat.com>
12 * invoke.texi (-finline-small-functions): Mention other optimization
14 (-findirect-inlining, -fpartial-inlining): Same.
15 (-finline-functions-called-once): Same.
16 (-freorder-blocks-and-partition): Same.
18 2018-04-09 Jan Hubicka <jh@suse.cz>
21 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
22 jumps; choose last target that matches the criteria (i.e.
23 no partition changes for non-crossing jumps).
24 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
25 support for redirecting crossing jumps to non-crossing.
27 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
29 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
30 also for naked functions.
32 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
34 * config/arc/arc.md (add_shift): New pattern.
35 (add_shift2): Likewise.
36 (sub_shift): Likewise.
37 (sub_shift_cmp0_noout): Likewise.
38 (compare_si_ashiftsi): Likewise.
39 (xbfu_cmp0_noout): New combine pattern.
40 (xbfu_cmp0"): Likewise.
41 (movsi_set_cc_insn): Place the predicable variant first.
42 (commutative_binary_cmp0_noout): Remove clobber.
43 (commutative_binary_cmp0): New pattern.
44 (noncommutative_binary_cmp0): Likewise.
45 (noncommutative_binary_cmp0_noout): Likewise.
46 (noncommutative_binary_comparison_result_used): Removed.
47 (rsub_cmp0): New pattern.
48 (rsub_cmp0_noout): Likewise.
49 (extzvsi): Changed, keep only meaningful variants.
50 (SQH, SEZ): New iterators.
51 (SQH_postfix): New mode attribute.
52 (SEZ_prefix): New code attribute.
53 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
54 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
55 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
57 (noncommutative_operator): Check the availability of barrel
60 2018-04-09 Richard Biener <rguenther@suse.de>
62 PR tree-optimization/85284
63 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
64 Only use the niter constraining form of simple_iv when the exit
67 2018-04-09 Tom de Vries <tom@codesourcery.com>
70 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
71 (define_expand "*memory_barrier"): New define_expand.
72 (define_insn "memory_barrier"): New insn.
74 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
76 PR rtl-optimization/80463
77 PR rtl-optimization/83972
78 PR rtl-optimization/83480
80 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
81 correct producer for the insn.
82 (tidy_control_flow): Fixup seqnos in case of debug insns.
84 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
86 PR rtl-optimization/83913
88 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
89 different sched-times when merging exprs.
91 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
93 PR rtl-optimization/83962
95 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
96 tidy_fallthru_edge and tidy_control_flow.
98 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
100 PR rtl-optimization/83530
102 * sel-sched.c (force_next_insn): New global variable.
103 (remove_insn_for_debug): When force_next_insn is true, also leave only
104 next insn in the ready list.
105 (sel_sched_region): When the region wasn't scheduled, make another pass
106 over it with force_next_insn set to 1.
108 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
110 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
112 * config/nds32/constants.md (unspec_volatile_element): Add enum values
113 for interrupt control.
114 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
115 functions for interrupt control.
116 * config/nds32/nds32-intrinsic.md: Likewise.
117 * config/nds32/nds32_intrinsic.h: Likewise.
118 * config/nds32/nds32.h (nds32_builtins): Likewise.
120 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
122 * config/nds32/nds32.c (nds32_init_machine_status,
123 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
124 strict_aligned_p field.
125 (nds32_expand_to_rtl_hook): New function.
126 (TARGET_EXPAND_TO_RTL_HOOK): Define.
127 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
129 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
130 Chung-Ju Wu <jasonwucj@gmail.com>
132 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
133 * config/nds32/nds32-n7.md: New file.
134 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
135 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
137 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
138 * config/nds32/nds32.md (pipeline_model): Add n7.
139 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
140 * config/nds32/pipelines.md: Include n7 settings.
142 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
143 Chung-Ju Wu <jasonwucj@gmail.com>
145 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
146 * config/nds32/nds32-e8.md: New file.
147 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
148 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
150 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
151 * config/nds32/nds32.md (pipeline_model): Add e8.
152 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
153 * config/nds32/pipelines.md: Include e8 settings.
155 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
156 Chung-Ju Wu <jasonwucj@gmail.com>
158 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
159 * config/nds32/nds32-n8.md: New file.
160 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
161 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
163 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
164 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
165 * config/nds32/nds32.md (pipeline_model): Add n8.
166 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
167 * config/nds32/pipelines.md: Include n8 settings.
169 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
170 Chung-Ju Wu <jasonwucj@gmail.com>
172 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
173 * config/nds32/nds32-n9-2r1w.md: New file.
174 * config/nds32/nds32-n9-3r2w.md: New file.
175 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
176 nds32_register_ports): New or modify for cpu n9.
177 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
179 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
180 * config/nds32/nds32-utils.c: New file.
181 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
182 TARGET_MUL_SLOW): Define.
183 * config/nds32/nds32.md (pipeline_model): New attribute.
184 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
185 New options that support cpu n9.
186 * config/nds32/pipelines.md: Include n9 settings.
187 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
189 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
191 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
192 information if necessary.
193 (output_cond_branch_compare_zero): Likewise.
194 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
195 (nds32_target_alignment): Refine for alignment.
196 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
197 (FUNCTION_BOUNDARY): Modify.
198 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
200 * config/nds32/nds32.opt (malways-align, malign-functions): New.
202 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
204 * config/nds32/constants.md (unspec_volatile_element): Add values for
205 TLB operation and data prefetch.
206 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
207 functions for TLB operation and data prefetch.
208 * config/nds32/nds32-intrinsic.md: Likewise.
209 * config/nds32/nds32_intrinsic.h: Likewise.
210 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
211 (nds32_print_operand): Likewise.
212 * config/nds32/nds32.h (nds32_builtins): Likewise.
214 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
215 Andrew Pinski <pinsika@gcc.gnu.org>
218 * match.pd: Use constant_boolean_node of correct type instead of
219 boolean_true_node or boolean_false_node for simplifying
220 pointer comparisons to zero.
222 2018-04-07 Jakub Jelinek <jakub@redhat.com>
224 PR tree-optimization/80021
225 * tree.c (verify_type_variant): Make error call in verify_variant_match
226 translatable and remove final full stop.
228 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
230 * config/nds32/constants.md (unspec_volatile_element): Add
231 UNSPEC_VOLATILE_EH_RETURN.
232 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
233 nds32_output_stack_pop): Support dwarf exception handling process.
234 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
235 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
236 exception handling process.
237 (nds32_compute_stack_frame): Likewise.
238 (nds32_return_addr_rtx): Likewise.
239 (nds32_initial_elimination_offset): Likewise.
240 (nds32_expand_prologue): Likewise.
241 (nds32_expand_epilogue): Likewise.
242 (nds32_dynamic_chain_address): New function.
243 * config/nds32/nds32.h (machine_function): Add fields for dwarf
245 (DYNAMIC_CHAIN_ADDRESS): Define.
246 (EH_RETURN_DATA_REGNO): Define.
247 (EH_RETURN_STACKADJ_RTX): Define.
248 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
249 patterns for dwarf exception handling.
251 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
253 * config/nds32/nds32.h: Clean up obsolete macros.
255 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
257 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
258 Add enum values for particular instructions.
259 * config/nds32/nds32-intrinsic.c: Implementation of expanding
260 particular intrinsic functions.
261 * config/nds32/nds32-intrinsic.md: Likewise.
262 * config/nds32/nds32_intrinsic.h: Likewise.
263 * config/nds32/nds32.h (nds32_builtins): Likewise.
264 * config/nds32/nds32.md (type): Add pbsad and pbsada.
265 (btst, ave): New patterns for particular instructions.
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 atomic load/store and memory sync.
271 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
273 * config/nds32/nds32-intrinsic.md: Likewise.
274 * config/nds32/nds32_intrinsic.h: Likewise.
275 * config/nds32/nds32.h (nds32_builtins): Likewise.
277 2018-04-07 Jakub Jelinek <jakub@redhat.com>
279 PR tree-optimization/85257
280 * fold-const.c (native_encode_vector): If not all elts could fit
281 and off is -1, return 0 rather than offset.
282 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
283 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
284 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
285 adjust buffer in native_interpret_expr call.
287 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
289 * config/nds32/constants.md (unspec_volatile_element): Add cache
291 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
292 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
293 * config/nds32/nds32.c (nds32_cctl_names): New.
294 (nds32_print_operand): Handle cache control register names.
295 * config/nds32/nds32.h (nds32_builtins): New enum values.
296 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
298 * config/nds32/nds32.md (type): Add mmu.
299 * config/nds32/pipelines.md (simple_insn): Add mmu.
301 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
303 * config/nds32/nds32.md (type): Remove call.
304 * config/nds32/pipelines.md (simple_insn): Likewise.
306 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
308 * config/nds32/constants.md (unspec_volatile_element): Add
309 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
310 UNSPEC_VOLATILE_FMFCFG.
311 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
312 description for fmfcfg and fmfcsr.
313 (bdesc_1arg): Add fmtcsr.
314 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
315 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
316 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
317 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
318 unspec_fmfcfg): New patterns.
319 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
320 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
321 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
322 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
323 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
324 __nds32__fmfcfg): Define.
326 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
328 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
329 intrinsic register names.
330 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
331 intrinsic register enum values and macros.
333 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
335 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
336 for load/store addressing form.
337 (nds32_print_operand_address): Likewise.
339 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
342 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
343 based on LABEL_REF. Remove useless assertion.
344 (pic_address_needs_scratch): Fix formatting.
345 (sparc_legitimize_pic_address): Minor tweaks.
346 (sparc_delegitimize_address): Adjust assertion accordingly.
347 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
348 into symbolic_operand.
349 (movsi_high_pic_label_ref): Likewise.
350 (movsi_lo_sum_pic_label_ref): Likewise.
351 (movdi_pic_label_ref): Likewise.
352 (movdi_high_pic_label_ref): Likewise.
353 (movdi_lo_sum_pic_label_ref): Likewise.
355 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
357 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
358 custom LIB_SPEC setup.
360 2018-04-06 Ruslan Bukin <br@bsdpad.com>
361 Kito Cheng <kito.cheng@gmail.com>
363 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
364 * config/riscv/freebsd.h: New.
366 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
368 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
369 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
372 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
373 Kito Cheng <kito.cheng@gmail.com>
375 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
376 nds32_output_call, nds32_symbol_binds_local_p): New functions.
377 * config/nds32/nds32-protos.h (nds32_output_call,
378 nds32_output_return): Declare.
379 * config/nds32/nds32.md: Refine all the call and return patterns.
381 2018-04-06 Jakub Jelinek <jakub@redhat.com>
384 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
385 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
387 PR rtl-optimization/84872
388 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
389 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
392 2018-04-06 Tamar Christina <tamar.christina@arm.com>
394 * expr.c (copy_blkmode_to_reg): Revert 254862.
395 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
397 2018-04-06 Richard Biener <rguenther@suse.de>
400 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
401 after seeing a component reference with an adjacent field. Treat
402 refs to arrays at struct end of external decls similar to
403 refs to unconstrained commons.
405 2018-04-06 Jakub Jelinek <jakub@redhat.com>
408 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
409 look through SAVE_EXPRs with non-side-effects argument. Adjust
411 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
414 2018-04-06 Richard Biener <rguenther@suse.de>
417 * alias.c (find_base_term): New wrapper around find_base_term
418 unwinding CSELIB_VAL_PTR changes.
419 (find_base_term): Do not restore CSELIB_VAL_PTR during the
422 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
424 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
426 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
427 constant definitions.
428 ("nop"): lr 0,0 -> nopr r0
429 ("nop_lr0", "nop_lr1"): New insn definitions.
431 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
433 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
434 NDS32_V3PUSH_AVAILABLE_P macro.
436 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
437 Chung-Ju Wu <jasonwucj@gmail.com>
439 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
440 (nds32*-*-*): Add float and fpu_config into supported_defaults.
441 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
442 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
443 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
444 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
445 * config/nds32/constraints.md: New constraints and checking for hard
447 * config/nds32/iterators.md: New mode iterator and attribute for hard
449 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
451 * config/nds32/nds32-fpu.md: New file.
452 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
453 deal with hard float code generation.
454 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
456 (abi_type, float_reg_number): New enum type.
457 * config/nds32/nds32-predicates.c: New predicates for hard float.
458 * config/nds32/nds32-protos.h: Declare functions for hard float.
459 * config/nds32/nds32.c: Implementation for hard float configuration.
460 * config/nds32/nds32.h: Definitions for hard float configuration.
461 * config/nds32/nds32.md: Include hard float machine description and
462 modify patterns for hard float configuration.
463 * config/nds32/nds32.opt: New options for hard float configuration.
464 * config/nds32/predicates.md: New predicates for hard float
467 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
469 * common/config/nds32/nds32-common.c
470 (nds32_option_optimization_table): Enable -mreleax-hint by default.
472 2018-04-05 Jakub Jelinek <jakub@redhat.com>
475 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
476 CONSTRUCTOR_ELT (ctor, ...)->value.
478 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
481 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
483 2018-04-05 Tom de Vries <tom@codesourcery.com>
486 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
489 2018-04-05 Shiva Chen <shiva0217@gmail.com>
490 Kito Cheng <kito.cheng@gmail.com>
492 * config/nds32/constraints.md (U33): Fine-tune checking condition.
493 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
494 * config/nds32/nds32.h (nds32_16bit_address_type): Add
495 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
497 2018-04-05 Shiva Chen <shiva0217@gmail.com>
498 Kito Cheng <kito.cheng@gmail.com>
500 * config/nds32/constraints.md (Ufe): New memory constraint.
501 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
502 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
503 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
505 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
506 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
508 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
510 * config/nds32/nds32.md: Use optimize_size in the condition for
511 alu-shift instructions.
513 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
515 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
517 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
519 * config/nds32/nds32.md (negsi2): Refine pattern.
521 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
522 Chung-Ju Wu <jasonwucj@gmail.com>
524 * config/nds32/iterators.md (shift_rotate): New code iterator.
525 (shift): New code attribute.
526 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
527 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
528 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
529 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
531 (andsi3, *andsi3): Ditto.
532 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
533 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
534 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
535 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
536 nds32_ior_operand, nds32_xor_operand): New predicates.
538 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
540 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
541 (addsi3, subsi3): ... this.
543 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
545 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
547 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
549 * config/nds32/nds32.md: Adjust indention.
551 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
553 * config/nds32/nds32.md (feature): New attribute.
555 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
557 * config/nds32/nds32.md (subtype): New attribute.
559 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
562 * config/arm/arm-builtins.c (arm_expand_builtin): Change
563 expansion to perform a bitwise AND of the argument followed by a
564 boolean negation of the result.
566 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
568 PR rtl-optimization/84878
569 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
570 the basic block. Assert the use reference is not artificial and that
571 it has an associated insn.
573 2018-04-04 Michael Matz <matz@suse.de>
575 * builtins.c (compute_objsize): Pass correct operand
576 to array_at_struct_end_p.
578 2018-04-04 Richard Biener <rguenther@suse.de>
581 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
582 from contexts for DINFO_LEVEL_TERSE and below.
584 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
586 * config/nds32/nds32-doubleword.md (move_<mode>): Require
587 resiter_operand condition.
588 * config/nds32/nds32.md (*move<mode>): Ditto.
590 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
591 Monk Chiang <sh.chiang04@gmail.com>
593 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
595 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
597 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
599 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
600 Kito Cheng <kito.cheng@gmail.com>
602 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
603 nds32_cond_code_str, output_cond_branch,
604 output_cond_branch_compare_zero, nds32_expand_cbranch,
605 nds32_expand_cstore, nds32_expand_movcc,
606 nds32_output_cbranchsi4_equality_zero,
607 nds32_output_cbranchsi4_equality_reg,
608 nds32_output_cbranchsi4_equality_reg_or_const_int,
609 nds32_output_cbranchsi4_greater_less_zero: New functions.
610 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
611 nds32_expand_cstore, nds32_expand_movcc,
612 nds32_output_cbranchsi4_equality_zero,
613 nds32_output_cbranchsi4_equality_reg,
614 nds32_output_cbranchsi4_equality_reg_or_const_int,
615 nds32_output_cbranchsi4_greater_less_zero): Declare.
616 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
617 nds32_rimm11s_operand): New predicates.
618 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
619 * config/nds32/nds32.md: Rewrite all the branch and conditional move
622 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
624 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
625 * config/nds32/nds32.md: Ditto.
626 * config/nds32/pipelines.md: Ditto.
628 2018-04-04 Richard Biener <rguenther@suse.de>
630 PR tree-optimization/85168
631 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
632 propagating abnormals.
634 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
636 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
638 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
639 Kito Cheng <kito.cheng@gmail.com>
641 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
642 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
643 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
644 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
645 * config/nds32/nds32.md (sibcall_internal): New.
646 (sibcall_register): Remove.
647 (sibcall_immediate): Remove.
648 (sibcall_value_internal): New.
649 (sibcall_value_register): Remove.
650 (sibcall_value_immediate): Remove.
651 * config/nds32/predicates.md (nds32_general_register_operand): New.
652 (nds32_call_address_operand): New.
654 2018-04-03 Jakub Jelinek <jakub@redhat.com>
656 PR rtl-optimization/85167
657 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
658 bb_defs if *split_p, instead preinitialize it to NULL.
660 PR tree-optimization/85156
661 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
662 evaluating the argument multiple times.
664 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
666 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
668 (_mm_cvtpd_ps): Likewise.
669 (_mm_cvttpd_epi32): Likewise.
670 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
671 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
672 vector, pixel, and bool following altivec.h include.
674 2018-04-03 Martin Sebor <msebor@redhat.com>
676 * doc/extend.texi (Common Function Attributes): Clarify.
677 (const attribute): Likewise.
678 (pure attribute): Likewise.
680 2018-04-03 Jakub Jelinek <jakub@redhat.com>
683 * config/i386/i386.c (ix86_expand_vector_set): Use
684 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
686 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
688 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
689 instructions when changing rounding bits to preserve precision bits
690 in the x87 control word.
692 2018-04-03 Martin Liska <mliska@suse.cz>
694 PR tree-optimization/82491
695 * rtl.h (strip_offset_and_add): Replace += suboffset with
696 poly_uint64 () + suboffset.
698 2018-03-29 Martin Liska <mliska@suse.cz>
699 Martin Jambor <mjambor@suse.cz>
702 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
703 param_type is not an integral or pointer type.
705 2018-04-03 Richard Biener <rguenther@suse.de>
707 * sese.h (recompute_all_dominators): Remove.
709 2018-04-02 Martin Sebor <msebor@redhat.com>
711 * doc/invoke.texi (-Wrestrict): Fix typos.
713 2018-04-02 Jim Wilson <jimw@sifive.com>
715 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
716 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
717 (<optab>di3, <optab>si3_extend): Likewise.
718 (<optab>si3_mask, <optab>si3_mask_1): New.
719 (<optab>di3_mask, <optab>di3_mask_1): New.
720 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
721 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
722 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
724 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
726 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
729 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
731 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
732 (nds32_canonicalize_comparison): New function.
734 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
735 Kito Cheng <kito.cheng@gmail.com>
736 Kuan-Lin Chen <kuanlinchentw@gmail.com>
738 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
739 * config/nds32/constants.md (unspec_volatile_element): Add
740 UNSPEC_VOLATILE_RELAX_GROUP.
741 * config/nds32/nds32-relax-opt.c: New file.
742 * config/nds32/nds32-predicates.c
743 (nds32_symbol_load_store_p): New function.
744 * config/nds32/nds32-protos.h
745 (nds32_symbol_load_store_p): Declare function.
746 (make_pass_nds32_relax_opt): Declare new rtl pass function.
747 * config/nds32/nds32.c
748 (nds32_register_pass): New function to register pass.
749 (nds32_register_passes): New function to register passes.
750 * config/nds32/nds32.md (relax_group): New pattern.
751 * config/nds32/nds32.opt (mrelax-hint): New option.
752 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
754 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
756 * config/nds32/t-nds32: Modify files dependency.
758 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
760 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
761 (PROFILE_HOOK): Define its implementation.
763 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
765 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
766 type and 32-bit size.
768 2018-04-01 Jakub Jelinek <jakub@redhat.com>
771 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
772 (V_128_256): New mode iterator.
773 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
774 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
775 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
777 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
778 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
780 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
783 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
784 NaN inputs correctly.
786 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
789 * config/rs6000/vsx.md (??r): New mode attribute.
790 (*vsx_mov<mode>_64bit): Use it.
791 (*vsx_mov<mode>_32bit): Likewise.
793 2018-03-30 Martin Sebor <msebor@redhat.com>
795 PR tree-optimization/84818
796 * builtins.c (check_access): Use warning_n.
798 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
801 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
803 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
806 2018-03-30 Julia Koval <julia.koval@intel.com>
809 * x86-tune.def (movx, partial_reg_dependency): Enable for
812 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
815 * lra-constraints.c (process_alt_operands): Move setting
816 this_alternative_matches below.
818 2018-03-29 Martin Liska <mliska@suse.cz>
821 * doc/invoke.texi: Document how LTO works with debug info.
822 Describe auto-load support of binutils. Mention 'x86-64'
823 as valid option value of -march option.
825 2018-03-29 Jakub Jelinek <jakub@redhat.com>
827 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
830 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
831 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
832 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
835 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
838 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
839 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
840 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
841 for RS6000_BTM_POWERPC64.
842 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
843 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
844 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
849 2018-03-28 Carl Love <cel@us.ibm.com>
852 2017-09-27 Carl Love <cel@us.ibm.com>
854 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
855 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
856 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
859 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
861 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
862 instead of __vector bool.
863 (_mm_max_pu8): Likewise.
864 (_mm_min_pi16): Likewise.
866 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
869 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
873 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
875 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
876 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
877 (div_extend): Likewise.
878 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
880 (__builtin_divweuo): Likewise.
881 (__builtin_divdeo): Likewise.
882 (__builtin_divdeuo): Likewise.
884 2018-03-28 Jakub Jelinek <jakub@redhat.com>
887 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
888 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
890 PR tree-optimization/82004
891 * gimple-match-head.c (optimize_pow_to_exp): New function.
892 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
893 Don't fold to exp if optimize_pow_to_exp is false.
895 2018-03-28 Martin Liska <mliska@suse.cz>
898 * calls.c (initialize_argument_information): Fix trailing space.
899 * common.opt: Fix typo and provide better explanation for
900 -fsanitize-coverage option.
901 * config/i386/i386.opt: Fix typo.
903 2018-03-28 Jakub Jelinek <jakub@redhat.com>
904 Martin Liska <mliska@suse.cz>
907 * gimplify.c (asan_poison_variable): Don't do the check for
908 gimplify_omp_ctxp here.
909 (gimplify_decl_expr): Do it here.
910 (gimplify_target_expr): Likewise.
912 2018-03-28 Martin Liska <mliska@suse.cz>
915 * config/i386/i386.c (ix86_function_arg_advance): Do not call
916 chkp_type_bounds_count if MPX is not enabled.
918 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
920 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
922 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
925 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
926 function to create the function decl for complex long double
927 multiply and divide for -mabi=ieeelongdouble.
928 (init_float128_ieee): Call it.
930 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
933 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
934 -fcf-protection=branch -mibt.
935 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
937 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
940 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
942 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
945 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
946 extern array declarations.
948 2018-03-27 Richard Biener <rguenther@suse.de>
951 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
952 explicit single_use checks.
954 2018-03-27 Richard Biener <rguenther@suse.de>
956 PR tree-optimization/85082
957 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
960 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
962 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
963 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
964 Turn on fasynchronous-unwind-tables and funwind-tables.
966 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
969 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
970 (*bmi_blsr_<mode>_ccz): Ditto.
972 2018-03-26 Tom de Vries <tom@codesourcery.com>
974 PR tree-optimization/85063
975 * omp-general.c (offloading_function_p): New function. Factor out
977 * omp-offload.c (pass_omp_target_link::gate): ... here.
978 * omp-general.h (offloading_function_p): Declare.
979 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
980 with attribute omp declare target for offloading functions.
982 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
984 PR tree-optimization/84005
985 * tree-data-ref.h (get_base_for_alignment): Declare.
986 * tree-data-ref.c (get_base_for_alignment_1): New function.
987 (get_base_for_alignment): Likewise.
988 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
989 get_base_for_alignment to find a suitable base object, instead
990 of always using drb->base_address.
992 2018-03-23 Jakub Jelinek <jakub@redhat.com>
995 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
996 known size by default.
998 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
1001 * lra-constraints.c (process_alt_operands): Don't match BLKmode
1002 and non BLKmode operands.
1004 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1007 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
1008 Clean up attributes.
1010 2018-03-23 Richard Biener <rguenther@suse.de>
1013 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
1014 we are going to emit early debug for LTO.
1016 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1019 * function.c (match_asm_constraints_1): Don't optimize if input
1020 doesn't satisfy general_operand predicate for output's mode.
1023 * alias.c (write_dependence_p): Don't require for x_canonicalized
1024 non-VOIDmode if x has VOIDmode.
1027 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
1028 just don't try to optimize it rather than assert it never happens.
1030 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1032 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
1033 macro expansions for definition of ST_INTERNAL_<mode> and
1034 LD_INTERNAL_<mode> builtins.
1035 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
1037 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
1039 (altivec_expand_st_builtin): Likewise.
1040 (altivec_expand_builtin): Remove calls to deleted functions.
1041 (rs6000_address_for_altivec): Delete this function.
1042 * config/rs6000/vector.md: Remove expands for
1043 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
1045 2018-03-22 Sudakshina Das <sudi.das@arm.com>
1048 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
1049 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
1050 re-computing once computed.
1051 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
1052 (arm_init_machine_status): Initialize
1053 machine->static_chain_stack_bytes.
1055 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1058 * doc/extend.texi: Add four new prototypes for vec_ld.
1059 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
1060 definitions for more logical presentation.
1061 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
1062 entries for V1TI variants of __builtin_altivec_ld builtin.
1063 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
1064 handling of V1TI variant of LVX icode pattern.
1065 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
1066 (rs6000_gimple_fold_builtin): Likewise.
1067 (altivec_init_builtins): Add code to define
1068 __builtin_altivec_lvx_v1ti function.
1070 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1073 * function.c (match_asm_constraints_1): Don't do the optimization
1074 if input isn't a REG, SUBREG, MEM or constant.
1076 2018-03-22 Tom de Vries <tom@codesourcery.com>
1078 PR tree-optimization/84956
1079 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
1080 bb_has_abnormal_pred.
1082 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1085 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
1086 DECL_INITIAL (decl) to decl at the end.
1087 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
1090 2018-03-21 Joseph Myers <joseph@codesourcery.com>
1092 * doc/extend.texi (__builtin_tgmath): Document when complex
1093 integer types are treated as _Complex _Float64.
1095 2018-03-21 Tom de Vries <tom@codesourcery.com>
1097 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
1099 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1101 PR tree-optimization/84960
1102 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
1103 if it is ENTRY block, move them into single succ of ENTRY in that case.
1105 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
1107 PR tree-optimization/84811
1108 * poly-int.h (poly_span_traits): Remove the T3 parameter and
1109 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
1110 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
1111 (known_subrange_p): Update accordingly. Cast each value involved
1112 in the size comparison, rather than casting the result of the
1115 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1117 PR tree-optimization/84982
1118 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
1119 by flipping the least significant bit rather than all bits from
1120 bitpos to bitpos + bitsize - 1.
1122 2018-03-21 Nathan Sidwell <nathan@acm.org>
1124 * doc/extend.texi (Deprecated Features): Remove mention of
1125 long-deleted deprecations.
1127 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1130 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
1131 * configure: Regenerate.
1133 2018-03-21 Tom de Vries <tom@codesourcery.com>
1135 PR tree-optimization/83126
1136 * tree-parloops.c (num_phis): New function.
1137 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
1139 2018-03-21 Nathan Sidwell <nathan@acm.org>
1141 * doc/extend.texi (Deprecated Features): Update deprecated flags,
1142 mention anon-struct/union members and trailing attributes.
1144 2018-03-21 Bin Cheng <bin.cheng@arm.com>
1146 PR tree-optimization/84969
1147 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
1148 builtin memset partitions if they set different rhs values.
1150 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1152 PR rtl-optimization/84989
1153 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
1154 VEC_DUPLICATE with scalar result mode.
1156 2018-03-21 Martin Liska <mliska@suse.cz>
1159 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
1160 not intended return statement.
1162 2018-03-21 Martin Liska <mliska@suse.cz>
1165 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
1166 (chkp_find_bound_slots_1): Limit number of iterations.
1168 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
1171 * Minor grammar fixes for x86 options.
1173 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1176 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
1177 holding REG_CFA_RESTORE notes, instead turn them into a USE.
1179 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
1182 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
1183 (altivec_lvx_<mode>_1op): Likewise.
1184 (altivec_stvx_<mode>_2op): Likewise.
1185 (altivec_stvx_<mode>_1op): Likewise.
1186 (altivec_lvx_<VM2:mode>): New define_expand.
1187 (altivec_stvx_<VM2:mode>): Likewise.
1188 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
1189 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1190 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
1191 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1192 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
1193 (rs6000_gen_lvx): Likewise.
1194 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
1195 (altivec_expand_stv_builtin): Likewise.
1196 (altivec_expand_builtin): Likewise.
1197 * config/rs6000/vector.md: Likewise.
1199 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1202 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
1205 2018-03-20 Richard Biener <rguenther@suse.de>
1208 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
1209 sign-conversions as zero, fall back to standard scalar_stmt
1212 2018-03-20 Martin Liska <mliska@suse.cz>
1215 * predict.c (rebuild_frequencies): Handle case when we have
1216 PROFILE_ABSENT, but flag_guess_branch_prob is false.
1218 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1221 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
1222 flag_section_anchors.
1223 * varasm.c (use_blocks_for_decl_p): Remove hack for
1224 dw2_force_const_mem.
1227 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
1229 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
1230 be created, use lowpart_subreg of operands[0] rather than operands[0]
1232 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
1233 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
1234 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
1235 and n constraint instead of aarch64_shift_imm_di and Usd.
1236 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
1237 (*aarch64_<optab>_reg_minus<mode>3): ... this.
1239 2018-03-20 Sudakshina Das <sudi.das@arm.com>
1242 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
1243 to favor GPR over NEON registers.
1244 (<shift>di3_neon): Likewise.
1246 2018-03-20 Tom de Vries <tom@codesourcery.com>
1249 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
1250 (nvptx_process_pars): Emit bar.sync asap and alap.
1252 2018-03-20 Tom de Vries <tom@codesourcery.com>
1255 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
1256 seen_label if seen_label is already set.
1258 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1261 * config/i386/i386.c (fold_builtin_cpu): For features above 31
1262 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
1263 Use 1U instead of 1. Formatting fixes.
1266 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
1267 instead of TREE_TYPE (s1) for the return value.
1269 2018-03-19 Jakub Jelinek <jakub@redhat.com>
1271 PR tree-optimization/84946
1272 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
1273 bitsize + bitsize in poly_uint64 rather than poly_int64.
1276 * dwarf2asm.c: Include fold-const.c.
1277 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
1278 of decl rather than decl itself.
1280 PR rtl-optimization/84643
1281 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
1283 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
1286 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
1287 calling assemble_variable.
1289 2018-03-19 Sudakshina Das <sudi.das@arm.com>
1292 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
1293 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
1295 2018-03-19 Jim Wilson <jimw@sifive.com>
1298 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
1299 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
1300 (riscv_first_stack_step): Likewise.
1301 (riscv_option_override): Use STACK_BOUNDARY instead of
1303 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
1305 (BIGGEST_ALIGNMENT): Set to 128.
1306 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
1307 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
1310 2018-03-19 Richard Biener <rguenther@suse.de>
1312 PR tree-optimization/84933
1313 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
1314 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
1316 2018-03-19 Richard Biener <rguenther@suse.de>
1318 PR tree-optimization/84859
1319 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
1320 (cond_if_else_store_replacement): Perform sinking operation on
1321 single-store BBs regardless of MAX_STORES_TO_SINK setting.
1322 Generalize what a BB with a single eligible store is.
1324 2018-03-19 Richard Biener <rguenther@suse.de>
1326 PR tree-optimization/84929
1327 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
1328 chrec_is_positive against non-chrec arg.
1330 2018-03-19 Tamar Christina <tamar.christina@arm.com>
1333 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
1335 2018-03-18 Martin Liska <mliska@suse.cz>
1337 PR rtl-optimization/84635
1338 * regrename.c (build_def_use): Use matches_mode only when
1341 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
1343 PR tree-optimization/84913
1344 * tree-vect-loop.c (vectorizable_reduction): Don't try to
1345 vectorize chains of COND_EXPRs.
1347 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1349 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
1351 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1353 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
1355 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1357 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
1359 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1360 Kito Cheng <kito.cheng@gmail.com>
1362 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
1363 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
1364 (nds32_adjust_reg_alloc_order): New function.
1365 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
1367 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
1369 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
1370 nds32_print_operand, nds32_print_operand_address): Use
1371 HOST_WIDE_INT_PRINT_DEC instead.
1373 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1375 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
1377 2018-03-17 Jakub Jelinek <jakub@redhat.com>
1380 * config/i386/i386.c (initial_ix86_tune_features,
1381 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
1383 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
1384 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
1385 rather than 1u << ix86_tune. Formatting fix.
1386 (ix86_option_override_internal): Change ix86_arch_mask from
1387 unsigned int to unsigned HOST_WIDE_INT, initialize to
1388 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
1389 (ix86_function_specific_restore): Likewise.
1391 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1394 * postreload.c (reload_combine_recognize_pattern): Perform
1395 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
1396 truncate_int_for_mode the result for the destination's mode.
1399 * hsa-gen.c (mem_type_for_type): Fix comment typo.
1400 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
1402 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1405 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
1408 * lra-assigns.c (lra_split_hard_reg_for): Don't use
1409 regno_allocno_class_array and sorted_pseudos.
1410 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
1411 insns where regno is used.
1413 2018-03-16 Martin Liska <mliska@suse.cz>
1416 * multiple_target.c (create_dispatcher_calls): Redirect
1417 reference in the symbol table.
1419 2018-03-16 Martin Liska <mliska@suse.cz>
1422 * multiple_target.c (create_dispatcher_calls): Redirect also
1425 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1429 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
1430 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
1431 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
1433 2018-03-16 Julia Koval <julia.koval@intel.com>
1435 * doc/invoke.texi (Skylake Server): Add CLWB.
1436 Cannonlake): Remove CLWB.
1438 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1440 PR tree-optimization/84841
1441 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
1443 (FLOAT_ONE_CONST_TYPE): Define.
1444 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
1445 (sort_by_operand_rank): Put entries with higher constant_type last
1446 rather than first to match comments.
1448 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
1450 * config/nios2/nios2.md (movsi_internal): Fix thinko in
1453 2018-03-15 Jakub Jelinek <jakub@redhat.com>
1456 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
1457 check and use address of target always.
1459 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
1462 * config/i386/i386.c (indirect_thunk_needed): Update comments.
1463 (indirect_thunk_bnd_needed): Likewise.
1464 (indirect_thunks_used): Likewise.
1465 (indirect_thunks_bnd_used): Likewise.
1466 (indirect_return_needed): New.
1467 (indirect_return_bnd_needed): Likewise.
1468 (output_indirect_thunk_function): Add a bool argument for
1470 (output_indirect_thunk_function): Don't generate alias for
1471 function return thunk.
1472 (ix86_code_end): Call output_indirect_thunk_function to generate
1473 function return thunks.
1474 (ix86_output_function_return): Set indirect_return_bnd_needed
1475 and indirect_return_needed instead of indirect_thunk_bnd_needed
1476 and indirect_thunk_needed.
1478 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
1480 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
1481 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
1482 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
1484 2018-03-15 David Malcolm <dmalcolm@redhat.com>
1485 Paul Hua <paul.hua.gm@gmail.com>
1488 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
1490 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
1492 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
1493 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
1496 2018-03-15 Tamar Christina <tamar.christina@arm.com>
1499 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
1500 instead of GET_MODE_SIZE when comparing Units.
1502 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1505 * varasm.c (hash_section): Return an unchangeble hash value
1506 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
1507 Return !aarch64_can_use_per_function_literal_pools_p ().
1509 2018-03-15 Jakub Jelinek <jakub@redhat.com>
1512 * optabs.c (emit_conditional_move): Pass address of cmode's copy
1513 rather than address of cmode as last argument to prepare_cmp_insn.
1515 2018-03-15 Julia Koval <julia.koval@intel.com>
1517 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
1518 F_AVX512VNNI, F_AVX512BITALG): New.
1520 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
1523 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
1524 insn for floating-point loads and stores.
1526 2018-03-14 Carl Love <cel@us.ibm.com>
1528 * config/rs6000/rs6000-c.c: Add macro definitions for
1529 ALTIVEC_BUILTIN_VEC_PERMXOR.
1530 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
1531 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
1532 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
1533 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
1535 * config/doc/extend.texi: Add prototypes for vec_permxor.
1537 2018-03-14 David Malcolm <dmalcolm@redhat.com>
1540 * diagnostic-show-locus.c (class layout_point): Convert m_line
1541 from int to linenum_type.
1542 (line_span::comparator): Use linenum "compare" function when
1543 comparing line numbers.
1544 (test_line_span): New function.
1545 (layout_range::contains_point): Convert param "row" from int to
1547 (layout_range::intersects_line_p): Likewise.
1548 (layout::will_show_line_p): Likewise.
1549 (layout::print_source_line): Likewise.
1550 (layout::should_print_annotation_line_p): Likewise.
1551 (layout::print_annotation_line): Likewise.
1552 (layout::print_leading_fixits): Likewise.
1553 (layout::annotation_line_showed_range_p): Likewise.
1554 (struct line_corrections): Likewise for field m_row.
1555 (line_corrections::line_corrections): Likewise for param "row".
1556 (layout::print_trailing_fixits): Likewise.
1557 (layout::get_state_at_point): Likewise.
1558 (layout::get_x_bound_for_row): Likewise.
1559 (layout::print_line): Likewise.
1560 (diagnostic_show_locus): Likewise for locals "last_line" and
1562 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
1563 * input.c (selftest::test_linenum_comparisons): New function.
1564 (selftest::input_c_tests): Call it.
1565 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
1566 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
1567 * selftest.h (ASSERT_GT): New macro.
1568 (ASSERT_GT_AT): New macro.
1569 (ASSERT_LT): New macro.
1570 (ASSERT_LT_AT): New macro.
1572 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
1574 PR rtl-optimization/84780
1575 * combine.c (distribute_links): Don't make a link based on pc_rtx.
1577 2018-03-14 Martin Liska <mliska@suse.cz>
1579 * tree.c (record_node_allocation_statistics): Use
1580 get_stats_node_kind.
1581 (get_stats_node_kind): New function extracted from
1582 record_node_allocation_statistics.
1583 (free_node): Use get_stats_node_kind.
1585 2018-03-14 Richard Biener <rguenther@suse.de>
1587 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
1588 that the value-set of ANTIC_IN doesn't grow.
1591 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1593 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1594 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1596 2018-03-14 Julia Koval <julia.koval@intel.com>
1598 * config.gcc (icelake-client, icelake-server): New.
1600 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
1601 (initial_ix86_arch_features): Ditto.
1602 (PTA_SKYLAKE): Add SGX.
1603 (PTA_ICELAKE): Remove.
1604 (PTA_ICELAKE_CLIENT): New.
1605 (PTA_ICELAKE_SERVER): New.
1606 (ix86_option_override_internal): Split up icelake on icelake client and
1608 (get_builtin_code_for_version): Ditto.
1609 (fold_builtin_cpu): Ditto.
1610 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
1611 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
1612 * config/i386/i386.h (processor_type): Ditto.
1613 * doc/invoke.texi: Ditto.
1615 2018-03-14 Jakub Jelinek <jakub@redhat.com>
1618 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
1619 INTEGER_CST offset, add it together with bitpos / 8 and
1620 sign extend based on POINTER_SIZE.
1624 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
1627 * config/i386/constraints.md (Yc): New register constraint.
1628 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
1629 Use Yc constraint for alternative 2 of operand 0. Remove
1630 preferred_for_speed attribute.
1632 2018-03-14 Richard Biener <rguenther@suse.de>
1634 PR tree-optimization/84830
1635 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
1636 with the old one to avoid oscillations.
1638 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
1641 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
1643 (assign_by_spills): Return a flag of reload assignment failure.
1644 Do not process the reload assignment failures. Do not spill other
1645 reload pseudos if they has the same reg class. Update n if
1647 (lra_assign): Add a return arg. Set up from the result of
1648 assign_by_spills call.
1649 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1650 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1651 usage_insns if it is not NULL.
1652 (spill_hard_reg_in_range): New function.
1653 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1654 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1655 function prototypes.
1656 (lra_assign): Change prototype.
1657 * lra.c (lra): Add code to deal with fails by splitting hard reg
1660 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
1662 * config/riscv/riscv.opt (mrelax): New option.
1663 * config/riscv/riscv.c (riscv_file_start): Emit ".option
1664 "norelax" when riscv_mrelax is disabled.
1665 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
1667 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1670 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
1671 reassociation for int modes.
1673 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1675 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
1676 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
1678 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
1679 * config/aarch64/aarch64-sve.md
1680 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
1681 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
1682 (*extend<mode><Vwide>2): Rename to...
1683 (aarch64_sve_extend<mode><Vwide>2): ...this.
1684 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
1685 renaming the old pattern to...
1686 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
1688 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
1689 define_expand, renaming the old pattern to...
1690 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
1691 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
1692 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
1693 account when deciding which SVE instruction the optab should use.
1694 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
1696 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1698 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
1699 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
1700 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
1701 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
1702 (tlsdesc_small_<mode>): Turn a define_expand and use
1703 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
1704 (tlsdesc_small_advsimd_<mode>): ...this.
1705 (tlsdesc_small_sve_<mode>): New pattern.
1707 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1709 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
1710 (UNSPEC_UMUL_HIGHPART): New constants.
1711 (MUL_HIGHPART): New int iteraor.
1712 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
1713 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
1715 (*<su>mul<mode>3_highpart): New define_insn.
1717 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1720 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1723 2018-03-13 Martin Liska <mliska@suse.cz>
1726 * (sem_item_optimizer::sem_item_optimizer): Initialize new
1728 (sem_item_optimizer::~sem_item_optimizer): Release it.
1729 (sem_item_optimizer::merge_classes): Register variable aliases.
1730 (sem_item_optimizer::fixup_pt_set): New function.
1731 (sem_item_optimizer::fixup_points_to_sets): Likewise.
1732 * ipa-icf.h: Declare new variables and functions.
1734 2018-03-13 Jakub Jelinek <jakub@redhat.com>
1737 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
1738 integer_pow2p@2 and test integer_pow2p in condition.
1739 (A < 0 ? C : 0): Similarly for @1.
1742 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
1743 characters starting at p contain '\0' character, don't look beyond
1747 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
1748 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
1751 * reg-stack.c (change_stack): Change update_end var from int to
1752 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
1753 also call set_block_for_insn on the newly added insns and rescan.
1756 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
1757 on the last operand.
1760 * tree.c (stabilize_reference_1): Return save_expr (e) for
1761 STATEMENT_LIST even if it doesn't have side-effects.
1763 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
1765 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
1767 2018-03-12 Renlin Li <renlin.li@arm.com>
1769 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
1770 aarch64_output_scalar_simd_mov_immediate.
1772 2018-03-12 Martin Sebor <msebor@redhat.com>
1774 PR tree-optimization/83456
1775 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
1776 for perfectly overlapping calls to memcpy.
1777 (gimple_fold_builtin_memory_chk): Same.
1778 (gimple_fold_builtin_strcpy): Handle no-warning.
1779 (gimple_fold_builtin_stxcpy_chk): Same.
1780 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
1782 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
1784 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
1785 parameter. Use it for SFmode.
1786 (rs6000_function_arg_advance_1): Adjust.
1787 (rs6000_function_arg): Adjust.
1788 (rs6000_gimplify_va_arg): Pass false for that new parameter.
1790 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
1792 PR rtl-optimization/84169
1793 PR rtl-optimization/84780
1794 * combine.c (can_combine_p): Check for a 2-insn combination whether
1795 the destination register is used between the two insns, too.
1797 2018-03-12 Richard Biener <rguenther@suse.de>
1799 PR tree-optimization/84803
1800 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
1801 for refs DR analysis didn't process.
1803 2018-03-12 Richard Biener <rguenther@suse.de>
1805 PR tree-optimization/84777
1806 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
1807 force-vectorize loops ignore whether we are optimizing for size.
1809 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
1811 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
1812 (TARGET_MD_ASM_ADJUST): Define.
1814 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
1815 Kito Cheng <kito.cheng@gmail.com>
1816 Chung-Ju Wu <jasonwucj@gmail.com>
1818 * config/nds32/nds32.c (nds32_compute_stack_frame,
1819 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
1820 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
1821 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
1822 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
1823 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
1824 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
1825 * config/nds32/nds32.md (prologue, epilogue): Use macro
1826 NDS32_V3PUSH_AVAILABLE_P to do checking.
1828 2018-03-11 Jakub Jelinek <jakub@redhat.com>
1831 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
1832 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
1833 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
1834 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
1835 addition of most attributes on !orig_type_die or the attribute not
1836 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
1838 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1839 Chung-Ju Wu <jasonwucj@gmail.com>
1841 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
1843 * config/nds32/nds32.opt (mvh): New option.
1845 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1846 Chung-Ju Wu <jasonwucj@gmail.com>
1848 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
1850 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
1851 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
1854 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1855 Chung-Ju Wu <jasonwucj@gmail.com>
1857 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
1859 * config/nds32/nds32-multiple.md (strlensi): New pattern.
1860 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
1862 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
1863 Kito Cheng <kito.cheng@gmail.com>
1864 Chung-Ju Wu <jasonwucj@gmail.com>
1866 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
1867 UNSPEC_FFMISM and UNSPEC_FLMISM.
1868 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
1869 for ffb, ffmism and flmism.
1870 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
1871 (unspec_ffmism): Ditto.
1872 (unspec_flmism): Ditto.
1873 (nds32_expand_builtin_impl): Check if string extension is available.
1874 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
1875 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
1877 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
1880 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1883 * lra-assigns.c (assign_by_spills): Return a flag of reload
1884 assignment failure. Do not process the reload assignment
1885 failures. Do not spill other reload pseudos if they has the same
1887 (lra_assign): Add a return arg. Set up from the result of
1888 assign_by_spills call.
1889 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1890 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1891 usage_insns if it is not NULL.
1892 (spill_hard_reg_in_range): New function.
1893 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1894 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1895 function prototypes.
1896 (lra_assign): Change prototype.
1897 * lra.c (lra): Add code to deal with fails by splitting hard reg
1900 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
1903 * config/i386/i386.opt: Replace Enforcment with Enforcement.
1905 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
1908 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
1909 (dw_val_node): Add val_symbolic_view.
1910 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
1911 (symview_upper_bound): New.
1912 (new_line_info_table): Initialize symviews_since_reset.
1913 (dwarf2out_source_line): Count symviews_since_reset and set
1914 symview_upper_bound.
1915 (dw_val_equal_p): Handle symview.
1916 (add_AT_symview): New.
1917 (print_dw_val): Handle symview.
1918 (attr_checksum, attr_checksum_ordered): Likewise.
1919 (same_dw_val_p, size_of_die): Likewise.
1920 (value_format, output_die): Likewise.
1921 (add_high_low_attributes): Use add_AT_symview for entry_view.
1922 (dwarf2out_finish): Reset symview_upper_bound, clear
1925 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
1928 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
1929 Add strict argument and use it.
1930 (rs6000_split_multireg_move): Update for new strict argument.
1931 (mem_operand_gpr): Disallow all non-offsettable addresses.
1932 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
1934 2018-03-09 Jakub Jelinek <jakub@redhat.com>
1937 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
1938 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
1939 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
1942 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
1943 decl, use remap_type if we want to use the type.
1945 2018-03-09 Martin Sebor <msebor@redhat.com>
1947 PR tree-optimization/84526
1948 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1950 (builtin_access::generic_overlap): Be prepared to handle non-array
1953 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
1955 PR rtl-optimization/84682
1956 * lra-constraints.c (process_address_1): Check is_address flag
1957 for address constraints.
1958 (process_alt_operands): Likewise.
1959 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
1960 preprocess_constraints.
1961 * recog.h (preprocess_constraints): Add oploc parameter.
1963 * recog.c (preprocess_constraints): Test address_operand for
1964 CT_ADDRESS constraints.
1966 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1969 * lra-assigns.c (assign_by_spills): Return a flag of reload
1970 assignment failure. Do not process the reload assignment
1971 failures. Do not spill other reload pseudos if they has the same
1973 (lra_assign): Add a return arg. Set up from the result of
1974 assign_by_spills call.
1975 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1976 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1977 usage_insns if it is not NULL.
1978 (spill_hard_reg_in_range): New function.
1979 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1980 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1981 function prototypes.
1982 (lra_assign): Change prototype.
1983 * lra.c (lra): Add code to deal with fails by splitting hard reg
1986 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1989 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
1990 Accept complain bool parameter. Only emit errors if it is true.
1991 (arm_parse_cpu_option_name): Likewise.
1992 (arm_target_thumb_only): Adjust callers of the above.
1993 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
1994 prototype to take a default true bool parameter.
1995 (arm_parse_arch_option_name): Likewise.
1997 2018-03-09 David Malcolm <dmalcolm@redhat.com>
1998 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
2002 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
2003 * configure: Regenerate.
2004 * configure.ac ("linker --version-script option"): New.
2005 ("linker soname option"): New.
2007 2018-03-09 Richard Biener <rguenther@suse.de>
2009 PR tree-optimization/84775
2010 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
2011 immediate uses of predicate stmts and mark them modified.
2014 PR tree-optimization/84178
2015 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2017 (version_loop_for_if_conversion): Delay update_ssa call.
2018 (tree_if_conversion): Delay update_ssa until after predicate
2021 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
2024 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
2025 when the function accesses prior frames.
2027 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2030 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
2031 gen_llsym, otherwise call maybe_gen_llsym.
2034 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
2035 has ',' character inside of it.
2037 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2040 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
2041 as clobbering CC_REGNUM.
2043 2018-03-08 Richard Biener <rguenther@suse.de>
2046 * tree-scalar-evolution.c: Include tree-into-ssa.h.
2047 (follow_copies_to_constant): Do not follow SSA names registered
2050 2018-03-08 Richard Biener <rguenther@suse.de>
2052 PR tree-optimization/84178
2053 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2055 (version_loop_for_if_conversion): Delay update_ssa call.
2056 (tree_if_conversion): Delay update_ssa until after predicate
2059 2018-03-08 David Malcolm <dmalcolm@redhat.com>
2061 PR tree-optimization/84178
2062 * tree-if-conv.c (release_bb_predicate): Remove the
2063 the assertion that the stmts have NULL use_ops.
2064 Discard the statements, asserting that they haven't
2065 yet been added to a BB.
2067 2018-03-08 Richard Biener <rguenther@suse.de>
2069 PR tree-optimization/84746
2070 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
2071 (phi_translate): Pass in destination ANTIC_OUT set.
2072 (phi_translate_1): Likewise. For a simplified result lookup
2073 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
2074 (phi_translate_set): Adjust.
2075 (do_pre_regular_insertion): Likewise.
2076 (do_pre_partial_partial_insertion): Likewise.
2078 2018-03-08 Martin Liska <mliska@suse.cz>
2080 PR gcov-profile/84735
2081 * doc/gcov.texi: Document usage of profile files.
2082 * gcov-io.h: Document changes in the format.
2084 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
2088 * dwarf2out.c (struct dw_line_info_table): Update comments for
2090 (FORCE_RESET_NEXT_VIEW): New.
2091 (FORCE_RESETTING_VIEW_P): New.
2092 (RESETTING_VIEW_P): Check for -1 too.
2093 (ZERO_VIEW_P): Likewise.
2094 (new_line_info_table): Force-reset next view.
2095 (dwarf2out_begin_function): Likewise.
2096 (dwarf2out_source_line): Simplify zero_view_p initialization.
2097 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
2098 view directly. Omit view when omitting .loc at line 0.
2100 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2102 PR tree-optimization/84740
2103 * tree-switch-conversion.c (process_switch): Call build_constructors
2104 only if info.phi_count is non-zero.
2106 PR tree-optimization/84739
2107 * tree-tailcall.c (find_tail_calls): Check call arguments against
2108 DECL_ARGUMENTS (current_function_decl) rather than
2109 DECL_ARGUMENTS (func) when checking for tail recursion.
2111 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2113 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
2114 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
2115 Volker Reichelt's entry and add entries for people that perform
2116 GCC fuzzy testing and report numerous bugs.
2118 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
2121 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
2122 readonly data in sdata, if that is disabled.
2123 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
2124 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
2125 -mreadonly-in-sdata option.
2127 2018-03-07 Martin Sebor <msebor@redhat.com>
2129 PR tree-optimization/84468
2130 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
2131 basic block when looking for nul assignment.
2133 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
2136 * except.h (output_function_exception_table): Adjust prototype.
2137 * except.c (output_function_exception_table): Remove FNNAME parameter
2138 and add SECTION parameter. Ouput one part of the table at a time.
2139 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
2140 the first part of the exception table and emit unwind directives.
2141 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
2142 (i386_pe_seh_cold_init): Likewise.
2143 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
2144 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
2145 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
2146 (ix86_output_call_insn): Emit a nop in one more case for SEH.
2147 * config/i386/winnt.c: Include except.h.
2148 (struct seh_frame_state): Add reg_offset, after_prologue and
2149 in_cold_section fields.
2150 (i386_pe_seh_end_prologue): Set seh->after_prologue.
2151 (i386_pe_seh_cold_init): New function.
2152 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
2153 to seh->in_cold_section.
2154 (seh_emit_push): Record the offset of the push.
2155 (seh_emit_save): Record the offet of the save.
2156 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
2157 Test seh->after_prologue to disregard the epilogue.
2158 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
2159 (i386_pe_end_cold_function): New function.
2161 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2164 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
2165 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
2168 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
2170 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
2171 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
2174 * multiple_target.c: Include tree-inline.h and intl.h.
2175 (expand_target_clones): Diagnose and fail if node->definition and
2176 !tree_versionable_function_p (node->decl).
2178 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
2180 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
2182 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
2183 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
2184 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
2186 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2189 * combine.c (try_combine): Use reg_or_subregno instead of handling
2190 just paradoxical SUBREGs and REGs.
2192 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
2194 * config/arc/arc.c (arc_finalize_pic): Remove function.
2195 (arc_must_save_register): We use single base PIC register, remove
2196 checks to save/restore the PIC register.
2197 (arc_expand_prologue): Likewise.
2198 * config/arc/arc-protos.h (arc_set_default_type_attributes):
2200 (arc_verify_short): Likewise.
2201 (arc_attr_type): Likewise.
2202 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
2203 (walk_stores): Likewise.
2204 (arc_address_cost): Make it static.
2205 (arc_verify_short): Likewise.
2206 (branch_dest): Likewise.
2207 (arc_attr_type): Likewise.
2208 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
2209 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
2210 (arc_final_prescan_insn): Remove inserting the nops due to
2211 hardware hazards. It is done in reorg step.
2212 (insn_length_variant_t): Remove.
2213 (insn_length_parameters_t): Likewise.
2214 (arc_insn_length_parameters): Likewise.
2215 (arc_get_insn_variants): Likewise.
2216 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
2218 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2221 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
2224 PR tree-optimization/84687
2225 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
2227 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
2229 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2231 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
2232 Rename to ppc_speculation_barrier.
2233 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
2234 __builtin_ppc_speculation_barrier.
2236 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2239 * combine.c (combine_simplify_rtx): Don't try to simplify if
2240 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
2243 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
2245 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
2246 to 32 bytes when compiling for POWER9.
2248 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2251 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
2252 regparm >= 3 with no arg reg available also for calls with
2253 flag_force_indirect_call. Pass decl to ix86_function_regparm.
2256 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
2258 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
2260 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
2263 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
2265 2018-03-05 Richard Biener <rguenther@suse.de>
2267 PR tree-optimization/84486
2268 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
2269 When inserting a __builtin_assume_aligned call set the LHS
2270 SSA name alignment info accordingly.
2272 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
2274 PR tree-optimization/84114
2275 * config/aarch64/aarch64.c (aarch64_reassociation_width)
2276 Avoid reassociation of FLOAT_MODE addition.
2278 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
2280 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
2281 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
2282 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
2283 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
2284 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
2285 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
2286 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
2288 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
2289 __builtin_ia32_wbinvd): New builtins.
2290 (SPECIAL_ARGS2): New.
2291 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
2292 (SPECIAL_ARGS2): New.
2293 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
2294 (ix86_valid_target_attribute_inner_p): Ditto.
2295 (ix86_init_mmx_sse_builtins): Add special_args2.
2296 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
2297 TARGET_WBNOINVD_P): New.
2298 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
2299 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
2300 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
2301 * config/i386/immintrin.h (_wbinvd): New intrinsic.
2302 * config/i386/pconfigintrin.h: New file.
2303 * config/i386/wbnoinvdintrin.h: Ditto.
2304 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
2305 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
2307 2018-03-05 Richard Biener <rguenther@suse.de>
2309 PR tree-optimization/84670
2310 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2312 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2313 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2314 (compute_antic_aux): Only assert the number of values in ANTIC_IN
2315 doesn't grow if all successors (recursively) were visited at least
2318 2018-03-05 Richard Biener <rguenther@suse.de>
2320 PR tree-optimization/84650
2321 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
2322 if executed in the loop pipeline.
2324 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
2326 * doc/configfiles.texi (Configuration Files): Move info about
2327 conditionalizing $target-protos.h to...
2328 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
2329 differs from $target-protos.h.
2331 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
2332 Chung-Ju Wu <jasonwucj@gmail.com>
2334 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
2335 * config/nds32/nds32-multiple.md (setmemsi): Define.
2336 * config/nds32/nds32-memory-manipulation.c
2337 (nds32_gen_dup_4_byte_to_word_value): New.
2338 (emit_setmem_word_loop): New.
2339 (emit_setmem_byte_loop): New.
2340 (nds32_expand_setmem_loop): New.
2341 (nds32_expand_setmem_loop_v3m): New.
2342 (nds32_expand_setmem_unroll): New.
2343 (nds32_expand_setmem): New.
2345 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2346 Chung-Ju Wu <jasonwucj@gmail.com>
2348 * config/nds32/nds32-memory-manipulation.c
2349 (nds32_emit_load_store): New.
2350 (nds32_emit_post_inc_load_store): New.
2351 (nds32_emit_mem_move): New.
2352 (nds32_emit_mem_move_block): New.
2353 (nds32_expand_movmemsi_loop_unknown_size): New.
2354 (nds32_expand_movmemsi_loop_known_size): New.
2355 (nds32_expand_movmemsi_loop): New.
2356 (nds32_expand_movmemsi_unroll): New.
2357 (nds32_expand_movmemqi): Rename ...
2358 (nds32_expand_movmemsi): ... to this.
2359 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
2360 (movmemsi): ... to this.
2361 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
2362 (nds32_expand_movmemsi): ... to this.
2364 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2365 Monk Chiang <sh.chiang04@gmail.com>
2366 Chung-Ju Wu <jasonwucj@gmail.com>
2368 * config/nds32/nds32-protos.h
2369 (nds32_expand_load_multiple): New arguments.
2370 (nds32_expand_store_multiple): Ditto.
2371 (nds32_valid_multiple_load_store): Rename ...
2372 (nds32_valid_multiple_load_store_p): ... to this.
2373 * config/nds32/nds32-memory-manipulation.c
2374 (nds32_expand_load_multiple): Refine implementation.
2375 (nds32_expand_store_multiple): Ditto.
2376 * config/nds32/nds32-multiple.md
2377 (load_multiple): Update nds32_expand_load_multiple interface.
2378 (store_multiple): Update nds32_expand_store_multiple interface.
2379 * config/nds32/nds32-predicates.c
2380 (nds32_valid_multiple_load_store): Rename ...
2381 (nds32_valid_multiple_load_store_p): ... to this and refine
2383 * config/nds32/predicates.md
2384 (nds32_load_multiple_and_update_address_operation): New predicate.
2385 (nds32_store_multiple_and_update_address_operation): New predicate.
2387 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2388 Chung-Ju Wu <jasonwucj@gmail.com>
2390 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
2391 (combo): New attribute.
2392 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
2394 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
2396 * config/nds32/nds32.opt: Change -mcmodel= default value.
2398 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
2399 Monk Chiang <sh.chiang04@gmail.com>
2400 Chung-Ju Wu <jasonwucj@gmail.com>
2402 * config/nds32/constants.md (unspec_element): New enum.
2403 * config/nds32/constraints.md (Umw): New constraint.
2404 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
2405 * config/nds32/nds32-intrinsic.md: Likewise.
2406 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
2407 (nds32_valid_smw_lwm_base_p): New.
2408 (nds32_output_smw_single_word): New.
2409 (nds32_output_lmw_single_word): New.
2410 (nds32_expand_unaligned_load): New.
2411 (nds32_expand_unaligned_store): New.
2412 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
2413 (nds32_output_smw_single_word): Declare.
2414 (nds32_output_lmw_single_word): Declare.
2415 (nds32_expand_unaligned_load): Declare.
2416 (nds32_expand_unaligned_store): Declare.
2417 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
2418 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
2419 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
2420 NDS32_BUILTIN_UASTORE_DW.
2421 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
2424 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
2425 Kito Cheng <kito.cheng@gmail.com>
2426 Chung-Ju Wu <jasonwucj@gmail.com>
2428 * config/nds32/nds32-intrinsic.c
2429 (nds32_expand_builtin_null_ftype_reg): Delete.
2430 (nds32_expand_builtin_reg_ftype_imm): Ditto.
2431 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
2432 (nds32_read_argument): New.
2433 (nds32_legitimize_target): Ditto.
2434 (nds32_legitimize_argument): Ditto.
2435 (nds32_check_constant_argument): Ditto.
2436 (nds32_expand_unop_builtin): Ditto.
2437 (nds32_expand_unopimm_builtin): Ditto.
2438 (nds32_expand_binop_builtin): Ditto.
2439 (nds32_builtin_decl_impl): Ditto.
2440 (builtin_description): Ditto.
2441 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
2442 (nds32_init_builtins_impl): Ditto.
2443 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
2444 (nds32_builtin_decl): New.
2445 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
2446 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
2448 2018-03-02 Jeff Law <law@redhat.com>
2450 * reorg.c (stop_search_p): Handle DEBUG_INSN.
2451 (redundant_insn, fill_simple_delay_slots): Likewise.
2452 (fill_slots_from_thread): Likewise.
2453 * resource.c (mark_referenced_resources): Likewise.
2454 (mark_set_resources, find_dead_or_set_registers): Likewise.
2456 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2458 * substring-locations.h (format_warning_va): Formatting fix for
2460 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
2462 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
2463 * substring-locations.c: Include intl.h.
2464 (format_warning_va): Turned into small wrapper around
2465 format_warning_n_va, renamed to ...
2466 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
2467 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
2469 (format_warning_at_substring_n): New function.
2470 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
2471 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
2472 format_warning_at_substring with just a shorter name instead of
2473 const function pointer.
2474 (fmtwarn_n): New function.
2475 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
2476 appropriate, get rid of all the fmtstr temporaries, move conditionals
2477 with G_() wrapped string literals directly into fmtwarn arguments,
2478 cast dir.len to (int), formatting fixes.
2480 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
2482 * doc/invoke.texi: Remove "Cilk Plus" references.
2484 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2485 Richard Biener <rguenther@suse.de>
2488 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
2489 for error or warning attributes if CALL_FROM_THUNK_P is set.
2492 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2495 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
2496 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
2499 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
2500 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
2502 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
2503 instead of -1U in last predictors element's probability member.
2505 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2508 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2509 arguments if they are comparable.
2511 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
2513 PR tree-optimization/84634
2514 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
2515 masks and masked_loop_p with a single loop_masks, making sure it's
2516 null for bb vectorization.
2518 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
2520 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
2521 (vect_analyze_data_ref_access): Use loop->safe_len rather than
2522 loop->force_vectorize to check whether there is no alias.
2524 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2527 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
2529 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
2531 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
2532 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
2533 instead of a loop around prev_real_insn.
2534 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
2538 * config/i386/i386.c (ix86_print_operand): Use conditional
2539 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
2542 2018-03-02 Richard Biener <rguenther@suse.de>
2544 PR tree-optimization/84427
2545 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
2546 (bitmap_set_subtract_values): Rewrite to handle multiple
2549 (prune_clobbered_mems): Likewise.
2550 (phi_translate): Take edge instead of pred/phiblock.
2551 (phi_translate_1): Likewise.
2552 (phi_translate_set): Likewise. Insert all translated
2553 exprs for a value into the set, keeping possibly multiple
2554 expressions per value.
2555 (compute_antic_aux): Adjust for phi_translate changes.
2556 When intersecting union the expressions and prune those
2557 not in the final value set, keeping possibly multiple
2558 expressions per value. Do not use value-insertion
2559 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
2560 all expressions. Add verification that the value-sets
2561 only shrink during iteration.
2562 (compute_partial_antic_aux): Adjust for the phi_translate changes.
2563 (do_pre_regular_insertion): Likewise.
2564 (do_pre_partial_partial_insertion): Likewise.
2566 2018-03-02 Richard Biener <rguenther@suse.de>
2569 * config/darwin.c (saved_debug_info_level): New static global.
2570 (darwin_asm_lto_start): Disable debug info generation for LTO out.
2571 (darwin_asm_lto_end): Restore debug info generation settings.
2573 2018-03-01 Martin Liska <mliska@suse.cz>
2576 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
2579 2018-03-01 Richard Biener <rguenther@suse.de>
2582 * dwarf2out.c (gen_variable_die): Properly handle late VLA
2583 type annotation with LTO when debug was disabled at compile-time.
2585 2018-03-01 Matthew Fortune <mfortune@gmail.com>
2587 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
2589 (mips_final_postscan_insn): Likewise.
2591 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
2593 PR rtl-optimization/84528
2594 * alias.c (init_alias_target): Add commentary.
2595 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
2596 a unique base value if the frame pointer is not eliminated
2597 to the stack pointer.
2599 2018-03-01 Tom de Vries <tom@codesourcery.com>
2601 PR rtl-optimization/83327
2602 * lra-int.h (hard_regs_spilled_into): Declare.
2603 * lra.c (hard_regs_spilled_into): Define.
2604 (init_reg_info): Init hard_regs_spilled_into.
2605 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
2606 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
2607 (process_bb_lives): Handle hard_regs_spilled_into.
2608 (lra_create_live_ranges_1): Before doing liveness propagation, clear
2609 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
2611 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
2613 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
2614 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
2615 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
2616 * config/rs6000/aix72.h: New file.
2618 2018-02-28 Jakub Jelinek <jakub@redhat.com>
2620 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
2621 instead of warning_at with conditional singular and plural messages
2625 * stor-layout.c (update_alignment_for_field): For
2626 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
2627 && !DECL_PACKED (field), do the alignment update, just use
2628 only desired_align instead of MAX (type_align, desired_align)
2630 (place_field): Don't do known_align < desired_align handling
2631 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
2632 is non-NULL, instead do it after rli->prev_field handling and
2633 only if not within a bitfield word. For DECL_PACKED (field)
2634 use type_align of BITS_PER_UNIT.
2636 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
2638 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
2639 superfluous parentheses and trailing spaces.
2641 2018-02-28 Richard Biener <rguenther@suse.de>
2643 PR tree-optimization/84584
2644 * graphite-scop-detection.c (scop_detection::add_scop): Discard
2645 SCoPs with fake exit edge.
2647 2018-02-28 Martin Liska <mliska@suse.cz>
2650 * timevar.c (timer::print): Fix format to properly print 100%
2653 2018-02-28 Richard Biener <rguenther@suse.de>
2656 * genmatch.c (capture_info::walk_match): Do not mark
2657 captured expressions without operands as expr_p given
2658 they act more like predicates and should be subject to
2659 "lost tail" side-effect preserving.
2661 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
2663 PR rtl-optimization/81611
2664 * auto-inc-dec.c (attempt_change): Move dead note from
2665 mem_insn if it's the next use of regno
2666 (find_address): Take address use of reg holding
2667 non-incremented value. Add parm to limit search to the named
2669 (merge_in_block): Attempt to use a mem insn that is the next
2670 use of the original regno.
2672 2018-02-27 Martin Sebor <msebor@redhat.com>
2675 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
2676 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
2678 2018-02-27 Martin Sebor <msebor@redhat.com>
2680 PR translation/84207
2681 * diagnostic-core.h (warning_n, error_n, inform_n): Change
2682 n argument to unsigned HOST_WIDE_INT.
2683 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
2684 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
2685 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
2686 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
2688 2018-02-27 Richard Biener <rguenther@suse.de>
2690 PR tree-optimization/84512
2691 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
2692 Do not use the estimate returned from record_stmt_cost for
2693 the scalar iteration cost but sum properly using add_stmt_cost.
2695 2018-02-27 Richard Biener <rguenther@suse.de>
2697 PR tree-optimization/84466
2698 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2699 Adjust last change to less strictly validate use operands.
2701 2018-02-27 Martin Liska <mliska@suse.cz>
2703 PR gcov-profile/84548
2704 * gcov.c (process_file): Allow partial overlap and consider it
2705 also as group functions.
2706 (output_lines): Properly calculate range of lines for a group.
2708 2018-02-27 Martin Liska <mliska@suse.cz>
2710 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
2711 'ggc' suffixes. Change first column width.
2712 (timer::print): Fix formatting of the column.
2714 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
2716 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
2717 preserve inline entry blocks for the sake of debug inline
2718 entry point markers alone.
2719 (remove_unused_locals): Suggest in comments a better place to
2720 force the preservation of inline entry blocks that are
2721 otherwise unused, but do not preserve them.
2723 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2725 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
2727 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2730 * config/i386/constraints.md (Bs): Replace
2731 ix86_indirect_branch_register with
2732 TARGET_INDIRECT_BRANCH_REGISTER.
2734 * config/i386/i386.md (indirect_jump): Likewise.
2735 (tablejump): Likewise.
2736 (*sibcall_memory): Likewise.
2737 (*sibcall_value_memory): Likewise.
2738 Peepholes of indirect call and jump via memory: Likewise.
2739 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
2740 (*sibcall_value_GOT_32): Likewise.
2741 * config/i386/predicates.md (indirect_branch_operand): Likewise.
2742 (GOT_memory_operand): Likewise.
2743 (call_insn_operand): Likewise.
2744 (sibcall_insn_operand): Likewise.
2745 (GOT32_symbol_operand): Likewise.
2746 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
2748 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2750 PR rtl-optimization/83496
2751 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
2752 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
2753 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
2754 redundant insn, if any.
2755 (relax_delay_slots): Likewise.
2756 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
2758 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
2760 PR tree-optimization/83965
2761 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
2762 that grouped statements are part of a reduction chain. Return
2763 true if the statement is not marked as a reduction itself but
2765 (vect_recog_dot_prod_pattern): Don't check whether the statement
2766 is part of a group here.
2767 (vect_recog_sad_pattern): Likewise.
2768 (vect_recog_widen_sum_pattern): Likewise.
2770 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2773 * final.c (rest_of_clean_state): Also look for calls inside sequences.
2775 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2778 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
2780 (ix86_output_indirect_function_return): New prototype.
2781 (ix86_split_simple_return_pop_internal): Likewise.
2782 * config/i386/i386.c (indirect_return_via_cx): New.
2783 (indirect_return_via_cx_bnd): Likewise.
2784 (indirect_thunk_name): Handle return va CX_REG.
2785 (output_indirect_thunk_function): Create alias for
2786 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
2787 (ix86_output_indirect_jmp): Remove the bool argument.
2788 (ix86_output_indirect_function_return): New function.
2789 (ix86_split_simple_return_pop_internal): Likewise.
2790 * config/i386/i386.md (*indirect_jump): Don't pass false
2791 to ix86_output_indirect_jmp.
2792 (*tablejump_1): Likewise.
2793 (simple_return_pop_internal): Change it to define_insn_and_split.
2794 Call ix86_split_simple_return_pop_internal to split it for
2796 (simple_return_indirect_internal): Call
2797 ix86_output_indirect_function_return instead of
2798 ix86_output_indirect_jmp.
2800 2018-02-26 Jakub Jelinek <jakub@redhat.com>
2803 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
2804 memset and value initialization afterwards.
2806 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
2808 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
2810 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2813 * common/config/aarch64/aarch64-common.c
2814 (aarch_option_optimization_table[]): Switch
2815 off fomit-frame-pointer
2817 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
2818 Chung-Ju Wu <jasonwucj@gmail.com>
2820 * config/nds32/nds32-multiple.md (load_multiple): Disallow
2822 (store_multiple): Ditto.
2824 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
2826 * config.gcc: Add --with-cpu support for nds32 target.
2827 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
2828 * config/nds32/nds32.opt: Add -mcpu= option.
2830 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
2832 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
2833 isel=yes): Warn for these deprecated options.
2835 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
2837 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
2838 ISA_2_5_MASKS_EMBEDDED.
2840 2018-02-23 Jakub Jelinek <jakub@redhat.com>
2842 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
2843 p->max as pointers rather than using iterative_hash_expr.
2845 2018-02-23 Carl Love <cel@us.ibm.com>
2847 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
2848 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
2850 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
2851 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
2852 P8V_BUILTIN_VEC_VUNSIGNED2.
2854 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
2857 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
2858 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
2860 * lra-constraints.c (curr_insn_transform): Set up
2861 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
2863 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
2864 * lra-eliminations.c (spill_pseudos): Ditto.
2865 (process_insn_for_elimination): Ditto.
2866 * lra-lives.c (reg_early_clobber_p): Use the new macros.
2867 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
2868 LRA_NON_CLOBBERED_ALT.
2870 2018-02-22 Martin Sebor <msebor@redhat.com>
2872 PR tree-optimization/84480
2873 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
2874 to maybe_diag_stxncpy_trunc. Call it.
2875 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
2876 from gimple_fold_builtin_strcpy. Print inlining stack.
2877 (handle_builtin_stxncpy): Print inlining stack.
2878 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
2880 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
2883 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
2884 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
2885 and -fcheck-pointer-bounds are used together.
2886 (indirect_thunk_prefix): New enum.
2887 (indirect_thunk_need_prefix): New function.
2888 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
2889 "_nt" instead of "_bnd" for NOTRACK prefix.
2890 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
2891 (output_indirect_thunk_function): Likewise.
2893 (ix86_code_end): Update output_indirect_thunk_function calls.
2894 (ix86_output_indirect_branch_via_reg): Replace
2895 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
2896 (ix86_output_indirect_branch_via_push): Likewise.
2897 (ix86_output_function_return): Likewise.
2898 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
2899 incompatible with -fcf-protection=branch and
2900 -fcheck-pointer-bounds.
2902 2018-02-22 Steve Ellcey <sellcey@cavium.com>
2905 * config/aarch64/aarch64.c (aarch64_print_address_internal):
2906 Change gcc_assert call to output_operand_lossage.
2908 2018-02-22 Steve Ellcey <sellcey@cavium.com>
2910 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
2912 2018-02-22 DJ Delorie <dj@redhat.com>
2913 Sebastian Perta <sebastian.perta@renesas.com>
2914 Oleg Endo <olegendo@gcc.gnu.org>
2916 * config/rx/rx.c (rx_rtx_costs): New function.
2917 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
2919 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
2921 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
2923 2018-02-22 Martin Liska <mliska@suse.cz>
2926 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
2927 Add "native" as a possible value.
2929 2018-02-22 Martin Liska <mliska@suse.cz>
2932 * config/i386/i386.c (ix86_option_override_internal):
2933 Add "native" as a possible value for -march and -mtune.
2935 2018-02-22 Jakub Jelinek <jakub@redhat.com>
2938 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
2939 to all type variants.
2941 PR tree-optimization/84503
2942 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
2943 width as info->bitpos + info->bitsize - start.
2944 (merged_store_group::merge_overlapping): Simplify width computation.
2945 (check_no_overlap): New function.
2946 (imm_store_chain_info::try_coalesce_bswap): Compute expected
2947 start + width and last_order of the group, fail if check_no_overlap
2949 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
2950 to group if check_no_overlap fails.
2952 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2954 * config/rs6000/altivec.md: Delete contraint arguments to
2955 define_expand, define_split, and define_peephole2, and in
2956 define_insn_and_split if always unused.
2957 * config/rs6000/darwin.md: Ditto.
2958 * config/rs6000/dfp.md: Ditto.
2959 * config/rs6000/rs6000.md: Ditto.
2960 * config/rs6000/sync.md: Ditto.
2961 * config/rs6000/vector.md: Ditto.
2962 * config/rs6000/vsx.md: Ditto.
2964 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2966 * config/rs6000/altivec.md: Write output control strings as braced
2967 blocks instead of double-quoted strings.
2968 * config/rs6000/darwin.md: Ditto.
2969 * config/rs6000/rs6000.md: Ditto.
2970 * config/rs6000/vector.md: Ditto.
2971 * config/rs6000/vsx.md: Ditto.
2973 2018-02-21 Jason Merrill <jason@redhat.com>
2975 PR c++/84314 - ICE with templates and fastcall attribute.
2976 * attribs.c (build_type_attribute_qual_variant): Remove assert.
2978 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2980 * ipa-cp.c (determine_versionability): Fix comment typos.
2982 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2985 * ipa-cp.c (determine_versionability): Do not version functions caling
2988 2018-02-21 Martin Liska <mliska@suse.cz>
2991 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
2992 Add "native" as a possible value.
2993 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
2994 the macro when native cpu detection is available.
2996 2018-02-21 Martin Liska <mliska@suse.cz>
2999 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
3000 Add "native" as a possible value.
3001 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
3002 when native cpu detection is available.
3004 2018-02-21 Jakub Jelinek <jakub@redhat.com>
3005 Martin Sebor <msebor@redhat.com>
3007 PR tree-optimization/84478
3008 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
3010 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
3011 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
3012 support which is conservatively correct, for 2 only stay conservative
3013 for maxlen. Formatting and comment capitalization fixes. Add STRICT
3014 argument to the 2 argument get_range_strlen, adjust 6 arg
3015 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
3017 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
3018 (gimple_fold_builtin_strlen): Pass true as last argument to
3021 2018-02-20 Martin Sebor <msebor@redhat.com>
3024 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
3025 (builtin_memref::set_base_and_offset): Same. Handle inner references.
3026 (builtin_memref::builtin_memref): Factor out parts into
3027 set_base_and_offset and call it.
3029 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3032 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
3033 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
3034 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
3035 search at the associated MODE_INT.
3037 2018-02-20 Jeff Law <law@redhat.com>
3040 PR tree-optimization/81592
3042 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
3043 for range data rather than using global data.
3044 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
3045 range data rather than using global data.
3046 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
3047 pass it to children as needed.
3048 (struct directive::fmtresult): Similarly.
3049 (struct directive::set_width): Similarly.
3050 (struct directive::set_precision): Similarly.
3051 (format_integer, format_directive, parse_directive): Similarly.
3052 (format_none): Accept unnamed vr_values parameter.
3053 (format_percent, format_floating, format_character): Similarly.
3054 (format_string, format_plain): Similarly.
3055 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
3056 the EVRP range analyzer for range data rather than using global data.
3057 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
3058 gimple-ssa-evrp-analyze.h
3059 (class sprintf_dom_walker): Add after_dom_children member function.
3060 Add evrp_range_analyzer member.
3061 (sprintf_dom_walker::before_dom_children): Call into the EVRP
3062 range analyzer as needed.
3063 (sprintf_dom_walker::after_dom_children): New member function.
3064 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
3066 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
3067 (evrp_range_analyzer::pop_to_marker): Likewise.
3069 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3071 PR tree-optimization/84419
3072 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
3073 with the required type if its current type is compatible but
3076 2018-02-20 Jakub Jelinek <jakub@redhat.com>
3079 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
3080 after vectorization.
3082 2018-02-20 Martin Liska <mliska@suse.cz>
3085 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
3086 possible values if we don't have a hint.
3088 2018-02-20 Martin Liska <mliska@suse.cz>
3092 * final.c (shorten_branches): Build align_tab array with one
3094 * opts.c (finish_options): Add alignment option limit check.
3095 (MAX_CODE_ALIGN): Likewise.
3096 (MAX_CODE_ALIGN_VALUE): Likewise.
3097 * doc/invoke.texi: Document maximum allowed option value for
3098 all -falign-* options.
3100 2018-02-19 Jakub Jelinek <jakub@redhat.com>
3103 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
3104 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
3105 * var-tracking.c (emit_note_insn_var_location): Remove all references
3106 to NOTE_INSN_CALL_ARG_LOCATION.
3107 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
3108 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
3109 Use copy_rtx_if_shared.
3110 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
3111 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
3112 (dwarf2out_var_location): Remove handling of
3113 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
3115 * final.c (final_scan_insn): Remove all references to
3116 NOTE_INSN_CALL_ARG_LOCATION.
3117 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
3118 before dumping final insns.
3119 * except.c (emit_note_eh_region_end): Remove all references to
3120 NOTE_INSN_CALL_ARG_LOCATION.
3121 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
3122 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
3123 * config/arc/arc.c (hwloop_optimize): Likewise.
3124 * config/arm/arm.c (create_fix_barrier): Likewise.
3125 * config/s390/s390.c (s390_chunkify_start): Likewise.
3126 * config/sh/sh.c (find_barrier): Likewise.
3127 * config/i386/i386.c (rest_of_insert_endbranch,
3128 ix86_seh_fixup_eh_fallthru): Likewise.
3129 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
3130 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
3131 * config/frv/frv.c (frv_function_prologue): Likewise.
3132 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
3134 (note_outside_basic_block_p): Remove all references to
3135 NOTE_INSN_CALL_ARG_LOCATION.
3136 * gengtype.c (adjust_field_rtx_def): Likewise.
3137 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
3139 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
3140 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
3143 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
3146 PR tree-optimization/84452
3147 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
3148 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
3151 2018-02-19 Martin Liska <mliska@suse.cz>
3154 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
3156 2018-02-19 Martin Liska <mliska@suse.cz>
3157 Richard Sandiford <richard.sandiford@linaro.org>
3159 PR tree-optimization/82491
3160 * gimple-fold.c (get_base_constructor): Make earlier bail out
3163 2018-02-19 Carl Love <cel@us.ibm.com>
3165 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
3166 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
3168 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
3169 P8V_BUILTIN_VEC_NEG.
3171 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
3173 * config/rl78/rl78.md (movdf): New define expand.
3175 2018-02-19 Martin Liska <mliska@suse.cz>
3178 * doc/invoke.texi: Fix typo.
3179 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
3181 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
3183 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
3184 handle rs6000_single_float and rs6000_double_float specially for
3187 2018-02-16 Jeff Law <law@redhat.com>
3189 * config/rx/rx.c (add_pop_cfi_notes): New function.;
3192 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3195 * ipa-inline.c (inline_small_functions): Fix a typo.
3197 2018-02-16 Nathan Sidwell <nathan@acm.org>
3199 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
3201 2018-02-16 Carl Love <cel@us.ibm.com>
3203 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
3204 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
3205 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
3206 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
3207 expansion to P8V_BUILTIN_VEC_FLOAT2.
3209 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
3211 PR rtl-optimization/70023
3212 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
3213 src_regno into account.
3215 2018-02-16 Carl Love <cel@us.ibm.com>
3217 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
3218 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
3219 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
3220 * config/rs6000/rs6000.c: Remove case statements for
3221 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
3222 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
3223 and P9V_BUILTIN_VEC_VINSERT4B.
3224 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
3225 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
3226 * config/rs6000/vsx.md:
3227 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
3230 2018-02-16 Carl Love <cel@us.ibm.com>
3232 * config/rs6000/altivec.h: Add builtin names vec_extract4b
3234 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
3236 * config/rs6000/rs6000-c.c: Add the definitions for
3237 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
3238 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
3239 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
3240 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
3241 definition for insert4b and define insn *insert3b_internal.
3242 * doc/extend.texi: Add documentation for vec_extract4b.
3244 2018-02-16 Nathan Sidwell <nathan@acm.org>
3246 * doc/extend.texi (Backwards Compatibility): Mention friend
3247 injection. Note for-scope is deprecated.
3248 * doc/invoke.texi (-ffriend-injection): Deprecate.
3250 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
3252 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
3253 that moved to I2, also allow destinations that are a paradoxical
3254 subreg (instead of a normal reg).
3256 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
3259 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
3262 2018-02-16 Richard Biener <rguenther@suse.de>
3264 PR tree-optimization/84037
3265 PR tree-optimization/84016
3267 * config/i386/i386.c (ix86_builtin_vectorization_cost):
3268 Adjust vec_construct for the fact we need additional higher latency
3269 128bit inserts for AVX256 and AVX512 vector builds.
3270 (ix86_add_stmt_cost): Scale vector construction cost for
3273 2018-02-16 Richard Biener <rguenther@suse.de>
3275 PR tree-optimization/84417
3276 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
3277 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
3278 (non_rewritable_lvalue_p): Likewise, use poly-ints.
3280 2018-02-16 Martin Liska <mliska@suse.cz>
3283 * internal-fn.def (ASAN_CHECK): Set proper flags.
3284 (ASAN_MARK): Likewise.
3286 2018-02-16 Julia Koval <julia.koval@intel.com>
3288 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
3289 from PTA_CANNONLAKE.
3291 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3294 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
3295 Use ++iter rather than iter++ for std::list iterators.
3296 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
3297 defer deleting them until all nodes in the forest are processed. Do
3298 free even leaf nodes. Change to_process into auto_vec.
3301 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
3302 * vec.h (vec_default_construct): Use memset instead of placement new
3303 if BROKEN_VALUE_INITIALIZATION is defined.
3304 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
3305 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
3308 PR rtl-optimization/83723
3309 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
3310 * lra.c (lra_substitute_pseudo): Likewise. If true, use
3311 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
3313 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
3315 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
3317 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
3319 PR rtl-optimization/81443
3320 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
3321 from inner REGs to paradoxical SUBREGs.
3323 2018-02-16 Richard Biener <rguenther@suse.de>
3325 PR tree-optimization/84399
3326 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3327 For operands we can analyze at their definition make sure we can
3328 analyze them at each use as well.
3330 2018-02-16 Richard Biener <rguenther@suse.de>
3332 PR tree-optimization/84190
3333 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
3334 volatile accesses if the decl isn't volatile.
3336 2018-02-15 Jason Merrill <jason@redhat.com>
3338 PR c++/84314 - ICE with templates and fastcall attribute.
3339 * attribs.c (build_type_attribute_qual_variant): Don't clobber
3340 TYPE_CANONICAL on an existing type.
3342 2018-02-15 Jakub Jelinek <jakub@redhat.com>
3344 PR tree-optimization/84383
3345 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
3346 dstoff nor call operand_equal_p if dstbase is NULL.
3348 PR tree-optimization/84334
3349 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
3350 also a CONSTANT_CLASS_P, punt.
3352 2018-02-14 Jim Wilson <jimw@sifive.com>
3354 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
3355 first SMALL_OPERAND check. New local min_second_step. Move assert
3356 to where locals are set. Add TARGET_RVC support.
3357 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
3359 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
3361 * doc/invoke.texi: Correct -Wformat-overflow code sample.
3363 2018-02-14 Martin Sebor <msebor@redhat.com>
3365 PR tree-optimization/83698
3366 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
3367 arrays constrain the offset range to their bounds.
3368 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
3369 (builtin_access::overlap): Avoid setting the size of overlap if it's
3371 (maybe_diag_overlap): Also consider arrays when deciding what values
3372 of offsets to include in diagnostics.
3374 2018-02-14 Martin Sebor <msebor@redhat.com>
3377 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
3378 that correspond to the kind of a declaration.
3380 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
3383 * config/pa/pa.md: Load address of PIC label using the linkage table
3384 if the label is nonlocal.
3386 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
3388 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
3389 warning message if user requests -maltivec=be.
3390 * doc/invoke.texi: Document deprecation of -maltivec=be.
3392 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
3395 * config/rs6000/rs6000-c.c: Update definitions for
3396 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
3397 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
3399 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3402 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
3403 add _get_ssp intrinsics. Remove argument from
3404 __builtin_ia32_rdssp[d|q].
3405 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
3406 * config/i386/i386-builtin.def: Remove argument from
3407 __builtin_ia32_rdssp[d|q].
3408 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
3409 ix86_expand_special_args_builtin for _rdssp[d|q].
3410 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
3411 Clear register before usage.
3412 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
3413 Add documentation for new _get_ssp and _inc_ssp intrinsics.
3415 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
3417 PR tree-optimization/84357
3418 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
3419 operand 1 of an ARRAY_REF too.
3421 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
3424 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
3425 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3427 (set_of_reg): New struct.
3428 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
3429 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
3430 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3432 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
3433 Split into bitclr, bitset, bitinvert patterns if appropriate.
3434 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
3435 use rx_fuse_in_memory_bitop.
3436 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
3437 to named insn, correct maximum insn length.
3439 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
3442 * machmode.def: Define a complex mode for PARTIAL_INT.
3443 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
3445 * doc/rtl.texi: Document CSPImode.
3446 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
3448 (msp430_hard_regno_nregs_with_padding): Likewise.
3450 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
3453 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
3455 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
3457 PR rtl-optimization/84169
3458 * combine.c (try_combine): New variable split_i2i3. Set it to true if
3459 we generated a parallel as new i3 and we split that to new i2 and i3
3460 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
3461 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
3462 those to i2, not i1. Partially rewrite this scan code.
3464 2018-02-13 Jakub Jelinek <jakub@redhat.com>
3467 * stor-layout.c (place_field): For variable length fields, adjust
3468 offset_align afterwards not just based on the field's alignment,
3469 but also on the size.
3472 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
3473 of exps and logs in the use_exp2 case.
3475 2018-02-13 Jeff Law <law@redhat.com>
3477 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
3480 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
3483 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
3487 * common.opt (gas-loc-support, gas-locview-support): New.
3488 (ginline-points, ginternal-reset-location-views): New.
3489 * doc/invoke.texi: Document them. Use @itemx where intended.
3490 (gvariable-location-views): Adjust.
3491 * target.def (reset_location_view): New.
3492 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
3493 (TARGET_RESET_LOCATION_VIEW): New.
3494 * doc/tm.texi: Rebuilt.
3495 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
3496 (dwarf2out_default_as_locview_support): New.
3497 (output_asm_line_debug_info): Use option variables.
3498 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
3499 (output_loc_list): Likewise.
3500 (add_high_low_attributes): Check option variables.
3501 Don't output entry view attribute in strict mode.
3502 (gen_inlined_subroutine_die): Check option variables.
3503 (dwarf2out_inline_entry): Likewise.
3504 (init_sections_and_labels): Likewise.
3505 (dwarf2out_early_finish): Likewise.
3506 (maybe_reset_location_view): New, from...
3507 (dwarf2out_var_location): ... here. Call it.
3508 * debug.h (dwarf2out_default_as_loc_support): Declare.
3509 (dwarf2out_default_as_locview_support): Declare.
3510 * hooks.c (hook_int_rtx_insn_0): New.
3511 * hooks.h (hook_int_rtx_insn_0): Declare.
3512 * toplev.c (process_options): Take -gas-loc-support and
3513 -gas-locview-support from dwarf2out. Enable
3514 -gvariable-location-views by default only with locview
3515 assembler support. Enable -ginternal-reset-location-views by
3516 default only if the target defines the corresponding hook.
3517 Enable -ginline-points by default if location views are
3518 enabled; force it disabled if statement frontiers are
3520 * tree-inline.c (expand_call_inline): Check option variables.
3521 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
3523 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
3525 PR tree-optimization/84321
3526 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
3527 handling. Also check whether the anti-range contains any values
3528 that satisfy the mask; switch to a VR_RANGE if not.
3530 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
3533 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
3535 2018-02-13 Martin Jambor <mjambor@suse.cz>
3538 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
3539 of call statements, also set location of a load to a temporary.
3541 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
3543 * config/rl78/rl78.c (add_vector_labels): New function.
3544 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
3545 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
3546 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
3547 which checks that no arguments are passed.
3548 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
3549 * doc/extend.texi: Documentation for the new attribute.
3551 2018-02-13 Andreas Schwab <schwab@suse.de>
3553 * config/riscv/linux.h (CPP_SPEC): Define.
3555 2018-02-13 Jakub Jelinek <jakub@redhat.com>
3558 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
3559 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
3560 OPTION_MASK_ISA_AES as first argument to def_builtin_const
3561 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
3562 instead of OPTION_MASK_ISA_PCLMUL as first argument to
3563 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
3564 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
3565 temporarily for AES and PCLMUL builtins.
3567 PR tree-optimization/84339
3568 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
3569 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
3573 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
3574 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
3575 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
3577 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
3579 * omp-simd-clone.h: New file.
3580 * omp-simd-clone.c: Include omp-simd-clone.h.
3581 (expand_simd_clones): No longer static.
3582 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
3583 cgraph.h and omp-simd-clone.h.
3584 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
3585 (vect_recog_widen_shift_pattern): Formatting fix.
3586 (vect_pattern_recog_1): Don't check optab for calls.
3589 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
3590 operands[2] into a REG before using gen_lowpart on it.
3592 2018-02-12 Jeff Law <law@redhat.com>
3595 * config/sh/sh.c (find_barrier): Consider a sibling call
3598 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
3599 successfully back substituting a reg.
3601 2018-02-12 Richard Biener <rguenther@suse.de>
3603 PR tree-optimization/84037
3604 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
3605 parameter, move visited init to caller.
3606 (vect_slp_analyze_operations): Separate cost from validity
3607 check, initialize visited once for all instances.
3608 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
3610 * tree-vect-stmts.c (vect_model_simple_cost): Make early
3612 (vect_model_promotion_demotion_cost): Likewise.
3613 (vectorizable_bswap): Guard cost modeling with !slp_node
3614 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
3616 (vectorizable_call): Likewise.
3617 (vectorizable_conversion): Likewise.
3618 (vectorizable_assignment): Likewise.
3619 (vectorizable_shift): Likewise.
3620 (vectorizable_operation): Likewise.
3621 (vectorizable_store): Likewise.
3622 (vectorizable_load): Likewise.
3623 (vectorizable_condition): Likewise.
3624 (vectorizable_comparison): Likewise.
3626 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
3629 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
3630 (ASAN_MARK): Fix fnspec to account for return value, change pointer
3631 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
3633 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3636 * params.def (inline-min-speedup): Increase from 8 to 15.
3637 (max-inline-insns-auto): Decrease from 40 to 30.
3638 * ipa-split.c (consider_split): Add some buffer for function to
3639 be considered inlining candidate.
3640 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
3643 2018-02-12 Richard Biener <rguenther@suse.de>
3645 PR tree-optimization/84037
3646 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
3647 matched stmts if we cannot swap the non-matched ones.
3649 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3651 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
3652 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
3653 _mm_maskz_scalef_round_ss): New intrinsics.
3654 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
3655 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
3656 __builtin_ia32_scalefss_round): Remove.
3657 (__builtin_ia32_scalefsd_mask_round,
3658 __builtin_ia32_scalefss_mask_round): New intrinsics.
3659 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
3660 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
3661 ((match_operand:VF_128 2 "<round_nimm_predicate>"
3662 "<round_constraint>")): Changed to ...
3663 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
3664 "<round_scalar_constraint>")): ... this.
3665 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
3666 %0, %1, %2<round_op3>}"): Changed to ...
3667 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
3668 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
3669 %2<round_scalar_mask_op3>}"): ... this.
3670 * config/i386/subst.md (round_scalar_nimm_predicate): New.
3672 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3674 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
3675 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
3676 (_mm_maskz_sqrt_round_ss): New intrinsics.
3677 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
3678 (__builtin_ia32_sqrtsd_mask_round)
3679 (__builtin_ia32_sqrtss_mask_round): New builtins.
3680 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
3681 (__builtin_ia32_sqrtss_round): Remove.
3682 (__builtin_ia32_sqrtsd_mask_round)
3683 (__builtin_ia32_sqrtss_mask_round): New builtins.
3684 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
3685 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
3686 ((match_operand:VF_128 1 "vector_operand"
3687 "xBm,<round_constraint>")): Changed to ...
3688 ((match_operand:VF_128 1 "vector_operand"
3689 "xBm,<round_scalar_constraint>")): ... this.
3690 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
3691 %0, %2, %<iptr>1<round_op3>}): Changed to ...
3692 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
3693 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
3694 %<iptr>1<round_scalar_mask_op3>}): ... this.
3695 ((set_attr "prefix" "<round_prefix>")): Changed to ...
3696 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
3698 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
3701 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
3702 Cast vec_cmpeq result to correct type.
3703 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
3704 Cast vec_cmpgt result to correct type.
3706 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
3708 * final.c (final_scan_insn_1): Renamed from...
3709 (final_scan_insn): ... this. New wrapper, to recover
3710 seen from the outermost call in recursive ones.
3711 * config/sparc/sparc.c (output_return): Drop seen from call.
3712 (output_sibcall): Likewise.
3713 * config/visium/visium.c (output_branch): Likewise.
3715 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
3717 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
3720 2018-02-10 Alan Modra <amodra@gmail.com>
3723 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
3724 Specify LR as an input.
3726 2018-02-10 Jakub Jelinek <jakub@redhat.com>
3729 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
3730 remove_member_access_dummy_vars): New functions.
3731 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
3732 lower_omp_1, execute_lower_omp): Use them.
3734 PR rtl-optimization/84308
3735 * shrink-wrap.c (spread_components): Release todo vector.
3737 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
3739 PR rtl-optimization/57193
3740 * ira-color.c (struct allocno_color_data): Add member
3741 conflict_allocno_hard_prefs.
3742 (update_conflict_allocno_hard_prefs): New.
3743 (bucket_allocno_compare_func): Add a preference based on
3744 conflict_allocno_hard_prefs.
3745 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
3746 (color_allocnos): Remove a dead code. Initiate
3747 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
3749 2018-02-09 Jakub Jelinek <jakub@redhat.com>
3752 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
3753 constraint from =wa to wa. Avoid a subreg on the output operand,
3754 instead use a pseudo and subreg it in a move.
3755 (p9_xxbrd_<mode>): Changed to ...
3756 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
3757 (p9_xxbrd_v2df): New expander.
3758 (p9_xxbrw_<mode>): Changed to ...
3759 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
3760 (p9_xxbrw_v4sf): New expander.
3762 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
3764 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
3766 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
3769 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
3770 multiply in 32-bit mode.
3771 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
3772 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
3775 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
3777 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
3778 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
3779 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
3780 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
3782 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
3785 * dwarf2out.c (is_trivial_indirect_ref): New function.
3786 (dwarf2out_late_global_decl): Do not generate a location
3787 attribute for variables that have a non-trivial DECL_VALUE_EXPR
3788 and that are not defined in the current unit.
3790 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
3792 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
3793 instead of a libcall for UNORDERED.
3795 2018-02-09 Tamar Christina <tamar.christina@arm.com>
3798 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
3799 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
3801 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3804 * config/s390/s390.c (s390_set_current_function): Invoke
3805 s390_indirect_branch_settings also if fndecl didn't change.
3807 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
3809 * config/rs6000/rs6000.md (blockage): Set length to zero.
3811 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
3813 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
3815 2018-02-09 Jakub Jelinek <jakub@redhat.com>
3818 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
3819 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
3823 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
3824 PARALLEL incoming that failed vt_get_decl_and_offset check.
3827 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
3828 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
3830 (get_variable_section): For decls in named .bss* sections pass true as
3831 second argument to bss_initializer_p.
3833 2018-02-09 Marek Polacek <polacek@redhat.com>
3834 Jakub Jelinek <jakub@redhat.com>
3837 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
3838 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
3839 Sync some changes from cxx_fold_indirect_ref.
3841 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
3843 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
3845 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
3846 (BLOCK_INLINE_ENTRY_LABEL): New.
3847 (dwarf2out_var_location): Disregard inline entry markers.
3848 (inline_entry_data): New struct.
3849 (inline_entry_data_hasher): New hashtable type.
3850 (inline_entry_data_hasher::hash): New.
3851 (inline_entry_data_hasher::equal): New.
3852 (inline_entry_data_table): New variable.
3853 (add_high_low_attributes): Add DW_AT_entry_pc and
3854 DW_AT_GNU_entry_view attributes if a pending entry is found
3855 in inline_entry_data_table. Add old entry_pc attribute only
3856 if debug nonbinding markers are disabled.
3857 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
3858 markers are enabled.
3859 (block_within_block_p, dwarf2out_inline_entry): New.
3860 (dwarf2out_finish): Check that no entries remained in
3861 inline_entry_data_table.
3862 * final.c (reemit_insn_block_notes): Handle inline entry notes.
3863 (final_scan_insn, notice_source_line): Likewise.
3864 (rest_of_clean_state): Skip inline entry markers.
3865 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
3867 * gimple.c (gimple_build_debug_inline_entry): New.
3868 * gimple.h (enum gimple_debug_subcode): Add
3869 GIMPLE_DEBUG_INLINE_ENTRY.
3870 (gimple_build_debug_inline_entry): Declare.
3871 (gimple_debug_inline_entry_p): New.
3872 (gimple_debug_nonbind_marker_p): Adjust.
3873 * insn-notes.def (INLINE_ENTRY): New.
3874 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
3875 inline entry marker notes.
3876 (print_insn): Likewise.
3877 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
3878 (INSN_DEBUG_MARKER_KIND): Likewise.
3879 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
3880 * tree-inline.c (expand_call_inline): Build and insert
3881 debug_inline_entry stmt.
3882 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
3883 inline entry blocks early, if nonbind markers are enabled.
3884 (dump_scope_block): Dump fragment info.
3885 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
3886 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
3887 (gimple_build_debug_inline_entry): New.
3888 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
3889 Enable/disable inline entry points too.
3890 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
3891 (DEBUG_INSN): Describe inline entry markers.
3893 * common.opt (gvariable-location-views): New.
3894 (gvariable-location-views=incompat5): New.
3895 * config.in: Rebuilt.
3896 * configure: Rebuilt.
3897 * configure.ac: Test assembler for view support.
3898 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
3899 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
3900 * dwarf2out.c (var_loc_view): New typedef.
3901 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
3902 (dwarf2out_locviews_in_attribute): New.
3903 (dwarf2out_locviews_in_loclist): New.
3904 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
3905 (enum dw_line_info_opcode): Add LI_adv_address.
3906 (struct dw_line_info_table): Add view.
3907 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
3908 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
3909 (zero_view_p): New variable.
3910 (ZERO_VIEW_P): New macro.
3911 (output_asm_line_debug_info): New.
3912 (struct var_loc_node): Add view.
3913 (add_AT_view_list, AT_loc_list): New.
3914 (add_var_loc_to_decl): Add view param. Test it against last.
3915 (new_loc_list): Add view params. Record them.
3916 (AT_loc_list_ptr): Handle loc and view lists.
3917 (view_list_to_loc_list_val_node): New.
3918 (print_dw_val): Handle dw_val_class_view_list.
3919 (size_of_die): Likewise.
3920 (value_format): Likewise.
3921 (loc_list_has_views): New.
3922 (gen_llsym): Set vl_symbol too.
3923 (maybe_gen_llsym, skip_loc_list_entry): New.
3924 (dwarf2out_maybe_output_loclist_view_pair): New.
3925 (output_loc_list): Output view list or entries too.
3926 (output_view_list_offset): New.
3927 (output_die): Handle dw_val_class_view_list.
3928 (output_dwarf_version): New.
3929 (output_compilation_unit_header): Use it.
3930 (output_skeleton_debug_sections): Likewise.
3931 (output_rnglists, output_line_info): Likewise.
3932 (output_pubnames, output_aranges): Update version comments.
3933 (output_one_line_info_table): Output view numbers in asm comments.
3934 (dw_loc_list): Determine current endview, pass it to new_loc_list.
3935 Call maybe_gen_llsym.
3936 (loc_list_from_tree_1): Adjust.
3937 (add_AT_location_description): Create view list attribute if
3938 needed, check it's absent otherwise.
3939 (convert_cfa_to_fb_loc_list): Adjust.
3940 (maybe_emit_file): Call output_asm_line_debug_info for test.
3941 (dwarf2out_var_location): Reset views as needed. Precompute
3942 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
3943 attribute. Set view.
3944 (new_line_info_table): Reset next view.
3945 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
3946 (dwarf2out_source_line): Likewise. Output view resets and labels to
3947 the assembler, or select appropriate line info opcodes.
3948 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
3949 (optimize_string_length): Catch it. Adjust.
3950 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
3951 dw_val_class_view_list, and remove it if no longer needed.
3952 (hash_loc_list): Hash view numbers.
3953 (loc_list_hasher::equal): Compare them.
3954 (optimize_location_lists): Check whether a view list symbol is
3955 needed, and whether the locview attribute is present, and
3956 whether they match. Remove the locview attribute if no longer
3958 (index_location_lists): Call skip_loc_list_entry for test.
3959 (dwarf2out_finish): Call output_asm_line_debug_info for test.
3960 Use output_dwarf_version.
3961 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
3962 (struct dw_val_node): Add val_view_list.
3963 * final.c (SEEN_NEXT_VIEW): New.
3964 (set_next_view_needed): New.
3965 (clear_next_view_needed): New.
3966 (maybe_output_next_view): New.
3967 (final_start_function): Rename to...
3968 (final_start_function_1): ... this. Take pointer to FIRST,
3969 add SEEN parameter. Emit param bindings in the initial view.
3970 (final_start_function): Reintroduce SEEN-less interface.
3971 (final): Rename to...
3972 (final_1): ... this. Take SEEN parameter. Output final pending
3973 next view at the end.
3974 (final): Reintroduce seen-less interface.
3975 (final_scan_insn): Output pending next view before switching
3976 sections or ending a block. Mark the next view as needed when
3977 outputting variable locations. Notify debug backend of section
3978 changes, and of location view changes.
3979 (rest_of_handle_final): Adjust.
3980 * toplev.c (process_options): Autodetect value for debug variable
3981 location views option. Warn on incompat5 without -gdwarf-5.
3982 * doc/invoke.texi (gvariable-location-views): New.
3983 (gvariable-location-views=incompat5): New.
3984 (gno-variable-location-views): New.
3986 2018-02-08 David Malcolm <dmalcolm@redhat.com>
3988 PR tree-optimization/84136
3989 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
3990 that the result of find_edge is non-NULL.
3992 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
3995 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
3996 storing integer register in SImode. Fix cost of 256 and 512
3997 byte aligned SSE register store.
3999 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4001 * config/i386/i386.c (ix86_multiplication_cost): Fix
4002 multiplication cost for TARGET_AVX512DQ.
4004 2018-02-08 Marek Polacek <polacek@redhat.com>
4006 PR tree-optimization/84238
4007 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
4010 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4012 PR tree-optimization/84265
4013 * tree-vect-stmts.c (vectorizable_store): Don't treat
4014 VMAT_CONTIGUOUS accesses as grouped.
4015 (vectorizable_load): Likewise.
4017 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4019 PR tree-optimization/81635
4020 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
4021 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
4022 (test_round_for_mask): New functions.
4023 (wide_int_cc_tests): Call test_round_for_mask.
4024 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
4025 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
4026 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
4027 range returned by get_range_info.
4029 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4032 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
4033 * symtab.c: Include builtins.h
4034 (symtab_node::output_to_lto_symbol_table_p): Move here
4035 from lto-streamer-out.c:output_symbol_p.
4036 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
4037 (output_symbol_p): Move all logic to symtab.c
4038 (produce_symtab): Update.
4040 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4042 * config/s390/s390-opts.h (enum indirect_branch): Define.
4043 * config/s390/s390-protos.h (s390_return_addr_from_memory)
4044 (s390_indirect_branch_via_thunk)
4045 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
4046 (enum s390_indirect_branch_type): Define.
4047 * config/s390/s390.c (struct s390_frame_layout, struct
4048 machine_function): Remove.
4049 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
4050 (indirect_branch_table_label_no, indirect_branch_table_name):
4052 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
4053 (enum s390_indirect_branch_option): Define.
4054 (s390_return_addr_from_memory): New function.
4055 (s390_handle_string_attribute): New function.
4056 (s390_attribute_table): Add new attribute handler.
4057 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
4058 (s390_indirect_branch_via_thunk): New function.
4059 (s390_indirect_branch_via_inline_thunk): New function.
4060 (s390_function_ok_for_sibcall): When jumping via thunk disallow
4061 sibling call optimization for non z10 compiles.
4062 (s390_emit_call): Force indirect branch target to be a single
4063 register. Add r1 clobber for non-z10 compiles.
4064 (s390_emit_epilogue): Emit return jump via return_use expander.
4065 (s390_reorg): Handle JUMP_INSNs as execute targets.
4066 (s390_option_override_internal): Perform validity checks for the
4067 new command line options.
4068 (s390_indirect_branch_attrvalue): New function.
4069 (s390_indirect_branch_settings): New function.
4070 (s390_set_current_function): Invoke s390_indirect_branch_settings.
4071 (s390_output_indirect_thunk_function): New function.
4072 (s390_code_end): Implement target hook.
4073 (s390_case_values_threshold): Implement target hook.
4074 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
4076 * config/s390/s390.h (struct s390_frame_layout)
4077 (struct machine_function): Move here from s390.c.
4078 (TARGET_INDIRECT_BRANCH_NOBP_RET)
4079 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
4080 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
4081 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
4082 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
4083 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
4084 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
4085 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
4086 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
4087 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
4088 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
4089 (mnemonic attribute): Add values which aren't recognized
4091 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
4092 pattern for branch conversion. Fix mnemonic attribute.
4093 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
4094 indirect branch via thunk if requested.
4095 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
4096 ("*indirect_jump"): Disable for branch conversion using out of
4098 ("indirect_jump_via_thunk<mode>_z10")
4099 ("indirect_jump_via_thunk<mode>")
4100 ("indirect_jump_via_inlinethunk<mode>_z10")
4101 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
4102 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
4103 ("casesi_jump_via_inlinethunk<mode>_z10")
4104 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
4105 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
4106 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
4107 ("*indirect2_jump"): Disable for branch conversion.
4108 ("casesi_jump"): Turn into expander and expand patterns for branch
4110 ("return_use"): New expander.
4111 ("*return"): Emit return via thunk and rename it to ...
4112 ("*return<mode>"): ... this one.
4113 * config/s390/s390.opt: Add new options and and enum for the
4116 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4118 * lra-constraints.c (match_reload): Unconditionally use
4119 gen_lowpart_SUBREG, rather than selecting between that
4120 and equivalent gen_rtx_SUBREG code.
4122 2018-02-08 Richard Biener <rguenther@suse.de>
4124 PR tree-optimization/84233
4125 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
4126 changed flag instead of boguously re-using phi_inserted.
4128 2018-02-08 Martin Jambor <mjambor@suse.cz>
4130 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
4131 static local variables.
4133 2018-02-08 Richard Biener <rguenther@suse.de>
4135 PR tree-optimization/84278
4136 * tree-vect-stmts.c (vectorizable_store): When looking for
4137 smaller vector types to perform grouped strided loads/stores
4138 make sure the mode is supported by the target.
4139 (vectorizable_load): Likewise.
4141 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4143 * config/aarch64/aarch64.c (aarch64_components_for_bb):
4144 Increase LDP/STP opportunities by adding adjacent callee-saves.
4146 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4148 PR rtl-optimization/84068
4149 PR rtl-optimization/83459
4150 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
4152 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
4154 PR tree-optimization/84224
4155 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
4156 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
4159 2018-02-07 Iain Sandoe <iain@codesourcery.com>
4162 * config/rs6000/altivec.md (*restore_world): Remove LR use.
4163 * config/rs6000/predicates.md (restore_world_operation): Adjust op
4164 count, remove one USE.
4166 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
4168 * doc/install.texi (Configuration): Document the
4169 --with-long-double-format={ibm,ieee} PowerPC configuration
4173 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
4174 Convert from define_expand to be define_insn_and_split. Rework
4175 float/double/_Float128 conversions to QI/HI/SImode to work with
4176 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
4177 conversions to QI/HImode types did a store and then a load to
4178 truncate the value. For conversions to VSX registers, don't split
4179 the insn, instead emit the code directly. Use the code iterator
4180 any_fix to combine signed and unsigned conversions.
4181 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
4182 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
4183 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4184 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
4185 (fix_<mode>di2_hw): Likewise.
4186 (fixuns_<mode>di2_hw): Likewise.
4187 (fix_<mode>si2_hw): Likewise.
4188 (fixuns_<mode>si2_hw): Likewise.
4189 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
4190 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4191 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
4192 fix<uns>_trunc<SFDF:mode>si2_p8.
4193 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
4195 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
4196 (fix<uns>_<mode>_mem): Likewise.
4197 (fctiw<u>z_<mode>_mem): Likewise.
4198 (fix<uns>_<mode>_mem): Likewise.
4199 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
4200 the register allocator from doing a direct move to the GPRs to do
4201 a store, and instead use the ISA 3.0 store byte/half-word from
4202 vector register instruction. For IEEE 128-bit floating point,
4203 also optimize stores of 32-bit ints.
4204 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
4206 2018-02-07 Alan Hayward <alan.hayward@arm.com>
4208 * genextract.c (push_pathstr_operand): New function to support
4210 (walk_rtx): Call push_pathstr_operand.
4211 (print_path): Support [a-zA-Z].
4213 2018-02-07 Richard Biener <rguenther@suse.de>
4215 PR tree-optimization/84037
4216 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
4217 (cse_and_gimplify_to_preheader): Declare.
4218 (vect_get_place_in_interleaving_chain): Likewise.
4219 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4221 (_loop_vec_info::~_loop_vec_info): Delete it.
4222 (cse_and_gimplify_to_preheader): New function.
4223 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
4224 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
4225 (vectorizable_load): Likewise. For grouped stores always base
4226 the IV on the first element.
4227 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
4228 condition before gimplifying.
4230 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4232 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
4233 *DIV_EXPR and *MOD_EXPR.
4235 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
4238 * config/i386/i386.c (ix86_option_override_internal): Mask out
4239 the CF_SET bit when checking -fcf-protection.
4241 2018-02-07 Tom de Vries <tom@codesourcery.com>
4244 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
4247 2018-02-07 Richard Biener <rguenther@suse.de>
4249 PR tree-optimization/84204
4250 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
4253 PR tree-optimization/84205
4254 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
4255 special-case isl_ast_op_zdiv_r.
4257 PR tree-optimization/84223
4258 * graphite-scop-detection.c (gather_bbs::before_dom_children):
4259 Only add conditions from within the region.
4260 (gather_bbs::after_dom_children): Adjust.
4262 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
4265 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
4266 * config/avr/avr.md: Only post-reload split REG-REG moves if
4267 either register is GENERAL_REG_P.
4269 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4271 PR tree-optimization/84235
4272 * tree-ssa-scopedtables.c
4273 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
4274 if the subtraction is performed in floating point type where NaNs are
4275 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
4276 build 1. Formatting fix.
4278 2018-02-06 Jakub Jelinek <jakub@redhat.com>
4281 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
4282 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
4283 and skip it regardless of bb boundaries. Use CALL_P macro,
4284 don't test INSN_P (insn) together with CALL_P or JUMP_P check
4285 unnecessarily, formatting fix.
4287 2018-02-06 Michael Collison <michael.collison@arm.com>
4289 * config/arm/thumb2.md:
4290 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
4291 (*thumb_mov_notscc): Ditto.
4293 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
4296 * config/rs6000/rs6000.md (su code attribute): Use "u" for
4297 unsigned_fix, not "s".
4299 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4301 * configure.ac (gcc_fn_eh_frame_ro): New function.
4302 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
4303 correct .eh_frame permissions.
4304 * configure: Regenerate.
4306 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
4308 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
4311 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4313 * config/rs6000/rs6000.c (rs6000_option_override_internal):
4314 Display warning message for -mno-speculate-indirect-jumps.
4316 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
4318 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
4320 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
4322 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
4324 PR tree-optimization/84225
4325 * tree-eh.c (find_trapping_overflow): Only call
4326 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
4328 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4331 * config/i386/i386.c: Reimplement the check of possible options
4332 -mibt/-mshstk conbination. Change error messages.
4333 * doc/invoke.texi: Fix a typo: remove extra '='.
4335 2018-02-06 Marek Polacek <polacek@redhat.com>
4337 PR tree-optimization/84228
4338 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
4340 2018-02-06 Tamar Christina <tamar.christina@arm.com>
4343 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
4344 emitted arch directives.
4345 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
4346 __ARM_FEATURE_COPROC before changing architectures.
4348 2018-02-06 Richard Biener <rguenther@suse.de>
4350 * config/i386/i386.c (print_reg): Fix typo.
4351 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
4353 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
4355 * configure: Regenerate.
4357 2018-02-05 Martin Sebor <msebor@redhat.com>
4359 PR tree-optimization/83369
4360 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
4363 2018-02-05 Martin Liska <mliska@suse.cz>
4365 * doc/invoke.texi: Cherry-pick upstream r323995.
4367 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
4369 * ira.c (ira_init_register_move_cost): Adjust comment.
4371 2018-02-05 Martin Liska <mliska@suse.cz>
4373 PR gcov-profile/84137
4374 * doc/gcov.texi: Fix typo in documentation.
4376 2018-02-05 Martin Liska <mliska@suse.cz>
4378 PR gcov-profile/83879
4379 * doc/gcov.texi: Document necessity of --dynamic-list-data when
4380 using dlopen functionality.
4382 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
4384 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
4385 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
4386 _mm_maskz_range_ss, _mm_mask_range_round_ss,
4387 _mm_maskz_range_round_ss): New intrinsics.
4388 (__builtin_ia32_rangesd128_round)
4389 (__builtin_ia32_rangess128_round): Remove.
4390 (__builtin_ia32_rangesd128_mask_round,
4391 __builtin_ia32_rangess128_mask_round): New builtins.
4392 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
4393 __builtin_ia32_rangess128_round): Remove.
4394 (__builtin_ia32_rangesd128_mask_round,
4395 __builtin_ia32_rangess128_mask_round): New builtins.
4396 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
4397 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
4398 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
4399 "<round_saeonly_constraint>")): Changed to ...
4400 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
4401 "<round_saeonly_scalar_constraint>")): ... this.
4402 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
4403 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
4404 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
4405 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
4406 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
4408 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
4410 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
4412 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
4413 Remove all values except native, 8540 and 8548.
4415 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
4417 * config/i386/i386.c (ix86_output_function_return): Pass
4418 INVALID_REGNUM, instead of -1, as invalid register number to
4419 indirect_thunk_name and output_indirect_thunk.
4421 2018-02-02 Julia Koval <julia.koval@intel.com>
4423 * config.gcc: Add -march=icelake.
4424 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
4425 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
4426 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
4427 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
4428 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
4429 (processor_target_table): Add icelake.
4430 (ix86_option_override_internal): Handle new PTAs.
4431 (get_builtin_code_for_version): Handle icelake.
4432 (M_INTEL_COREI7_ICELAKE): New.
4433 (fold_builtin_cpu): Handle icelake.
4434 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
4435 * doc/invoke.texi: Add -march=icelake.
4437 2018-02-02 Julia Koval <julia.koval@intel.com>
4439 * config/i386/i386.c (ix86_option_override_internal): Change flags type
4440 to wide_int_bitmask.
4441 * wide-int-bitmask.h: New.
4443 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4446 * config/i386/i386.md: Replace Pmode with word_mode in
4447 builtin_setjmp_setup and builtin_longjmp to support x32.
4449 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
4453 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
4455 (rs6000_supported_cpu_names): New static variable.
4456 (linux_cpu_translation_table): Likewise.
4457 (elf_platform) <cpu>: Define new static variable and use it.
4458 Translate kernel AT_PLATFORM name to canonical name if needed.
4459 Error if platform name is unknown.
4461 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
4464 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
4466 2018-02-01 Jeff Law <law@redhat.com>
4469 * config/i386/i386.c (release_scratch_register_on_entry): Add new
4470 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
4471 the scratch if RELEASE_VIA_POP is false.
4472 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
4473 If we have to save a temporary register, decrement SIZE appropriately.
4474 Pass new arguments to release_scratch_register_on_entry.
4475 (ix86_adjust_stack_and_probe): Likewise.
4476 (ix86_emit_probe_stack_range): Pass new arguments to
4477 release_scratch_register_on_entry.
4479 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
4481 PR rtl-optimization/84157
4482 * combine.c (change_zero_ext): Use REG_P predicate in
4483 front of HARD_REGISTER_P predicate.
4485 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
4487 * config/avr/avr.c (avr_option_override): Move disabling of
4488 -fdelete-null-pointer-checks to...
4489 * common/config/avr/avr-common.c (avr_option_optimization_table):
4492 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4494 PR tree-optimization/81635
4495 * tree-data-ref.c (split_constant_offset_1): For types that
4496 wrap on overflow, try to use range info to prove that wrapping
4499 2018-02-01 Renlin Li <renlin.li@arm.com>
4502 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
4504 (aarch64_register_move_cost): Likewise.
4505 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
4507 (REG_CLASS_NAMES): Likewise.
4508 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
4509 TAILCALL_ADDR_REGS. Remove IP registers.
4510 * config/aarch64/aarch64.md (Ucs): Update register constraint.
4512 2018-02-01 Richard Biener <rguenther@suse.de>
4514 * domwalk.h (dom_walker::dom_walker): Add additional constructor
4515 for specifying RPO order and allow NULL for that.
4516 * domwalk.c (dom_walker::dom_walker): Likewise.
4517 (dom_walker::walk): Handle NULL RPO order.
4518 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
4520 (rewrite_update_dom_walker): Likewise.
4521 (mark_def_dom_walker): Likewise.
4523 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4525 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
4526 (aarch64_maybe_expand_sve_subreg_move): Declare.
4527 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
4528 * config/aarch64/predicates.md (aarch64_any_register_operand): New
4530 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
4531 that are semantically a reverse operation.
4532 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
4533 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
4534 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
4536 (aarch64_can_change_mode_class): For big-endian, forbid changes
4537 between two SVE modes if they have different element sizes.
4539 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4541 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
4542 the TImode handling for big-endian targets.
4544 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4546 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
4547 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
4549 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
4550 Remove BSWAP handing for big-endian targets and use the form of
4551 LD1RQ appropariate for the mode.
4553 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4555 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
4556 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
4559 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4562 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
4563 check for operands that need to go through aarch64_sve_reload_be.
4565 2018-02-01 Jakub Jelinek <jakub@redhat.com>
4567 PR tree-optimization/81661
4568 PR tree-optimization/84117
4569 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
4570 * tree-eh.c: Include gimplify.h.
4571 (find_trapping_overflow, replace_trapping_overflow,
4572 rewrite_to_non_trapping_overflow): New functions.
4573 * tree-vect-loop.c: Include tree-eh.h.
4574 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
4575 * tree-data-ref.c: Include tree-eh.h.
4576 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
4578 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
4580 PR rtl-optimization/84123
4581 * combine.c (change_zero_ext): Check if hard register satisfies
4582 can_change_dest_mode before calling gen_lowpart_SUBREG.
4584 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
4587 * ira.c (ira_init_register_move_cost): Remove assert.
4589 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4591 PR rtl-optimization/84071
4592 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
4593 * doc/tm.texi: Regenerate.
4595 2018-01-31 Richard Biener <rguenther@suse.de>
4597 PR tree-optimization/84132
4598 * tree-data-ref.c (analyze_miv_subscript): Properly
4599 check whether evolution_function_is_affine_multivariate_p
4600 before calling gcd_of_steps_may_divide_p.
4602 2018-01-31 Julia Koval <julia.koval@intel.com>
4605 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
4606 * config/i386/i386.md (rdpid_rex64) New.
4607 (rdpid): Make 32bit only.
4609 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
4612 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
4613 an IDENTIFIER_NODE for FUNCTION_TYPE's.
4615 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4618 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4620 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
4622 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4624 PR rtl-optimization/84071
4625 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
4626 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
4628 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4630 * config/arc/arc.c (arc_handle_aux_attribute): New function.
4631 (arc_attribute_table): Add 'aux' attribute.
4632 (arc_in_small_data_p): Consider aux like variables.
4633 (arc_is_aux_reg_p): New function.
4634 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
4635 (arc_get_aux_arg): New function.
4636 (prepare_move_operands): Handle aux-register access.
4637 (arc_handle_aux_attribute): New function.
4638 * doc/extend.texi (ARC Variable attributes): Add subsection.
4640 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4642 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
4643 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
4644 (arc_attribute_table): Add 'uncached' attribute.
4645 (arc_print_operand): Print '.di' flag for uncached memory
4647 (arc_in_small_data_p): Do not consider for small data the uncached
4649 (arc_is_uncached_mem_p): New function.
4650 * config/arc/predicates.md (compact_store_memory_operand): Check
4651 for uncached memory accesses.
4652 (nonvol_nonimm_operand): Likewise.
4653 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
4655 2018-01-31 Jakub Jelinek <jakub@redhat.com>
4658 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
4659 falign-loops=): Add Optimization flag.
4661 2018-01-30 Jeff Law <law@redhat.com>
4664 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
4665 INT_REGISTERS_SAVED. Check it prior to calling
4666 get_scratch_register_on_entry.
4667 (ix86_adjust_stack_and_probe): Similarly.
4668 (ix86_emit_probe_stack_range): Similarly.
4669 (ix86_expand_prologue): Corresponding changes.
4671 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4674 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
4675 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
4677 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
4680 * lra-constraints.c (curr_insn_transform): Process AND in the
4683 2018-01-30 Jakub Jelinek <jakub@redhat.com>
4685 PR rtl-optimization/83986
4686 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
4687 dependence against last_pending_memory_flush in addition to
4690 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
4692 PR tree-optimization/81611
4693 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
4696 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4699 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
4702 2018-01-30 Richard Biener <rguenther@suse.de>
4703 Jakub Jelinek <jakub@redhat.com>
4705 PR tree-optimization/84111
4706 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
4707 inner loops added during recursion, as they don't have up-to-date
4710 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4713 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
4714 (can_inline_edge_by_limits_p): ... here.
4715 (can_early_inline_edge_p, check_callers,
4716 update_caller_keys, update_callee_keys, recursive_inlining,
4717 add_new_edges_to_heap, speculation_useful_p,
4718 inline_small_functions,
4719 inline_small_functions, flatten_function,
4720 inline_to_all_callers_1): Update.
4722 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4724 * profile-count.c (profile_count::combine_with_ipa_count): Handle
4727 2018-01-30 Richard Biener <rguenther@suse.de>
4729 PR tree-optimization/83008
4730 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
4731 invariant and constant vector uses in stmts when they need
4734 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4737 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
4738 * configure: Regenerate.
4740 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4742 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
4744 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
4745 Use gen_rtx_REG rather than gen_lowpart.
4747 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4749 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
4750 rather than 0 when creating partial subregs.
4752 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4754 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
4757 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
4760 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
4761 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
4762 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
4763 flags. This restores the settings used before the 2017-07-24.
4764 Turning off pre increment/decrement/modify allows IVOPTS to
4765 optimize DF/SF loops where the index is an int.
4767 2018-01-29 Richard Biener <rguenther@suse.de>
4768 Kelvin Nilsen <kelvin@gcc.gnu.org>
4771 * tree-vect-stmts.c (vectorizable_call): Don't call
4772 targetm.vectorize_builtin_md_vectorized_function if callee is
4775 2018-01-22 Carl Love <cel@us.ibm.com>
4777 * doc/extend.tex: Fix typo in second arg in
4778 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
4780 2018-01-29 Richard Biener <rguenther@suse.de>
4782 PR tree-optimization/84086
4783 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
4784 (flush_ssaname_freelist): When SSA names were released reset
4785 the SCEV hash table.
4787 2018-01-29 Richard Biener <rguenther@suse.de>
4789 PR tree-optimization/84057
4790 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
4791 removed paths when removing edges.
4793 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
4795 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
4796 -mfunction-return=@var{choice}.
4798 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
4801 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
4803 (layout::print_source_line): Likewise.
4804 (test_get_line_width_without_trailing_whitespace): Add test cases.
4806 2018-01-27 Jakub Jelinek <jakub@redhat.com>
4809 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
4812 2018-01-26 Jim Wilson <jimw@sifive.com>
4814 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
4816 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
4819 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4821 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
4822 and CMP + SUB-immediate -> SUBS.
4824 2018-01-26 Martin Sebor <msebor@redhat.com>
4826 PR tree-optimization/83896
4827 * tree-ssa-strlen.c (get_string_len): Rename...
4828 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
4829 Avoid assuming length is constant.
4830 (handle_char_store): Use HOST_WIDE_INT for string length.
4832 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
4835 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
4836 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
4838 2018-01-26 Richard Biener <rguenther@suse.de>
4840 PR rtl-optimization/84003
4841 * dse.c (record_store): Only record redundant stores when
4842 the earlier store aliases at least all accesses the later one does.
4844 2018-01-26 Jakub Jelinek <jakub@redhat.com>
4846 PR rtl-optimization/83985
4847 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
4848 REG_CFA_RESTORE insns.
4849 (delete_unmarked_insns): Don't ignore separate shrink wrapping
4850 REG_CFA_RESTORE insns here.
4853 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
4854 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
4856 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4858 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
4859 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
4860 (arc_init): Likewise.
4861 (arc_override_options): Likewise.
4862 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
4864 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
4866 * config/arc/arc.h (TARGET_DBNZ): Define.
4867 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
4868 properly set the tune attribute.
4869 (dbnz): Use TARGET_DBNZ guard.
4870 * config/arc/arc.opt (mtune): Add core3 option.
4872 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4874 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
4875 recognize new pic like addresses.
4876 (arc_delegitimize_address): Clean up.
4878 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4880 * config/arc/arc-arches.def: Option mrf16 valid for all
4882 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
4883 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
4884 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
4885 * config/arc/arc-tables.opt: Regenerate.
4886 * config/arc/arc.c (arc_conditional_register_usage): Handle
4887 reduced register file case.
4888 (arc_file_start): Set must have build attributes.
4889 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
4891 * config/arc/arc.opt (mrf16): Add new option.
4892 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
4893 * config/arc/genmultilib.awk: Handle new mrf16 option.
4894 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
4895 * config/arc/t-multilib: Regenerate.
4896 * doc/invoke.texi (ARC Options): Document mrf16 option.
4898 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4900 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
4901 * config/arc/arc.c (arc_handle_secure_attribute): New function.
4902 (arc_attribute_table): Add 'secure_call' attribute.
4903 (arc_print_operand): Print secure call operand.
4904 (arc_function_ok_for_sibcall): Don't optimize tail calls when
4906 (arc_is_secure_call_p): New function. * config/arc/arc.md
4907 (call_i): Add support for sjli instruction.
4908 (call_value_i): Likewise.
4909 * config/arc/constraints.md (Csc): New constraint.
4911 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4912 John Eric Martin <John.Martin@emmicro-us.com>
4914 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
4915 * config/arc/arc.c (_arc_jli_section): New struct.
4916 (arc_jli_section): New type.
4917 (rc_jli_sections): New static variable.
4918 (arc_handle_jli_attribute): New function.
4919 (arc_attribute_table): Add jli_always and jli_fixed attribute.
4920 (arc_file_end): New function.
4921 (TARGET_ASM_FILE_END): Define.
4922 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
4923 (arc_add_jli_section): New function.
4924 (jli_call_scan): Likewise.
4925 (arc_reorg): Call jli_call_scan.
4926 (arc_output_addsi): Remove 'S' from printing asm operand.
4927 (arc_is_jli_call_p): New function.
4928 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
4930 (movhi_insn): Likewise.
4931 (movsi_insn): Likewise.
4932 (movsi_set_cc_insn): Likewise.
4933 (loadqi_update): Likewise.
4934 (load_zeroextendqisi_update): Likewise.
4935 (load_signextendqisi_update): Likewise.
4936 (loadhi_update): Likewise.
4937 (load_zeroextendhisi_update): Likewise.
4938 (load_signextendhisi_update): Likewise.
4939 (loadsi_update): Likewise.
4940 (loadsf_update): Likewise.
4941 (movsicc_insn): Likewise.
4942 (bset_insn): Likewise.
4943 (bxor_insn): Likewise.
4944 (bclr_insn): Likewise.
4945 (bmsk_insn): Likewise.
4946 (bicsi3_insn): Likewise.
4947 (cmpsi_cc_c_insn): Likewise.
4948 (movsi_ne): Likewise.
4949 (movsi_cond_exec): Likewise.
4950 (clrsbsi2): Likewise.
4961 (call_i): Remove 'S' asm letter, add jli instruction.
4962 (call_value_i): Likewise.
4963 * config/arc/arc.op (mjli-always): New option.
4964 * config/arc/constraints.md (Cji): New constraint.
4965 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
4967 (subsf3_fpx): Likewise.
4968 (mulsf3_fpx): Likewise.
4969 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
4971 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
4973 * doc/invoke.texi (ARC): Document mjli-always option.
4975 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
4977 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
4978 avoid addition with 0 and use incw and decw where possible.
4980 2018-01-26 Richard Biener <rguenther@suse.de>
4982 PR tree-optimization/81082
4983 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
4984 association if it requires casting to unsigned.
4985 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
4986 from fold_plusminus_mult_expr to catch important cases late when
4987 range info is available.
4989 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4991 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
4992 * configure.ac (hidden_linkonce): New test.
4993 * configure: Regenerate.
4994 * config.in: Regenerate.
4996 2018-01-26 Julia Koval <julia.koval@intel.com>
4998 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
4999 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
5000 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
5001 _mm_mask_bitshuffle_epi64_mask): Fix type.
5002 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
5003 USI_FTYPE_V4DI_V4DI_USI): Remove.
5004 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
5005 __builtin_ia32_vpshufbitqmb256_mask,
5006 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
5007 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
5008 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
5010 2018-01-26 Alan Modra <amodra@gmail.com>
5013 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
5014 UNSPEC_VBPERMQ. Sort other unspecs.
5016 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
5018 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
5020 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
5023 * predict.c (drop_profile): Do not push/pop cfun; update also
5025 (handle_missing_profiles): Fix logic looking for zero profiles.
5027 2018-01-25 Jakub Jelinek <jakub@redhat.com>
5030 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
5031 on functions with #pragma omp declare simd or functions with simd
5033 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
5034 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
5035 Remove trailing \n from warning_at calls.
5037 2018-01-25 Tom de Vries <tom@codesourcery.com>
5040 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5041 for neutered workers.
5043 2018-01-24 Joseph Myers <joseph@codesourcery.com>
5046 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
5047 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
5049 2018-01-24 Jeff Law <law@redhat.com>
5052 * i386.c (get_probe_interval): Move to earlier point.
5053 (ix86_compute_frame_layout): If -fstack-clash-protection and
5054 the frame is larger than the probe interval, then use pushes
5055 to save registers rather than reg->mem moves.
5056 (ix86_expand_prologue): Remove conditional for int_registers_saved
5059 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
5062 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
5063 min/max for never referenced object.
5065 2018-01-24 Jakub Jelinek <jakub@redhat.com>
5068 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
5070 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
5071 attributes from DECL_ATTRIBUTES (decl) without affecting
5072 DECL_ATTRIBUTES (current_function_decl).
5073 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
5074 functions with non-NULL DECL_ABSTRACT_ORIGIN.
5076 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
5078 PR tree-optimization/83979
5079 * fold-const.c (fold_comparison): Use constant_boolean_node
5080 instead of boolean_{true,false}_node.
5082 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
5084 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
5087 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5089 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5090 Simplify the clause that sets the length attribute.
5091 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5092 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
5093 clause that sets the length attribute.
5094 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5096 2018-01-24 Tom de Vries <tom@codesourcery.com>
5099 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
5100 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
5101 Add strict parameter.
5102 (prevent_branch_around_nothing): Insert dummy insn between branch to
5103 label and label with no ptx insn inbetween.
5104 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
5106 2018-01-24 Tom de Vries <tom@codesourcery.com>
5109 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5110 for neutered threads in warp.
5111 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
5113 2018-01-24 Richard Biener <rguenther@suse.de>
5115 PR tree-optimization/83176
5116 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
5119 2018-01-24 Richard Biener <rguenther@suse.de>
5121 PR tree-optimization/82819
5122 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
5123 code generating pluses that are no-ops in the target precision.
5125 2018-01-24 Richard Biener <rguenther@suse.de>
5128 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
5130 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5132 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
5133 to merge probabilities.
5134 * predict.c (probably_never_executed): Also mark as cold functions
5135 with global 0 profile and guessed local profile.
5136 * profile-count.c (profile_probability::combine_with_count): New
5138 * profile-count.h (profile_probability::operator*,
5139 profile_probability::operator*=, profile_probability::operator/,
5140 profile_probability::operator/=): Reduce precision to adjusted
5141 and set value to guessed on contradictory divisions.
5142 (profile_probability::combine_with_freq): Remove.
5143 (profile_probability::combine_wiht_count): Declare.
5144 (profile_count::force_nonzero):: Set to adjusted.
5145 (profile_count::probability_in):: Set quality to adjusted.
5146 * tree-ssa-tail-merge.c (replace_block_by): Use
5149 2018-01-23 Andrew Waterman <andrew@sifive.com>
5150 Jim Wilson <jimw@sifive.com>
5152 * config/riscv/riscv.c (riscv_stack_boundary): New.
5153 (riscv_option_override): Set riscv_stack_boundary. Handle
5154 riscv_preferred_stack_boundary_arg.
5155 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
5156 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
5157 (STACK_BOUNDARY): Set to riscv_stack_boundary.
5158 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
5159 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
5160 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
5162 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
5165 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
5166 of struct ix86_frame.
5167 (ix86_expand_epilogue): Likewise. Add a local variable for
5168 the reg_save_offset field in struct ix86_frame.
5170 2018-01-23 Bin Cheng <bin.cheng@arm.com>
5172 PR tree-optimization/82604
5173 * tree-loop-distribution.c (enum partition_kind): New enum item
5174 PKIND_PARTIAL_MEMSET.
5175 (partition_builtin_p): Support above new enum item.
5176 (generate_code_for_partition): Ditto.
5177 (compute_access_range): Differentiate cases that equality can be
5178 proven at all loops, the innermost loops or no loops.
5179 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
5180 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
5181 (finalize_partitions, distribute_loop): Don't fuse partition of
5182 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
5183 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
5186 2018-01-23 Martin Liska <mliska@suse.cz>
5188 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
5189 order to ignore the predictor.
5190 (PRED_POLYMORPHIC_CALL): Likewise.
5191 (PRED_RECURSIVE_CALL): Likewise.
5193 2018-01-23 Martin Liska <mliska@suse.cz>
5195 * tree-profile.c (tree_profiling): Print function header to
5196 aware reader which function we are working on.
5197 * value-prof.c (gimple_find_values_to_profile): Do not print
5198 not interesting value histograms.
5200 2018-01-23 Martin Liska <mliska@suse.cz>
5202 * profile-count.h (enum profile_quality): Add
5203 profile_uninitialized as the first value. Do not number values
5204 as they are zero based.
5205 (profile_count::verify): Update sanity check.
5206 (profile_probability::verify): Likewise.
5208 2018-01-23 Nathan Sidwell <nathan@acm.org>
5210 * doc/invoke.texi (ffor-scope): Deprecate.
5212 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5214 PR tree-optimization/83510
5215 * domwalk.c (set_all_edges_as_executable): New function.
5216 (dom_walker::dom_walker): Convert bool param
5217 "skip_unreachable_blocks" to enum reachability. Move setup of
5218 edge flags to set_all_edges_as_executable and only do it when
5219 reachability is REACHABLE_BLOCKS.
5220 * domwalk.h (enum dom_walker::reachability): New enum.
5221 (dom_walker::dom_walker): Convert bool param
5222 "skip_unreachable_blocks" to enum reachability.
5223 (set_all_edges_as_executable): New decl.
5224 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
5225 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
5227 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
5228 but converting true to REACHABLE_BLOCKS.
5229 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
5231 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
5232 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
5233 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
5235 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
5236 if check_all_array_refs will be called.
5238 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5240 * tree.c (selftest::test_location_wrappers): Add more test
5243 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5245 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
5246 (selftest::test_bit_in_range): Likewise.
5248 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5251 * doc/sourcebuild.texi (vect_float): Say that the selector
5252 only describes the situation when -funsafe-math-optimizations is on.
5253 (vect_float_strict): Document.
5255 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5257 PR tree-optimization/83965
5258 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
5259 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
5260 instead of checking only for a reduction.
5261 (vect_recog_widen_sum_pattern): Likewise.
5263 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5265 * predict.c (probably_never_executed): Only use precise profile info.
5266 (compute_function_frequency): Skip after inlining hack since we now
5267 have quality checking.
5269 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5271 * profile-count.h (profile_probability::very_unlikely,
5272 profile_probability::unlikely, profile_probability::even): Set
5273 precision to guessed.
5275 2018-01-23 Richard Biener <rguenther@suse.de>
5277 PR tree-optimization/83963
5278 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
5279 Properly terminate dominator walk when crossing the exit edge not
5280 when visiting its source block.
5282 2018-01-23 Jakub Jelinek <jakub@redhat.com>
5285 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
5286 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
5288 2018-01-22 Jakub Jelinek <jakub@redhat.com>
5290 PR tree-optimization/83957
5291 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
5292 semicolon after for body surrounded by braces.
5294 PR tree-optimization/83081
5295 * profile-count.h (profile_probability::split): New method.
5296 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
5297 Use profile_probability::split.
5298 (do_compare_rtx_and_jump): Fix adjustment of probabilities
5299 when splitting a single conditional jump into 2.
5301 2018-01-22 David Malcolm <dmalcolm@redhat.com>
5303 PR tree-optimization/69452
5304 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
5307 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5309 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
5310 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
5311 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
5313 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5315 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
5316 * config/rl78/rl78.md (movdi): New define_expand.
5317 * config/rl78/rl78.c (rl78_split_movdi): New function.
5319 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
5322 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
5324 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
5325 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
5326 128-bit to produce an UNSPEC move to get the double word with the
5327 signbit and then a shift directly to do signbit.
5328 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
5329 implementation with a new version that just does either a direct
5330 move or a regular move. Move memory interface to separate insns.
5331 Move insns so they are next to the expander.
5332 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
5333 with signbit move. Split big and little endian case.
5334 (signbit<mode>2_dm_mem_le): Likewise.
5335 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
5336 (signbit<mode>2_dm2): Likewise.
5338 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5340 * config/rl78/rl78.md (anddi3): New define_expand.
5342 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5344 * config/rl78/rl78.md (umindi3): New define_expand.
5346 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5348 * config/rl78/rl78.md (smindi3): New define_expand.
5350 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5352 * config/rl78/rl78.md (smaxdi3): New define_expand.
5354 2018-01-22 Carl Love <cel@us.ibm.com>
5356 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
5357 LVX_V1TI): Add macro expansion.
5358 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
5359 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
5360 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
5361 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
5362 Change check to determine if the instruction is a byte reversing
5363 entry. Fix typo in comment.
5364 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
5365 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
5366 Add def_builtin calls for new builtins.
5367 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
5368 Add define_insn expansion.
5370 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5372 * config/rl78/rl78.md (umaxdi3): New define_expand.
5374 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5376 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
5377 for non-QImode registers.
5379 2018-01-22 Richard Biener <rguenther@suse.de>
5381 PR tree-optimization/83963
5382 * graphite-scop-detection.c (scop_detection::get_sese): Delay
5383 including the loop exit block.
5384 (scop_detection::merge_sese): Likewise.
5385 (scop_detection::add_scop): Do it here instead.
5387 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5389 * doc/sourcebuild.texi (arm_softfloat): Document.
5391 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
5394 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
5395 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
5396 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
5398 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5399 David Edelsohn <dje.gcc@gmail.com>
5402 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5403 Change "crset eq" to "crset 2".
5404 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5405 (*call_indirect_aix<mode>_nospec): Likewise.
5406 (*call_value_indirect_aix<mode>_nospec): Likewise.
5407 (*call_indirect_elfv2<mode>_nospec): Likewise.
5408 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
5409 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
5410 change assembly output from . to $.
5411 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5412 (indirect_jump<mode>_nospec): Change assembly output from . to $.
5413 (*tablejump<mode>_internal1_nospec): Likewise.
5415 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
5418 * config/sh/sh_optimize_sett_clrt.cc:
5419 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
5421 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
5423 PR tree-optimization/83940
5424 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
5425 offset_dt to vect_constant_def rather than vect_unknown_def_type.
5426 (vect_check_load_store_mask): Add a mask_dt_out parameter and
5427 use it to pass back the definition type.
5428 (vect_check_store_rhs): Likewise rhs_dt_out.
5429 (vect_build_gather_load_calls): Add a mask_dt argument and use
5430 it instead of a call to vect_is_simple_use.
5431 (vectorizable_store): Update calls to vect_check_load_store_mask
5432 and vect_check_store_rhs. Use the dt returned by the latter instead
5433 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
5434 instead of calls to vect_is_simple_use. Pass the scalar rather
5435 than the vector operand to vect_is_simple_use when handling
5436 second and subsequent copies of an rhs value.
5437 (vectorizable_load): Update calls to vect_check_load_store_mask
5438 and vect_build_gather_load_calls. Use the cached mask_dt and
5439 gs_info.offset_dt instead of calls to vect_is_simple_use.
5441 2018-01-20 Jakub Jelinek <jakub@redhat.com>
5444 * tree-emutls.c: Include gimplify.h.
5445 (lower_emutls_2): New function.
5446 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
5447 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
5448 it before further processing.
5451 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
5452 UINTVAL (trueop1) instead of INTVAL (op1).
5454 2018-01-19 Jakub Jelinek <jakub@redhat.com>
5458 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
5459 INCOMING_FRAME_SP_OFFSET if not defined.
5460 (scan_trace): Add ENTRY argument. If true and
5461 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
5462 emit a note to adjust the CFA offset.
5463 (create_cfi_notes): Adjust scan_trace callers.
5464 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
5465 INCOMING_FRAME_SP_OFFSET in the CIE.
5466 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
5467 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
5469 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
5470 * doc/tm.texi: Regenerated.
5472 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5474 PR rtl-optimization/83147
5475 * lra-constraints.c (remove_inheritance_pseudos): Use
5476 lra_substitute_pseudo_within_insn.
5478 2018-01-19 Tom de Vries <tom@codesourcery.com>
5479 Cesar Philippidis <cesar@codesourcery.com>
5482 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
5484 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
5487 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
5488 spaces for function labels.
5490 2018-01-19 Martin Liska <mliska@suse.cz>
5492 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
5493 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
5494 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
5495 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
5496 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
5497 (PRED_CONST_RETURN): Change from 69 to 65.
5498 (PRED_NULL_RETURN): Change from 91 to 71.
5499 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
5500 (PRED_LOOP_GUARD): Change from 66 to 73.
5502 2018-01-19 Martin Liska <mliska@suse.cz>
5504 * predict.c (predict_insn_def): Add new assert.
5505 (struct branch_predictor): Change type to signed integer.
5506 (test_prediction_value_range): Amend test to cover
5508 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
5509 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
5510 (PRED_LOOP_ITERATIONS_MAX): Likewise.
5511 (PRED_LOOP_IV_COMPARE): Likewise.
5512 * predict.h (PROB_UNINITIALIZED): Define new constant.
5514 2018-01-19 Martin Liska <mliska@suse.cz>
5516 * predict.c (dump_prediction): Add new format for
5517 analyze_brprob.py script which is enabled with -details
5519 * profile-count.h (precise_p): New function.
5521 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
5523 PR tree-optimization/83922
5524 * tree-vect-loop.c (vect_verify_full_masking): Return false if
5525 there are no statements that need masking.
5526 (vect_active_double_reduction_p): New function.
5527 (vect_analyze_loop_operations): Use it when handling phis that
5528 are not in the loop header.
5530 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
5532 PR tree-optimization/83914
5533 * tree-vect-loop.c (vectorizable_induction): Don't convert
5534 init_expr or apply the peeling adjustment for inductions
5535 that are nested within the vectorized loop.
5537 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5539 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
5542 2018-01-18 Jakub Jelinek <jakub@redhat.com>
5546 * function.h (gimplify_parameters): Add gimple_seq * argument.
5547 * function.c: Include gimple.h and options.h.
5548 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
5549 for the added local temporaries if needed.
5550 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
5551 if there are any parameter cleanups, wrap whole body into a
5552 try/finally with the cleanups.
5554 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
5557 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
5558 Use GET_MODE_CLASS for scalar floating point.
5560 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
5564 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
5565 Fix call of call_cgraph_insertion_hooks.
5567 2018-01-18 Martin Sebor <msebor@redhat.com>
5569 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
5571 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
5574 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
5577 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
5580 * common.opt: (-ffile-prefix-map): New option.
5581 * opts.c (common_handle_option): Defer it.
5582 * opts-global.c (handle_common_deferred_options): Handle it.
5583 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
5584 * file-prefix-map.h: New file.
5585 (remap_debug_filename, add_debug_prefix_map): ...here.
5586 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
5587 * final.c (debug_prefix_map, add_debug_prefix_map
5588 remap_debug_filename): Move to...
5589 * file-prefix-map.c: New file.
5590 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
5591 generalize, get rid of alloca(), use strrchr() instead of strchr().
5592 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
5593 Implement in terms of add_prefix_map().
5594 (remap_macro_filename, remap_debug_filename): Implement in term of
5596 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
5597 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
5598 * dbxout.c: Include file-prefix-map.h.
5599 * varasm.c: Likewise.
5600 * vmsdbgout.c: Likewise.
5601 * xcoffout.c: Likewise.
5602 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
5603 * doc/cppopts.texi (-fmacro-prefix-map): Document.
5604 * doc/invoke.texi (-ffile-prefix-map): Document.
5605 (-fdebug-prefix-map): Update description.
5607 2018-01-18 Martin Liska <mliska@suse.cz>
5609 * config/i386/i386.c (indirect_thunk_name): Document that also
5611 (output_indirect_thunk): Document why both instructions
5612 (pause and lfence) are generated.
5614 2018-01-18 Richard Biener <rguenther@suse.de>
5616 PR tree-optimization/83887
5617 * graphite-scop-detection.c
5618 (scop_detection::get_nearest_dom_with_single_entry): Remove.
5619 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
5620 (scop_detection::merge_sese): Re-implement with a flood-fill
5621 algorithm that properly finds a SESE region if it exists.
5623 2018-01-18 Jakub Jelinek <jakub@redhat.com>
5626 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
5627 pointer_diff optimizations use view_convert instead of convert.
5629 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5631 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5632 Generate different code for -mno-speculate-indirect-jumps.
5633 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5634 (*call_indirect_aix<mode>): Disable for
5635 -mno-speculate-indirect-jumps.
5636 (*call_indirect_aix<mode>_nospec): New define_insn.
5637 (*call_value_indirect_aix<mode>): Disable for
5638 -mno-speculate-indirect-jumps.
5639 (*call_value_indirect_aix<mode>_nospec): New define_insn.
5640 (*sibcall_nonlocal_sysv<mode>): Generate different code for
5641 -mno-speculate-indirect-jumps.
5642 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5644 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
5646 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
5647 long double type, set the flags for noting the default long double
5648 type, even if we don't pass or return a long double type.
5650 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
5653 * ipa-inline.c (flatten_function): Do not overwrite final inlining
5656 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
5658 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
5659 support for merge[hl].
5660 (fold_mergehl_helper): New helper function.
5661 (tree-vector-builder.h): New #include for tree_vector_builder usage.
5662 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
5663 (altivec_vmrglw_direct): Add xxmrglw insn.
5665 2018-01-17 Andrew Waterman <andrew@sifive.com>
5667 * config/riscv/riscv.c (riscv_conditional_register_usage): If
5668 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
5670 2018-01-17 David Malcolm <dmalcolm@redhat.com>
5673 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
5674 call the lto_location_cache before reading the
5675 DECL_SOURCE_LOCATION of the types.
5677 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
5678 Richard Sandiford <richard.sandiford@linaro.org>
5680 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
5681 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
5682 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
5683 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
5684 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
5686 * config/aarch64/constraints.md (aarch64_movti_operand):
5688 * config/aarch64/predicates.md (Uti): Add new constraint.
5690 2018-01-17 Carl Love <cel@us.ibm.com>
5692 * config/rs6000/vsx.md (define_expand xl_len_r,
5693 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
5694 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
5696 (define_expand, define_insn): Move the shift left from the
5697 define_insn to the define_expand for lxvl and stxvl instructions.
5698 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
5699 and XL_LEN_R definitions to PURE.
5701 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
5703 * config/i386/i386.c (indirect_thunk_name): Declare regno
5704 as unsigned int. Compare regno with INVALID_REGNUM.
5705 (output_indirect_thunk): Ditto.
5706 (output_indirect_thunk_function): Ditto.
5707 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
5708 in the call to output_indirect_thunk_function.
5710 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
5713 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
5714 rather than the size of inner_type to determine the stack slot size
5715 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
5717 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
5720 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
5723 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
5725 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
5726 endian Linux systems to optionally enable multilibs for selecting
5727 the long double type if the user configured an explicit type.
5728 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
5729 have no long double multilibs if not defined.
5730 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
5731 warn if the user used -mabi={ieee,ibm}longdouble and we built
5732 multilibs for long double.
5733 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
5734 appropriate multilib option.
5735 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
5737 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
5738 for building long double multilibs.
5739 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
5741 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
5743 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
5746 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
5748 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
5751 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
5754 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
5757 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5759 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
5760 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
5762 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
5764 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
5765 different rtl trees depending on TARGET_64BIT.
5766 (rs6000_gen_lvx): Likewise.
5768 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5770 * config/visium/visium.md (nop): Tweak comment.
5771 (hazard_nop): Likewise.
5773 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5775 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
5776 -mspeculate-indirect-jumps.
5777 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
5778 for -mno-speculate-indirect-jumps.
5779 (*call_indirect_elfv2<mode>_nospec): New define_insn.
5780 (*call_value_indirect_elfv2<mode>): Disable for
5781 -mno-speculate-indirect-jumps.
5782 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
5783 (indirect_jump): Emit different RTL for
5784 -mno-speculate-indirect-jumps.
5785 (*indirect_jump<mode>): Disable for
5786 -mno-speculate-indirect-jumps.
5787 (*indirect_jump<mode>_nospec): New define_insn.
5788 (tablejump): Emit different RTL for
5789 -mno-speculate-indirect-jumps.
5790 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
5791 (tablejumpsi_nospec): New define_expand.
5792 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
5793 (tablejumpdi_nospec): New define_expand.
5794 (*tablejump<mode>_internal1): Disable for
5795 -mno-speculate-indirect-jumps.
5796 (*tablejump<mode>_internal1_nospec): New define_insn.
5797 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
5800 2018-01-16 Artyom Skrobov tyomitch@gmail.com
5802 * caller-save.c (insert_save): Drop unnecessary parameter. All
5805 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5806 Richard Biener <rguenth@suse.de>
5809 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
5810 return early, inline manually is_gimple_sizepos. Make sure if we
5811 call gimplify_expr we don't end up with a gimple constant.
5812 * tree.c (variably_modified_type_p): Don't return true for
5813 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
5814 * gimplify.h (is_gimple_sizepos): Remove.
5816 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5818 PR tree-optimization/83857
5819 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
5820 vectorizable_live_operation for pure SLP statements.
5821 (vectorizable_live_operation): Handle PHIs.
5823 2018-01-16 Richard Biener <rguenther@suse.de>
5825 PR tree-optimization/83867
5826 * tree-vect-stmts.c (vect_transform_stmt): Precompute
5827 nested_in_vect_loop_p since the scalar stmt may get invalidated.
5829 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5832 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
5833 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
5834 If off is not INTEGER_CST, issue a may not be aligned warning
5835 rather than isn't aligned. Use isn%'t rather than isn't.
5836 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
5838 <case MULT_EXPR>: Improve the case when bottom and one of the
5839 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
5840 operand, in that case check if the other operand is multiple of
5841 bottom divided by the INTEGER_CST operand.
5843 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5846 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
5847 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
5848 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
5849 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
5850 * config/pa/pa.c (pa_function_arg_advance): Likewise.
5851 (pa_function_arg, pa_arg_partial_bytes): Likewise.
5852 (pa_function_arg_size): New function.
5854 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5856 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
5857 in a separate statement.
5859 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5861 PR tree-optimization/83847
5862 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
5863 group gathers and scatters.
5865 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5867 PR rtl-optimization/86620
5868 * params.def (max-sched-ready-insns): Bump minimum value to 1.
5870 PR rtl-optimization/83213
5871 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
5872 to last if both are JUMP_INSNs.
5874 PR tree-optimization/83843
5875 * gimple-ssa-store-merging.c
5876 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
5877 store_immediate_info for bswap/nop orig_stores.
5879 2018-01-15 Andrew Waterman <andrew@sifive.com>
5881 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
5883 <UDIV>: Increase cost if !TARGET_DIV.
5885 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
5887 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
5888 (define_attr "cr_logical_3op"): New.
5889 (cceq_ior_compare): Adjust.
5890 (cceq_ior_compare_complement): Adjust.
5891 (*cceq_rev_compare): Adjust.
5892 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
5893 (is_cracked_insn): Adjust.
5894 (insn_must_be_first_in_group): Adjust.
5895 * config/rs6000/40x.md: Adjust.
5896 * config/rs6000/440.md: Adjust.
5897 * config/rs6000/476.md: Adjust.
5898 * config/rs6000/601.md: Adjust.
5899 * config/rs6000/603.md: Adjust.
5900 * config/rs6000/6xx.md: Adjust.
5901 * config/rs6000/7450.md: Adjust.
5902 * config/rs6000/7xx.md: Adjust.
5903 * config/rs6000/8540.md: Adjust.
5904 * config/rs6000/cell.md: Adjust.
5905 * config/rs6000/e300c2c3.md: Adjust.
5906 * config/rs6000/e500mc.md: Adjust.
5907 * config/rs6000/e500mc64.md: Adjust.
5908 * config/rs6000/e5500.md: Adjust.
5909 * config/rs6000/e6500.md: Adjust.
5910 * config/rs6000/mpc.md: Adjust.
5911 * config/rs6000/power4.md: Adjust.
5912 * config/rs6000/power5.md: Adjust.
5913 * config/rs6000/power6.md: Adjust.
5914 * config/rs6000/power7.md: Adjust.
5915 * config/rs6000/power8.md: Adjust.
5916 * config/rs6000/power9.md: Adjust.
5917 * config/rs6000/rs64.md: Adjust.
5918 * config/rs6000/titan.md: Adjust.
5920 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5922 * config/i386/predicates.md (indirect_branch_operand): Rewrite
5923 ix86_indirect_branch_register logic.
5925 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5927 * config/i386/constraints.md (Bs): Update
5928 ix86_indirect_branch_register check. Don't check
5929 ix86_indirect_branch_register with GOT_memory_operand.
5931 * config/i386/predicates.md (GOT_memory_operand): Don't check
5932 ix86_indirect_branch_register here.
5933 (GOT32_symbol_operand): Likewise.
5935 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5937 * config/i386/predicates.md (constant_call_address_operand):
5938 Rewrite ix86_indirect_branch_register logic.
5939 (sibcall_insn_operand): Likewise.
5941 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5943 * config/i386/constraints.md (Bs): Replace
5944 ix86_indirect_branch_thunk_register with
5945 ix86_indirect_branch_register.
5947 * config/i386/i386.md (indirect_jump): Likewise.
5948 (tablejump): Likewise.
5949 (*sibcall_memory): Likewise.
5950 (*sibcall_value_memory): Likewise.
5951 Peepholes of indirect call and jump via memory: Likewise.
5952 * config/i386/i386.opt: Likewise.
5953 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5954 (GOT_memory_operand): Likewise.
5955 (call_insn_operand): Likewise.
5956 (sibcall_insn_operand): Likewise.
5957 (GOT32_symbol_operand): Likewise.
5959 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5962 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
5963 type rather than type addr's type points to.
5964 (expand_omp_atomic_mutex): Likewise.
5965 (expand_omp_atomic): Likewise.
5967 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5970 * config/i386/i386.c (output_indirect_thunk_function): Use
5971 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
5972 for __x86_return_thunk.
5974 2018-01-15 Richard Biener <rguenther@suse.de>
5977 * expmed.c (extract_bit_field_1): Fix typo.
5979 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5982 * config/arm/iterators.md (VF): New mode iterator.
5983 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
5984 Remove integer-related logic from pattern.
5985 (neon_vabd<mode>_3): Likewise.
5987 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5990 * common.opt (fstrict-overflow): No longer an alias.
5991 (fwrapv-pointer): New option.
5992 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
5993 also for pointer types based on flag_wrapv_pointer.
5994 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
5995 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
5996 opts->x_flag_wrapv got set.
5997 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
5998 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
5999 POINTER_TYPE_OVERFLOW_UNDEFINED.
6000 * match.pd: Likewise in address comparison pattern.
6001 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
6003 2018-01-15 Richard Biener <rguenther@suse.de>
6006 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
6007 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
6008 Reset type names to their identifier if their TYPE_DECL doesn't
6009 have linkage (and thus is used for ODR and devirt).
6010 (save_debug_info_for_decl): Remove.
6011 (save_debug_info_for_type): Likewise.
6012 (add_tree_to_fld_list): Adjust.
6013 * tree-pretty-print.c (dump_generic_node): Make dumping of
6014 type names more robust.
6016 2018-01-15 Richard Biener <rguenther@suse.de>
6018 * BASE-VER: Bump to 8.0.1.
6020 2018-01-14 Martin Sebor <msebor@redhat.com>
6023 * builtins.c (check_access): Avoid warning when the no-warning bit
6026 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
6028 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
6029 * ira-color (allocno_hard_regs_compare): Likewise.
6031 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
6034 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
6035 Use .pushsection/.popsection.
6037 2018-01-14 Martin Sebor <msebor@redhat.com>
6040 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
6042 2018-01-14 Jakub Jelinek <jakub@redhat.com>
6044 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
6045 entry from extra_headers.
6046 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
6047 extra_headers, make the list bitwise identical to the i?86-*-* one.
6049 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6051 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
6052 -mcmodel=large with -mindirect-branch=thunk,
6053 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
6054 -mfunction-return=thunk-extern.
6055 * doc/invoke.texi: Document -mcmodel=large is incompatible with
6056 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
6057 -mfunction-return=thunk and -mfunction-return=thunk-extern.
6059 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6061 * config/i386/i386.c (print_reg): Print the name of the full
6062 integer register without '%'.
6063 (ix86_print_operand): Handle 'V'.
6064 * doc/extend.texi: Document 'V' modifier.
6066 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6068 * config/i386/constraints.md (Bs): Disallow memory operand for
6069 -mindirect-branch-register.
6071 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6072 (GOT_memory_operand): Likewise.
6073 (call_insn_operand): Likewise.
6074 (sibcall_insn_operand): Likewise.
6075 (GOT32_symbol_operand): Likewise.
6076 * config/i386/i386.md (indirect_jump): Call convert_memory_address
6077 for -mindirect-branch-register.
6078 (tablejump): Likewise.
6079 (*sibcall_memory): Likewise.
6080 (*sibcall_value_memory): Likewise.
6081 Disallow peepholes of indirect call and jump via memory for
6082 -mindirect-branch-register.
6083 (*call_pop): Replace m with Bw.
6084 (*call_value_pop): Likewise.
6085 (*sibcall_pop_memory): Replace m with Bs.
6086 * config/i386/i386.opt (mindirect-branch-register): New option.
6087 * doc/invoke.texi: Document -mindirect-branch-register option.
6089 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6091 * config/i386/i386-protos.h (ix86_output_function_return): New.
6092 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
6093 set function_return_type.
6094 (indirect_thunk_name): Add ret_p to indicate thunk for function
6096 (output_indirect_thunk_function): Pass false to
6097 indirect_thunk_name.
6098 (ix86_output_indirect_branch_via_reg): Likewise.
6099 (ix86_output_indirect_branch_via_push): Likewise.
6100 (output_indirect_thunk_function): Create alias for function
6101 return thunk if regno < 0.
6102 (ix86_output_function_return): New function.
6103 (ix86_handle_fndecl_attribute): Handle function_return.
6104 (ix86_attribute_table): Add function_return.
6105 * config/i386/i386.h (machine_function): Add
6106 function_return_type.
6107 * config/i386/i386.md (simple_return_internal): Use
6108 ix86_output_function_return.
6109 (simple_return_internal_long): Likewise.
6110 * config/i386/i386.opt (mfunction-return=): New option.
6111 (indirect_branch): Mention -mfunction-return=.
6112 * doc/extend.texi: Document function_return function attribute.
6113 * doc/invoke.texi: Document -mfunction-return= option.
6115 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6117 * config/i386/i386-opts.h (indirect_branch): New.
6118 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
6119 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
6120 with local indirect jump when converting indirect call and jump.
6121 (ix86_set_indirect_branch_type): New.
6122 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
6123 (indirectlabelno): New.
6124 (indirect_thunk_needed): Likewise.
6125 (indirect_thunk_bnd_needed): Likewise.
6126 (indirect_thunks_used): Likewise.
6127 (indirect_thunks_bnd_used): Likewise.
6128 (INDIRECT_LABEL): Likewise.
6129 (indirect_thunk_name): Likewise.
6130 (output_indirect_thunk): Likewise.
6131 (output_indirect_thunk_function): Likewise.
6132 (ix86_output_indirect_branch_via_reg): Likewise.
6133 (ix86_output_indirect_branch_via_push): Likewise.
6134 (ix86_output_indirect_branch): Likewise.
6135 (ix86_output_indirect_jmp): Likewise.
6136 (ix86_code_end): Call output_indirect_thunk_function if needed.
6137 (ix86_output_call_insn): Call ix86_output_indirect_branch if
6139 (ix86_handle_fndecl_attribute): Handle indirect_branch.
6140 (ix86_attribute_table): Add indirect_branch.
6141 * config/i386/i386.h (machine_function): Add indirect_branch_type
6142 and has_local_indirect_jump.
6143 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
6145 (tablejump): Likewise.
6146 (*indirect_jump): Use ix86_output_indirect_jmp.
6147 (*tablejump_1): Likewise.
6148 (simple_return_indirect_internal): Likewise.
6149 * config/i386/i386.opt (mindirect-branch=): New option.
6150 (indirect_branch): New.
6153 (thunk-inline): Likewise.
6154 (thunk-extern): Likewise.
6155 * doc/extend.texi: Document indirect_branch function attribute.
6156 * doc/invoke.texi: Document -mindirect-branch= option.
6158 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
6161 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
6163 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
6165 * ipa-inline.c (want_inline_small_function_p): Return false if
6166 inlining has already failed with CIF_FINAL_ERROR.
6167 (update_caller_keys): Call want_inline_small_function_p before
6169 (update_callee_keys): Likewise.
6171 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6173 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
6175 (rs6000_quadword_masked_address_p): Likewise.
6176 (quad_aligned_load_p): Likewise.
6177 (quad_aligned_store_p): Likewise.
6178 (const_load_sequence_p): Add comment to describe the outer-most loop.
6179 (mimic_memory_attributes_and_flags): New function.
6180 (rs6000_gen_stvx): Likewise.
6181 (replace_swapped_aligned_store): Likewise.
6182 (rs6000_gen_lvx): Likewise.
6183 (replace_swapped_aligned_load): Likewise.
6184 (replace_swapped_load_constant): Capitalize argument name in
6185 comment describing this function.
6186 (rs6000_analyze_swaps): Add a third pass to search for vector loads
6187 and stores that access quad-word aligned addresses and replace
6188 with stvx or lvx instructions when appropriate.
6189 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
6190 New function prototype.
6191 (rs6000_quadword_masked_address_p): Likewise.
6192 (rs6000_gen_lvx): Likewise.
6193 (rs6000_gen_stvx): Likewise.
6194 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
6195 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
6196 when memory address is aligned.
6197 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
6198 this split to select lvx instruction when memory address is aligned.
6199 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
6200 instruction when memory address is aligned.
6201 (*vsx_le_perm_load_v16qi): Likewise.
6202 (four unnamed splitters): Modify to select the stvx instruction
6203 when memory is aligned.
6205 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
6207 * predict.c (determine_unlikely_bbs): Handle correctly BBs
6208 which appears in the queue multiple times.
6210 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6211 Alan Hayward <alan.hayward@arm.com>
6212 David Sherwood <david.sherwood@arm.com>
6214 * tree-vectorizer.h (vec_lower_bound): New structure.
6215 (_loop_vec_info): Add check_nonzero and lower_bounds.
6216 (LOOP_VINFO_CHECK_NONZERO): New macro.
6217 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
6218 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
6219 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
6220 fields. Make seg_len the distance travelled, not including the
6222 (dr_direction_indicator): Declare.
6223 (dr_zero_step_indicator): Likewise.
6224 (dr_known_forward_stride_p): Likewise.
6225 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
6227 (runtime_alias_check_p): Allow runtime alias checks with
6229 (operator ==): Compare access_size and align.
6230 (prune_runtime_alias_test_list): Rework for new distinction between
6231 the access_size and seg_len.
6232 (create_intersect_range_checks_index): Likewise. Cope with polynomial
6234 (get_segment_min_max): New function.
6235 (create_intersect_range_checks): Use it.
6236 (dr_step_indicator): New function.
6237 (dr_direction_indicator): Likewise.
6238 (dr_zero_step_indicator): Likewise.
6239 (dr_known_forward_stride_p): Likewise.
6240 * tree-loop-distribution.c (data_ref_segment_size): Return
6241 DR_STEP * (niters - 1).
6242 (compute_alias_check_pairs): Update call to the dr_with_seg_len
6244 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
6245 (vect_preserves_scalar_order_p): New function, split out from...
6246 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
6247 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
6248 (vect_vfa_access_size): New function.
6249 (vect_vfa_align): Likewise.
6250 (vect_compile_time_alias): Take access_size_a and access_b arguments.
6251 (dump_lower_bound): New function.
6252 (vect_check_lower_bound): Likewise.
6253 (vect_small_gap_p): Likewise.
6254 (vectorizable_with_step_bound_p): Likewise.
6255 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
6256 depencies if the vectorization factor is 1. Convert the checks
6257 for nonzero steps into checks on the bounds of DR_STEP. Try using
6258 a bunds check for variable steps if the minimum required step is
6259 relatively small. Update calls to the dr_with_seg_len
6260 constructor and to vect_compile_time_alias.
6261 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
6263 (vect_loop_versioning): Call it.
6264 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
6266 (vect_estimate_min_profitable_iters): Account for any bounds checks.
6268 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6269 Alan Hayward <alan.hayward@arm.com>
6270 David Sherwood <david.sherwood@arm.com>
6272 * doc/sourcebuild.texi (vect_scatter_store): Document.
6273 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
6275 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
6277 * genopinit.c (main): Add supports_vec_scatter_store and
6278 supports_vec_scatter_store_cached to target_optabs.
6279 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
6280 IFN_MASK_SCATTER_STORE.
6281 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
6283 * internal-fn.h (internal_store_fn_p): Declare.
6284 (internal_fn_stored_value_index): Likewise.
6285 * internal-fn.c (scatter_store_direct): New macro.
6286 (expand_scatter_store_optab_fn): New function.
6287 (direct_scatter_store_optab_supported_p): New macro.
6288 (internal_store_fn_p): New function.
6289 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
6290 IFN_MASK_SCATTER_STORE.
6291 (internal_fn_mask_index): Likewise.
6292 (internal_fn_stored_value_index): New function.
6293 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
6295 * optabs-query.h (supports_vec_scatter_store_p): Declare.
6296 * optabs-query.c (supports_vec_scatter_store_p): New function.
6297 * tree-vectorizer.h (vect_get_store_rhs): Declare.
6298 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
6299 true for scatter stores.
6300 (vect_gather_scatter_fn_p): Handle scatter stores too.
6301 (vect_check_gather_scatter): Consider using scatter stores if
6302 supports_vec_scatter_store_p.
6303 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
6305 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6306 internal_fn_stored_value_index.
6307 (check_load_store_masking): Handle scatter stores too.
6308 (vect_get_store_rhs): Make public.
6309 (vectorizable_call): Use internal_store_fn_p.
6310 (vectorizable_store): Handle scatter store internal functions.
6311 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
6312 when deciding whether the end of the group has been reached.
6313 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
6314 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
6315 (mask_scatter_store<mode>): New insns.
6317 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6318 Alan Hayward <alan.hayward@arm.com>
6319 David Sherwood <david.sherwood@arm.com>
6321 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
6322 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
6323 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
6325 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
6326 Use vect_truncate_gather_scatter_offset if we can't treat the
6327 operation as a normal gather load or scatter store.
6328 (get_group_load_store_type): Take the gather_scatter_info
6329 as argument. Try using a gather load or scatter store for
6330 single-element groups.
6331 (get_load_store_type): Update calls to get_group_load_store_type
6332 and vect_use_strided_gather_scatters_p.
6334 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6335 Alan Hayward <alan.hayward@arm.com>
6336 David Sherwood <david.sherwood@arm.com>
6338 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
6339 optional tree argument.
6340 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
6342 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
6343 but continue to use the current value as a fallback.
6344 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
6345 to compare the updates.
6346 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
6347 (get_load_store_type): Use it when handling a strided access.
6348 (vect_get_strided_load_store_ops): New function.
6349 (vect_get_data_ptr_increment): Likewise.
6350 (vectorizable_load): Handle strided gather loads. Always pass
6351 a step to vect_create_data_ref_ptr and bump_vector_ptr.
6353 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6354 Alan Hayward <alan.hayward@arm.com>
6355 David Sherwood <david.sherwood@arm.com>
6357 * doc/md.texi (gather_load@var{m}): Document.
6358 (mask_gather_load@var{m}): Likewise.
6359 * genopinit.c (main): Add supports_vec_gather_load and
6360 supports_vec_gather_load_cached to target_optabs.
6361 * optabs-tree.c (init_tree_optimization_optabs): Use
6362 ggc_cleared_alloc to allocate target_optabs.
6363 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
6364 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
6366 * internal-fn.h (internal_load_fn_p): Declare.
6367 (internal_gather_scatter_fn_p): Likewise.
6368 (internal_fn_mask_index): Likewise.
6369 (internal_gather_scatter_fn_supported_p): Likewise.
6370 * internal-fn.c (gather_load_direct): New macro.
6371 (expand_gather_load_optab_fn): New function.
6372 (direct_gather_load_optab_supported_p): New macro.
6373 (direct_internal_fn_optab): New function.
6374 (internal_load_fn_p): Likewise.
6375 (internal_gather_scatter_fn_p): Likewise.
6376 (internal_fn_mask_index): Likewise.
6377 (internal_gather_scatter_fn_supported_p): Likewise.
6378 * optabs-query.c (supports_at_least_one_mode_p): New function.
6379 (supports_vec_gather_load_p): Likewise.
6380 * optabs-query.h (supports_vec_gather_load_p): Declare.
6381 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
6382 and memory_type field.
6383 (NUM_PATTERNS): Bump to 15.
6384 * tree-vect-data-refs.c: Include internal-fn.h.
6385 (vect_gather_scatter_fn_p): New function.
6386 (vect_describe_gather_scatter_call): Likewise.
6387 (vect_check_gather_scatter): Try using internal functions for
6388 gather loads. Recognize existing calls to a gather load function.
6389 (vect_analyze_data_refs): Consider using gather loads if
6390 supports_vec_gather_load_p.
6391 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
6392 (vect_get_gather_scatter_offset_type): Likewise.
6393 (vect_convert_mask_for_vectype): Likewise.
6394 (vect_add_conversion_to_patterm): Likewise.
6395 (vect_try_gather_scatter_pattern): Likewise.
6396 (vect_recog_gather_scatter_pattern): New pattern recognizer.
6397 (vect_vect_recog_func_ptrs): Add it.
6398 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6399 internal_fn_mask_index and internal_gather_scatter_fn_p.
6400 (check_load_store_masking): Take the gather_scatter_info as an
6401 argument and handle gather loads.
6402 (vect_get_gather_scatter_ops): New function.
6403 (vectorizable_call): Check internal_load_fn_p.
6404 (vectorizable_load): Likewise. Handle gather load internal
6406 (vectorizable_store): Update call to check_load_store_masking.
6407 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
6408 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
6409 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
6410 (aarch64_gather_scale_operand_d): New predicates.
6411 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
6412 (mask_gather_load<mode>): New insns.
6414 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6415 Alan Hayward <alan.hayward@arm.com>
6416 David Sherwood <david.sherwood@arm.com>
6418 * optabs.def (fold_left_plus_optab): New optab.
6419 * doc/md.texi (fold_left_plus_@var{m}): Document.
6420 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
6421 * internal-fn.c (fold_left_direct): Define.
6422 (expand_fold_left_optab_fn): Likewise.
6423 (direct_fold_left_optab_supported_p): Likewise.
6424 * fold-const-call.c (fold_const_fold_left): New function.
6425 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
6426 * tree-parloops.c (valid_reduction_p): New function.
6427 (gather_scalar_reductions): Use it.
6428 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
6429 (vect_finish_replace_stmt): Declare.
6430 * tree-vect-loop.c (fold_left_reduction_fn): New function.
6431 (needs_fold_left_reduction_p): New function, split out from...
6432 (vect_is_simple_reduction): ...here. Accept reductions that
6433 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
6434 (vect_force_simple_reduction): Also store the reduction type in
6435 the assignment's STMT_VINFO_REDUC_TYPE.
6436 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
6437 (merge_with_identity): New function.
6438 (vect_expand_fold_left): Likewise.
6439 (vectorize_fold_left_reduction): Likewise.
6440 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
6441 scalar phi in place for it. Check for target support and reject
6442 cases that would reassociate the operation. Defer the transform
6443 phase to vectorize_fold_left_reduction.
6444 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
6445 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
6446 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
6448 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6450 * tree-if-conv.c (predicate_mem_writes): Remove redundant
6451 call to ifc_temp_var.
6453 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6454 Alan Hayward <alan.hayward@arm.com>
6455 David Sherwood <david.sherwood@arm.com>
6457 * target.def (legitimize_address_displacement): Take the original
6458 offset as a poly_int.
6459 * targhooks.h (default_legitimize_address_displacement): Update
6461 * targhooks.c (default_legitimize_address_displacement): Likewise.
6462 * doc/tm.texi: Regenerate.
6463 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
6464 as an argument, moving assert of ad->disp == ad->disp_term to...
6465 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
6466 Try calling targetm.legitimize_address_displacement before expanding
6467 the address rather than afterwards, and adjust for the new interface.
6468 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
6469 Match the new hook interface. Handle SVE addresses.
6470 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
6473 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6475 * Makefile.in (OBJS): Add early-remat.o.
6476 * target.def (select_early_remat_modes): New hook.
6477 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
6478 * doc/tm.texi: Regenerate.
6479 * targhooks.h (default_select_early_remat_modes): Declare.
6480 * targhooks.c (default_select_early_remat_modes): New function.
6481 * timevar.def (TV_EARLY_REMAT): New timevar.
6482 * passes.def (pass_early_remat): New pass.
6483 * tree-pass.h (make_pass_early_remat): Declare.
6484 * early-remat.c: New file.
6485 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
6487 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
6489 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6490 Alan Hayward <alan.hayward@arm.com>
6491 David Sherwood <david.sherwood@arm.com>
6493 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
6494 vfm1 with a bound_epilog parameter.
6495 (vect_do_peeling): Update calls accordingly, and move the prologue
6496 call earlier in the function. Treat the base bound_epilog as 0 for
6497 fully-masked loops and retain vf - 1 for other loops. Add 1 to
6498 this base when peeling for gaps.
6499 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
6500 with fully-masked loops.
6501 (vect_estimate_min_profitable_iters): Handle the single peeled
6502 iteration in that case.
6504 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6505 Alan Hayward <alan.hayward@arm.com>
6506 David Sherwood <david.sherwood@arm.com>
6508 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
6509 single-element interleaving even if the size is not a power of 2.
6510 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
6511 accesses for single-element interleaving if the group size is
6514 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6515 Alan Hayward <alan.hayward@arm.com>
6516 David Sherwood <david.sherwood@arm.com>
6518 * doc/md.texi (fold_extract_last_@var{m}): Document.
6519 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
6520 * optabs.def (fold_extract_last_optab): New optab.
6521 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
6522 * internal-fn.c (fold_extract_direct): New macro.
6523 (expand_fold_extract_optab_fn): Likewise.
6524 (direct_fold_extract_optab_supported_p): Likewise.
6525 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
6526 * tree-vect-loop.c (vect_model_reduction_cost): Handle
6527 EXTRACT_LAST_REDUCTION.
6528 (get_initial_def_for_reduction): Do not create an initial vector
6529 for EXTRACT_LAST_REDUCTION reductions.
6530 (vectorizable_reduction): Leave the scalar phi in place for
6531 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
6532 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
6533 epilogue code for EXTRACT_LAST_REDUCTION and defer the
6534 transform phase to vectorizable_condition.
6535 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
6537 (vect_finish_stmt_generation): ...here.
6538 (vect_finish_replace_stmt): New function.
6539 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
6540 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
6542 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
6544 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6545 Alan Hayward <alan.hayward@arm.com>
6546 David Sherwood <david.sherwood@arm.com>
6548 * doc/md.texi (extract_last_@var{m}): Document.
6549 * optabs.def (extract_last_optab): New optab.
6550 * internal-fn.def (EXTRACT_LAST): New internal function.
6551 * internal-fn.c (cond_unary_direct): New macro.
6552 (expand_cond_unary_optab_fn): Likewise.
6553 (direct_cond_unary_optab_supported_p): Likewise.
6554 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
6555 loops using EXTRACT_LAST.
6556 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
6557 (extract_last_<mode>): ...this optab.
6558 (vec_extract<mode><Vel>): Update accordingly.
6560 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6561 Alan Hayward <alan.hayward@arm.com>
6562 David Sherwood <david.sherwood@arm.com>
6564 * target.def (empty_mask_is_expensive): New hook.
6565 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
6566 * doc/tm.texi: Regenerate.
6567 * targhooks.h (default_empty_mask_is_expensive): Declare.
6568 * targhooks.c (default_empty_mask_is_expensive): New function.
6569 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
6570 if the target says that empty masks are expensive.
6571 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
6573 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
6575 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6576 Alan Hayward <alan.hayward@arm.com>
6577 David Sherwood <david.sherwood@arm.com>
6579 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
6580 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
6581 (vect_use_loop_mask_for_alignment_p): New function.
6582 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
6583 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
6584 niters_skip argument. Make sure that the first niters_skip elements
6585 of the first iteration are inactive.
6586 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
6587 Update call to vect_set_loop_masks_directly.
6588 (get_misalign_in_elems): New function, split out from...
6589 (vect_gen_prolog_loop_niters): ...here.
6590 (vect_update_init_of_dr): Take a code argument that specifies whether
6591 the adjustment should be added or subtracted.
6592 (vect_update_init_of_drs): Likewise.
6593 (vect_prepare_for_masked_peels): New function.
6594 (vect_do_peeling): Skip prologue peeling if we're using a mask
6595 instead. Update call to vect_update_inits_of_drs.
6596 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6598 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
6599 alignment. Do not include the number of peeled iterations in
6600 the minimum threshold in that case.
6601 (vectorizable_induction): Adjust the start value down by
6602 LOOP_VINFO_MASK_SKIP_NITERS iterations.
6603 (vect_transform_loop): Call vect_prepare_for_masked_peels.
6604 Take the number of skipped iterations into account when calculating
6606 * tree-vect-stmts.c (vect_gen_while_not): New function.
6608 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6609 Alan Hayward <alan.hayward@arm.com>
6610 David Sherwood <david.sherwood@arm.com>
6612 * doc/sourcebuild.texi (vect_fully_masked): Document.
6613 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
6615 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
6617 (vect_analyze_loop_2): ...here. Don't check the vectorization
6618 factor against the number of loop iterations if the loop is
6621 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6622 Alan Hayward <alan.hayward@arm.com>
6623 David Sherwood <david.sherwood@arm.com>
6625 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
6626 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
6627 (dump_groups): Update accordingly.
6628 (iv_use::mem_type): New member variable.
6629 (address_p): New function.
6630 (record_use): Add a mem_type argument and initialize the new
6632 (record_group_use): Add a mem_type argument. Use address_p.
6633 Remove obsolete null checks of base_object. Update call to record_use.
6634 (find_interesting_uses_op): Update call to record_group_use.
6635 (find_interesting_uses_cond): Likewise.
6636 (find_interesting_uses_address): Likewise.
6637 (get_mem_type_for_internal_fn): New function.
6638 (find_address_like_use): Likewise.
6639 (find_interesting_uses_stmt): Try find_address_like_use before
6640 calling find_interesting_uses_op.
6641 (addr_offset_valid_p): Use the iv mem_type field as the type
6642 of the addressed memory.
6643 (add_autoinc_candidates): Likewise.
6644 (get_address_cost): Likewise.
6645 (split_small_address_groups_p): Use address_p.
6646 (split_address_groups): Likewise.
6647 (add_iv_candidate_for_use): Likewise.
6648 (autoinc_possible_for_pair): Likewise.
6649 (rewrite_groups): Likewise.
6650 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
6651 (determine_group_iv_cost): Update after split of USE_ADDRESS.
6652 (get_alias_ptr_type_for_ptr_address): New function.
6653 (rewrite_use_address): Rewrite address uses in calls that were
6654 identified by find_address_like_use.
6656 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6657 Alan Hayward <alan.hayward@arm.com>
6658 David Sherwood <david.sherwood@arm.com>
6660 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
6662 * gimple-expr.h (is_gimple_addressable: Likewise.
6663 * gimple-expr.c (is_gimple_address): Likewise.
6664 * internal-fn.c (expand_call_mem_ref): New function.
6665 (expand_mask_load_optab_fn): Use it.
6666 (expand_mask_store_optab_fn): Likewise.
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 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
6673 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
6674 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
6675 (cond_umax@var{mode}): Document.
6676 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
6677 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
6678 (cond_umin_optab, cond_umax_optab): New optabs.
6679 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
6680 (COND_IOR, COND_XOR): New internal functions.
6681 * internal-fn.h (get_conditional_internal_fn): Declare.
6682 * internal-fn.c (cond_binary_direct): New macro.
6683 (expand_cond_binary_optab_fn): Likewise.
6684 (direct_cond_binary_optab_supported_p): Likewise.
6685 (get_conditional_internal_fn): New function.
6686 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
6687 Cope with reduction statements that are vectorized as calls rather
6689 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
6690 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
6691 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
6692 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
6693 (UNSPEC_COND_EOR): New unspecs.
6694 (optab): Add mappings for them.
6695 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
6696 (sve_int_op, sve_fp_op): New int attributes.
6698 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6699 Alan Hayward <alan.hayward@arm.com>
6700 David Sherwood <david.sherwood@arm.com>
6702 * optabs.def (while_ult_optab): New optab.
6703 * doc/md.texi (while_ult@var{m}@var{n}): Document.
6704 * internal-fn.def (WHILE_ULT): New internal function.
6705 * internal-fn.h (direct_internal_fn_supported_p): New override
6706 that takes two types as argument.
6707 * internal-fn.c (while_direct): New macro.
6708 (expand_while_optab_fn): New function.
6709 (convert_optab_supported_p): Likewise.
6710 (direct_while_optab_supported_p): New macro.
6711 * wide-int.h (wi::udiv_ceil): New function.
6712 * tree-vectorizer.h (rgroup_masks): New structure.
6713 (vec_loop_masks): New typedef.
6714 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
6716 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
6717 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
6718 (vect_max_vf): New function.
6719 (slpeel_make_loop_iterate_ntimes): Delete.
6720 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
6721 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
6722 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
6723 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
6724 internal-fn.h, stor-layout.h and optabs-query.h.
6725 (vect_set_loop_mask): New function.
6726 (add_preheader_seq): Likewise.
6727 (add_header_seq): Likewise.
6728 (interleave_supported_p): Likewise.
6729 (vect_maybe_permute_loop_masks): Likewise.
6730 (vect_set_loop_masks_directly): Likewise.
6731 (vect_set_loop_condition_masked): Likewise.
6732 (vect_set_loop_condition_unmasked): New function, split out from
6733 slpeel_make_loop_iterate_ntimes.
6734 (slpeel_make_loop_iterate_ntimes): Rename to..
6735 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
6736 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
6737 (vect_do_peeling): Update call accordingly.
6738 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
6740 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6741 mask_compare_type, can_fully_mask_p and fully_masked_p.
6742 (release_vec_loop_masks): New function.
6743 (_loop_vec_info): Use it to free the loop masks.
6744 (can_produce_all_loop_masks_p): New function.
6745 (vect_get_max_nscalars_per_iter): Likewise.
6746 (vect_verify_full_masking): Likewise.
6747 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
6748 retries, and free the mask rgroups before retrying. Check loop-wide
6749 reasons for disallowing fully-masked loops. Make the final decision
6750 about whether use a fully-masked loop or not.
6751 (vect_estimate_min_profitable_iters): Do not assume that peeling
6752 for the number of iterations will be needed for fully-masked loops.
6753 (vectorizable_reduction): Disable fully-masked loops.
6754 (vectorizable_live_operation): Likewise.
6755 (vect_halve_mask_nunits): New function.
6756 (vect_double_mask_nunits): Likewise.
6757 (vect_record_loop_mask): Likewise.
6758 (vect_get_loop_mask): Likewise.
6759 (vect_transform_loop): Handle the case in which the final loop
6760 iteration might handle a partial vector. Call vect_set_loop_condition
6761 instead of slpeel_make_loop_iterate_ntimes.
6762 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
6763 (check_load_store_masking): New function.
6764 (prepare_load_store_mask): Likewise.
6765 (vectorizable_store): Handle fully-masked loops.
6766 (vectorizable_load): Likewise.
6767 (supportable_widening_operation): Use vect_halve_mask_nunits for
6769 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
6770 (vect_gen_while): New function.
6771 * config/aarch64/aarch64.md (umax<mode>3): New expander.
6772 (aarch64_uqdec<mode>): New insn.
6774 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6775 Alan Hayward <alan.hayward@arm.com>
6776 David Sherwood <david.sherwood@arm.com>
6778 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
6779 (reduc_xor_scal_optab): New optabs.
6780 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
6781 (reduc_xor_scal_@var{m}): Document.
6782 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
6783 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
6785 * fold-const-call.c (fold_const_call): Handle them.
6786 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
6787 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
6788 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
6789 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
6790 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
6791 (UNSPEC_XORV): New unspecs.
6792 (optab): Add entries for them.
6793 (BITWISEV): New int iterator.
6794 (bit_reduc_op): New int attributes.
6796 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6797 Alan Hayward <alan.hayward@arm.com>
6798 David Sherwood <david.sherwood@arm.com>
6800 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
6801 * internal-fn.def (VEC_SHL_INSERT): New internal function.
6802 * optabs.def (vec_shl_insert_optab): New optab.
6803 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
6804 (duplicate_and_interleave): Likewise.
6805 * tree-vect-loop.c: Include internal-fn.h.
6806 (neutral_op_for_slp_reduction): New function, split out from
6807 get_initial_defs_for_reduction.
6808 (get_initial_def_for_reduction): Handle option 2 for variable-length
6809 vectors by loading the neutral value into a vector and then shifting
6810 the initial value into element 0.
6811 (get_initial_defs_for_reduction): Replace the code argument with
6812 the neutral value calculated by neutral_op_for_slp_reduction.
6813 Use gimple_build_vector for constant-length vectors.
6814 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
6815 but the first group_size elements have a neutral value.
6816 Use duplicate_and_interleave otherwise.
6817 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
6818 Update call to get_initial_defs_for_reduction. Handle SLP
6819 reductions for variable-length vectors by creating one vector
6820 result for each scalar result, with the elements associated
6821 with other scalar results stubbed out with the neutral value.
6822 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
6823 Require IFN_VEC_SHL_INSERT for double reductions on
6824 variable-length vectors, or SLP reductions that have
6825 a neutral value. Require can_duplicate_and_interleave_p
6826 support for variable-length unchained SLP reductions if there
6827 is no neutral value, such as for MIN/MAX reductions. Also require
6828 the number of vector elements to be a multiple of the number of
6829 SLP statements when doing variable-length unchained SLP reductions.
6830 Update call to vect_create_epilog_for_reduction.
6831 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
6832 and remove initial values.
6833 (duplicate_and_interleave): Make public.
6834 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
6835 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
6837 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6838 Alan Hayward <alan.hayward@arm.com>
6839 David Sherwood <david.sherwood@arm.com>
6841 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
6842 (can_duplicate_and_interleave_p): New function.
6843 (vect_get_and_check_slp_defs): Take the vector of statements
6844 rather than just the current one. Remove excess parentheses.
6845 Restriction rejectinon of vect_constant_def and vect_external_def
6846 for variable-length vectors to boolean types, or types for which
6847 can_duplicate_and_interleave_p is false.
6848 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
6849 (duplicate_and_interleave): New function.
6850 (vect_get_constant_vectors): Use gimple_build_vector for
6851 constant-length vectors and suitable variable-length constant
6852 vectors. Use duplicate_and_interleave for other variable-length
6853 vectors. Don't defer the update when inserting new statements.
6855 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6856 Alan Hayward <alan.hayward@arm.com>
6857 David Sherwood <david.sherwood@arm.com>
6859 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
6860 min_profitable_iters doesn't go negative.
6862 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6863 Alan Hayward <alan.hayward@arm.com>
6864 David Sherwood <david.sherwood@arm.com>
6866 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
6867 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
6868 * optabs.def (vec_mask_load_lanes_optab): New optab.
6869 (vec_mask_store_lanes_optab): Likewise.
6870 * internal-fn.def (MASK_LOAD_LANES): New internal function.
6871 (MASK_STORE_LANES): Likewise.
6872 * internal-fn.c (mask_load_lanes_direct): New macro.
6873 (mask_store_lanes_direct): Likewise.
6874 (expand_mask_load_optab_fn): Handle masked operations.
6875 (expand_mask_load_lanes_optab_fn): New macro.
6876 (expand_mask_store_optab_fn): Handle masked operations.
6877 (expand_mask_store_lanes_optab_fn): New macro.
6878 (direct_mask_load_lanes_optab_supported_p): Likewise.
6879 (direct_mask_store_lanes_optab_supported_p): Likewise.
6880 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
6882 (vect_load_lanes_supported): Likewise.
6883 * tree-vect-data-refs.c (strip_conversion): New function.
6884 (can_group_stmts_p): Likewise.
6885 (vect_analyze_data_ref_accesses): Use it instead of checking
6886 for a pair of assignments.
6887 (vect_store_lanes_supported): Take a masked_p parameter.
6888 (vect_load_lanes_supported): Likewise.
6889 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
6890 vect_store_lanes_supported and vect_load_lanes_supported.
6891 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6892 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
6893 parameter. Don't allow gaps for masked accesses.
6894 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
6895 and vect_load_lanes_supported.
6896 (get_load_store_type): Take a masked_p parameter and update
6897 call to get_group_load_store_type.
6898 (vectorizable_store): Update call to get_load_store_type.
6899 Handle IFN_MASK_STORE_LANES.
6900 (vectorizable_load): Update call to get_load_store_type.
6901 Handle IFN_MASK_LOAD_LANES.
6903 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6904 Alan Hayward <alan.hayward@arm.com>
6905 David Sherwood <david.sherwood@arm.com>
6907 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
6909 * config/aarch64/aarch64-protos.h
6910 (aarch64_sve_struct_memory_operand_p): Declare.
6911 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
6912 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
6913 (VPRED, vpred): Handle SVE structure modes.
6914 * config/aarch64/constraints.md (Utx): New constraint.
6915 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
6916 (aarch64_sve_struct_nonimmediate_operand): New predicates.
6917 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
6918 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
6919 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
6920 structure modes. Split into pieces after RA.
6921 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
6922 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
6924 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
6925 SVE structure modes.
6926 (aarch64_classify_address): Likewise.
6927 (sizetochar): Move earlier in file.
6928 (aarch64_print_operand): Handle SVE register lists.
6929 (aarch64_array_mode): New function.
6930 (aarch64_sve_struct_memory_operand_p): Likewise.
6931 (TARGET_ARRAY_MODE): Redefine.
6933 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6934 Alan Hayward <alan.hayward@arm.com>
6935 David Sherwood <david.sherwood@arm.com>
6937 * target.def (array_mode): New target hook.
6938 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
6939 * doc/tm.texi: Regenerate.
6940 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
6941 * hooks.c (hook_optmode_mode_uhwi_none): New function.
6942 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
6944 * stor-layout.c (mode_for_array): Likewise. Support polynomial
6947 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6948 Alan Hayward <alan.hayward@arm.com>
6949 David Sherwood <david.sherwood@arm.com>
6951 * fold-const.c (fold_binary_loc): Check the argument types
6952 rather than the result type when testing for a vector operation.
6954 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6956 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
6957 * doc/tm.texi: Regenerate.
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 * doc/invoke.texi (-msve-vector-bits=): Document new option.
6964 (sve): Document new AArch64 extension.
6965 * doc/md.texi (w): Extend the description of the AArch64
6966 constraint to include SVE vectors.
6967 (Upl, Upa): Document new AArch64 predicate constraints.
6968 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
6970 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
6971 (msve-vector-bits=): New option.
6972 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
6973 SVE when these are disabled.
6974 (sve): New extension.
6975 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
6976 modes. Adjust their number of units based on aarch64_sve_vg.
6977 (MAX_BITSIZE_MODE_ANY_MODE): Define.
6978 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
6979 aarch64_addr_query_type.
6980 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
6981 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
6982 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
6983 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
6984 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
6985 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
6986 (aarch64_simd_imm_zero_p): Delete.
6987 (aarch64_check_zero_based_sve_index_immediate): Declare.
6988 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6989 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6990 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6991 (aarch64_sve_float_mul_immediate_p): Likewise.
6992 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6994 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
6995 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
6996 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
6997 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
6998 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
6999 (aarch64_regmode_natural_size): Likewise.
7000 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
7001 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
7003 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
7004 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
7005 for VG and the SVE predicate registers.
7006 (V_ALIASES): Add a "z"-prefixed alias.
7007 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
7008 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
7009 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
7010 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
7011 (REG_CLASS_NAMES): Add entries for them.
7012 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
7013 and the predicate registers.
7014 (aarch64_sve_vg): Declare.
7015 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
7016 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
7017 (REGMODE_NATURAL_SIZE): Define.
7018 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
7020 * config/aarch64/aarch64.c: Include cfgrtl.h.
7021 (simd_immediate_info): Add a constructor for series vectors,
7022 and an associated step field.
7023 (aarch64_sve_vg): New variable.
7024 (aarch64_dbx_register_number): Handle VG and the predicate registers.
7025 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
7026 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
7027 (VEC_ANY_DATA, VEC_STRUCT): New constants.
7028 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
7029 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
7030 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
7031 (aarch64_get_mask_mode): New functions.
7032 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
7033 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
7034 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
7035 predicate modes and predicate registers. Explicitly restrict
7036 GPRs to modes of 16 bytes or smaller. Only allow FP registers
7037 to store a vector mode if it is recognized by
7038 aarch64_classify_vector_mode.
7039 (aarch64_regmode_natural_size): New function.
7040 (aarch64_hard_regno_caller_save_mode): Return the original mode
7042 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
7043 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
7044 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
7045 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
7047 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
7048 does not overlap dest if the function is frame-related. Handle
7050 (aarch64_split_add_offset): New function.
7051 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
7052 them aarch64_add_offset.
7053 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
7054 and update call to aarch64_sub_sp.
7055 (aarch64_add_cfa_expression): New function.
7056 (aarch64_expand_prologue): Pass extra temporary registers to the
7057 functions above. Handle the case in which we need to emit new
7058 DW_CFA_expressions for registers that were originally saved
7059 relative to the stack pointer, but now have to be expressed
7060 relative to the frame pointer.
7061 (aarch64_output_mi_thunk): Pass extra temporary registers to the
7063 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
7064 IP0 and IP1 values for SVE frames.
7065 (aarch64_expand_vec_series): New function.
7066 (aarch64_expand_sve_widened_duplicate): Likewise.
7067 (aarch64_expand_sve_const_vector): Likewise.
7068 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
7069 Handle SVE constants. Use emit_move_insn to move a force_const_mem
7070 into the register, rather than emitting a SET directly.
7071 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
7072 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
7073 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
7074 (offset_9bit_signed_scaled_p): New functions.
7075 (aarch64_replicate_bitmask_imm): New function.
7076 (aarch64_bitmask_imm): Use it.
7077 (aarch64_cannot_force_const_mem): Reject expressions involving
7078 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
7079 (aarch64_classify_index): Handle SVE indices, by requiring
7080 a plain register index with a scale that matches the element size.
7081 (aarch64_classify_address): Handle SVE addresses. Assert that
7082 the mode of the address is VOIDmode or an integer mode.
7083 Update call to aarch64_classify_symbol.
7084 (aarch64_classify_symbolic_expression): Update call to
7085 aarch64_classify_symbol.
7086 (aarch64_const_vec_all_in_range_p): New function.
7087 (aarch64_print_vector_float_operand): Likewise.
7088 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
7089 "vN" for FP registers with SVE modes. Handle (const ...) vectors
7090 and the FP immediates 1.0 and 0.5.
7091 (aarch64_print_address_internal): Handle SVE addresses.
7092 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
7093 (aarch64_regno_regclass): Handle predicate registers.
7094 (aarch64_secondary_reload): Handle big-endian reloads of SVE
7096 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
7097 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
7098 (aarch64_convert_sve_vector_bits): New function.
7099 (aarch64_override_options): Use it to handle -msve-vector-bits=.
7100 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7102 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
7103 Handle SVE vector and predicate modes. Accept VL-based constants
7104 that need only one temporary register, and VL offsets that require
7105 no temporary registers.
7106 (aarch64_conditional_register_usage): Mark the predicate registers
7107 as fixed if SVE isn't available.
7108 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
7109 Return true for SVE vector and predicate modes.
7110 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
7111 rather than an unsigned int. Handle SVE modes.
7112 (aarch64_preferred_simd_mode): Update call accordingly. Handle
7114 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
7116 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7117 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7118 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7119 (aarch64_sve_float_mul_immediate_p): New functions.
7120 (aarch64_sve_valid_immediate): New function.
7121 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
7122 Explicitly reject structure modes. Check for INDEX constants.
7123 Handle PTRUE and PFALSE constants.
7124 (aarch64_check_zero_based_sve_index_immediate): New function.
7125 (aarch64_simd_imm_zero_p): Delete.
7126 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
7127 vector modes. Accept constants in the range of CNT[BHWD].
7128 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
7129 ask for an Advanced SIMD mode.
7130 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
7131 (aarch64_simd_vector_alignment): Handle SVE predicates.
7132 (aarch64_vectorize_preferred_vector_alignment): New function.
7133 (aarch64_simd_vector_alignment_reachable): Use it instead of
7135 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
7136 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
7138 (MAX_VECT_LEN): Delete.
7139 (expand_vec_perm_d): Add a vec_flags field.
7140 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
7141 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
7142 (aarch64_evpc_ext): Don't apply a big-endian lane correction
7144 (aarch64_evpc_rev): Rename to...
7145 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
7146 (aarch64_evpc_rev_global): New function.
7147 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
7148 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
7150 (aarch64_evpc_sve_tbl): New function.
7151 (aarch64_expand_vec_perm_const_1): Update after rename of
7152 aarch64_evpc_rev. Handle SVE permutes too, trying
7153 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
7154 than aarch64_evpc_tbl.
7155 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
7156 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
7157 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
7158 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
7159 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
7160 (aarch64_expand_sve_vcond): New functions.
7161 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
7162 of aarch64_vector_mode_p.
7163 (aarch64_dwarf_poly_indeterminate_value): New function.
7164 (aarch64_compute_pressure_classes): Likewise.
7165 (aarch64_can_change_mode_class): Likewise.
7166 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
7167 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
7168 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
7169 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
7170 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
7171 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
7172 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
7173 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
7175 (Dn, Dl, Dr): Accept const as well as const_vector.
7176 (Dz): Likewise. Compare against CONST0_RTX.
7177 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
7178 of "vector" where appropriate.
7179 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
7180 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
7181 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
7182 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
7183 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
7184 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
7185 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
7186 (v_int_equiv): Extend to SVE modes.
7187 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
7189 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
7190 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
7191 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
7192 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
7193 (SVE_COND_FP_CMP): New int iterators.
7194 (perm_hilo): Handle the new unpack unspecs.
7195 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
7197 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
7198 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
7199 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
7200 (aarch64_equality_operator, aarch64_constant_vector_operand)
7201 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
7202 (aarch64_sve_nonimmediate_operand): Likewise.
7203 (aarch64_sve_general_operand): Likewise.
7204 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
7205 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
7206 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
7207 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
7208 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
7209 (aarch64_sve_float_arith_immediate): Likewise.
7210 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
7211 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
7212 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
7213 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
7214 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
7215 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
7216 (aarch64_sve_float_arith_operand): Likewise.
7217 (aarch64_sve_float_arith_with_sub_operand): Likewise.
7218 (aarch64_sve_float_mul_operand): Likewise.
7219 (aarch64_sve_vec_perm_operand): Likewise.
7220 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
7221 (aarch64_mov_operand): Accept const_poly_int and const_vector.
7222 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
7223 as well as const_vector.
7224 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
7225 in file. Use CONST0_RTX and CONSTM1_RTX.
7226 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
7227 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
7228 Use aarch64_simd_imm_zero.
7229 * config/aarch64/aarch64-sve.md: New file.
7230 * config/aarch64/aarch64.md: Include it.
7231 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
7232 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
7233 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
7234 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
7235 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
7236 (sve): New attribute.
7237 (enabled): Disable instructions with the sve attribute unless
7239 (movqi, movhi): Pass CONST_POLY_INT operaneds through
7240 aarch64_expand_mov_immediate.
7241 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
7242 CNT[BHSD] immediates.
7243 (movti): Split CONST_POLY_INT moves into two halves.
7244 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
7245 Split additions that need a temporary here if the destination
7246 is the stack pointer.
7247 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
7248 (*add<mode>3_poly_1): New instruction.
7249 (set_clobber_cc): New expander.
7251 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7253 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
7254 parameter and use it instead of GET_MODE_SIZE (innermode). Use
7255 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
7256 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
7257 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
7258 Change innermode from fixed_mode_size to machine_mode.
7259 (simplify_subreg): Update call accordingly. Handle a constant-sized
7260 subreg of a variable-length CONST_VECTOR.
7262 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7263 Alan Hayward <alan.hayward@arm.com>
7264 David Sherwood <david.sherwood@arm.com>
7266 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
7267 (add_offset_to_base): New function, split out from...
7268 (create_mem_ref): ...here. When handling a scale other than 1,
7269 check first whether the address is valid without the offset.
7270 Add it into the base if so, leaving the index and scale as-is.
7272 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7275 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
7276 fold_for_warn before checking if arg2 is INTEGER_CST.
7278 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
7280 * config/rs6000/predicates.md (load_multiple_operation): Delete.
7281 (store_multiple_operation): Delete.
7282 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
7283 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
7284 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
7285 guarded by TARGET_STRING.
7286 (rs6000_output_load_multiple): Delete.
7287 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
7288 OPTION_MASK_STRING / TARGET_STRING handling.
7289 (print_operand) <'N', 'O'>: Add comment that these are unused now.
7290 (const rs6000_opt_masks) <"string">: Change mask to 0.
7291 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
7292 (MASK_STRING): Delete.
7293 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
7295 (load_multiple): Delete.
7302 (store_multiple): Delete.
7309 (movmemsi_8reg): Delete.
7310 (corresponding unnamed define_insn): Delete.
7311 (movmemsi_6reg): Delete.
7312 (corresponding unnamed define_insn): Delete.
7313 (movmemsi_4reg): Delete.
7314 (corresponding unnamed define_insn): Delete.
7315 (movmemsi_2reg): Delete.
7316 (corresponding unnamed define_insn): Delete.
7317 (movmemsi_1reg): Delete.
7318 (corresponding unnamed define_insn): Delete.
7319 * config/rs6000/rs6000.opt (mno-string): New.
7320 (mstring): Replace by deprecation warning stub.
7321 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
7323 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7325 * regrename.c (regrename_do_replace): If replacing the same
7326 reg multiple times, try to reuse last created gen_raw_REG.
7329 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
7330 main to workaround a bug in GDB.
7332 2018-01-12 Tom de Vries <tom@codesourcery.com>
7335 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
7337 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
7339 PR rtl-optimization/80481
7340 * ira-color.c (get_cap_member): New function.
7341 (allocnos_conflict_by_live_ranges_p): Use it.
7342 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
7343 (setup_slot_coalesced_allocno_live_ranges): Ditto.
7345 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
7348 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
7349 (*saddl_se_1): Ditto.
7351 (*ssubl_se_1): Ditto.
7353 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7355 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
7356 rather than wi::to_widest for DR_INITs.
7357 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
7358 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
7359 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
7361 (vect_analyze_group_access_1): Note that here.
7363 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7365 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
7366 polynomial type sizes.
7368 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7370 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
7371 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
7372 (gimple_add_tmp_var): Likewise.
7374 2018-01-12 Martin Liska <mliska@suse.cz>
7376 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
7377 (gimple_alloc_sizes): Likewise.
7378 (dump_gimple_statistics): Use PRIu64 in printf format.
7379 * gimple.h: Change uint64_t to int.
7381 2018-01-12 Martin Liska <mliska@suse.cz>
7383 * tree-core.h: Use uint64_t instead of int.
7384 * tree.c (tree_node_counts): Likewise.
7385 (tree_node_sizes): Likewise.
7386 (dump_tree_statistics): Use PRIu64 in printf format.
7388 2018-01-12 Martin Liska <mliska@suse.cz>
7390 * Makefile.in: As qsort_chk is implemented in vec.c, add
7391 vec.o to linkage of gencfn-macros.
7392 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
7393 passing the info to record_node_allocation_statistics.
7394 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
7396 * ggc-common.c (struct ggc_usage): Add operator== and use
7397 it in operator< and compare function.
7398 * mem-stats.h (struct mem_usage): Likewise.
7399 * vec.c (struct vec_usage): Remove operator< and compare
7400 function. Can be simply inherited.
7402 2018-01-12 Martin Jambor <mjambor@suse.cz>
7405 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
7406 * tree-ssa-math-opts.c: Include domwalk.h.
7407 (convert_mult_to_fma_1): New function.
7408 (fma_transformation_info): New type.
7409 (fma_deferring_state): Likewise.
7410 (cancel_fma_deferring): New function.
7411 (result_of_phi): Likewise.
7412 (last_fma_candidate_feeds_initial_phi): Likewise.
7413 (convert_mult_to_fma): Added deferring logic, split actual
7414 transformation to convert_mult_to_fma_1.
7415 (math_opts_dom_walker): New type.
7416 (math_opts_dom_walker::after_dom_children): New method, body moved
7417 here from pass_optimize_widening_mul::execute, added deferring logic
7419 (pass_optimize_widening_mul::execute): Moved most of code to
7420 math_opts_dom_walker::after_dom_children.
7421 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
7422 * config/i386/i386.c (ix86_option_override_internal): Added
7423 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
7425 2018-01-12 Richard Biener <rguenther@suse.de>
7428 * dwarf2out.c (gen_variable_die): Do not reset old_die for
7429 inline instance vars.
7431 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
7434 * config/rx/rx.c (rx_is_restricted_memory_address):
7437 2018-01-12 Richard Biener <rguenther@suse.de>
7439 PR tree-optimization/80846
7440 * target.def (split_reduction): New target hook.
7441 * targhooks.c (default_split_reduction): New function.
7442 * targhooks.h (default_split_reduction): Declare.
7443 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
7444 target requests first reduce vectors by combining low and high
7446 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
7447 (get_vectype_for_scalar_type_and_size): Export.
7448 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
7449 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
7450 * doc/tm.texi: Regenerate.
7451 * config/i386/i386.c (ix86_split_reduction): Implement
7452 TARGET_VECTORIZE_SPLIT_REDUCTION.
7454 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7457 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
7458 in PIC mode except for TARGET_VXWORKS_RTP.
7459 * config/sparc/sparc.c: Include cfgrtl.h.
7460 (TARGET_INIT_PIC_REG): Define.
7461 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
7462 (sparc_pic_register_p): New predicate.
7463 (sparc_legitimate_address_p): Use it.
7464 (sparc_legitimize_pic_address): Likewise.
7465 (sparc_delegitimize_address): Likewise.
7466 (sparc_mode_dependent_address_p): Likewise.
7467 (gen_load_pcrel_sym): Remove 4th parameter.
7468 (load_got_register): Adjust call to above. Remove obsolete stuff.
7469 (sparc_expand_prologue): Do not call load_got_register here.
7470 (sparc_flat_expand_prologue): Likewise.
7471 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
7472 (sparc_use_pseudo_pic_reg): New function.
7473 (sparc_init_pic_reg): Likewise.
7474 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7475 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
7477 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
7479 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
7480 Add item for branch_cost.
7482 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7484 PR rtl-optimization/83565
7485 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
7486 not extend the result to a larger mode for rotate operations.
7487 (num_sign_bit_copies1): Likewise.
7489 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7492 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
7494 Use values-Xc.o for -pedantic.
7495 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
7497 2018-01-12 Martin Liska <mliska@suse.cz>
7500 * ipa-devirt.c (final_warning_record::grow_type_warnings):
7502 (possible_polymorphic_call_targets): Use it.
7503 (ipa_devirt): Likewise.
7505 2018-01-12 Martin Liska <mliska@suse.cz>
7507 * profile-count.h (enum profile_quality): Use 0 as invalid
7508 enum value of profile_quality.
7510 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
7512 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
7513 -mext-string options.
7515 2018-01-12 Richard Biener <rguenther@suse.de>
7517 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
7518 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
7519 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
7521 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
7523 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
7525 * configure.ac (--with-long-double-format): Add support for the
7526 configuration option to change the default long double format on
7528 * config.gcc (powerpc*-linux*-*): Likewise.
7529 * configure: Regenerate.
7530 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
7531 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
7532 used without modification.
7534 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7536 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
7537 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
7538 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
7539 MISC_BUILTIN_SPEC_BARRIER.
7540 (rs6000_init_builtins): Likewise.
7541 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
7543 (speculation_barrier): New define_insn.
7544 * doc/extend.texi: Document __builtin_speculation_barrier.
7546 2018-01-11 Jakub Jelinek <jakub@redhat.com>
7549 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
7550 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
7551 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
7553 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
7554 integral modes instead of "ss" and "sd".
7555 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
7556 vectors with 32-bit and 64-bit elements.
7557 (vecdupssescalarmodesuffix): New mode attribute.
7558 (vec_dup<mode>): Use it.
7560 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
7563 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
7564 frame if argument is passed on stack.
7566 2018-01-11 Jakub Jelinek <jakub@redhat.com>
7569 * ree.c (combine_reaching_defs): Optimize also
7570 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
7571 reg2=any_extend(exp); reg1=reg2;, formatting fix.
7573 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
7576 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
7578 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
7581 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
7582 after they are computed.
7584 2018-01-11 Bin Cheng <bin.cheng@arm.com>
7586 PR tree-optimization/83695
7587 * gimple-loop-linterchange.cc
7588 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
7589 reset cached scev information after interchange.
7590 (pass_linterchange::execute): Remove call to scev_reset_htab.
7592 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7594 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
7595 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
7596 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
7597 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
7598 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
7599 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
7600 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
7601 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
7602 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
7603 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
7604 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
7605 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
7606 (V_lane_reg): Likewise.
7607 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
7609 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
7610 (vfmal_lane_low<mode>_intrinsic,
7611 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
7612 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
7613 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
7614 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
7615 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
7616 vfmsl_lane_high<mode>_intrinsic): New define_insns.
7618 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7620 * config/arm/arm-cpus.in (fp16fml): New feature.
7621 (ALL_SIMD): Add fp16fml.
7622 (armv8.2-a): Add fp16fml as an option.
7623 (armv8.3-a): Likewise.
7624 (armv8.4-a): Add fp16fml as part of fp16.
7625 * config/arm/arm.h (TARGET_FP16FML): Define.
7626 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
7628 * config/arm/arm-modes.def (V2HF): Define.
7629 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
7630 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
7631 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
7632 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
7633 vfmsl_low, vfmsl_high): New set of builtins.
7634 * config/arm/iterators.md (PLUSMINUS): New code iterator.
7635 (vfml_op): New code attribute.
7636 (VFMLHALVES): New int iterator.
7637 (VFML, VFMLSEL): New mode attributes.
7638 (V_reg): Define mapping for V2HF.
7639 (V_hi, V_lo): New mode attributes.
7640 (VF_constraint): Likewise.
7641 (vfml_half, vfml_half_selector): New int attributes.
7642 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
7644 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
7645 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
7647 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
7648 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
7649 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
7650 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
7652 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
7653 Document new effective target and option set.
7655 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7657 * config/arm/arm-cpus.in (armv8_4): New feature.
7658 (ARMv8_4a): New fgroup.
7659 (armv8.4-a): New arch.
7660 * config/arm/arm-tables.opt: Regenerate.
7661 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
7662 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
7663 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
7664 Add matching rules for -march=armv8.4-a and extensions.
7665 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
7667 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
7670 * config/rx/rx.md (BW): New mode attribute.
7671 (sync_lock_test_and_setsi): Add mode suffix to insn output.
7673 2018-01-11 Richard Biener <rguenther@suse.de>
7675 PR tree-optimization/83435
7676 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
7677 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
7678 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
7680 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7681 Alan Hayward <alan.hayward@arm.com>
7682 David Sherwood <david.sherwood@arm.com>
7684 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
7686 (aarch64_classify_address): Initialize it. Track polynomial offsets.
7687 (aarch64_print_address_internal): Use it to check for a zero offset.
7689 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7690 Alan Hayward <alan.hayward@arm.com>
7691 David Sherwood <david.sherwood@arm.com>
7693 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
7694 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
7695 Return a poly_int64 rather than a HOST_WIDE_INT.
7696 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
7697 rather than a HOST_WIDE_INT.
7698 * config/aarch64/aarch64.h (aarch64_frame): Protect with
7699 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
7700 hard_fp_offset, frame_size, initial_adjust, callee_offset and
7701 final_offset from HOST_WIDE_INT to poly_int64.
7702 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
7703 to_constant when getting the number of units in an Advanced SIMD
7705 (aarch64_builtin_vectorized_function): Check for a constant number
7707 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
7709 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
7710 attribute instead of GET_MODE_NUNITS.
7711 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
7712 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
7713 GET_MODE_SIZE for fixed-size registers.
7714 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
7715 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
7716 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
7717 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
7718 (aarch64_print_operand, aarch64_print_address_internal)
7719 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
7720 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
7721 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
7722 Handle polynomial GET_MODE_SIZE.
7723 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
7724 wider than SImode without modification.
7725 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
7726 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
7727 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
7728 passing and returning SVE modes.
7729 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
7730 rather than GEN_INT.
7731 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
7732 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
7733 (aarch64_allocate_and_probe_stack_space): Likewise.
7734 (aarch64_layout_frame): Cope with polynomial offsets.
7735 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
7736 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
7738 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
7739 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
7740 poly_int64 rather than a HOST_WIDE_INT.
7741 (aarch64_get_separate_components, aarch64_process_components)
7742 (aarch64_expand_prologue, aarch64_expand_epilogue)
7743 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
7744 (aarch64_anchor_offset): New function, split out from...
7745 (aarch64_legitimize_address): ...here.
7746 (aarch64_builtin_vectorization_cost): Handle polynomial
7747 TYPE_VECTOR_SUBPARTS.
7748 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
7750 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
7751 number of elements from the PARALLEL rather than the mode.
7752 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
7753 rather than GET_MODE_BITSIZE.
7754 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
7755 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
7756 (aarch64_expand_vec_perm_const_1): Handle polynomial
7757 d->perm.length () and d->perm elements.
7758 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
7759 Apply to_constant to d->perm elements.
7760 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
7761 polynomial CONST_VECTOR_NUNITS.
7762 (aarch64_move_pointer): Take amount as a poly_int64 rather
7764 (aarch64_progress_pointer): Avoid temporary variable.
7765 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
7766 the mode attribute instead of GET_MODE.
7768 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7769 Alan Hayward <alan.hayward@arm.com>
7770 David Sherwood <david.sherwood@arm.com>
7772 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
7773 x exists before using it.
7774 (aarch64_add_constant_internal): Rename to...
7775 (aarch64_add_offset_1): ...this. Replace regnum with separate
7776 src and dest rtxes. Handle the case in which they're different,
7777 including when the offset is zero. Replace scratchreg with an rtx.
7778 Use 2 additions if there is no spare register into which we can
7779 move a 16-bit constant.
7780 (aarch64_add_constant): Delete.
7781 (aarch64_add_offset): Replace reg with separate src and dest
7782 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
7783 Use aarch64_add_offset_1.
7784 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
7785 an rtx rather than an int. Take the delta as a poly_int64
7786 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
7787 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
7788 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
7789 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
7790 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
7792 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
7793 aarch64_add_constant.
7795 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7797 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
7798 Use scalar_float_mode.
7800 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7802 * config/aarch64/aarch64-simd.md
7803 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
7804 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
7805 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
7806 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
7807 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
7808 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
7809 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
7810 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
7811 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
7812 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
7814 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7817 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
7818 targ_options->x_arm_arch_string is non NULL.
7820 2018-01-11 Tamar Christina <tamar.christina@arm.com>
7822 * config/aarch64/aarch64.h
7823 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
7825 2018-01-11 Sudakshina Das <sudi.das@arm.com>
7828 * expmed.c (emit_store_flag_force): Swap if const op0
7829 and change VOIDmode to mode of op0.
7831 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7833 PR rtl-optimization/83761
7834 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
7835 than bytes to mode_for_size.
7837 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7840 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
7841 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
7844 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7847 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
7848 when in layout mode.
7849 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
7850 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
7853 2018-01-10 Michael Collison <michael.collison@arm.com>
7855 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
7856 * config/aarch64/aarch64-option-extension.def: Add
7857 AARCH64_OPT_EXTENSION of 'fp16fml'.
7858 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7859 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
7860 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
7861 * config/aarch64/constraints.md (Ui7): New constraint.
7862 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
7863 (VFMLA_SEL_W): Ditto.
7866 (VFMLA16_LOW): New int iterator.
7867 (VFMLA16_HIGH): Ditto.
7868 (UNSPEC_FMLAL): New unspec.
7869 (UNSPEC_FMLSL): Ditto.
7870 (UNSPEC_FMLAL2): Ditto.
7871 (UNSPEC_FMLSL2): Ditto.
7872 (f16mac): New code attribute.
7873 * config/aarch64/aarch64-simd-builtins.def
7874 (aarch64_fmlal_lowv2sf): Ditto.
7875 (aarch64_fmlsl_lowv2sf): Ditto.
7876 (aarch64_fmlalq_lowv4sf): Ditto.
7877 (aarch64_fmlslq_lowv4sf): Ditto.
7878 (aarch64_fmlal_highv2sf): Ditto.
7879 (aarch64_fmlsl_highv2sf): Ditto.
7880 (aarch64_fmlalq_highv4sf): Ditto.
7881 (aarch64_fmlslq_highv4sf): Ditto.
7882 (aarch64_fmlal_lane_lowv2sf): Ditto.
7883 (aarch64_fmlsl_lane_lowv2sf): Ditto.
7884 (aarch64_fmlal_laneq_lowv2sf): Ditto.
7885 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
7886 (aarch64_fmlalq_lane_lowv4sf): Ditto.
7887 (aarch64_fmlsl_lane_lowv4sf): Ditto.
7888 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
7889 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
7890 (aarch64_fmlal_lane_highv2sf): Ditto.
7891 (aarch64_fmlsl_lane_highv2sf): Ditto.
7892 (aarch64_fmlal_laneq_highv2sf): Ditto.
7893 (aarch64_fmlsl_laneq_highv2sf): Ditto.
7894 (aarch64_fmlalq_lane_highv4sf): Ditto.
7895 (aarch64_fmlsl_lane_highv4sf): Ditto.
7896 (aarch64_fmlalq_laneq_highv4sf): Ditto.
7897 (aarch64_fmlsl_laneq_highv4sf): Ditto.
7898 * config/aarch64/aarch64-simd.md:
7899 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
7900 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7901 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
7902 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7903 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
7904 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
7905 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
7906 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
7907 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
7908 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
7909 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
7910 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
7911 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
7912 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
7913 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
7914 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
7915 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
7916 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
7917 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
7918 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
7919 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
7920 (vfmlsl_low_u32): Ditto.
7921 (vfmlalq_low_u32): Ditto.
7922 (vfmlslq_low_u32): Ditto.
7923 (vfmlal_high_u32): Ditto.
7924 (vfmlsl_high_u32): Ditto.
7925 (vfmlalq_high_u32): Ditto.
7926 (vfmlslq_high_u32): Ditto.
7927 (vfmlal_lane_low_u32): Ditto.
7928 (vfmlsl_lane_low_u32): Ditto.
7929 (vfmlal_laneq_low_u32): Ditto.
7930 (vfmlsl_laneq_low_u32): Ditto.
7931 (vfmlalq_lane_low_u32): Ditto.
7932 (vfmlslq_lane_low_u32): Ditto.
7933 (vfmlalq_laneq_low_u32): Ditto.
7934 (vfmlslq_laneq_low_u32): Ditto.
7935 (vfmlal_lane_high_u32): Ditto.
7936 (vfmlsl_lane_high_u32): Ditto.
7937 (vfmlal_laneq_high_u32): Ditto.
7938 (vfmlsl_laneq_high_u32): Ditto.
7939 (vfmlalq_lane_high_u32): Ditto.
7940 (vfmlslq_lane_high_u32): Ditto.
7941 (vfmlalq_laneq_high_u32): Ditto.
7942 (vfmlslq_laneq_high_u32): Ditto.
7943 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
7944 (AARCH64_FL_FOR_ARCH8_4): New.
7945 (AARCH64_ISA_F16FML): New ISA flag.
7946 (TARGET_F16FML): New feature flag for fp16fml.
7947 (doc/invoke.texi): Document new fp16fml option.
7949 2018-01-10 Michael Collison <michael.collison@arm.com>
7951 * config/aarch64/aarch64-builtins.c:
7952 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
7953 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7954 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
7955 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
7956 (AARCH64_ISA_SHA3): New ISA flag.
7957 (TARGET_SHA3): New feature flag for sha3.
7958 * config/aarch64/iterators.md (sha512_op): New int attribute.
7959 (CRYPTO_SHA512): New int iterator.
7960 (UNSPEC_SHA512H): New unspec.
7961 (UNSPEC_SHA512H2): Ditto.
7962 (UNSPEC_SHA512SU0): Ditto.
7963 (UNSPEC_SHA512SU1): Ditto.
7964 * config/aarch64/aarch64-simd-builtins.def
7965 (aarch64_crypto_sha512hqv2di): New builtin.
7966 (aarch64_crypto_sha512h2qv2di): Ditto.
7967 (aarch64_crypto_sha512su0qv2di): Ditto.
7968 (aarch64_crypto_sha512su1qv2di): Ditto.
7969 (aarch64_eor3qv8hi): Ditto.
7970 (aarch64_rax1qv2di): Ditto.
7971 (aarch64_xarqv2di): Ditto.
7972 (aarch64_bcaxqv8hi): Ditto.
7973 * config/aarch64/aarch64-simd.md:
7974 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
7975 (aarch64_crypto_sha512su0qv2di): Ditto.
7976 (aarch64_crypto_sha512su1qv2di): Ditto.
7977 (aarch64_eor3qv8hi): Ditto.
7978 (aarch64_rax1qv2di): Ditto.
7979 (aarch64_xarqv2di): Ditto.
7980 (aarch64_bcaxqv8hi): Ditto.
7981 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
7982 (vsha512h2q_u64): Ditto.
7983 (vsha512su0q_u64): Ditto.
7984 (vsha512su1q_u64): Ditto.
7985 (veor3q_u16): Ditto.
7986 (vrax1q_u64): Ditto.
7988 (vbcaxq_u16): Ditto.
7989 * config/arm/types.md (crypto_sha512): New type attribute.
7990 (crypto_sha3): Ditto.
7991 (doc/invoke.texi): Document new sha3 option.
7993 2018-01-10 Michael Collison <michael.collison@arm.com>
7995 * config/aarch64/aarch64-builtins.c:
7996 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
7997 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7998 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
7999 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
8000 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
8001 (AARCH64_ISA_SM4): New ISA flag.
8002 (TARGET_SM4): New feature flag for sm4.
8003 * config/aarch64/aarch64-simd-builtins.def
8004 (aarch64_sm3ss1qv4si): Ditto.
8005 (aarch64_sm3tt1aq4si): Ditto.
8006 (aarch64_sm3tt1bq4si): Ditto.
8007 (aarch64_sm3tt2aq4si): Ditto.
8008 (aarch64_sm3tt2bq4si): Ditto.
8009 (aarch64_sm3partw1qv4si): Ditto.
8010 (aarch64_sm3partw2qv4si): Ditto.
8011 (aarch64_sm4eqv4si): Ditto.
8012 (aarch64_sm4ekeyqv4si): Ditto.
8013 * config/aarch64/aarch64-simd.md:
8014 (aarch64_sm3ss1qv4si): Ditto.
8015 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
8016 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
8017 (aarch64_sm4eqv4si): Ditto.
8018 (aarch64_sm4ekeyqv4si): Ditto.
8019 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
8020 (sm3part_op): Ditto.
8021 (CRYPTO_SM3TT): Ditto.
8022 (CRYPTO_SM3PART): Ditto.
8023 (UNSPEC_SM3SS1): New unspec.
8024 (UNSPEC_SM3TT1A): Ditto.
8025 (UNSPEC_SM3TT1B): Ditto.
8026 (UNSPEC_SM3TT2A): Ditto.
8027 (UNSPEC_SM3TT2B): Ditto.
8028 (UNSPEC_SM3PARTW1): Ditto.
8029 (UNSPEC_SM3PARTW2): Ditto.
8030 (UNSPEC_SM4E): Ditto.
8031 (UNSPEC_SM4EKEY): Ditto.
8032 * config/aarch64/constraints.md (Ui2): New constraint.
8033 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
8034 * config/arm/types.md (crypto_sm3): New type attribute.
8035 (crypto_sm4): Ditto.
8036 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
8037 (vsm3tt1aq_u32): Ditto.
8038 (vsm3tt1bq_u32): Ditto.
8039 (vsm3tt2aq_u32): Ditto.
8040 (vsm3tt2bq_u32): Ditto.
8041 (vsm3partw1q_u32): Ditto.
8042 (vsm3partw2q_u32): Ditto.
8043 (vsm4eq_u32): Ditto.
8044 (vsm4ekeyq_u32): Ditto.
8045 (doc/invoke.texi): Document new sm4 option.
8047 2018-01-10 Michael Collison <michael.collison@arm.com>
8049 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
8050 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
8051 (AARCH64_FL_FOR_ARCH8_4): New.
8052 (AARCH64_FL_V8_4): New flag.
8053 (doc/invoke.texi): Document new armv8.4-a option.
8055 2018-01-10 Michael Collison <michael.collison@arm.com>
8057 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8058 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
8059 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
8060 * config/aarch64/aarch64-option-extension.def: Add
8061 AARCH64_OPT_EXTENSION of 'sha2'.
8062 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
8063 (crypto): Disable sha2 and aes if crypto disabled.
8064 (crypto): Enable aes and sha2 if enabled.
8065 (simd): Disable sha2 and aes if simd disabled.
8066 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
8068 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
8069 (TARGET_SHA2): New feature flag for sha2.
8070 (TARGET_AES): New feature flag for aes.
8071 * config/aarch64/aarch64-simd.md:
8072 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
8073 conditional on TARGET_AES.
8074 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
8075 (aarch64_crypto_sha1hsi): Make pattern conditional
8077 (aarch64_crypto_sha1hv4si): Ditto.
8078 (aarch64_be_crypto_sha1hv4si): Ditto.
8079 (aarch64_crypto_sha1su1v4si): Ditto.
8080 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
8081 (aarch64_crypto_sha1su0v4si): Ditto.
8082 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
8083 (aarch64_crypto_sha256su0v4si): Ditto.
8084 (aarch64_crypto_sha256su1v4si): Ditto.
8085 (doc/invoke.texi): Document new aes and sha2 options.
8087 2018-01-10 Martin Sebor <msebor@redhat.com>
8089 PR tree-optimization/83781
8090 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
8093 2018-01-11 Martin Sebor <msebor@gmail.com>
8094 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8096 PR tree-optimization/83501
8097 PR tree-optimization/81703
8099 * tree-ssa-strlen.c (get_string_cst): Rename...
8100 (get_string_len): ...to this. Handle global constants.
8101 (handle_char_store): Adjust.
8103 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
8104 Jim Wilson <jimw@sifive.com>
8106 * config/riscv/riscv-protos.h (riscv_output_return): New.
8107 * config/riscv/riscv.c (struct machine_function): New naked_p field.
8108 (riscv_attribute_table, riscv_output_return),
8109 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
8110 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
8111 (riscv_compute_frame_info): Only compute frame->mask if not a naked
8113 (riscv_expand_prologue): Add early return for naked function.
8114 (riscv_expand_epilogue): Likewise.
8115 (riscv_function_ok_for_sibcall): Return false for naked function.
8116 (riscv_set_current_function): New.
8117 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
8118 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
8119 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
8120 * doc/extend.texi (RISC-V Function Attributes): New.
8122 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
8124 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
8125 check for 128-bit long double before checking TCmode.
8126 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
8127 128-bit long doubles before checking TFmode or TCmode.
8128 (FLOAT128_IBM_P): Likewise.
8130 2018-01-10 Martin Sebor <msebor@redhat.com>
8132 PR tree-optimization/83671
8133 * builtins.c (c_strlen): Unconditionally return zero for the empty
8135 Use -Warray-bounds for warnings.
8136 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
8137 for non-constant array indices with COMPONENT_REF, arrays of
8138 arrays, and pointers to arrays.
8139 (gimple_fold_builtin_strlen): Determine and set length range for
8140 non-constant character arrays.
8142 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
8145 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
8148 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
8150 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
8152 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8155 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
8156 VECTOR_MEM_ALTIVEC_OR_VSX_P.
8157 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
8158 indexed_or_indirect_operand predicate.
8159 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
8160 (*vsx_le_perm_load_v8hi): Likewise.
8161 (*vsx_le_perm_load_v16qi): Likewise.
8162 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
8163 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
8164 (*vsx_le_perm_store_v8hi): Likewise.
8165 (*vsx_le_perm_store_v16qi): Likewise.
8166 (eight unnamed splitters): Likewise.
8168 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8170 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
8171 * config/rs6000/emmintrin.h: Likewise.
8172 * config/rs6000/mmintrin.h: Likewise.
8173 * config/rs6000/xmmintrin.h: Likewise.
8175 2018-01-10 David Malcolm <dmalcolm@redhat.com>
8178 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
8180 * tree.c (tree_nop_conversion): Return true for location wrapper
8182 (maybe_wrap_with_location): New function.
8183 (selftest::check_strip_nops): New function.
8184 (selftest::test_location_wrappers): New function.
8185 (selftest::tree_c_tests): Call it.
8186 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
8187 (maybe_wrap_with_location): New decl.
8188 (EXPR_LOCATION_WRAPPER_P): New macro.
8189 (location_wrapper_p): New inline function.
8190 (tree_strip_any_location_wrapper): New inline function.
8192 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
8195 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
8196 stack_realign_offset for the largest alignment of stack slot
8198 (ix86_find_max_used_stack_alignment): New function.
8199 (ix86_finalize_stack_frame_flags): Use it. Set
8200 max_used_stack_alignment if we don't realign stack.
8201 * config/i386/i386.h (machine_function): Add
8202 max_used_stack_alignment.
8204 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
8206 * config/arm/arm.opt (-mbranch-cost): New option.
8207 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
8210 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
8213 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
8214 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
8216 2018-01-10 Richard Biener <rguenther@suse.de>
8219 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
8220 early out so it also covers the case where we have a non-NULL
8223 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8225 PR tree-optimization/83753
8226 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
8227 for non-strided grouped accesses if the number of elements is 1.
8229 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8232 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
8233 * i386.h (TARGET_USE_GATHER): Define.
8234 * x86-tune.def (X86_TUNE_USE_GATHER): New.
8236 2018-01-10 Martin Liska <mliska@suse.cz>
8239 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
8240 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
8242 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
8243 CLEANUP_NO_PARTITIONING is not set.
8245 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8247 * doc/rtl.texi: Remove documentation of (const ...) wrappers
8248 for vectors, as a partial revert of r254296.
8249 * rtl.h (const_vec_p): Delete.
8250 (const_vec_duplicate_p): Don't test for vector CONSTs.
8251 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
8252 * expmed.c (make_tree): Likewise.
8255 * common.md (E, F): Use CONSTANT_P instead of checking for
8257 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
8258 checking for CONST_VECTOR.
8260 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8263 * predict.c (force_edge_cold): Handle in more sane way edges
8266 2018-01-09 Carl Love <cel@us.ibm.com>
8268 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
8270 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
8271 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
8272 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
8273 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
8274 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
8275 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
8276 * config/rs6000/rs6000-protos.h: Add extern defition for
8277 rs6000_generate_float2_double_code.
8278 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
8280 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
8281 (float2_v2df): Add define_expand.
8283 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
8286 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
8287 op_mode in the force_to_mode call.
8289 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8291 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
8292 instead of checking each element individually.
8293 (aarch64_evpc_uzp): Likewise.
8294 (aarch64_evpc_zip): Likewise.
8295 (aarch64_evpc_ext): Likewise.
8296 (aarch64_evpc_rev): Likewise.
8297 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
8298 instead of checking each element individually. Return true without
8300 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
8301 whether all selected elements come from the same input, instead of
8302 checking each element individually. Remove calls to gen_rtx_REG,
8303 start_sequence and end_sequence and instead assert that no rtl is
8306 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8308 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
8309 order of HIGH and CONST checks.
8311 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8313 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
8314 if the destination isn't an SSA_NAME.
8316 2018-01-09 Richard Biener <rguenther@suse.de>
8318 PR tree-optimization/83668
8319 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
8321 (canonicalize_loop_form): ... here, renamed from ...
8322 (canonicalize_loop_closed_ssa_form): ... this and amended to
8323 swap successor edges for loop exit blocks to make us use
8324 the RPO order we need for initial schedule generation.
8326 2018-01-09 Joseph Myers <joseph@codesourcery.com>
8328 PR tree-optimization/64811
8329 * match.pd: When optimizing comparisons with Inf, avoid
8330 introducing or losing exceptions from comparisons with NaN.
8332 2018-01-09 Martin Liska <mliska@suse.cz>
8335 * asan.c (shadow_mem_size): Add gcc_assert.
8337 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
8339 Don't save registers in main().
8342 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
8343 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
8344 * config/avr/avr.c (avr_set_current_function): Don't error if
8345 naked, OS_task or OS_main are specified at the same time.
8346 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
8348 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
8350 * common/config/avr/avr-common.c (avr_option_optimization_table):
8351 Switch on -mmain-is-OS_task for optimizing compilations.
8353 2018-01-09 Richard Biener <rguenther@suse.de>
8355 PR tree-optimization/83572
8356 * graphite.c: Include cfganal.h.
8357 (graphite_transform_loops): Connect infinite loops to exit
8358 and remove fake edges at the end.
8360 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8362 * ipa-inline.c (edge_badness): Revert accidental checkin.
8364 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8367 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
8368 symbols; not inline clones.
8370 2018-01-09 Jakub Jelinek <jakub@redhat.com>
8373 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
8374 hard registers. Formatting fixes.
8376 PR preprocessor/83722
8377 * gcc.c (try_generate_repro): Pass
8378 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
8379 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
8382 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
8383 Kito Cheng <kito.cheng@gmail.com>
8385 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
8386 (riscv_leaf_function_p): Delete.
8387 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
8389 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8391 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
8393 (do_ifelse): New function.
8394 (do_isel): New function.
8395 (do_sub3): New function.
8396 (do_add3): New function.
8397 (do_load_mask_compare): New function.
8398 (do_overlap_load_compare): New function.
8399 (expand_compare_loop): New function.
8400 (expand_block_compare): Call expand_compare_loop() when appropriate.
8401 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
8403 (-mblock-compare-inline-loop-limit): New option.
8405 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8408 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
8409 Reverse order of second and third operands in first alternative.
8410 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
8411 of first and second elements in UNSPEC_VPERMR vector.
8412 (altivec_expand_vec_perm_le): Likewise.
8414 2018-01-08 Jeff Law <law@redhat.com>
8416 PR rtl-optimizatin/81308
8417 * tree-switch-conversion.c (cfg_altered): New file scoped static.
8418 (process_switch): If group_case_labels makes a change, then set
8420 (pass_convert_switch::execute): If a switch is converted, then
8421 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
8423 PR rtl-optimization/81308
8424 * recog.c (split_all_insns): Conditionally cleanup the CFG after
8427 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
8429 PR target/83663 - Revert r255946
8430 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
8431 generation for cases where splatting a value is not useful.
8432 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
8433 across a vec_duplicate and a paradoxical subreg forming a vector
8434 mode to a vec_concat.
8436 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8438 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
8439 -march=armv8.3-a variants.
8440 * config/arm/t-multilib: Likewise.
8441 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
8443 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8445 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
8447 (cceq_ior_compare_complement): Give it a name so I can use it, and
8448 change boolean_or_operator predicate to boolean_operator so it can
8449 be used to generate a crand.
8450 (eqne): New code iterator.
8451 (bd/bd_neg): New code_attrs.
8452 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
8453 a single define_insn.
8454 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
8455 decrement (bdnzt/bdnzf/bdzt/bdzf).
8456 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
8457 with the new names of the branch decrement patterns, and added the
8458 names of the branch decrement conditional patterns.
8460 2018-01-08 Richard Biener <rguenther@suse.de>
8462 PR tree-optimization/83563
8463 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
8466 2018-01-08 Richard Biener <rguenther@suse.de>
8469 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
8471 2018-01-08 Richard Biener <rguenther@suse.de>
8473 PR tree-optimization/83685
8474 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
8475 references to abnormals.
8477 2018-01-08 Richard Biener <rguenther@suse.de>
8480 * dwarf2out.c (output_indirect_strings): Handle empty
8481 skeleton_debug_str_hash.
8482 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
8484 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
8486 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
8487 (emit_store_direct): Likewise.
8488 (arc_trampoline_adjust_address): Likewise.
8489 (arc_asm_trampoline_template): New function.
8490 (arc_initialize_trampoline): Use asm_trampoline_template.
8491 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
8492 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
8493 * config/arc/arc.md (flush_icache): Delete pattern.
8495 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
8497 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
8498 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
8501 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
8504 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
8505 by not USED_FOR_TARGET.
8506 (make_pass_resolve_sw_modes): Likewise.
8508 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
8510 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
8513 2018-01-08 Richard Biener <rguenther@suse.de>
8516 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
8518 2018-01-08 Richard Biener <rguenther@suse.de>
8521 * match.pd ((t * 2) / 2) -> t): Add missing :c.
8523 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
8526 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
8527 basic blocks with a small number of successors.
8528 (convert_control_dep_chain_into_preds): Improve handling of
8530 (dump_predicates): Split apart into...
8531 (dump_pred_chain): ...here...
8532 (dump_pred_info): ...and here.
8533 (can_one_predicate_be_invalidated_p): Add debugging printfs.
8534 (can_chain_union_be_invalidated_p): Improve check for invalidation
8536 (uninit_uses_cannot_happen): Avoid unnecessary if
8537 convert_control_dep_chain_into_preds yielded nothing.
8539 2018-01-06 Martin Sebor <msebor@redhat.com>
8541 PR tree-optimization/83640
8542 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
8543 subtracting negative offset from size.
8544 (builtin_access::overlap): Adjust offset bounds of the access to fall
8545 within the size of the object if possible.
8547 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
8549 PR rtl-optimization/83699
8550 * expmed.c (extract_bit_field_1): Restrict the vector usage of
8551 extract_bit_field_as_subreg to cases in which the extracted
8552 value is also a vector.
8554 * lra-constraints.c (process_alt_operands): Test for the equivalence
8555 substitutions when detecting a possible reload cycle.
8557 2018-01-06 Jakub Jelinek <jakub@redhat.com>
8560 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
8561 by default if flag_selective_schedling{,2}. Formatting fixes.
8563 PR rtl-optimization/83682
8564 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
8565 if it has non-VECTOR_MODE element mode.
8566 (vec_duplicate_p): Likewise.
8569 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
8570 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
8572 2018-01-05 Jakub Jelinek <jakub@redhat.com>
8575 * config/i386/i386-builtin.def
8576 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
8577 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
8578 Require also OPTION_MASK_ISA_AVX512F in addition to
8579 OPTION_MASK_ISA_GFNI.
8580 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
8581 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
8582 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
8583 to OPTION_MASK_ISA_GFNI.
8584 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
8585 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
8586 OPTION_MASK_ISA_AVX512BW.
8587 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
8588 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
8589 addition to OPTION_MASK_ISA_GFNI.
8590 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
8591 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
8592 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
8593 to OPTION_MASK_ISA_GFNI.
8594 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
8595 a requirement for all ISAs rather than any of them with a few
8597 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
8599 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
8600 bitmasks to be enabled with 3 exceptions, instead of requiring any
8601 enabled ISA with lots of exceptions.
8602 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
8603 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
8604 Change avx512bw in isa attribute to avx512f.
8605 * config/i386/sgxintrin.h: Add license boilerplate.
8606 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
8607 to __AVX512F__ and __AVX512VL to __AVX512VL__.
8608 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
8609 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
8611 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
8612 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
8613 temporarily sse2 rather than sse if not enabled already.
8616 * config/i386/sse.md (VI248_VLBW): Rename to ...
8617 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
8618 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
8619 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
8620 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
8621 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
8622 mode iterator instead of VI248_VLBW.
8624 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
8626 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
8627 (record_modified): Skip clobbers; add debug output.
8628 (param_change_prob): Use sreal frequencies.
8630 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8632 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
8633 punt for user-aligned variables.
8635 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8637 * tree-chrec.c (chrec_contains_symbols): Return true for
8640 2018-01-05 Sudakshina Das <sudi.das@arm.com>
8643 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
8644 of (x|y) == x for BICS pattern.
8646 2018-01-05 Jakub Jelinek <jakub@redhat.com>
8648 PR tree-optimization/83605
8649 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
8650 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
8653 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
8655 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
8656 * config/epiphany/rtems.h: New file.
8658 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8659 Uros Bizjak <ubizjak@gmail.com>
8662 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
8663 QIreg_operand instead of register_operand predicate.
8664 * config/i386/i386.c (ix86_rop_should_change_byte_p,
8665 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
8666 comments instead of -fmitigate[-_]rop.
8668 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8671 * cgraphunit.c (symbol_table::compile): Switch to text_section
8672 before calling assembly_start debug hook.
8673 * run-rtl-passes.c (run_rtl_passes): Likewise.
8676 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8678 * tree-vrp.c (extract_range_from_binary_expr_1): Check
8679 range_int_cst_p rather than !symbolic_range_p before calling
8680 extract_range_from_multiplicative_op_1.
8682 2018-01-04 Jeff Law <law@redhat.com>
8684 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
8685 redundant test in assertion.
8687 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8689 * doc/rtl.texi: Document machine_mode wrapper classes.
8691 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8693 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
8696 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8698 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
8699 the VEC_PERM_EXPR fold to fail.
8701 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8704 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
8705 to switched_sections.
8707 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8710 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
8713 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
8716 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
8717 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
8719 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8722 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
8723 is BLKmode and bitpos not zero or mode change is needed.
8725 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8728 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
8731 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
8734 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
8735 instead of MULT rtx. Update all corresponding splitters.
8737 (*ssub<modesuffix>): Ditto.
8739 (*cmp_sadd_di): Update split patterns.
8740 (*cmp_sadd_si): Ditto.
8741 (*cmp_sadd_sidi): Ditto.
8742 (*cmp_ssub_di): Ditto.
8743 (*cmp_ssub_si): Ditto.
8744 (*cmp_ssub_sidi): Ditto.
8745 * config/alpha/predicates.md (const23_operand): New predicate.
8746 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
8747 Look for ASHIFT, not MULT inner operand.
8748 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
8750 2018-01-04 Martin Liska <mliska@suse.cz>
8752 PR gcov-profile/83669
8753 * gcov.c (output_intermediate_file): Add version to intermediate
8755 * doc/gcov.texi: Document new field 'version' in intermediate
8756 file format. Fix location of '-k' option of gcov command.
8758 2018-01-04 Martin Liska <mliska@suse.cz>
8761 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
8763 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8765 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
8767 2018-01-03 Martin Sebor <msebor@redhat.com>
8769 PR tree-optimization/83655
8770 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
8771 checking calls with invalid arguments.
8773 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8775 * tree-vect-stmts.c (vect_get_store_rhs): New function.
8776 (vectorizable_mask_load_store): Delete.
8777 (vectorizable_call): Return false for masked loads and stores.
8778 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
8779 instead of gimple_assign_rhs1.
8780 (vectorizable_load): Handle IFN_MASK_LOAD.
8781 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
8783 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8785 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
8787 (vectorizable_mask_load_store): ...here.
8788 (vectorizable_load): ...and here.
8790 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8792 * tree-vect-stmts.c (vect_build_all_ones_mask)
8793 (vect_build_zero_merge_argument): New functions, split out from...
8794 (vectorizable_load): ...here.
8796 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8798 * tree-vect-stmts.c (vect_check_store_rhs): New function,
8800 (vectorizable_mask_load_store): ...here.
8801 (vectorizable_store): ...and here.
8803 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8805 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
8807 (vectorizable_mask_load_store): ...here.
8809 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8811 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
8812 (vect_model_store_cost): Take a vec_load_store_type instead of a
8814 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
8815 (vect_model_store_cost): Take a vec_load_store_type instead of a
8817 (vectorizable_mask_load_store): Update accordingly.
8818 (vectorizable_store): Likewise.
8819 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
8821 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8823 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
8824 IFN_MASK_LOAD calls here rather than...
8825 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
8827 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8828 Alan Hayward <alan.hayward@arm.com>
8829 David Sherwood <david.sherwood@arm.com>
8831 * expmed.c (extract_bit_field_1): For vector extracts,
8832 fall back to extract_bit_field_as_subreg if vec_extract
8835 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8836 Alan Hayward <alan.hayward@arm.com>
8837 David Sherwood <david.sherwood@arm.com>
8839 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
8840 they are variable or constant sized.
8841 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
8842 slots for constant-sized data.
8844 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8845 Alan Hayward <alan.hayward@arm.com>
8846 David Sherwood <david.sherwood@arm.com>
8848 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
8849 handling COND_EXPRs with boolean comparisons, try to find a better
8850 basis for the mask type than the boolean itself.
8852 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8854 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
8855 is calculated and how it can be overridden.
8856 * genmodes.c (max_bitsize_mode_any_mode): New variable.
8857 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
8859 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
8862 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8863 Alan Hayward <alan.hayward@arm.com>
8864 David Sherwood <david.sherwood@arm.com>
8866 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
8867 Remove the mode argument.
8868 (aarch64_simd_valid_immediate): Remove the mode and inverse
8870 * config/aarch64/iterators.md (bitsize): New iterator.
8871 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
8872 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
8873 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
8874 aarch64_simd_valid_immediate.
8875 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
8876 (aarch64_reg_or_bic_imm): Likewise.
8877 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
8878 with an insn_type enum and msl with a modifier_type enum.
8879 Replace element_width with a scalar_mode. Change the shift
8880 to unsigned int. Add constructors for scalar_float_mode and
8881 scalar_int_mode elements.
8882 (aarch64_vect_float_const_representable_p): Delete.
8883 (aarch64_can_const_movi_rtx_p)
8884 (aarch64_simd_scalar_immediate_valid_for_move)
8885 (aarch64_simd_make_constant): Update call to
8886 aarch64_simd_valid_immediate.
8887 (aarch64_advsimd_valid_immediate_hs): New function.
8888 (aarch64_advsimd_valid_immediate): Likewise.
8889 (aarch64_simd_valid_immediate): Remove mode and inverse
8890 arguments. Rewrite to use the above. Use const_vec_duplicate_p
8891 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
8892 and aarch64_float_const_representable_p on the result.
8893 (aarch64_output_simd_mov_immediate): Remove mode argument.
8894 Update call to aarch64_simd_valid_immediate and use of
8895 simd_immediate_info.
8896 (aarch64_output_scalar_simd_mov_immediate): Update call
8899 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8900 Alan Hayward <alan.hayward@arm.com>
8901 David Sherwood <david.sherwood@arm.com>
8903 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
8904 (mode_nunits): Likewise CONST_MODE_NUNITS.
8905 * machmode.def (ADJUST_NUNITS): Document.
8906 * genmodes.c (mode_data::need_nunits_adj): New field.
8907 (blank_mode): Update accordingly.
8908 (adj_nunits): New variable.
8909 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
8911 (emit_mode_size_inline): Set need_bytesize_adj for all modes
8912 listed in adj_nunits.
8913 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
8914 listed in adj_nunits. Don't emit case statements for such modes.
8915 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
8916 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
8917 nothing if adj_nunits is nonnull.
8918 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
8919 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
8920 (emit_mode_fbit): Update use of print_maybe_const_decl.
8921 (emit_move_size): Likewise. Treat the array as non-const
8923 (emit_mode_adjustments): Handle adj_nunits.
8925 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8927 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
8928 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
8929 (VECTOR_MODES): Use it.
8930 (make_vector_modes): Take the prefix as an argument.
8932 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8933 Alan Hayward <alan.hayward@arm.com>
8934 David Sherwood <david.sherwood@arm.com>
8936 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
8937 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
8938 for MODE_VECTOR_BOOL.
8939 * machmode.def (VECTOR_BOOL_MODE): Document.
8940 * genmodes.c (VECTOR_BOOL_MODE): New macro.
8941 (make_vector_bool_mode): New function.
8942 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
8944 * lto-streamer-in.c (lto_input_mode_table): Likewise.
8945 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
8947 * stor-layout.c (int_mode_for_mode): Likewise.
8948 * tree.c (build_vector_type_for_mode): Likewise.
8949 * varasm.c (output_constant_pool_2): Likewise.
8950 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
8951 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
8952 for MODE_VECTOR_BOOL.
8953 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
8954 of mode class checks.
8955 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
8956 instead of a list of mode class checks.
8957 (expand_vector_scalar_condition): Likewise.
8958 (type_for_widest_vector_mode): Handle BImode as an inner mode.
8960 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8961 Alan Hayward <alan.hayward@arm.com>
8962 David Sherwood <david.sherwood@arm.com>
8964 * machmode.h (mode_size): Change from unsigned short to
8966 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
8967 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8968 or if measurement_type is not polynomial.
8969 (fixed_size_mode::includes_p): Check for constant-sized modes.
8970 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
8971 return a poly_uint16 rather than an unsigned short.
8972 (emit_mode_size): Change the type of mode_size from unsigned short
8973 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
8974 (emit_mode_adjustments): Cope with polynomial vector sizes.
8975 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8977 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8979 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
8980 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
8981 * caller-save.c (setup_save_areas): Likewise.
8982 (replace_reg_with_saved_mem): Likewise.
8983 * calls.c (emit_library_call_value_1): Likewise.
8984 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
8985 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
8986 (gen_lowpart_for_combine): Likewise.
8987 * convert.c (convert_to_integer_1): Likewise.
8988 * cse.c (equiv_constant, cse_insn): Likewise.
8989 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
8990 (cselib_subst_to_values): Likewise.
8991 * dce.c (word_dce_process_block): Likewise.
8992 * df-problems.c (df_word_lr_mark_ref): Likewise.
8993 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
8994 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
8995 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
8996 (rtl_for_decl_location): Likewise.
8997 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
8998 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
8999 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
9000 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
9001 (expand_expr_real_1): Likewise.
9002 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
9003 (pad_below): Likewise.
9004 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9005 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
9006 * ira.c (get_subreg_tracking_sizes): Likewise.
9007 * ira-build.c (ira_create_allocno_objects): Likewise.
9008 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
9009 (ira_sort_regnos_for_alter_reg): Likewise.
9010 * ira-costs.c (record_operand_costs): Likewise.
9011 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
9012 (resolve_simple_move): Likewise.
9013 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
9014 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
9015 (lra_constraints): Likewise.
9016 (CONST_POOL_OK_P): Reject variable-sized modes.
9017 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
9018 (add_pseudo_to_slot, lra_spill): Likewise.
9019 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9020 * optabs-query.c (get_best_extraction_insn): Likewise.
9021 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9022 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
9023 (expand_mult_highpart, valid_multiword_target_p): Likewise.
9024 * recog.c (offsettable_address_addr_space_p): Likewise.
9025 * regcprop.c (maybe_mode_change): Likewise.
9026 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
9027 * regrename.c (build_def_use): Likewise.
9028 * regstat.c (dump_reg_info): Likewise.
9029 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
9030 (find_reloads, find_reloads_subreg_address): Likewise.
9031 * reload1.c (eliminate_regs_1): Likewise.
9032 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
9033 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
9034 (simplify_binary_operation_1, simplify_subreg): Likewise.
9035 * targhooks.c (default_function_arg_padding): Likewise.
9036 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
9037 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
9038 (verify_gimple_assign_ternary): Likewise.
9039 * tree-inline.c (estimate_move_cost): Likewise.
9040 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9041 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
9042 (get_address_cost_ainc): Likewise.
9043 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
9044 (vect_supportable_dr_alignment): Likewise.
9045 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9046 (vectorizable_reduction): Likewise.
9047 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
9048 (vectorizable_operation, vectorizable_load): Likewise.
9049 * tree.c (build_same_sized_truth_vector_type): Likewise.
9050 * valtrack.c (cleanup_auto_inc_dec): Likewise.
9051 * var-tracking.c (emit_note_insn_var_location): Likewise.
9052 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
9053 (ADDR_VEC_ALIGN): Likewise.
9055 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9056 Alan Hayward <alan.hayward@arm.com>
9057 David Sherwood <david.sherwood@arm.com>
9059 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
9061 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9062 or if measurement_type is polynomial.
9063 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
9064 * combine.c (make_extraction): Likewise.
9065 * dse.c (find_shift_sequence): Likewise.
9066 * dwarf2out.c (mem_loc_descriptor): Likewise.
9067 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
9068 (extract_bit_field, extract_low_bits): Likewise.
9069 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
9070 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
9071 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
9072 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
9073 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9074 * reload.c (find_reloads): Likewise.
9075 * reload1.c (alter_reg): Likewise.
9076 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
9077 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
9078 * tree-if-conv.c (predicate_mem_writes): Likewise.
9079 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
9080 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
9081 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
9082 * valtrack.c (dead_debug_insert_temp): Likewise.
9083 * varasm.c (mergeable_constant_section): Likewise.
9084 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
9086 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9087 Alan Hayward <alan.hayward@arm.com>
9088 David Sherwood <david.sherwood@arm.com>
9090 * expr.c (expand_assignment): Cope with polynomial mode sizes
9091 when assigning to a CONCAT.
9093 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9094 Alan Hayward <alan.hayward@arm.com>
9095 David Sherwood <david.sherwood@arm.com>
9097 * machmode.h (mode_precision): Change from unsigned short to
9099 (mode_to_precision): Return a poly_uint16 rather than an unsigned
9101 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
9102 or if measurement_type is not polynomial.
9103 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
9104 in which the mode is already known to be a scalar_int_mode.
9105 * genmodes.c (emit_mode_precision): Change the type of mode_precision
9106 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
9108 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9109 for GET_MODE_PRECISION.
9110 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9111 for GET_MODE_PRECISION.
9112 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
9114 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
9115 (expand_field_assignment, make_extraction): Likewise.
9116 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
9117 (get_last_value): Likewise.
9118 * convert.c (convert_to_integer_1): Likewise.
9119 * cse.c (cse_insn): Likewise.
9120 * expr.c (expand_expr_real_1): Likewise.
9121 * lra-constraints.c (simplify_operand_subreg): Likewise.
9122 * optabs-query.c (can_atomic_load_p): Likewise.
9123 * optabs.c (expand_atomic_load): Likewise.
9124 (expand_atomic_store): Likewise.
9125 * ree.c (combine_reaching_defs): Likewise.
9126 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
9127 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
9128 * tree.h (type_has_mode_precision_p): Likewise.
9129 * ubsan.c (instrument_si_overflow): Likewise.
9131 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9132 Alan Hayward <alan.hayward@arm.com>
9133 David Sherwood <david.sherwood@arm.com>
9135 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
9136 polynomial numbers of units.
9137 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
9138 (valid_vector_subparts_p): New function.
9139 (build_vector_type): Remove temporary shim and take the number
9140 of units as a poly_uint64 rather than an int.
9141 (build_opaque_vector_type): Take the number of units as a
9142 poly_uint64 rather than an int.
9143 * tree.c (build_vector_from_ctor): Handle polynomial
9144 TYPE_VECTOR_SUBPARTS.
9145 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
9146 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
9147 (build_vector_from_val): If the number of units is variable,
9148 use build_vec_duplicate_cst for constant operands and
9149 VEC_DUPLICATE_EXPR otherwise.
9150 (make_vector_type): Remove temporary is_constant ().
9151 (build_vector_type, build_opaque_vector_type): Take the number of
9152 units as a poly_uint64 rather than an int.
9153 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
9155 * cfgexpand.c (expand_debug_expr): Likewise.
9156 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
9157 (store_constructor, expand_expr_real_1): Likewise.
9158 (const_scalar_mask_from_tree): Likewise.
9159 * fold-const-call.c (fold_const_reduction): Likewise.
9160 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
9161 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
9162 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
9163 (fold_relational_const): Likewise.
9164 (native_interpret_vector): Likewise. Change the size from an
9165 int to an unsigned int.
9166 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
9167 TYPE_VECTOR_SUBPARTS.
9168 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
9169 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
9170 duplicating a non-constant operand into a variable-length vector.
9171 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
9172 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
9173 * ipa-icf.c (sem_variable::equals): Likewise.
9174 * match.pd: Likewise.
9175 * omp-simd-clone.c (simd_clone_subparts): Likewise.
9176 * print-tree.c (print_node): Likewise.
9177 * stor-layout.c (layout_type): Likewise.
9178 * targhooks.c (default_builtin_vectorization_cost): Likewise.
9179 * tree-cfg.c (verify_gimple_comparison): Likewise.
9180 (verify_gimple_assign_binary): Likewise.
9181 (verify_gimple_assign_ternary): Likewise.
9182 (verify_gimple_assign_single): Likewise.
9183 * tree-pretty-print.c (dump_generic_node): Likewise.
9184 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9185 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
9186 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
9187 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
9188 (vect_shift_permute_load_chain): Likewise.
9189 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
9190 (expand_vector_condition, optimize_vector_constructor): Likewise.
9191 (lower_vec_perm, get_compute_type): Likewise.
9192 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9193 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
9194 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
9195 (vect_recog_mask_conversion_pattern): Likewise.
9196 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
9197 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
9198 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9199 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
9200 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
9201 (vectorizable_shift, vectorizable_operation, vectorizable_store)
9202 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
9203 (supportable_widening_operation): Likewise.
9204 (supportable_narrowing_operation): Likewise.
9205 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
9207 * varasm.c (output_constant): Likewise.
9209 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9210 Alan Hayward <alan.hayward@arm.com>
9211 David Sherwood <david.sherwood@arm.com>
9213 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
9214 so that both the length == 3 and length != 3 cases set up their
9215 own permute vectors. Add comments explaining why we know the
9216 number of elements is constant.
9217 (vect_permute_load_chain): Likewise.
9219 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9220 Alan Hayward <alan.hayward@arm.com>
9221 David Sherwood <david.sherwood@arm.com>
9223 * machmode.h (mode_nunits): Change from unsigned char to
9225 (ONLY_FIXED_SIZE_MODES): New macro.
9226 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
9227 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
9228 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
9230 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
9231 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
9232 or if measurement_type is not polynomial.
9233 * genmodes.c (ZERO_COEFFS): New macro.
9234 (emit_mode_nunits_inline): Make mode_nunits_inline return a
9236 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
9237 Use ZERO_COEFFS when emitting initializers.
9238 * data-streamer.h (bp_pack_poly_value): New function.
9239 (bp_unpack_poly_value): Likewise.
9240 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9241 for GET_MODE_NUNITS.
9242 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9243 for GET_MODE_NUNITS.
9244 * tree.c (make_vector_type): Remove temporary shim and make
9245 the real function take the number of units as a poly_uint64
9247 (build_vector_type_for_mode): Handle polynomial nunits.
9248 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
9249 * emit-rtl.c (const_vec_series_p_1): Likewise.
9250 (gen_rtx_CONST_VECTOR): Likewise.
9251 * fold-const.c (test_vec_duplicate_folding): Likewise.
9252 * genrecog.c (validate_pattern): Likewise.
9253 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
9254 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9255 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
9256 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
9257 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
9258 * rtlanal.c (subreg_get_info): Likewise.
9259 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9260 (vect_grouped_load_supported): Likewise.
9261 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
9262 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
9263 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
9264 (simplify_const_unary_operation, simplify_binary_operation_1)
9265 (simplify_const_binary_operation, simplify_ternary_operation)
9266 (test_vector_ops_duplicate, test_vector_ops): Likewise.
9267 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
9268 instead of CONST_VECTOR_NUNITS.
9269 * varasm.c (output_constant_pool_2): Likewise.
9270 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
9271 explicit-encoded elements in the XVEC for variable-length vectors.
9273 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9275 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
9277 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9278 Alan Hayward <alan.hayward@arm.com>
9279 David Sherwood <david.sherwood@arm.com>
9281 * coretypes.h (fixed_size_mode): Declare.
9282 (fixed_size_mode_pod): New typedef.
9283 * builtins.h (target_builtins::x_apply_args_mode)
9284 (target_builtins::x_apply_result_mode): Change type to
9285 fixed_size_mode_pod.
9286 * builtins.c (apply_args_size, apply_result_size, result_vector)
9287 (expand_builtin_apply_args_1, expand_builtin_apply)
9288 (expand_builtin_return): Update accordingly.
9290 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9292 * cse.c (hash_rtx_cb): Hash only the encoded elements.
9293 * cselib.c (cselib_hash_rtx): Likewise.
9294 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
9295 CONST_VECTOR encoding.
9297 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9298 Jeff Law <law@redhat.com>
9301 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
9302 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
9303 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
9304 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
9307 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
9308 explicitly probe *sp in a noreturn function if there were any callee
9309 register saves or frame pointer is needed.
9311 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9314 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
9315 BLKmode for ternary, binary or unary expressions.
9318 * var-tracking.c (delete_vta_debug_insn): New inline function.
9319 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
9320 insns from get_insns () to NULL instead of each bb separately.
9321 Use delete_vta_debug_insn. No longer static.
9322 (vt_debug_insns_local, variable_tracking_main_1): Adjust
9323 delete_vta_debug_insns callers.
9324 * rtl.h (delete_vta_debug_insns): Declare.
9325 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
9326 instead of variable_tracking_main.
9328 2018-01-03 Martin Sebor <msebor@redhat.com>
9330 PR tree-optimization/83603
9331 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
9332 arguments past the endof the argument list in functions declared
9333 without a prototype.
9334 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
9335 Avoid checking when arguments are null.
9337 2018-01-03 Martin Sebor <msebor@redhat.com>
9340 * doc/extend.texi (attribute const): Fix a typo.
9341 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
9342 issuing -Wsuggest-attribute for void functions.
9344 2018-01-03 Martin Sebor <msebor@redhat.com>
9346 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
9347 offset_int::from instead of wide_int::to_shwi.
9348 (maybe_diag_overlap): Remove assertion.
9349 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
9350 * gimple-ssa-sprintf.c (format_directive): Same.
9351 (parse_directive): Same.
9352 (sprintf_dom_walker::compute_format_length): Same.
9353 (try_substitute_return_value): Same.
9355 2018-01-03 Jeff Law <law@redhat.com>
9358 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
9359 non-constant residual for zero at runtime and avoid probing in
9360 that case. Reorganize code for trailing problem to mirror handling
9363 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9365 PR tree-optimization/83501
9366 * tree-ssa-strlen.c (get_string_cst): New.
9367 (handle_char_store): Call get_string_cst.
9369 2018-01-03 Martin Liska <mliska@suse.cz>
9371 PR tree-optimization/83593
9372 * tree-ssa-strlen.c: Include tree-cfg.h.
9373 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
9374 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
9375 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
9377 (strlen_dom_walker::before_dom_children): Call
9378 gimple_purge_dead_eh_edges. Dump tranformation with details
9380 (strlen_dom_walker::before_dom_children): Update call by adding
9381 new argument cleanup_eh.
9382 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
9384 2018-01-03 Martin Liska <mliska@suse.cz>
9387 * cif-code.def (VARIADIC_THUNK): New enum value.
9388 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
9391 2018-01-03 Jan Beulich <jbeulich@suse.com>
9393 * sse.md (mov<mode>_internal): Tighten condition for when to use
9394 vmovdqu<ssescalarsize> for TI and OI modes.
9396 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9398 Update copyright years.
9400 2018-01-03 Martin Liska <mliska@suse.cz>
9403 * ipa-visibility.c (function_and_variable_visibility): Skip
9404 functions with noipa attribure.
9406 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9408 * gcc.c (process_command): Update copyright notice dates.
9409 * gcov-dump.c (print_version): Ditto.
9410 * gcov.c (print_version): Ditto.
9411 * gcov-tool.c (print_version): Ditto.
9412 * gengtype.c (create_file): Ditto.
9413 * doc/cpp.texi: Bump @copying's copyright year.
9414 * doc/cppinternals.texi: Ditto.
9415 * doc/gcc.texi: Ditto.
9416 * doc/gccint.texi: Ditto.
9417 * doc/gcov.texi: Ditto.
9418 * doc/install.texi: Ditto.
9419 * doc/invoke.texi: Ditto.
9421 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9423 * vector-builder.h (vector_builder::m_full_nelts): Change from
9424 unsigned int to poly_uint64.
9425 (vector_builder::full_nelts): Update prototype accordingly.
9426 (vector_builder::new_vector): Likewise.
9427 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
9428 (vector_builder::operator ==): Likewise.
9429 (vector_builder::finalize): Likewise.
9430 * int-vector-builder.h (int_vector_builder::int_vector_builder):
9431 Take the number of elements as a poly_uint64 rather than an
9433 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
9434 from unsigned int to poly_uint64.
9435 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
9436 (vec_perm_indices::new_vector): Likewise.
9437 (vec_perm_indices::length): Likewise.
9438 (vec_perm_indices::nelts_per_input): Likewise.
9439 (vec_perm_indices::input_nelts): Likewise.
9440 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
9441 number of elements per input as a poly_uint64 rather than an
9442 unsigned int. Use the original encoding for variable-length
9443 vectors, rather than clamping each individual element.
9444 For the second and subsequent elements in each pattern,
9445 clamp the step and base before clamping their sum.
9446 (vec_perm_indices::series_p): Handle polynomial element counts.
9447 (vec_perm_indices::all_in_range_p): Likewise.
9448 (vec_perm_indices_to_tree): Likewise.
9449 (vec_perm_indices_to_rtx): Likewise.
9450 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
9451 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
9452 (tree_vector_builder::new_binary_operation): Handle polynomial
9453 element counts. Return false if we need to know the number
9454 of elements at compile time.
9455 * fold-const.c (fold_vec_perm): Punt if the number of elements
9456 isn't known at compile time.
9458 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9460 * vec-perm-indices.h (vec_perm_builder): Change element type
9461 from HOST_WIDE_INT to poly_int64.
9462 (vec_perm_indices::element_type): Update accordingly.
9463 (vec_perm_indices::clamp): Handle polynomial element_types.
9464 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9465 (vec_perm_indices::all_in_range_p): Likewise.
9466 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
9468 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
9469 polynomial vec_perm_indices element types.
9470 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
9471 * fold-const.c (fold_vec_perm): Likewise.
9472 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
9473 * tree-vect-generic.c (lower_vec_perm): Likewise.
9474 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9475 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
9476 element type to HOST_WIDE_INT.
9478 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9479 Alan Hayward <alan.hayward@arm.com>
9480 David Sherwood <david.sherwood@arm.com>
9482 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
9483 rather than an int. Use plus_constant.
9484 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
9485 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
9487 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9488 Alan Hayward <alan.hayward@arm.com>
9489 David Sherwood <david.sherwood@arm.com>
9491 * calls.c (emit_call_1, expand_call): Change struct_value_size from
9492 a HOST_WIDE_INT to a poly_int64.
9494 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9495 Alan Hayward <alan.hayward@arm.com>
9496 David Sherwood <david.sherwood@arm.com>
9498 * calls.c (load_register_parameters): Cope with polynomial
9499 mode sizes. Require a constant size for BLKmode parameters
9500 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
9501 forces a parameter to be padded at the lsb end in order to
9502 fill a complete number of words, require the parameter size
9503 to be ordered wrt UNITS_PER_WORD.
9505 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9506 Alan Hayward <alan.hayward@arm.com>
9507 David Sherwood <david.sherwood@arm.com>
9509 * reload1.c (spill_stack_slot_width): Change element type
9510 from unsigned int to poly_uint64_pod.
9511 (alter_reg): Treat mode sizes as polynomial.
9513 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9514 Alan Hayward <alan.hayward@arm.com>
9515 David Sherwood <david.sherwood@arm.com>
9517 * reload.c (complex_word_subreg_p): New function.
9518 (reload_inner_reg_of_subreg, push_reload): Use it.
9520 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9521 Alan Hayward <alan.hayward@arm.com>
9522 David Sherwood <david.sherwood@arm.com>
9524 * lra-constraints.c (process_alt_operands): Reject matched
9525 operands whose sizes aren't ordered.
9526 (match_reload): Refer to this check here.
9528 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9529 Alan Hayward <alan.hayward@arm.com>
9530 David Sherwood <david.sherwood@arm.com>
9532 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
9533 that the mode size is in the set {1, 2, 4, 8, 16}.
9535 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9536 Alan Hayward <alan.hayward@arm.com>
9537 David Sherwood <david.sherwood@arm.com>
9539 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
9540 Use plus_constant instead of gen_rtx_PLUS.
9542 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9543 Alan Hayward <alan.hayward@arm.com>
9544 David Sherwood <david.sherwood@arm.com>
9546 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
9547 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
9548 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
9549 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
9550 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
9551 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
9552 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
9553 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
9554 * config/i386/i386.c (ix86_push_rounding): ...this new function.
9555 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
9557 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
9558 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
9559 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
9560 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
9561 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
9562 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
9563 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
9564 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
9565 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
9566 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
9568 * expr.c (emit_move_resolve_push): Treat the input and result
9569 of PUSH_ROUNDING as a poly_int64.
9570 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
9571 (emit_push_insn): Likewise.
9572 * lra-eliminations.c (mark_not_eliminable): Likewise.
9573 * recog.c (push_operand): Likewise.
9574 * reload1.c (elimination_effects): Likewise.
9575 * rtlanal.c (nonzero_bits1): Likewise.
9576 * calls.c (store_one_arg): Likewise. Require the padding to be
9577 known at compile time.
9579 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9580 Alan Hayward <alan.hayward@arm.com>
9581 David Sherwood <david.sherwood@arm.com>
9583 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
9584 Use plus_constant instead of gen_rtx_PLUS.
9586 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9587 Alan Hayward <alan.hayward@arm.com>
9588 David Sherwood <david.sherwood@arm.com>
9590 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
9593 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9594 Alan Hayward <alan.hayward@arm.com>
9595 David Sherwood <david.sherwood@arm.com>
9597 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
9598 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
9599 via stack temporaries. Treat the mode size as polynomial too.
9601 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9602 Alan Hayward <alan.hayward@arm.com>
9603 David Sherwood <david.sherwood@arm.com>
9605 * expr.c (expand_expr_real_2): When handling conversions involving
9606 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
9607 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
9608 as a poly_uint64 too.
9610 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9611 Alan Hayward <alan.hayward@arm.com>
9612 David Sherwood <david.sherwood@arm.com>
9614 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
9616 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9617 Alan Hayward <alan.hayward@arm.com>
9618 David Sherwood <david.sherwood@arm.com>
9620 * combine.c (can_change_dest_mode): Handle polynomial
9621 REGMODE_NATURAL_SIZE.
9622 * expmed.c (store_bit_field_1): Likewise.
9623 * expr.c (store_constructor): Likewise.
9624 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
9625 and polynomial REGMODE_NATURAL_SIZE.
9626 (gen_lowpart_common): Likewise.
9627 * reginfo.c (record_subregs_of_mode): Likewise.
9628 * rtlanal.c (read_modify_subreg_p): Likewise.
9630 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9631 Alan Hayward <alan.hayward@arm.com>
9632 David Sherwood <david.sherwood@arm.com>
9634 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
9635 numbers of elements.
9637 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9638 Alan Hayward <alan.hayward@arm.com>
9639 David Sherwood <david.sherwood@arm.com>
9641 * match.pd: Cope with polynomial numbers of vector elements.
9643 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9644 Alan Hayward <alan.hayward@arm.com>
9645 David Sherwood <david.sherwood@arm.com>
9647 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
9648 in a POINTER_PLUS_EXPR.
9650 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9651 Alan Hayward <alan.hayward@arm.com>
9652 David Sherwood <david.sherwood@arm.com>
9654 * omp-simd-clone.c (simd_clone_subparts): New function.
9655 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
9656 (ipa_simd_modify_function_body): Likewise.
9658 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9659 Alan Hayward <alan.hayward@arm.com>
9660 David Sherwood <david.sherwood@arm.com>
9662 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
9663 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
9664 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
9665 (expand_vector_condition, vector_element): Likewise.
9666 (subparts_gt): New function.
9667 (get_compute_type): Use subparts_gt.
9668 (count_type_subparts): Delete.
9669 (expand_vector_operations_1): Use subparts_gt instead of
9670 count_type_subparts.
9672 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9673 Alan Hayward <alan.hayward@arm.com>
9674 David Sherwood <david.sherwood@arm.com>
9676 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
9677 (vect_compile_time_alias): ...this new function. Do the calculation
9678 on poly_ints rather than trees.
9679 (vect_prune_runtime_alias_test_list): Update call accordingly.
9681 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9682 Alan Hayward <alan.hayward@arm.com>
9683 David Sherwood <david.sherwood@arm.com>
9685 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
9687 (vect_schedule_slp_instance): Likewise.
9689 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9690 Alan Hayward <alan.hayward@arm.com>
9691 David Sherwood <david.sherwood@arm.com>
9693 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
9694 constant and extern definitions for variable-length vectors.
9695 (vect_get_constant_vectors): Note that the number of units
9696 is known to be constant.
9698 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9699 Alan Hayward <alan.hayward@arm.com>
9700 David Sherwood <david.sherwood@arm.com>
9702 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
9703 of units as polynomial. Choose between WIDE and NARROW based
9706 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9707 Alan Hayward <alan.hayward@arm.com>
9708 David Sherwood <david.sherwood@arm.com>
9710 * tree-vect-stmts.c (simd_clone_subparts): New function.
9711 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
9713 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9714 Alan Hayward <alan.hayward@arm.com>
9715 David Sherwood <david.sherwood@arm.com>
9717 * tree-vect-stmts.c (vectorizable_call): Treat the number of
9718 vectors as polynomial. Use build_index_vector for
9721 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9722 Alan Hayward <alan.hayward@arm.com>
9723 David Sherwood <david.sherwood@arm.com>
9725 * tree-vect-stmts.c (get_load_store_type): Treat the number of
9726 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
9727 for variable-length vectors.
9728 (vectorizable_mask_load_store): Treat the number of units as
9729 polynomial, asserting that it is constant if the condition has
9730 already been enforced.
9731 (vectorizable_store, vectorizable_load): Likewise.
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-loop.c (vectorizable_live_operation): Treat the number
9738 of units as polynomial. Punt if we can't tell at compile time
9739 which vector contains the final result.
9741 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9742 Alan Hayward <alan.hayward@arm.com>
9743 David Sherwood <david.sherwood@arm.com>
9745 * tree-vect-loop.c (vectorizable_induction): Treat the number
9746 of units as polynomial. Punt on SLP inductions. Use an integer
9747 VEC_SERIES_EXPR for variable-length integer reductions. Use a
9748 cast of such a series for variable-length floating-point
9751 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9752 Alan Hayward <alan.hayward@arm.com>
9753 David Sherwood <david.sherwood@arm.com>
9755 * tree.h (build_index_vector): Declare.
9756 * tree.c (build_index_vector): New function.
9757 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
9758 of units as polynomial, forcibly converting it to a constant if
9759 vectorizable_reduction has already enforced the condition.
9760 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
9761 to create a {1,2,3,...} vector.
9762 (vectorizable_reduction): Treat the number of units as polynomial.
9763 Choose vectype_in based on the largest scalar element size rather
9764 than the smallest number of units. Enforce the restrictions
9767 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9768 Alan Hayward <alan.hayward@arm.com>
9769 David Sherwood <david.sherwood@arm.com>
9771 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
9772 number of units as polynomial.
9774 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9775 Alan Hayward <alan.hayward@arm.com>
9776 David Sherwood <david.sherwood@arm.com>
9778 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
9779 * target.def (autovectorize_vector_sizes): Return the vector sizes
9780 by pointer, using vector_sizes rather than a bitmask.
9781 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
9782 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
9783 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
9785 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
9786 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
9787 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
9788 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
9789 * omp-general.c (omp_max_vf): Likewise.
9790 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9791 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
9792 * tree-vect-loop.c (vect_analyze_loop): Likewise.
9793 * tree-vect-slp.c (vect_slp_bb): Likewise.
9794 * doc/tm.texi: Regenerate.
9795 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
9797 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
9798 the vector size as a poly_uint64 rather than an unsigned int.
9799 (current_vector_size): Change from an unsigned int to a poly_uint64.
9800 (get_vectype_for_scalar_type): Update accordingly.
9801 * tree.h (build_truth_vector_type): Take the size and number of
9802 units as a poly_uint64 rather than an unsigned int.
9803 (build_vector_type): Add a temporary overload that takes
9804 the number of units as a poly_uint64 rather than an unsigned int.
9805 * tree.c (make_vector_type): Likewise.
9806 (build_truth_vector_type): Take the number of units as a poly_uint64
9807 rather than an unsigned int.
9809 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9810 Alan Hayward <alan.hayward@arm.com>
9811 David Sherwood <david.sherwood@arm.com>
9813 * target.def (get_mask_mode): Take the number of units and length
9814 as poly_uint64s rather than unsigned ints.
9815 * targhooks.h (default_get_mask_mode): Update accordingly.
9816 * targhooks.c (default_get_mask_mode): Likewise.
9817 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
9818 * doc/tm.texi: Regenerate.
9820 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9821 Alan Hayward <alan.hayward@arm.com>
9822 David Sherwood <david.sherwood@arm.com>
9824 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
9825 * omp-general.c (omp_max_vf): Likewise.
9826 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
9827 (expand_omp_simd): Handle polynomial safelen.
9828 * omp-low.c (omplow_simd_context): Add a default constructor.
9829 (omplow_simd_context::max_vf): Change from int to poly_uint64.
9830 (lower_rec_simd_input_clauses): Update accordingly.
9831 (lower_rec_input_clauses): Likewise.
9833 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9834 Alan Hayward <alan.hayward@arm.com>
9835 David Sherwood <david.sherwood@arm.com>
9837 * tree-vectorizer.h (vect_nunits_for_cost): New function.
9838 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
9839 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
9840 (vect_analyze_slp_cost): Likewise.
9841 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
9842 (vect_model_load_cost): Likewise.
9844 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9845 Alan Hayward <alan.hayward@arm.com>
9846 David Sherwood <david.sherwood@arm.com>
9848 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
9849 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
9850 from an unsigned int * to a poly_uint64_pod *.
9851 (calculate_unrolling_factor): New function.
9852 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
9854 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9855 Alan Hayward <alan.hayward@arm.com>
9856 David Sherwood <david.sherwood@arm.com>
9858 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
9859 from an unsigned int to a poly_uint64.
9860 (_loop_vec_info::slp_unrolling_factor): Likewise.
9861 (_loop_vec_info::vectorization_factor): Change from an int
9863 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
9864 (vect_get_num_vectors): New function.
9865 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
9866 (vect_get_num_copies): Use vect_get_num_vectors.
9867 (vect_analyze_data_ref_dependences): Change max_vf from an int *
9868 to an unsigned int *.
9869 (vect_analyze_data_refs): Change min_vf from an int * to a
9871 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9872 than an unsigned HOST_WIDE_INT.
9873 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
9874 (vect_analyze_data_ref_dependence): Change max_vf from an int *
9875 to an unsigned int *.
9876 (vect_analyze_data_ref_dependences): Likewise.
9877 (vect_compute_data_ref_alignment): Handle polynomial vf.
9878 (vect_enhance_data_refs_alignment): Likewise.
9879 (vect_prune_runtime_alias_test_list): Likewise.
9880 (vect_shift_permute_load_chain): Likewise.
9881 (vect_supportable_dr_alignment): Likewise.
9882 (dependence_distance_ge_vf): Take the vectorization factor as a
9883 poly_uint64 rather than an unsigned HOST_WIDE_INT.
9884 (vect_analyze_data_refs): Change min_vf from an int * to a
9886 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
9887 vfm1 as a poly_uint64 rather than an int. Make the same change
9888 for the returned bound_scalar.
9889 (vect_gen_vector_loop_niters): Handle polynomial vf.
9890 (vect_do_peeling): Likewise. Update call to
9891 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
9892 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
9894 * tree-vect-loop.c (vect_determine_vectorization_factor)
9895 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
9896 (vect_get_known_peeling_cost): Likewise.
9897 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
9898 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
9899 (vect_transform_loop): Likewise. Use the lowest possible VF when
9900 updating the upper bounds of the loop.
9901 (vect_min_worthwhile_factor): Make static. Return an unsigned int
9903 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
9904 polynomial unroll factors.
9905 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
9906 (vect_make_slp_decision): Likewise.
9907 (vect_supported_load_permutation_p): Likewise, and polynomial
9909 (vect_analyze_slp_cost): Handle polynomial vf.
9910 (vect_slp_analyze_node_operations): Likewise.
9911 (vect_slp_analyze_bb_1): Likewise.
9912 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9913 than an unsigned HOST_WIDE_INT.
9914 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
9915 (vectorizable_load): Handle polynomial vf.
9916 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
9918 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
9920 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9921 Alan Hayward <alan.hayward@arm.com>
9922 David Sherwood <david.sherwood@arm.com>
9924 * match.pd: Handle bit operations involving three constants
9925 and try to fold one pair.
9927 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9929 * tree-vect-loop-manip.c: Include gimple-fold.h.
9930 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
9931 niters_maybe_zero parameters. Handle other cases besides a step of 1.
9932 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
9933 Add a path that uses a step of VF instead of 1, but disable it
9935 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
9936 and niters_no_overflow parameters. Update calls to
9937 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
9938 Create a new SSA name if the latter choses to use a ste other
9939 than zero, and return it via niters_vector_mult_vf_var.
9940 * tree-vect-loop.c (vect_transform_loop): Update calls to
9941 vect_do_peeling, vect_gen_vector_loop_niters and
9942 slpeel_make_loop_iterate_ntimes.
9943 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
9944 (vect_gen_vector_loop_niters): Update declarations after above changes.
9946 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
9948 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
9949 128-bit round to integer instructions.
9950 (ceil<mode>2): Likewise.
9951 (btrunc<mode>2): Likewise.
9952 (round<mode>2): Likewise.
9954 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9956 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
9957 unaligned VSX load/store on P8/P9.
9958 (expand_block_clear): Allow the use of unaligned VSX
9959 load/store on P8/P9.
9961 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9963 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
9965 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
9966 swap associated with both a load and a store.
9968 2018-01-02 Andrew Waterman <andrew@sifive.com>
9970 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
9971 * config/riscv/riscv.md (clear_cache): Use it.
9973 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
9975 * web.c: Remove out-of-date comment.
9977 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9979 * expr.c (fixup_args_size_notes): Check that any existing
9980 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
9981 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
9982 (emit_single_push_insn): ...here.
9984 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9986 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
9987 (const_vector_encoded_nelts): New function.
9988 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
9989 (const_vector_int_elt, const_vector_elt): Declare.
9990 * emit-rtl.c (const_vector_int_elt_1): New function.
9991 (const_vector_elt): Likewise.
9992 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
9993 of CONST_VECTOR_ELT.
9995 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9997 * expr.c: Include rtx-vector-builder.h.
9998 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
9999 directly on the tree encoding.
10000 (const_vector_from_tree): Likewise.
10001 * optabs.c: Include rtx-vector-builder.h.
10002 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
10003 sequence of "u" values.
10004 * vec-perm-indices.c: Include rtx-vector-builder.h.
10005 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
10006 directly on the vec_perm_indices encoding.
10008 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10010 * doc/rtl.texi (const_vector): Describe new encoding scheme.
10011 * Makefile.in (OBJS): Add rtx-vector-builder.o.
10012 * rtx-vector-builder.h: New file.
10013 * rtx-vector-builder.c: Likewise.
10014 * rtl.h (rtx_def::u2): Add a const_vector field.
10015 (CONST_VECTOR_NPATTERNS): New macro.
10016 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
10017 (CONST_VECTOR_DUPLICATE_P): Likewise.
10018 (CONST_VECTOR_STEPPED_P): Likewise.
10019 (CONST_VECTOR_ENCODED_ELT): Likewise.
10020 (const_vec_duplicate_p): Check for a duplicated vector encoding.
10021 (unwrap_const_vec_duplicate): Likewise.
10022 (const_vec_series_p): Check for a non-duplicated vector encoding.
10023 Say that the function only returns true for integer vectors.
10024 * emit-rtl.c: Include rtx-vector-builder.h.
10025 (gen_const_vec_duplicate_1): Delete.
10026 (gen_const_vector): Call gen_const_vec_duplicate instead of
10027 gen_const_vec_duplicate_1.
10028 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
10029 (gen_const_vec_duplicate): Use rtx_vector_builder.
10030 (gen_const_vec_series): Likewise.
10031 (gen_rtx_CONST_VECTOR): Likewise.
10032 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
10033 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10034 Build a new vector rather than modifying a CONST_VECTOR in-place.
10035 (handle_special_swappables): Update call accordingly.
10036 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
10037 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10038 Build a new vector rather than modifying a CONST_VECTOR in-place.
10039 (handle_special_swappables): Update call accordingly.
10041 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10043 * simplify-rtx.c (simplify_const_binary_operation): Use
10044 CONST_VECTOR_ELT instead of XVECEXP.
10046 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10048 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
10049 the selector elements to be different from the data elements
10050 if the selector is a VECTOR_CST.
10051 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
10052 ssizetype for the selector.
10054 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10056 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
10057 before testing each element individually.
10058 * tree-vect-generic.c (lower_vec_perm): Likewise.
10060 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10062 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
10063 * selftest-run-tests.c (selftest::run_tests): Call it.
10064 * vector-builder.h (vector_builder::operator ==): New function.
10065 (vector_builder::operator !=): Likewise.
10066 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
10067 (vec_perm_indices::all_from_input_p): New function.
10068 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10069 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
10070 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
10071 instead of reading the VECTOR_CST directly. Detect whether both
10072 vector inputs are the same before constructing the vec_perm_indices,
10073 and update the number of inputs argument accordingly. Use the
10074 utility functions added above. Only construct sel2 if we need to.
10076 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10078 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
10079 the broadcast of the low byte.
10080 (expand_mult_highpart): Use an explicit encoding for the permutes.
10081 * optabs-query.c (can_mult_highpart_p): Likewise.
10082 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
10083 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10084 (vectorizable_bswap): Likewise.
10085 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
10086 explicit encoding for the power-of-2 permutes.
10087 (vect_permute_store_chain): Likewise.
10088 (vect_grouped_load_supported): Likewise.
10089 (vect_permute_load_chain): Likewise.
10091 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10093 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
10094 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
10095 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
10096 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10097 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
10098 (vect_gen_perm_mask_any): Likewise.
10100 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10102 * int-vector-builder.h: New file.
10103 * vec-perm-indices.h: Include int-vector-builder.h.
10104 (vec_perm_indices): Redefine as an int_vector_builder.
10105 (auto_vec_perm_indices): Delete.
10106 (vec_perm_builder): Redefine as a stand-alone class.
10107 (vec_perm_indices::vec_perm_indices): New function.
10108 (vec_perm_indices::clamp): Likewise.
10109 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
10110 (vec_perm_indices::new_vector): New function.
10111 (vec_perm_indices::new_expanded_vector): Update for new
10112 vec_perm_indices class.
10113 (vec_perm_indices::rotate_inputs): New function.
10114 (vec_perm_indices::all_in_range_p): Operate directly on the
10115 encoded form, without computing elided elements.
10116 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
10117 encoding. Update for new vec_perm_indices class.
10118 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
10119 the given vec_perm_builder.
10120 (expand_vec_perm_var): Update vec_perm_builder constructor.
10121 (expand_mult_highpart): Use vec_perm_builder instead of
10122 auto_vec_perm_indices.
10123 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
10124 vec_perm_indices instead of auto_vec_perm_indices. Use a single
10125 or double series encoding as appropriate.
10126 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
10127 vec_perm_indices instead of auto_vec_perm_indices.
10128 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10129 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10130 (vect_permute_store_chain): Likewise.
10131 (vect_grouped_load_supported): Likewise.
10132 (vect_permute_load_chain): Likewise.
10133 (vect_shift_permute_load_chain): Likewise.
10134 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10135 (vect_transform_slp_perm_load): Likewise.
10136 (vect_schedule_slp_instance): Likewise.
10137 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10138 (vectorizable_mask_load_store): Likewise.
10139 (vectorizable_bswap): Likewise.
10140 (vectorizable_store): Likewise.
10141 (vectorizable_load): Likewise.
10142 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
10143 vec_perm_indices instead of auto_vec_perm_indices. Use
10144 tree_to_vec_perm_builder to read the vector from a tree.
10145 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
10146 vec_perm_builder instead of a vec_perm_indices.
10147 (have_whole_vector_shift): Use vec_perm_builder and
10148 vec_perm_indices instead of auto_vec_perm_indices. Leave the
10149 truncation to calc_vec_perm_mask_for_shift.
10150 (vect_create_epilog_for_reduction): Likewise.
10151 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
10152 from auto_vec_perm_indices to vec_perm_indices.
10153 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10154 instead of changing individual elements.
10155 (aarch64_vectorize_vec_perm_const): Use new_vector to install
10156 the vector in d.perm.
10157 * config/arm/arm.c (expand_vec_perm_d::perm): Change
10158 from auto_vec_perm_indices to vec_perm_indices.
10159 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10160 instead of changing individual elements.
10161 (arm_vectorize_vec_perm_const): Use new_vector to install
10162 the vector in d.perm.
10163 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
10164 Update vec_perm_builder constructor.
10165 (rs6000_expand_interleave): Likewise.
10166 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
10167 (rs6000_expand_interleave): Likewise.
10169 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10171 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
10172 to qimode could truncate the indices.
10173 * optabs.c (expand_vec_perm_var): Likewise.
10175 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10177 * Makefile.in (OBJS): Add vec-perm-indices.o.
10178 * vec-perm-indices.h: New file.
10179 * vec-perm-indices.c: Likewise.
10180 * target.h (vec_perm_indices): Replace with a forward class
10182 (auto_vec_perm_indices): Move to vec-perm-indices.h.
10183 * optabs.h: Include vec-perm-indices.h.
10184 (expand_vec_perm): Delete.
10185 (selector_fits_mode_p, expand_vec_perm_var): Declare.
10186 (expand_vec_perm_const): Declare.
10187 * target.def (vec_perm_const_ok): Replace with...
10188 (vec_perm_const): ...this new hook.
10189 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
10190 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
10191 * doc/tm.texi: Regenerate.
10192 * optabs.def (vec_perm_const): Delete.
10193 * doc/md.texi (vec_perm_const): Likewise.
10194 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
10195 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
10196 expand_vec_perm for constant permutation vectors. Assert that
10197 the mode of variable permutation vectors is the integer equivalent
10198 of the mode that is being permuted.
10199 * optabs-query.h (selector_fits_mode_p): Declare.
10200 * optabs-query.c: Include vec-perm-indices.h.
10201 (selector_fits_mode_p): New function.
10202 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
10203 is defined, instead of checking whether the vec_perm_const_optab
10204 exists. Use targetm.vectorize.vec_perm_const instead of
10205 targetm.vectorize.vec_perm_const_ok. Check whether the indices
10206 fit in the vector mode before using a variable permute.
10207 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
10208 vec_perm_indices instead of an rtx.
10209 (expand_vec_perm): Replace with...
10210 (expand_vec_perm_const): ...this new function. Take the selector
10211 as a vec_perm_indices rather than an rtx. Also take the mode of
10212 the selector. Update call to shift_amt_for_vec_perm_mask.
10213 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
10214 Use vec_perm_indices::new_expanded_vector to expand the original
10215 selector into bytes. Check whether the indices fit in the vector
10216 mode before using a variable permute.
10217 (expand_vec_perm_var): Make global.
10218 (expand_mult_highpart): Use expand_vec_perm_const.
10219 * fold-const.c: Includes vec-perm-indices.h.
10220 * tree-ssa-forwprop.c: Likewise.
10221 * tree-vect-data-refs.c: Likewise.
10222 * tree-vect-generic.c: Likewise.
10223 * tree-vect-loop.c: Likewise.
10224 * tree-vect-slp.c: Likewise.
10225 * tree-vect-stmts.c: Likewise.
10226 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
10228 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
10229 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
10230 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
10231 (aarch64_vectorize_vec_perm_const): ...this new function.
10232 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10233 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10234 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
10235 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
10236 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10237 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10238 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
10240 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
10241 check for NEON modes.
10242 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
10243 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
10244 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
10245 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
10247 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
10248 the old VEC_PERM_CONST conditions.
10249 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
10250 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
10251 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
10252 (ia64_vectorize_vec_perm_const_ok): Merge into...
10253 (ia64_vectorize_vec_perm_const): ...this new function.
10254 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
10255 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
10256 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
10257 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
10258 * config/mips/mips.c (mips_expand_vec_perm_const)
10259 (mips_vectorize_vec_perm_const_ok): Merge into...
10260 (mips_vectorize_vec_perm_const): ...this new function.
10261 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
10262 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
10263 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
10264 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
10265 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
10266 (rs6000_expand_vec_perm_const): Delete.
10267 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
10269 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10270 (altivec_expand_vec_perm_const_le): Take each operand individually.
10271 Operate on constant selectors rather than rtxes.
10272 (altivec_expand_vec_perm_const): Likewise. Update call to
10273 altivec_expand_vec_perm_const_le.
10274 (rs6000_expand_vec_perm_const): Delete.
10275 (rs6000_vectorize_vec_perm_const_ok): Delete.
10276 (rs6000_vectorize_vec_perm_const): New function.
10277 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10278 an element count and rtx array.
10279 (rs6000_expand_extract_even): Update call accordingly.
10280 (rs6000_expand_interleave): Likewise.
10281 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
10282 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
10283 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
10284 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
10285 (rs6000_expand_vec_perm_const): Delete.
10286 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10287 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10288 (altivec_expand_vec_perm_const_le): Take each operand individually.
10289 Operate on constant selectors rather than rtxes.
10290 (altivec_expand_vec_perm_const): Likewise. Update call to
10291 altivec_expand_vec_perm_const_le.
10292 (rs6000_expand_vec_perm_const): Delete.
10293 (rs6000_vectorize_vec_perm_const_ok): Delete.
10294 (rs6000_vectorize_vec_perm_const): New function. Remove stray
10295 reference to the SPE evmerge intructions.
10296 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10297 an element count and rtx array.
10298 (rs6000_expand_extract_even): Update call accordingly.
10299 (rs6000_expand_interleave): Likewise.
10300 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
10301 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
10303 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10305 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10307 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
10308 vector mode and that that mode matches the mode of the data
10310 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
10311 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
10312 directly using expand_vec_perm_1 when forcing selectors into
10314 (expand_vec_perm_var): New function, split out from expand_vec_perm.
10316 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10318 * optabs-query.h (can_vec_perm_p): Delete.
10319 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
10320 * optabs-query.c (can_vec_perm_p): Split into...
10321 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
10322 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
10323 particular selector is valid.
10324 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10325 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10326 (vect_grouped_load_supported): Likewise.
10327 (vect_shift_permute_load_chain): Likewise.
10328 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10329 (vect_transform_slp_perm_load): Likewise.
10330 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10331 (vectorizable_bswap): Likewise.
10332 (vect_gen_perm_mask_checked): Likewise.
10333 * fold-const.c (fold_ternary_loc): Likewise. Don't take
10334 implementations of variable permutation vectors into account
10335 when deciding which selector to use.
10336 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
10337 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
10338 with a false third argument.
10339 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
10340 to test whether the constant selector is valid and can_vec_perm_var_p
10341 to test whether a variable selector is valid.
10343 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10345 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
10346 * optabs-query.c (can_vec_perm_p): Likewise.
10347 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
10348 instead of vec_perm_indices.
10349 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
10350 (vect_gen_perm_mask_checked): Likewise,
10351 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
10352 (vect_gen_perm_mask_checked): Likewise,
10354 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10356 * optabs-query.h (qimode_for_vec_perm): Declare.
10357 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
10358 (qimode_for_vec_perm): ...this new function.
10359 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
10361 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10363 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
10364 does not have a conditional at the top.
10366 2018-01-02 Richard Biener <rguenther@suse.de>
10368 * ipa-inline.c (big_speedup_p): Fix expression.
10370 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10373 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
10376 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10380 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
10381 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
10382 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
10383 cond_taken_branch_cost 3->4.
10385 2018-01-01 Jakub Jelinek <jakub@redhat.com>
10387 PR tree-optimization/83581
10388 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
10389 TODO_cleanup_cfg if any changes have been made.
10391 PR middle-end/83608
10392 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
10393 convert_modes if target mode has the right side, but different mode
10396 PR middle-end/83609
10397 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
10398 last argument when extracting from CONCAT. If either from_real or
10399 from_imag is NULL, use expansion through memory. If result is not
10400 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
10401 the parts directly to inner mode, if even that fails, use expansion
10404 PR middle-end/83623
10405 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
10406 check for bswap in mode rather than HImode and use that in expand_unop
10409 Copyright (C) 2018 Free Software Foundation, Inc.
10411 Copying and distribution of this file, with or without modification,
10412 are permitted in any medium without royalty provided the copyright
10413 notice and this notice are preserved.